CN113568386B - Thermal power generating unit all-working-condition big data analysis method based on interval estimation - Google Patents

Abstract

The invention discloses a large data analysis method for thermal power units under full working conditions based on interval estimation. The load of thermal power units is used as a characteristic variable for steady state discrimination. Within a certain time window, the confidence of the overall mean value of the load difference samples of thermal power units is If the interval contains zero value at the significance level α, and the absolute value of the difference between the lower confidence limit and the upper confidence limit is stable under the limited conditions, then it is determined that the unit is in a steady state within the time window, otherwise it is determined that it is in an unsteady state Working conditions: when the present invention analyzes the steady-state and non-steady-state working conditions of thermal power units, it can accurately determine the operating state of the unit under the condition that the steady-state and non-steady-state working conditions frequently alternately occur.

Description

A Big Data Analysis Method Based on Interval Estimation for Full Working Conditions of Thermal Power Units

technical field

The invention relates to the field of intelligent data processing of thermal power units, in particular to a large data analysis method for thermal power units under full working conditions based on interval estimation.

Background technique

With the promotion of digital transformation of thermal power units, intelligent process analysis and control technology has been applied in more and more thermal power generation units. In terms of long-term development trend, leading digital technology will be deeply integrated with power generation production management and control process. It has become an important driving force for power generation enterprises to practice intelligent transformation. On the one hand, technologies such as big data and artificial intelligence are playing an increasingly important role in the field of energy conservation and emission reduction in the power production process. Using big data, artificial intelligence and other technologies to conduct in-depth modeling and mining of the generation process and thermal systems is a An important means to break through the bottleneck of unit energy saving and emission reduction. On the other hand, thermal power generating units are gradually changing from electricity-based power sources to regulated power sources. The units are in the steady-state and unsteady-state operating modes for a long time, and the physical characteristics of the units under steady-state and unsteady-state conditions are very different. There are great differences in the operating status of the units, and correspondingly, the operating data show problems such as large fluctuation range, irregular deviation, very few occurrences of some features, and large differences in data distribution. As far as artificial intelligence algorithms are concerned, the instability of modeling results will increase. Therefore, distinguishing between steady-state and unsteady-state conditions, using big data and artificial intelligence algorithms to model the steady-state and unsteady-state conditions separately and organically integrating them will help improve the safety and stability of the power production process.

At present, there are mainly two methods for dividing the working conditions of thermal power units:

1) According to the load variation range of the thermal power unit, the equal width division method is used to divide the working conditions of the thermal power unit into several types. Although this method can cover all operating conditions of the unit, it cannot fully reflect the actual distribution characteristics of the operating conditions of the thermal power unit. In addition, when unsteady-state conditions and steady-state conditions alternately appear in the unit’s historical operation data, it is impossible to effectively distinguish the operation mode from the data, which will greatly affect the construction of big data and artificial intelligence modeling algorithms. stability and accuracy of model results.

2) The operating conditions of thermal power units are divided through historical data mining, and the unsupervised machine learning clustering algorithm is used to divide the historical operation data sets of the units through the similarity between samples, so as to divide the operating conditions of the units. Although this method overcomes the limitations of the equal-width partition method, the clustering algorithm cannot effectively determine the specific types of working conditions. For example, the number of clusters in the K-means clustering algorithm needs to be manually set, requiring more prior knowledge. The relevant professional background requirements for modelers are relatively high, and most big data and artificial intelligence algorithm professionals do not have thermal power professional background.

In summary, the existing thermal power unit working condition division method cannot effectively divide steady-state and unsteady-state working conditions. In order to solve this problem, the present invention proposes a universal steady-state and unsteady-state working condition analysis method , which effectively makes up for the shortcomings of other current methods, realizes the effective identification and differentiation of data in different operating modes, and provides a good data processing method for accurate modeling of the system.

Contents of the invention

The present invention proposes a large data analysis method based on interval estimation for thermal power units under full operating conditions. The unit load is selected as the characteristic variable for steady-state and unsteady-state discrimination. According to the nature of time series data, the interval estimation method is used to estimate the The confidence interval of the overall mean of the load difference sample can effectively distinguish the steady state/unsteady state of the unit.

