CN119051047B

CN119051047B - An automatic voltage control method based on active power trend determination

Info

Publication number: CN119051047B
Application number: CN202411229144.4A
Authority: CN
Inventors: 曹珲茹; 张俊; 尹文静; 罗易刚; 刘欣洁; 王艳萍; 洪小雁; 梅响
Original assignee: Wuhan Huayuan Electric Power Design Institute Co ltd
Current assignee: Wuhan Huayuan Electric Power Design Institute Co ltd
Priority date: 2024-09-03
Filing date: 2024-09-03
Publication date: 2025-03-07
Anticipated expiration: 2044-09-03
Also published as: CN119051047A

Abstract

The invention relates to the technical field of voltage control, in particular to an automatic voltage control method based on active trend judgment. The method comprises the steps of collecting multi-scale monitoring data according to a power grid management system to generate multi-scale power matrix data, predicting new energy output power according to the multi-scale power matrix data to generate new energy power prediction data, performing space-time node similarity processing according to the multi-scale power matrix data, performing node voltage change trend analysis to obtain node voltage trend data, performing power grid voltage stability margin assessment according to the new energy power prediction data and the node voltage trend data to generate voltage stability margin assessment data, and performing intelligent voltage adjustment instruction generation based on the voltage stability margin assessment data to obtain intelligent voltage adjustment instruction data. According to the invention, the automatic control of the power grid voltage is realized through the power grid voltage stability margin evaluation, and the problem of voltage stability caused by new energy output fluctuation is effectively solved.

Description

Automatic voltage control method based on active trend determination

Technical Field

The invention relates to the technical field of voltage control, in particular to an automatic voltage control method based on active trend judgment.

Background

With the ever-increasing scale of power systems and the ever-increasing load, the voltage control problem of power systems has become increasingly complex and important. The voltage control aims to maintain the voltage of each node of the power system within an allowable range, ensure the safe and stable operation of the power system and improve the power supply quality. The duty ratio of distributed new energy sources (such as solar energy, wind energy and the like) in the power system is gradually increased, so that the diversification of power production is promoted, but the voltage control is more complicated due to the uncertainty and fluctuation of the power generation characteristics. The access of the distributed new energy source causes the voltage level of the power grid to obviously fluctuate at different times and places, and the problems of voltage overrun, equipment overload and the like can be caused. Therefore, how to effectively perform voltage control and ensure safe and stable operation of the power system becomes a technical problem to be solved. However, conventional voltage control methods rely primarily on local voltage deviations to regulate, for example, when a node voltage is below a set point, by regulating reactive compensation devices (e.g., shunt capacitors, reactors) or generator excitation near the node to boost the voltage. The method cannot cope with voltage fluctuation caused by transient disturbance, and under the condition of high new energy grid-connected proportion, the method is difficult to meet the requirement of real-time response, so that the control effect is poor.

Disclosure of Invention

Based on the above, the present invention provides an automatic voltage control method based on active trend determination, so as to solve at least one of the above technical problems.

To achieve the above object, an automatic voltage control method based on active trend determination includes the steps of:

step S1, carrying out power grid topological structure processing according to a power grid management system to generate power grid topological structure data;

Step S2, extracting new energy node states of the multi-scale power matrix data to obtain new energy monitoring state data;

Step S3, line impedance analysis is carried out according to the power grid topological structure data to obtain power grid impedance topological structure data;

Step S4, carrying out voltage change sensitive node identification on the space-time similarity vector data through the power grid impedance topological data to generate key voltage sensitive node data;

step S5, carrying out grid voltage stability margin assessment according to the new energy power prediction data and the node voltage trend data to generate voltage stability margin assessment data;

And S6, generating an intelligent voltage regulation command according to the node voltage control signal data to obtain intelligent voltage regulation command data.

The invention can clearly draw the structure of the power grid and the connection relation between each node through processing the topological structure of the power grid under the support of the power grid management system. The multi-scale monitoring data acquisition is carried out based on the power grid topological structure data, so that the power data from different layers and areas can be comprehensively collected, the acquisition of the scale monitoring data covers different time and space dimensions of the power grid, and comprehensive power matrix data is formed. The new energy node state is extracted from the multi-scale power matrix data, so that the influence of the energy with large fluctuation on the power grid can be independently analyzed and understood, and the future new energy power generation amount can be accurately predicted. By performing space-time node similarity processing on the multi-scale power matrix data, the running state similarity of each node in the power grid in different time periods can be revealed, and the identification of potential risk areas and nodes is facilitated. The voltage change sensitive node identification is carried out on the space-time similarity vector data through the power grid impedance topological data, so that key nodes causing voltage instability in the power grid can be effectively identified, and the power grid is ensured to maintain a stable voltage level in the operation process. And carrying out power grid voltage stability margin assessment according to the new energy power prediction data and the node voltage trend data, providing safety margin for power grid operation, ensuring that the power grid can maintain normal operation under emergency conditions, and providing timely voltage regulation instructions for the power grid. According to the node voltage control signal data, intelligent voltage adjustment instruction generation is carried out, automatic adjustment of the power grid voltage can be achieved, automatic and intelligent management capacity of the power grid is improved, and efficient utilization and consumption of renewable energy sources are further promoted. Therefore, the automatic voltage control method based on active trend judgment provided by the invention has the advantages that the key nodes sensitive to voltage change and the active power change trend thereof are identified by constructing multi-scale power matrix data and fusing space-time associated information, and voltage stability margin evaluation is carried out by combining new energy power prediction, so that a prospective voltage control strategy is generated, voltage fluctuation caused by transient disturbance is effectively predicted and dealt with, cooperative control is carried out on new energy output fluctuation, the instantaneity and the effectiveness of voltage control are finally improved, and the problem of poor control effect under a new energy high permeability scene is solved.

Preferably, step S1 comprises the steps of:

Step S11, monitoring power node analysis is carried out according to a power grid management system to obtain monitoring power node data, wherein the monitoring power node data comprises new energy node data and power load node data;

Step S12, carrying out power grid topological structure processing based on the monitored power node data to generate power grid topological structure data;

step S13, performing time scale association analysis according to the power grid topological structure data, and performing self-adaptive time scale selection to obtain multi-scale time window data;

Step S14, multi-scale monitoring data acquisition is carried out on the monitoring power node data through multi-scale time window data to obtain multi-scale power monitoring data;

step S15, carrying out data cleaning processing on the multi-scale power monitoring data, and extracting monitoring power characteristics to generate monitoring node power characteristic data;

And S16, performing monitoring state matrix reconstruction on the monitoring node power characteristic data by using the multi-scale monitoring vector data to generate multi-scale power matrix data.

According to the invention, the monitoring power node analysis is carried out according to the power grid management system, so that the state of each monitoring node in the power grid can be identified, and the relationship between new energy power generation and power load can be identified. The power grid topological structure is processed based on the monitored power node data, so that an actual operation network of the power grid can be constructed, and the physical characteristics of the power grid can be reflected. And carrying out time scale association analysis according to the power grid topological structure data, and identifying the relation between different time scales in the power grid operation. The process of adaptive time scale selection ensures that the analysis is able to select the most appropriate time window according to the actual operating situation. The multi-scale monitoring data acquisition is carried out on the monitoring power node data through the multi-scale time window data, so that the comprehensive monitoring of the running state of the power grid can be realized. By monitoring the power characteristic extraction, important characteristic information in the operation of the power grid can be identified, including power load change trend, new energy power generation characteristics and the like. And the multi-scale monitoring vector data is utilized to reconstruct the monitoring state matrix of the monitoring node power characteristic data, so that the power characteristic information under different time scales can be effectively integrated.

Preferably, step S2 comprises the steps of:

S21, extracting a new energy node state according to the multi-scale power matrix data to obtain new energy monitoring state data;

s22, carrying out multi-scale time sequence feature fusion according to the new energy monitoring state data to generate time sequence monitoring feature fusion data;

step S23, historical power generation calculation is carried out based on the time sequence monitoring characteristic fusion data, and multi-scale power generation data are obtained;

S24, performing space-time new energy output analysis on the time sequence monitoring characteristic fusion data through the multi-scale power generation data to generate new energy output related data;

step S25, predicting the long-term power and the short-term power of the new energy according to the new energy output related data and the power grid topological structure data, and respectively generating short-term power prediction data and long-term power prediction data;

And S26, performing multi-time scale integration processing on the short-term power prediction data and the long-term power prediction data to generate new energy power prediction data.

According to the method, the state extraction of the new energy nodes is carried out according to the multi-scale power matrix data, the operation condition of the new energy nodes can be effectively identified and monitored, and the access and operation of the new energy can be matched with the power grid requirements. And carrying out multi-scale time sequence feature fusion according to the new energy monitoring state data, so that feature information under different time scales can be effectively integrated. Historical power generation calculation is performed based on the time sequence monitoring feature fusion data, so that the power generation capacity of the new energy node in the past time period can be accurately estimated. And the change rule of new energy power generation under different time and space conditions can be revealed by carrying out space-time new energy output analysis on the time-sequence monitoring characteristic fusion data through the multi-scale power generation data. And the new energy long-term and short-term power prediction is carried out according to the new energy output related data and the power grid topological structure data, the short-term power prediction can provide real-time power generation capacity prediction for power grid dispatching, and the long-term power prediction provides prospective reference for power grid planning and strategy formulation. The short-term power prediction data and the long-term power prediction data are integrated in a multi-time scale, prediction results in different time scales can be effectively combined, and the accuracy and the practicability of prediction are improved.

Preferably, step S25 comprises the steps of:

S251, constructing a multiple regression model of the new energy output related data by utilizing a gradient descent algorithm to obtain a short-term new energy output prediction model;

step S252, model training is carried out on the short-term new energy output prediction model through the multi-scale power generation data, and an optimized short-term new energy prediction model is generated;

step S253, short-term energy power prediction is carried out on the time sequence monitoring characteristic fusion data by utilizing an optimized short-term new energy prediction model, and short-term power prediction data are generated;

step S254, obtaining new energy node weather forecast based on the power grid topological structure data to obtain weather forecast data;

S255, carrying out influence factor analysis on the new energy output related data through weather forecast data to generate weather influence factor data;

step S256, performing migration learning on weather effect factor data and time sequence monitoring feature fusion data by using a preset long-short-term neural network model, and constructing a long-term new energy prediction model;

and S257, carrying out long-term energy power prediction on the short-term power prediction data and the weather influence factor data through a long-term new energy prediction model to generate long-term power prediction data.

The invention utilizes the gradient descent algorithm to construct the multiple regression model of the new energy output associated data, and can effectively capture the relation between the new energy output and related factors. By establishing a short-term new energy output prediction model, accurate short-term power generation capacity prediction can be provided for power grid dispatching. And short-term energy power prediction is carried out on the time-series monitoring characteristic fusion data by using an optimized short-term new energy prediction model, so that the power generation capacity change of the new energy can be reflected in real time. And the new energy node weather forecast is acquired based on the power grid topological structure data, so that weather information, such as wind speed, temperature, radiation and the like, closely related to new energy power generation can be collected. And the influence factor analysis is carried out on the new energy output related data through the weather forecast data, so that the influence rule of weather change on the new energy output can be revealed. And the existing data resources are utilized, so that the prediction capability and the application range of the model are improved. The constructed long-term new energy prediction model not only can consider short-term fluctuation, but also can reasonably predict long-term trend. The short-term power prediction data and the weather influence factor data are subjected to long-term energy power prediction, so that the power generation capacity prediction of a long time period in the future can be provided for the power grid.

Preferably, step S3 comprises the steps of:

s31, extracting node geographic coordinates according to the power grid topological structure data to generate power grid node coordinate data;

S32, performing inter-node distance matrix processing according to the grid node coordinate data to generate inter-node distance matrix data;

Step S33, line impedance matrix processing is carried out according to the power grid topological structure data, and line impedance matrix data and power grid impedance topological data are generated;

step S34, calculating a spatial correlation coefficient of the line impedance matrix data through the inter-node distance matrix data to generate spatial correlation matrix data;

step S35, performing time sequence mutual information calculation on the multi-scale power matrix data to obtain time correlation matrix data;

step S36, carrying out space-time correlation fusion according to the space-time correlation matrix data and the time correlation matrix data to generate space-time correlation tensor data;

And step S37, performing space-time weighted similarity processing on the multi-scale power matrix data through the space-time correlation tensor data to generate space-time similarity vector data.

According to the method, the node geographic coordinates are extracted according to the power grid topological structure data, the specific positions of all the nodes in the power grid can be obtained, and the physical relations among the nodes can be identified. The distance relation among all nodes of the power grid can be calculated by carrying out the distance matrix processing according to the coordinate data of the nodes of the power grid, so that the relative position and the connectivity among all nodes in the power grid can be recognized, and the power grid dispatching and the optimization can be assisted. The circuit impedance matrix processing is carried out according to the topological structure data of the power grid, so that the influence of each circuit in the power grid on the current can be revealed. And the space correlation coefficient calculation is carried out on the line impedance matrix data through the inter-node distance matrix data, so that the electric characteristics of which nodes in the power grid are influenced by the space distance can be identified. And the time sequence mutual information calculation is carried out on the multi-scale power matrix data, so that the time characteristics of the power load and the power generation capacity can be effectively identified. And carrying out space-time correlation fusion according to the space correlation matrix data and the time correlation matrix data, effectively combining the information of two dimensions of space and time, and comprehensively considering the interaction of the power grid state in time and space. The spatial-temporal weighted similarity processing is carried out on the multi-scale power matrix data through the spatial-temporal correlation tensor data, so that the similarity of each node in the power grid in space and time can be quantified.

Preferably, step S33 includes the steps of:

step S331, extracting power line parameters from the power grid topological structure data through a power grid management system to obtain power line parameter data;

step S332, constructing a distributed circuit model according to the power circuit parameter data to obtain electric circuit model data;

S333, performing impedance frequency simulation on the electric circuit model data to generate impedance frequency response data;

Step S334, line impedance attribute mapping is carried out on the power grid topological structure data through impedance frequency response data, and power grid impedance topological data is generated;

Step S335, carrying out inter-node harmonic impedance calculation according to the power grid impedance topology data to generate line harmonic impedance data;

step S336, performing impedance sensitivity analysis on the power line parameter data by using the line harmonic impedance data to generate line impedance matrix data.

According to the invention, the power line parameter extraction is carried out on the power grid topological structure data through the power grid management system, so that the basic physical characteristics and parameter information of the power line can be collected. And constructing a distributed circuit model according to the power circuit parameter data, and converting the physical characteristics of the power circuit into electric model data. Impedance frequency simulation is carried out on the electric circuit model data, so that impedance characteristics of the electric circuit under different frequencies can be analyzed, frequency response characteristics of the circuit can be revealed, and harmonic waves and unstable problems under different frequency conditions can be recognized. The impedance frequency response data is used for carrying out line impedance attribute mapping on the power grid topological structure data, so that detailed description can be provided for the overall impedance characteristic of the power grid, and the electrical relation between each node and each line in the power grid is revealed. And calculating the harmonic impedance among nodes according to the power grid impedance topology data, and evaluating the response capability of each node in the power grid to the harmonic current. The power line parameter data is subjected to impedance sensitivity analysis by utilizing the line harmonic impedance data, so that the sensitivity of the power line to different harmonic frequencies can be evaluated, and the stability and reliability of the power grid in the face of distributed new energy access and active trend variation are ensured.