Further explain as follows, the present invention adopts following technical scheme:

A big data analysis method based on interval estimation for thermal power units under full operating conditions, using thermal power unit load as the characteristic variable for steady-state discrimination, within a certain time window, the confidence interval of the overall mean value of the load difference of thermal power units is within the If the level α contains zero value, and the absolute value of the difference between the lower confidence limit and the upper confidence limit is stable under the limited conditions, then it is determined that the unit is in a steady state within the time window, otherwise it is determined that it is in an unsteady state; Including the following steps:

Step 1: Obtain the unit load sequence data {X ₀ ,X ₁ ,X ₂ ,…,X _n } from the thermal power unit DCS system; use the first-order difference method to calculate the unit load difference sequence data {d ₁ ,d ₂ , d ₃ ,…,d _n }; the specific calculation formula of the load difference sequence data of the computer group by the first-order difference method is as follows:

d _j ＝X _j -X _j-1 ,j=1,2,3,…,n Formula 1

In formula 1, X _j represents the unit load at time j, and d _j represents the difference between unit load at time j and time j-1;

表示对

取整，对于k≤j≤(n-k+1),在机组负荷差值序列数据{d₁,d₂,d₃,…,d_n}上，以步长为1，从左到右移动时间窗口；Step 2: Assume that the time window size is

express yes

Rounding, for k≤j≤(n-k+1), on the unit load difference sequence data {d ₁ ,d ₂ ,d ₃ ,…,d _n }, with a step size of 1, from left to right moving time window;

和

满足

在给定显著水平α下，首先计算时间窗口内的机组负荷差值样本的均值：Step 3: For each time window, using the interval estimation method, under a given significance level α, estimate the confidence interval of the overall mean m of the unit load difference within the time window

and

satisfy

Under a given significance level α, first calculate the mean value of the unit load difference samples in the time window:

代表时间窗口内的机组负荷差值样本的均值，d_j代表时刻j和时刻j-1的机组负荷差值；In formula 2,

Represents the mean value of the unit load difference samples in the time window, d _j represents the unit load difference between time j and time j-1;

Compute the standard deviation of unit load differences over a time window:

Calculate the upper and lower confidence limits of the confidence interval of the overall mean m of the unit load difference sample within the time window:

代表t(n-1)分布的

分位数；In formula 4 and formula 5,

Represents the t(n-1) distribution

quantile;

且

则判定机组在该时间窗口内处于稳定工况，否则判定机组在该时间窗口内处于非稳定工况；当k<j<(n-k+1)时，如果

且

则判定机组在时刻j处于稳定工况,否则，判定机组在时刻j处于非稳定工况。

和显著性水平α越小，则判定结果的不确定性越小，否则，则判定结果的不确定性越大。Step 4: Judgment of unit operating conditions: when j=k or j=n-k+1, if

and

Then it is determined that the unit is in a stable working condition within this time window, otherwise it is determined that the unit is in an unstable working condition within this time window; when k<j<(n-k+1), if

and

Then it is determined that the unit is in a stable condition at time j, otherwise, it is determined that the unit is in an unsteady condition at time j.

The smaller the sum significance level α, the smaller the uncertainty of the judgment result, otherwise, the greater the uncertainty of the judgment result.

The significance level α is 0.05 or 0.01.

Compared with the prior art, the present invention has the following characteristics:

The present invention can accurately determine the operating state of the unit under the condition that the steady-state and unstable-state conditions frequently alternately occur when analyzing the steady-state and unsteady-state working conditions of the thermal power unit. Not only can the operating conditions of the thermal power unit be judged, but also the uncertainty of the judgment result can be quantitatively analyzed by using the confidence degree and confidence interval of the interval estimation. In addition, the present invention does not require a specific professional background, and only needs to set the size of the sliding window and the significance level, and the present invention can be used for discriminant analysis of the steady-state and unsteady-state working conditions of thermal power units. This has strong practical significance for the application of artificial intelligence and big data algorithms in power generation enterprises.

Description of drawings

Fig. 1 is a flowchart of the present invention.

Fig. 2 is an analysis result of the steady-state and unsteady-state working conditions of a certain 1000MW unit in a certain period of time according to the embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with specific embodiments.