Preferably, step S4 comprises the steps of:

step S41, voltage change sensitive node identification is carried out on space-time similarity vector data through power grid impedance topological data, and key voltage sensitive node data are generated;

step S42, calculating the active power change rate of the key voltage sensitive node data to generate active power change rate data;

Step S43, calculating the change amplitude according to the change rate data of the active power to generate the change amplitude data of the active power;

s44, carrying out trend change labeling on the active power change amplitude data through preset trend evolution label data to generate active trend evolution label data;

step S45, real-time active trend mode classification is carried out according to the active trend evolution label data, voltage influence degree evaluation is carried out, and active trend influence factor data are generated;

And S46, carrying out node voltage change trend analysis on the key voltage sensitive node data through the active trend influence factor data to obtain node voltage trend data.

According to the invention, voltage change sensitive nodes are identified on space-time similarity vector data through power grid impedance topological data, so that key nodes sensitive to voltage change reaction can be effectively identified, and timely response and adjustment can be ensured when voltage fluctuates. The active power change rate calculation is performed on the key voltage sensitive node data, so that the power change rate of the sensitive nodes can be quantified. And calculating the change amplitude according to the data of the change rate of the active power, and reflecting the response degree of the power grid when the load or the power generation amount is changed severely. The trend change marking is carried out on the active power change amplitude data through the preset trend evolution label data, so that the power changes with different amplitudes can be classified. And evaluating the real-time mode of active power change in the power grid, and identifying factors with larger influence degree on voltage change. And carrying out node voltage change trend analysis on the key voltage sensitive node data through the active trend influence factor data, and directly indicating the voltage change directions of the nodes in the future.

Preferably, step S41 comprises the steps of:

Step S411, performing Z-score standardization processing on the space-time similarity vector data to generate standard space-time similarity vector data, and performing similarity threshold processing according to the standard space-time similarity vector data to obtain node similarity threshold data;

step S412, performing similar characteristic node grouping processing according to the space-time similarity vector data, and performing potential sensitive node screening according to the node similarity threshold data to obtain potential power grid sensitive node data;

step S413, calculating voltage change rate of the potential power grid sensitive node data to obtain sensitive node voltage rate data;

Step S414, carrying out topology importance assessment on the potential grid sensitive node data by utilizing the grid impedance topology data to generate topology importance assessment data;

step S415, dynamic response characteristic analysis is carried out on the potential power grid sensitive node data, and dynamic response characteristic data are generated;

Step S416, performing multi-criterion comprehensive evaluation processing on the potential power grid sensitive node data based on the sensitive node voltage rate data, the topology importance evaluation data and the dynamic response characteristic data to generate multi-criterion scoring data;

And S417, performing sensitive node verification processing on the potential power grid sensitive node data through the multi-criterion scoring data, and performing key voltage sensitive node screening to obtain key voltage sensitive node data.

According to the invention, the Z-score standardization processing is carried out on the time space similarity vector data, so that the dimensional influence among different features can be eliminated, and each feature is compared under the same standard. The node which causes the unstable power grid under the conditions of voltage change and load fluctuation can be effectively identified through grouping and screening by performing similar characteristic node grouping processing according to the space-time similarity vector data, so that the potential sensitive nodes can be intensively monitored and regulated in power grid management, and the safety of the power grid is improved. The voltage change rate of the potential power grid sensitive node data is calculated, and the voltage fluctuation condition of the nodes in the power grid operation process can be quantified. The topology importance of the potential power grid sensitive node data is evaluated by utilizing the power grid impedance topology data, so that the relative importance of the nodes in the power grid can be evaluated, and the influence of the nodes on the overall performance of the power grid is revealed. Dynamic response characteristic analysis is carried out on the data of the potential power grid sensitive nodes, and the response characteristics of the nodes under different running conditions can be revealed. The response characteristics of the nodes under different running conditions can be revealed, and key nodes with great influence on the stability of the power grid can be accurately identified in a complex power grid environment. The sensitive node verification processing is carried out on the potential power grid sensitive node data through the multi-criterion scoring data, so that the finally confirmed sensitive node can be ensured to have higher voltage change risk and importance, and effective monitoring and regulation objects are provided for power grid management.

Preferably, step S5 comprises the steps of:

step S51, carrying out grid voltage stability margin assessment according to new energy power prediction data and node voltage trend data, and generating voltage stability margin assessment data;

Step S52, performing control threshold dynamic adjustment on a preset initial voltage control threshold through voltage stability margin evaluation data to obtain a dynamic voltage control threshold;

Step S53, voltage fluctuation analysis is carried out on the node voltage trend data by utilizing the voltage stability margin evaluation data, and voltage deviation pre-judgment value calculation is carried out, so as to generate a voltage deviation pre-judgment value;

step S54, voltage regulation target value calculation is carried out according to the voltage deviation pre-judgment value, and voltage regulation target value data are obtained;

and step S55, voltage control signal generation is carried out on the voltage regulation target value data through the dynamic voltage control threshold value, and node voltage control signal data is generated.

According to the invention, the power grid voltage stability margin is evaluated according to the new energy power prediction data and the node voltage trend data, so that the fluctuation of new energy power generation and the change condition of node voltage can be comprehensively considered, and the power grid can maintain good voltage level under the scene of large-scale access of new energy, and system faults caused by unstable voltage are prevented. And (3) according to a real-time voltage stability evaluation result, timely adjusting a control threshold value to adapt to the change of the running condition of the power grid. The voltage stability margin evaluation data is utilized to carry out voltage fluctuation analysis on the node voltage trend data, so that the deviation condition of future voltage can be quantitatively predicted, and the method is beneficial to formulating corresponding regulation and control strategies in advance under the condition of aggravation of voltage fluctuation, so that the power grid can rapidly respond when facing emergency, and the voltage stability is maintained. The voltage regulation target value is calculated according to the voltage deviation pre-judging value, a clear voltage regulation target can be provided for power grid dispatching, and a reasonable regulation scheme can be formulated when the voltage deviation occurs. The calculated voltage regulation target value is converted into an actual control instruction, so that the control instruction is ensured to meet the safety operation requirement, the change of the power grid state can be responded rapidly, and the response speed and the control precision of the whole automatic voltage control system are improved.

Preferably, step S6 comprises the steps of:

step S61, obtaining distribution data of controllable resources of a power grid;

Step S62, carrying out regional voltage control target decomposition on the power grid controllable resource distribution data through the power grid impedance topology data to generate regional voltage control resource data;

step S63, carrying out regional controllable resource optimization scheduling on node voltage control signal data by utilizing regional voltage control resource data, and generating a new energy cooperative control strategy to generate an automatic voltage control strategy;

and S64, generating an intelligent voltage regulation command according to the automatic voltage control strategy to obtain intelligent voltage regulation command data.

The method and the system can comprehensively know the specific position and the capability of the controllable resources in the power grid by acquiring the distribution data of the controllable resources of the power grid. And decomposing the regional voltage control targets of the power grid controllable resource distribution data through the power grid impedance topology data, and refining the overall voltage control targets into specific control targets of all regions so as to ensure that the voltage control strategy has more pertinence and operability. The regional voltage control resource data is utilized to perform regional controllable resource optimization scheduling on the node voltage control signal data, so that reasonable allocation of controllable resources can be realized, and the voltage level of a power grid can be effectively maintained under the condition of new energy power generation fluctuation. The intelligent voltage regulation command is generated according to the automatic voltage control strategy, so that the formulated voltage control strategy can be converted into a specific operation command, and the power grid can respond to the voltage change in real time.

Drawings

FIG. 1 is a flow chart of steps of an automatic voltage control method based on active trend determination according to the present invention;

FIG. 2 is a flowchart illustrating the detailed implementation of step S2 in FIG. 1;

FIG. 3 is a flowchart illustrating the detailed implementation of step S5 in FIG. 1;

The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The following is a clear and complete description of the technical method of the present invention, taken in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

In order to achieve the above objective, referring to fig. 1 to 3, the present invention provides an automatic voltage control method based on active trend determination, comprising the following steps:

In the embodiment of the present invention, as described with reference to fig. 1, a step flow diagram of an automatic voltage control method based on active trend determination according to the present invention is provided, and in the embodiment, the automatic voltage control method based on active trend determination includes the following steps:

In the embodiment of the invention, a power grid management system (such as a SCADA system) is utilized to acquire real-time power grid equipment state information, including a breaker state, a disconnecting switch state and the like. Then, based on the obtained equipment state information and pre-stored power grid basic data (such as line parameters, transformer parameters and the like), a power grid topological structure diagram is constructed by utilizing a graph theory method, and is converted into a data format for storage, such as an adjacent matrix or a node-branch correlation matrix, so that the power grid topological structure diagram is the power grid topological structure data. Then, based on the constructed power grid topological structure, a multi-scale power sensor network (such as a smart meter, a miniature synchronous phasor measurement unit and the like) is utilized to collect power data such as voltage, current, power and the like of different time scales (such as seconds, minutes and hours), and the power data is organized into a multi-dimensional matrix form according to the collection time and node information, namely the multi-scale power matrix data.

In the embodiment of the invention, the corresponding positions of new energy nodes (such as photovoltaic power stations, wind power plants and the like) in the multi-scale power matrix data are identified according to the power grid topological structure data, and the information of the voltages, currents, power and the like of the nodes is extracted to construct new energy monitoring state data. Then, using the extracted new energy monitoring state data, combining with meteorological data (such as irradiance, wind speed, etc.) and historical output data, selecting a suitable new energy output prediction model, for example, for photovoltaic output prediction, a neural network model based on the historical data and weather forecast may be selected, and for wind power output prediction, a physical model based on numerical weather forecast and wind turbine generator characteristics may be selected. And generating new energy power prediction data of different time scales in a future period by training and testing the selected prediction model.

In the embodiment of the invention, according to the power grid topological structure data and the line parameter information generated in the step S1, the line impedance is calculated by using a circuit analysis method (such as a node voltage method, a loop current method and the like), and the calculation result is associated with the power grid topological structure data to generate the power grid impedance topological structure data, for example, the resistance and reactance value of each line can be stored in a matrix form and correspond to the line information in the power grid topological structure data. And then, calculating the similarity of power data such as voltage, current, power and the like of different nodes in the multi-scale power matrix data by using a Dynamic Time Warping (DTW) algorithm or other similarity measurement method based on the power grid topological structure data, and constructing a calculation result into space-time similarity vector data, wherein each element represents the similarity of two nodes in a specific time period.

In the embodiment of the invention, the key nodes which are sensitive to voltage change are identified by using a sensitivity analysis method (such as node importance index, network betweenness centrality and the like) based on graph theory by combining the power grid impedance topological data and the space-time similarity vector data, so that key voltage sensitive node data are generated. For example, a node having a greater impact on the propagation of voltage fluctuations may be identified by calculating the electrical distance of each node to other nodes, and combining the spatio-temporal similarity vector data of the nodes. Then, according to the key voltage sensitive node data, extracting voltage time sequence data of a corresponding node in the multi-scale power matrix data, analyzing the voltage change trend by using a time sequence analysis method (such as a moving average method, an exponential smoothing method, an ARIMA model and the like), and generating node voltage trend data, for example, the voltage change trend of the key node in a future period of time including the trend of voltage rising, falling or keeping stable can be predicted.

According to the embodiment of the invention, according to the new energy power prediction data and the node voltage trend data, the power grid voltage stability margin is evaluated by utilizing a static or dynamic voltage stability analysis method (such as P-V curve analysis, continuous power flow calculation and the like) in combination with power grid operation constraint conditions (such as upper and lower voltage limits, line transmission capacity and the like), voltage stability margin evaluation data is generated, and for example, voltage stability margin indexes of each node under different new energy output scenes, such as the difference value between a voltage collapse critical point and the current voltage, can be calculated. Then, based on the voltage stability margin evaluation data, a voltage control strategy is designed, for example, node voltage control signal data can be generated by adopting methods such as fuzzy control, predictive control and the like according to the size and the change trend of the voltage stability margin, for example, a voltage regulation signal which is pre-controlled in advance can be generated according to the reduction speed of the voltage stability margin, so that voltage out-of-limit is avoided.

According to the embodiment of the invention, specific intelligent voltage regulation instruction data is generated according to node voltage control signal data and combining the actual running state of the power grid and the regulation characteristics (such as response speed, regulation precision and the like) of the control equipment, for example, a switching instruction of reactive compensation equipment (such as a capacitor bank, a static var compensator and the like) or a regulation instruction of an on-load voltage regulating transformer, flexible alternating current transmission equipment and the like can be generated according to the size and the direction of the node voltage control signal, so that accurate control of the power grid voltage is realized, for example, an instruction of switching or cutting off a specific capacitor bank or an instruction of gear regulation of the on-load voltage regulating transformer can be generated according to the node voltage control signal, and finally, automatic control of the voltage in a distributed new energy consumption scene is realized.

Preferably, step S1 comprises the steps of:

In the embodiment of the invention, the real-time information of each power node in the power grid is obtained by carrying out data interaction with the power grid management system (such as the SCADA system, the power distribution management system and the like). And constructing a power grid topological structure diagram by combining power grid Geographic Information System (GIS) data and power system network parameter information through a graph theory method. First, each monitoring power node is mapped into a geographic space, and a node-branch association matrix is constructed according to the connection relation between the nodes. Each branch is then given a corresponding electrical parameter, such as resistance, reactance, line length, etc., in combination with the line parameter information. And finally, integrating the node-branch connection matrix and the line parameter information to form complete power grid topological structure data. According to the operation characteristics and analysis targets of the power system, a plurality of time scales are preset, such as a second level, a minute level, an hour level and the like, a time scale combination capable of effectively reflecting dynamic characteristics of the power system is selected, the size of a time window corresponding to each time scale is determined, multi-scale monitoring data acquisition is carried out on monitoring power node data through multi-scale time window data, each time scale comprises data of past T moments, for example, the second level time scale comprises data of past 10 seconds, the minute level time scale comprises data of past 10 minutes, and the hour level time scale comprises data of past 10 hours. And arranging the power data of each node under different time scales according to a time sequence, and constructing multi-scale monitoring vector data. And performing data cleaning processing on the multi-scale power monitoring data, wherein the data cleaning processing comprises missing value processing, abnormal value detection and processing and the like. The cleaned data is subjected to feature extraction, so that statistical features such as average value, standard deviation, peak-valley difference, power factor and the like of the power load can be extracted, and frequency features, waveform features and the like of the power load can be extracted by utilizing signal processing methods such as wavelet transformation, empirical mode decomposition and the like. And combining the extracted multiple characteristics to construct monitoring node power characteristic data. And respectively arranging the multi-scale monitoring vector data and the monitoring node power characteristic data under each time scale into a matrix form, extracting main characteristic vectors by using a PCA method, and projecting the original data into a characteristic vector space to obtain a low-dimensional characteristic matrix. And finally, arranging the feature matrixes under different time scales according to a time sequence to construct multi-scale power matrix data.

As an example of the present invention, referring to fig. 2, a detailed implementation step flow diagram of step S2 in fig. 1 is shown, where step S2 includes:

In the embodiment of the invention, the corresponding positions of new energy nodes (such as photovoltaic power stations, wind power plants and the like) in the multi-scale power matrix data are identified according to the power grid topological structure data. Power data, such as active power, reactive power, voltage, current, etc., at the nodes at different time scales, and other relevant information, such as irradiance of the photovoltaic power plant, wind speed of the wind farm, etc., are then extracted. And integrating the extracted data to form new energy monitoring state data.

In the embodiment of the invention, a multi-scale time sequence feature fusion method is adopted, for example, a method based on deep learning, such as a combined model of a Convolutional Neural Network (CNN) and a long-short-term memory network (LSTM), can be used. Firstly, the spatial characteristics of new energy monitoring state data under different time scales are extracted by utilizing a CNN network, for example, data of minute level, hour level and day level can be respectively input into different CNN branches for characteristic extraction. And then, inputting the extracted multi-scale feature map into an LSTM network, learning time sequence dependency relations among different time scale features, and finally generating feature vectors fused with multi-scale time sequence information, namely, instant monitoring feature fusion data.

In the embodiment of the invention, the new energy power generation power under different time scales is calculated by utilizing known historical meteorological data (such as irradiance, wind speed and the like) and a new energy output model and combining the time sequence monitoring characteristic fusion data generated in the step S22. For photovoltaic power generation, for example, a photovoltaic output prediction model based on a physical model can be adopted, weather data such as historical irradiance, temperature and the like and photovoltaic module parameters are input, historical photovoltaic output data is calculated, and for wind power generation, a wind power output prediction model based on weather data such as wind speed, wind direction, air density and the like and wind turbine generator set parameters can be adopted, and historical wind power output data is calculated. And integrating the historical power generation data obtained by calculation under different time scales to generate multi-scale power generation data.

In the embodiment of the invention, the power grid is divided into a plurality of geographic areas according to the geographic position information of the new energy nodes. And then, aiming at each region, adopting a sliding window method to align the multi-scale power generation data and the time sequence monitoring characteristic fusion data in the time dimension so as to form data samples of a plurality of time windows. Each sample contains multi-scale generated power data of the area in a specific time window and time sequence monitoring characteristic fusion data of all nodes. And calculating a correlation coefficient, such as a Pearson correlation coefficient, a Spearman correlation coefficient and the like, between the time sequence monitoring characteristic fusion data of each node and the regional multi-scale generated power data. The larger the absolute value of the correlation coefficient is, the stronger the correlation between the monitoring data of the node and the new energy output of the area is. Features which are obviously related to the output of the new energy node are identified, for example, the output of a photovoltaic power station is obviously positive with sunlight intensity and air temperature, and the output of a wind driven generator is obviously positive with wind speed. The new energy output related data comprises the correlation between the new energy node output and different characteristics, and the new energy node output change rules at different time periods and different positions.

In the embodiment of the invention, for short-term prediction, a method based on time series analysis, such as ARIMA model, prophet model and the like, or a method based on machine learning, such as a support vector machine, a neural network and the like, can be selected. For long-term prediction, a method based on numerical weather forecast data and a physical model, or a method based on deep learning, such as a recurrent neural network, a long-term and short-term memory network, etc., may be selected. And selecting a proper prediction model according to the prediction target and the data characteristics, and training and optimizing the model to respectively generate short-term power prediction data and long-term power prediction data.

In the embodiment of the invention, a time weighting method can be adopted for data fusion aiming at the overlapping time period of the short-term prediction result and the long-term prediction result. The more recent the current time, the greater the weight of the prediction result, and conversely, the less the weight. For example, the predicted value of 1 hour in the future may be given a weight of 0.8 for the short-term predicted result and 0.2 for the long-term predicted result, and the predicted value of 24 hours in the future may be given a weight of 0.2 for the short-term predicted result and 0.8 for the long-term predicted result. The accuracy advantage of short-term prediction and the trend guidance of long-term prediction are combined through a time weighting method, so that more accurate and smoother new energy power prediction data are generated.

Preferably, step S25 comprises the steps of:

In the embodiment of the invention, the historical new energy power generation data under the corresponding time scale is used as an output variable to construct a multiple regression model. The model can take various forms such as linear regression, ridge regression, lasso regression and the like into consideration, and a proper model is selected according to the characteristics of data. The model is constructed by adopting a gradient descent algorithm to perform parameter optimization, for example, a method of batch gradient descent, random gradient descent or small batch gradient descent can be used for iteratively updating model parameters until a loss function of the model is converged to a preset threshold range, and a preliminary short-term new energy output prediction model is obtained. And dividing the multi-scale power generation data into a training set and a verification set by adopting a sliding window mode, finely adjusting model parameters by utilizing the training set data, evaluating the prediction performance of the model by utilizing the verification set data, and finally obtaining an optimized short-term new energy prediction model. And inputting the time sequence monitoring characteristic fusion data into an optimized short-term new energy prediction model to obtain a new energy output prediction result in a future period (for example, 2 hours in the future), namely short-term power prediction data. And calling an external weather forecast interface (such as an API interface of a national weather service, an API interface of a commercial weather data provider and the like) according to the geographic position information of the new energy node in the power grid topological structure data, and acquiring weather forecast data of the position of the new energy node in a future period (such as 5 days in the future). And identifying weather elements with obvious influence on new energy output and influence degree thereof. For example, statistical indexes such as correlation coefficients, partial correlation coefficients and the like between different weather elements and new energy output can be calculated, or a model based on machine learning is constructed, and the influence degree of different weather elements on the new energy output is quantitatively analyzed. And taking the weather effect factor data and the time sequence monitoring characteristic fusion data as input, and performing migration learning on the LSTM model. For example, the weather-influencing factor data can be used as additional input of the LSTM model, or the weather-influencing factor data and the time sequence monitoring feature fusion data are spliced to be used as the input of the LSTM model, and the parameters of the LSTM model are finely adjusted according to the new energy output related data so as to adapt to a specific new energy prediction scene. And inputting the short-term power prediction data and the weather effect factor data into a long-term new energy prediction model to obtain new energy output prediction results in a longer time (for example, 5 days in the future), namely, long-term power prediction data.

Preferably, step S3 comprises the steps of:

In the embodiment of the present invention, the power grid topology structure data generally includes geographic location information, such as longitude and latitude coordinates, of each node. And analyzing the power grid topological structure data according to the data format, extracting longitude and latitude coordinates of each node, and storing the longitude and latitude coordinates as power grid node coordinate data. The geographic distance between any two nodes is calculated using a geographic distance calculation formula (e.g., HAVERSINE formula or Vincenty formula). And storing the calculated distance values in a matrix, wherein rows and columns of the matrix respectively represent different nodes, and matrix elements represent distances between corresponding nodes to form inter-node distance matrix data. And constructing a matrix corresponding to the power grid topological structure diagram according to the line connection information, wherein the rows and the columns of the matrix represent different nodes, if the line connection exists between the two nodes, the corresponding matrix element is 1, and otherwise, the matrix element is 0. And then, according to the line parameter information, replacing non-zero elements in the matrix with impedance values of corresponding lines to form line impedance matrix data. Meanwhile, the line impedance information is related to a power grid topological structure diagram, and power grid impedance topological data are constructed. And calculating a correlation coefficient, such as a pearson correlation coefficient, a spearman correlation coefficient and the like, between the distance between every two nodes and the corresponding line impedance by using the distance matrix data between the nodes and the line impedance matrix data. And storing the calculated correlation coefficients in a matrix, wherein rows and columns of the matrix respectively represent different nodes, and matrix elements represent the spatial correlation degree between corresponding nodes to form spatial correlation degree matrix data. For example, active power, voltage, etc., mutual information of time-series data between any two nodes is calculated. The mutual information can measure linear or nonlinear statistical dependency relationship between two time sequences to form time correlation matrix data. And respectively taking the space correlation degree matrix and the time correlation degree matrix as two dimensions of tensors to form three-dimensional space-time correlation degree tensor data. The space association degree matrix and the time association degree matrix can be subjected to linear weighted fusion to obtain two-dimensional space-time association degree matrix data. And calculating the space-time weighted similarity of any two nodes in the multi-scale power matrix data under different time scales. For example, a euclidean distance or Dynamic Time Warping (DTW) based method may be used to calculate the similarity of two node power data and weight the similarity according to the spatio-temporal correlation tensor data to obtain a spatio-temporal weighted similarity. And storing the calculated space-time weighted similarity in a vector, wherein each element in the vector represents the space-time similarity of two nodes under a specific time scale, and forming space-time similarity vector data.

Preferably, step S33 includes the steps of:

In the embodiment of the invention, the power grid management system generally stores detailed power line parameter information, such as line length, wire type, phase line structure and the like. And inquiring and acquiring various parameter information of a required power line through data interaction with a power grid management system, and storing the parameter information as power line parameter data. And selecting a proper distributed circuit model, such as a pi-type equivalent circuit model, a multi-segment pi-type equivalent circuit model and the like, and modeling each segment of power circuit. The model parameters can be obtained by consulting related power system analysis books or standards, and can also be corrected according to actual conditions. For example, the resistance, inductance and capacitance parameters of the line can be calculated according to the parameters such as the length of the line and the model of the wire, and the parameters are used as parameters of a pi-type equivalent circuit model. And integrating the model parameters of all the circuits to form electric circuit model data. And constructing a simulation model corresponding to the actual power grid topological structure, and importing the electric circuit model data into simulation software. Simulation parameters such as simulation time, simulation step size, frequency range, etc. are set and a simulation program is run. In the simulation process, excitation signals with different frequencies are applied to the line model, and voltage and current responses at two ends of the line are recorded. From the voltage and current responses, the line impedance is calculated and an impedance-frequency curve, i.e., impedance frequency response data, is plotted. Line impedance values at a particular frequency, such as 50Hz or at a particular harmonic frequency, are extracted from the impedance frequency response data. The extracted impedance values are used as line attributes and added to the power grid topological structure data, for example, the impedance values of each line at different frequencies can be stored as attribute information and added to corresponding lines of a power grid topological structure diagram. A node impedance matrix method, a davienan equivalent circuit method, and the like are used. And calculating a harmonic impedance value between any two nodes by using the power grid impedance topology data, and generating line harmonic impedance data. The extent of the effect of the line parameter change on the harmonic impedance is analyzed, for example, the rate of change of the harmonic impedance before and after the line parameter change can be calculated, or other sensitivity analysis methods can be used. And constructing line impedance matrix data according to the analysis result.

Preferably, step S4 comprises the steps of:

In the embodiment of the invention, according to the power grid impedance topology data, the electrical distance of each node, such as the shortest path distance, the resistance distance and the like, is calculated. The closer the electrical distance is, the higher the correlation between the voltage changes is generally. Then, the nodes with higher space-time similarity with other nodes are screened out by combining the space-time similarity vector data, and the voltage change of the nodes is more easily influenced by the other nodes. And finally, according to the electrical distance and the space-time similarity, grouping the nodes by using a clustering algorithm (such as a K-means algorithm, a DBSCAN algorithm and the like), identifying the node group with a short electrical distance and high space-time similarity as a key voltage sensitive node, generating key voltage sensitive node data, calculating the active power variation of each key voltage sensitive node in a time window according to the preset size of the time window (such as 1 minute, 5 minutes and the like), and dividing the active power variation by the size of the time window to obtain the active power variation rate. And calculating the maximum value and the minimum value of the active power change rate of each key voltage sensitive node in the time window according to the preset size of the time window, and subtracting the minimum value from the maximum value to obtain the active power change amplitude. For example, a sliding window approach may be used to calculate the very poor rate of change of active power within each time window. And correlating the calculated active power variation amplitude with the corresponding time stamp and node information to generate active power variation amplitude data. A set of trend evolution labels, such as "fast rise", "slow rise", "steady", "slow fall", "fast fall", etc., are predefined and corresponding thresholds are set. And comparing the active power change amplitude data with a preset threshold value according to the active power change amplitude data, and endowing each key voltage sensitive node in each time window with a corresponding trend evolution label according to a comparison result to generate active trend evolution label data. Different active power change trend modes are identified, for example, "fast rising" and "slow rising" may be categorized as "rising trend", "fast falling" and "slow falling" may be categorized as "falling trend". And then analyzing the influence degree of different trend modes on the voltage, giving corresponding weight coefficients for each trend mode, representing the influence degree of the trend modes on the voltage, and finally generating active trend influence factor data. And predicting the change trend of the node voltage in a future period of time by combining the historical voltage data of the key voltage sensitive node. For example, according to the active power change trend mode at the current moment and the corresponding weight coefficient, the voltage change direction and amplitude in a future period can be predicted, or by using a time sequence analysis method and combining historical voltage data and active trend influence factor data, the voltage change trend in a future period can be predicted. And correlating the predicted voltage change trend with the corresponding time stamp and node information to generate node voltage trend data.

Preferably, step S41 comprises the steps of:

In the embodiment of the invention, the space-time similarity vector data is standardized by using a Z-score standardization method, each element is converted into standard normal distribution data with the mean value of 0 and the standard deviation of 1, and the standard space-time similarity vector data is generated. And grouping node pairs with similarity higher than a threshold value according to the space-time similarity vector data to obtain a plurality of similar characteristic node groups. Then, for each node group, nodes with similarity higher than a threshold value with other nodes in the group are screened out according to the node similarity threshold value data, and the nodes are considered as potential grid sensitive nodes. According to the preset time window size (for example, 1 minute, 5 minutes and the like), calculating the voltage variation of each potential grid sensitive node in the time window, and dividing the voltage variation by the time window size to obtain the voltage variation rate. According to the power grid impedance topology data, calculating network centrality indexes of each potential power grid sensitive node, such as centrality, medium centrality, near centrality and the like. And building a simulation model corresponding to the actual power grid topological structure by using power system simulation software, and applying disturbance to each potential power grid sensitive node, such as step change of active power or reactive power. After disturbance occurs, the change process of the electrical quantity such as node voltage, current and the like is recorded, the dynamic response characteristic is analyzed, the dynamic response characteristic index obtained through analysis is associated with node information, and dynamic response characteristic data are generated. And comprehensively scoring the sensitive nodes of the potential power grid by utilizing a multi-criterion decision method (such as a hierarchical analysis method, an entropy weight method, a TOPSIS method and the like) according to the voltage rate data, the topology importance evaluation data and the dynamic response characteristic data of the sensitive nodes. First, the weight of each evaluation index needs to be determined according to the actual situation, for example, the weight may be determined according to expert experience, historical data analysis, and the like. Then, a composite score for each node is calculated based on the determined weights and the evaluation index value for each node. A screening threshold is determined based on a predetermined threshold or based on a statistical method, such as a percentile method based on a multi-criteria scoring data distribution. The multi-criterion scoring data is compared with a screening threshold value, and nodes with scores higher than the threshold value are screened out and are considered as final key voltage sensitive nodes because the nodes show higher sensitivity in terms of voltage change rate, topological importance, dynamic response characteristics and the like.

As an example of the present invention, referring to fig. 3, a detailed implementation step flow diagram of step S5 in fig. 1 is shown, where step S5 includes:

in the embodiment of the invention, new energy power prediction data and node voltage trend data are input into power system analysis software, such as PSS/E, PSAT and the like, so as to construct a power system power flow calculation model. Then, according to preset fault or disturbance scenes, such as new energy output fluctuation, load mutation and the like, simulating the steady-state operation state of the power system under different operation scenes. And finally, according to the load flow calculation result, calculating a voltage stability margin index of each node, such as a difference value between a voltage breakdown critical point and the current voltage, a voltage stability index and the like, and generating voltage stability margin evaluation data.

In the embodiment of the invention, the triggering condition of voltage control is determined according to a preset voltage stability margin target, for example, the difference between the voltage breakdown critical point and the current voltage is not lower than a certain safety value. And then, judging whether the current voltage stability margin meets a preset target according to the voltage stability margin evaluation data. If the voltage stability margin does not meet the target, the preset initial voltage control threshold is dynamically adjusted according to the deviation degree between the voltage stability margin and the target value, for example, the voltage control threshold can be linearly or non-linearly adjusted according to the deviation, and the larger the deviation is, the larger the adjustment amplitude is. The resulting dynamic voltage control threshold may be a personalized threshold for different nodes, different time periods, for finer voltage control. For example, a fuzzy logic control algorithm is used to dynamically adjust the control threshold according to the evaluation result of the voltage stability margin, and when the evaluation result of the voltage stability margin is lower than a preset threshold, the control threshold is lowered to improve the sensitivity of the voltage control, and when the evaluation result of the voltage stability margin is higher than the preset threshold, the control threshold is raised to reduce unnecessary voltage control operation.

according to the embodiment of the invention, the maximum value and the minimum value of the voltage of each node in a future period are calculated according to the predicted voltage change trend, and the maximum value and the minimum value are compared with a preset voltage allowable fluctuation range to judge whether the voltage of the node exceeds the allowable range. If the predicted maximum value or minimum value of the voltage exceeds the allowable range, calculating the value of the exceeding part as a voltage deviation pre-judging value.

In the embodiment of the invention, a preset voltage control strategy is combined, for example, the voltage deviation is controlled to be a specific value in an allowable range, and the voltage regulation target value of each node in a future period of time is calculated. For example, if the voltage deviation pre-determined value of a certain node in the future 15 minutes is 0.05p.u., the preset voltage control strategy is to control the voltage deviation within 0.02p.u., the voltage regulation target value of the node in the future 15 minutes is the current voltage value plus 0.03p.u. And correlating the calculated voltage regulation target value with the corresponding time stamp and node information to generate voltage regulation target value data.

In an embodiment of the invention, the voltage regulation target value is compared with the dynamic voltage control threshold value. If the voltage regulation target value exceeds the dynamic voltage control threshold value, a corresponding control signal is generated, for example, a step-up or step-down command is sent, and relevant equipment (such as a transformer tap, reactive compensation equipment and the like) is controlled to perform regulation operation, so that the node voltage tends to the target value. The control signal may be a continuous adjustment signal, such as a percentage of an adjustment amount, or a discrete switching signal, such as a switching command for the device. And associating the generated control signal with the corresponding timestamp, node information and control equipment information to generate node voltage control signal data. For example, it may be defined that a step-up measure is required when the "voltage deviation pre-determination value" exceeds a positive threshold, a step-down measure is required when the "voltage deviation pre-determination value" is lower than a negative threshold, and a voltage control operation is not required when the "voltage deviation pre-determination value" is within a threshold range.

Preferably, step S6 comprises the steps of:

In the embodiment of the invention, the distribution condition of controllable resources in the power grid, including information such as type, position, capacity, response characteristics and the like, is obtained through data interaction with the power grid management system. The controllable resources include conventional controllable resources such as transformer taps, capacitor banks, static Var Generators (SVGs), etc., as well as new energy controllable resources such as photovoltaic inverters, wind turbines, etc. with active/reactive regulation capability. And finishing the acquired data to form structured power grid controllable resource distribution data. According to the power grid impedance topology data, a clustering algorithm (such as spectral clustering, K-means clustering and the like) is utilized to divide the power grid into a plurality of voltage control areas, and node voltage changes in each area have higher relevance. When dividing the area, the factors such as the electrical distance, the line impedance, the geographic position and the like can be considered. Then, based on the node voltage control signal data in each region, the overall voltage control requirements of that region, such as voltage regulation, reactive compensation capacity, etc., are calculated. For each voltage control region, an optimal scheduling model, such as a linear programming model, a mixed integer programming model, and the like, is constructed according to the region voltage control resource data and the node voltage control signal data. The objective function of the model may be voltage bias minimization, control cost minimization, new energy consumption maximization, etc. Constraints of the model include output limits of controllable resources, voltage stability constraints, line transmission capacity constraints, and the like. For example, tap position adjustment commands may be generated for transformer taps, capacitor switching commands may be generated for capacitor banks, active and reactive power adjustment commands may be generated for photovoltaic inverters with active/reactive adjustment capability, etc. The instruction data needs to contain detailed control information such as control device ID, adjustment amount, adjustment time, execution order, etc., to ensure that the control instruction can be properly parsed and executed.

The method has the beneficial effects that the comprehensive basic information is provided for the operation of the power grid by acquiring the controllable resource distribution data of the power grid. The position, capacity and characteristics of the controllable resources are accurately mastered, so that the power grid can rapidly identify available resources when the power grid needs to be regulated, effective management of the controllable resources is ensured, the regulation and control capability of the power grid is exerted to the greatest extent, and particularly under the condition of generating fluctuation in the face of new energy, voltage change can be effectively treated, and stable operation of the power grid is ensured. By analyzing the power grid impedance topology data, the controllable resource distribution data is decomposed into the regional voltage control targets, so that the voltage control strategy can be more refined and accurate. The voltage control target of each area is clarified, the flexibility and pertinence of power grid management are improved, corresponding voltage regulation measures can be implemented in each area according to actual conditions, and potential risks caused by unstable voltage are reduced. The regional management mode not only improves the response speed of the power grid, but also enhances the adaptability of each region to the change of the power demand. Through voltage stability margin evaluation and dynamic voltage control threshold adjustment, the system can flexibly adjust the control strategy according to the real-time state and future trend of the power grid, and the voltage level is ensured to be always in a safe and stable range. The voltage deviation is prejudged and the adjustment target value is calculated, so that the voltage control is more active and accurate, unnecessary adjustment actions are reduced, and the running efficiency of the power grid is improved. Through the fine management of controllable resources and the cooperative control of new energy, the effective decomposition and optimized scheduling of the regional voltage control targets are realized, and the efficient execution of control measures is ensured. The intelligent voltage regulation command is generated, so that the intelligent voltage regulation command can not only quickly respond to the state change of the power grid, but also adapt to the trend of the active power, and the self-adaptive capacity and the fault recovery speed of the power grid are obviously enhanced. Not only improves the acceptance and utilization efficiency of the power grid to the distributed new energy, but also strengthens the dynamic balance adjustment capability of the power grid, and effectively solves the problem of voltage stability caused by the access of the new energy.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automatic voltage control method based on active power trend determination, characterized in that it comprises the following steps:

Step S1: Processing the power grid topology structure according to the power grid management system to generate power grid topology structure data; collecting multi-scale monitoring data based on the power grid topology structure data to generate multi-scale power matrix data;

Step S2: extracting new energy node status from the multi-scale power matrix data to obtain new energy monitoring status data; predicting new energy output power based on the new energy monitoring status data to generate new energy power prediction data;

Step S3: performing line impedance analysis according to the power grid topology data to obtain power grid impedance topology data; performing spatiotemporal node similarity processing on the multi-scale power matrix data based on the power grid topology data to generate spatiotemporal similarity vector data;

Step S4: identifying voltage change sensitive nodes on the spatiotemporal similarity vector data through the grid impedance topology data to generate key voltage sensitive node data; performing node voltage change trend analysis based on the key voltage sensitive node data to obtain node voltage trend data;

Step S5: evaluating the voltage stability margin of the power grid according to the new energy power prediction data and the node voltage trend data, and generating voltage stability margin evaluation data; generating a voltage control signal based on the voltage stability margin evaluation data, and generating node voltage control signal data;

Step S6: generating an intelligent voltage regulation instruction according to the node voltage control signal data to obtain intelligent voltage regulation instruction data.

2. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S1 comprises the following steps:

Step S11: performing monitoring power node analysis according to the power grid management system to obtain monitoring power node data, wherein the monitoring power node data includes new energy node data and power load node data;

Step S12: Processing the power grid topology structure based on the monitored power node data to generate power grid topology structure data;

Step S13: performing time scale correlation analysis according to the power grid topology data, and performing adaptive time scale selection to obtain multi-scale time window data;

Step S14: performing multi-scale monitoring data collection on the monitoring power node data through the multi-scale time window data to obtain multi-scale power monitoring data; performing multi-scale vector construction according to the multi-scale power monitoring data to generate multi-scale monitoring vector data;

Step S15: performing data cleaning processing on the multi-scale power monitoring data, and extracting monitoring power features to generate monitoring node power feature data;

Step S16: Reconstruct the monitoring state matrix of the monitoring node power characteristic data using the multi-scale monitoring vector data to generate multi-scale power matrix data.

3. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S2 comprises the following steps:

Step S21: extracting new energy node status according to multi-scale power matrix data to obtain new energy monitoring status data;

Step S22: performing multi-scale time series feature fusion according to the new energy monitoring status data to generate time series monitoring feature fusion data;

Step S23: Calculate the historical power generation based on the time series monitoring feature fusion data to obtain multi-scale power generation data;

Step S24: performing spatiotemporal new energy output analysis on the time series monitoring feature fusion data through multi-scale power generation data to generate new energy output related data;

Step S25: performing long-term and short-term power forecasts of new energy sources according to the new energy output correlation data and the power grid topology data, and generating short-term power forecast data and long-term power forecast data respectively;

Step S26: Perform multi-time scale integration processing on the short-term power prediction data and the long-term power prediction data to generate new energy power prediction data.

4. The automatic voltage control method based on active power trend determination according to claim 3, characterized in that step S25 comprises the following steps:

Step S251: construct a multivariate regression model for the new energy output related data using a gradient descent algorithm to obtain a short-term new energy output prediction model;

Step S252: training the short-term renewable energy output prediction model through multi-scale power generation data to generate an optimized short-term renewable energy prediction model;

Step S253: using the optimized short-term new energy prediction model to perform short-term energy power prediction on the time series monitoring feature fusion data to generate short-term power prediction data;

Step S254: acquiring weather forecast data for new energy nodes based on the power grid topology data;

Step S255: performing an influencing factor analysis on the new energy output related data through weather forecast data to generate weather influencing factor data;

Step S256: Using a preset long-term and short-term neural network model to perform transfer learning on weather influencing factor data and time series monitoring feature fusion data, and construct a long-term new energy prediction model;

Step S257: Perform long-term energy power prediction on the short-term power prediction data and the weather influencing factor data through the long-term new energy prediction model to generate long-term power prediction data.

5. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S3 comprises the following steps:

Step S31: extracting node geographic coordinates according to the power grid topology data to generate power grid node coordinate data;

Step S32: performing inter-node distance matrix processing according to the grid node coordinate data to generate inter-node distance matrix data;

Step S33: performing line impedance matrix processing according to the power grid topology data to generate line impedance matrix data and power grid impedance topology data;

Step S34: calculating the spatial correlation coefficient of the line impedance matrix data through the node distance matrix data to generate spatial correlation matrix data;

Step S35: Calculate the time series mutual information of the multi-scale power matrix data to obtain time correlation matrix data;

Step S36: performing spatiotemporal correlation fusion according to the spatial correlation matrix data and the temporal correlation matrix data to generate spatiotemporal correlation tensor data;

Step S37: Performing spatiotemporal weighted similarity processing on the multi-scale power matrix data through the spatiotemporal correlation tensor data to generate spatiotemporal similarity vector data.

6. The automatic voltage control method based on active power trend determination according to claim 5, characterized in that step S33 comprises the following steps:

Step S331: extracting power line parameters from the power grid topology data through the power grid management system to obtain power line parameter data;

Step S332: constructing a distributed line model according to the power line parameter data to obtain electrical line model data;

Step S333: performing impedance frequency simulation on the electrical circuit model data to generate impedance frequency response data;

Step S334: mapping the line impedance attributes of the power grid topology data through the impedance frequency response data to generate power grid impedance topology data;

Step S335: Calculate the inter-node harmonic impedance according to the grid impedance topology data to generate line harmonic impedance data;

Step S336: Use the line harmonic impedance data to perform impedance sensitivity analysis on the power line parameter data to generate line impedance matrix data.

7. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S4 comprises the following steps:

Step S41: identifying voltage change sensitive nodes using the time-space similarity vector data through the power grid impedance topology data to generate key voltage sensitive node data;

Step S42: Calculate the active power change rate of the key voltage sensitive node data to generate active power change rate data;

Step S43: Calculating the change amplitude according to the active power change rate data to generate active power change amplitude data;

Step S44: marking the active power variation amplitude data with trend changes by using preset trend evolution label data to generate active power trend evolution label data;

Step S45: classifying the real-time active power trend pattern according to the active power trend evolution label data, and evaluating the voltage impact degree to generate active power trend impact factor data;

Step S46: Analyze the node voltage change trend of key voltage sensitive node data using the active power trend influencing factor data to obtain node voltage trend data.

8. The automatic voltage control method based on active power trend determination according to claim 7, characterized in that step S41 comprises the following steps:

Step S411: performing Z-score standardization processing on the spatiotemporal similarity vector data to generate standard spatiotemporal similarity vector data; performing similarity threshold processing on the standard spatiotemporal similarity vector data to obtain node similarity threshold data;

Step S412: performing grouping processing of similar feature nodes according to the spatiotemporal similarity vector data, and screening potential sensitive nodes according to the node similarity threshold data, to obtain potential power grid sensitive node data;

Step S413: Calculate the voltage change rate of the potential power grid sensitive node data to obtain the sensitive node voltage rate data;

Step S414: using the power grid impedance topology data to perform topology importance assessment on potential power grid sensitive node data to generate topology importance assessment data;

Step S415: performing dynamic response characteristic analysis on potential power grid sensitive node data to generate dynamic response characteristic data;

Step S416: performing a multi-criteria comprehensive evaluation process on the potential power grid sensitive node data based on the sensitive node voltage rate data, the topology importance evaluation data, and the dynamic response characteristic data to generate multi-criteria scoring data;

Step S417: Perform sensitive node verification processing on potential power grid sensitive node data through multi-criteria scoring data, and screen key voltage sensitive nodes to obtain key voltage sensitive node data.

9. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S5 comprises the following steps:

Step S51: evaluating the voltage stability margin of the power grid according to the new energy power prediction data and the node voltage trend data, and generating voltage stability margin evaluation data;

Step S52: dynamically adjusting the preset initial voltage control threshold value according to the voltage stability margin evaluation data to obtain a dynamic voltage control threshold value;

Step S53: using the voltage stability margin assessment data to perform voltage fluctuation analysis on the node voltage trend data, and calculating the voltage deviation prediction value to generate the voltage deviation prediction value;

Step S54: Calculating the voltage regulation target value according to the voltage deviation prediction value to obtain voltage regulation target value data;

Step S55: generating a voltage control signal for the voltage regulation target value data through a dynamic voltage control threshold, thereby generating node voltage control signal data.

10. The automatic voltage control method based on active power trend determination according to claim 1, characterized in that step S6 comprises the following steps:

Step S61: Acquire power grid controllable resource distribution data;

Step S62: Decomposing the controllable resource distribution data of the power grid into regional voltage control targets by using the power grid impedance topology data to generate regional voltage control resource data;

Step S63: Utilizing the regional voltage control resource data to optimize the scheduling of the node voltage control signal data, and generating a new energy coordinated control strategy to generate an automatic voltage control strategy;

Step S64: generating intelligent voltage regulation instructions according to the automatic voltage control strategy to obtain intelligent voltage regulation instruction data.