Example:

As shown in Figure 1, for a 1000MW unit in a power plant, the specific steps to determine the working condition of the unit through the big data analysis method based on the interval estimation of the thermal power unit's steady-state and unsteady-state working conditions are as follows:

Step 1. Obtain unit load sequence data {X ₀ ,X ₁ ,X ₂ ,…,X _n } from the unit DCS system; use the first-order difference method to calculate the unit load difference sequence data {d ₁ ,d ₂ , d ₃ ,…,d _n }; the specific calculation formula of the load difference sequence data of the computer group by the first-order difference method is as follows:

d _j ＝X _j -X _j-1 ,j=1,2,3,…,n Formula 1

In formula 1, X _j represents the unit load at time j, and d _j represents the difference between unit load at time j and time j-1.

表示对

取整。对于k≤j≤(n-k+1),在机组负荷差值序列数据{d₁,d₂,d₃,…,d_n}上，以步长为1，从左到右移动时间窗口。Step 2, assuming that the time window size is

express yes

Rounding. For k≤j≤(n-k+1), on the unit load difference sequence data {d ₁ ,d ₂ ,d ₃ ,…,d _n }, move the time window from left to right with a step size of 1 .

和

满足

Step 3. For each time window, use the interval estimation method to estimate the confidence interval of the overall mean m of the unit load difference in the time window under a given significance level α

and

satisfy

Step 3.1, first calculate the mean value of the unit load difference samples in the time window:

代表时间窗口内的机组负荷差值样本的均值，d_j代表时刻j和时刻j-1的机组负荷差值。In formula 2,

Represents the mean value of the unit load difference samples in the time window, and d _j represents the unit load difference between time j and time j-1.

Step 3.2, calculate the standard deviation of unit load difference within the time window:

Step 3.3, calculate the upper confidence limit and the lower confidence limit of the confidence interval of the overall mean m of the unit load difference sample within the time window:

代表t(n-1)分布的

分位数。In formula 4 and formula 5,

Represents the t(n-1) distribution

quantile.

且

则可以判定机组在该时间窗口内处于稳定工况，否则判定机组在该时间窗口内处于非稳定工况；当k<j<(n-k+1)时，如果

且

则判定机组在时刻j处于稳定工况,否则，判定机组在时刻j处于非稳定工况；Step 4, when j=k or j=n-k+1, if

and

Then it can be determined that the unit is in a stable working condition within this time window, otherwise it can be determined that the unit is in an unstable working condition within this time window; when k<j<(n-k+1), if

and

Then it is determined that the unit is in a stable working condition at time j, otherwise, it is determined that the unit is in an unstable working condition at time j;

In this embodiment, the time window size k=20, the significance level α=0.05, and the confidence level is

As shown in Figure 2, the operating conditions of a 1000MW unit in a certain power plant are judged for a period of time. The experimental results show that this method can effectively determine whether the unit is in a steady state/unsteady state.

When the present invention analyzes the steady-state and unsteady-state working conditions of thermal power units, it can not only judge the operating conditions of thermal power units, but also use the confidence degree and confidence interval of the interval estimation to quantitatively analyze the uncertainty of the judgment results. It has strong practical significance for the application of artificial intelligence and big data algorithms in power generation enterprises.

Claims (2)

1. A large data analysis method based on interval estimation for thermal power units under full operating conditions, characterized in that: the load of thermal power units is used as the characteristic variable for steady-state discrimination, and within a certain time window, the overall mean value of the load difference samples of thermal power units The confidence interval of α contains zero value at the significance level α, and the absolute value of the difference between the lower confidence limit and the upper confidence limit is stable under the limited conditions, then it is determined that the unit is in a steady state within the time window, otherwise it is determined that it is in a non- Steady-state working conditions; specifically include the following steps: Step 1: Obtain the unit load sequence data {X ₀ , X ₁ , X ₂ ,..., X _n } from the thermal power unit DCS system; use the first-order difference method to calculate the unit load difference sequence data {d ₁ , d ₂ , d ₃ ,..., d _n }; the specific calculation formula of the load difference series data of the computer group by the first-order difference method is as follows: d _j ＝X _j -X _j-1 , j=1, 2, 3,..., n Formula 1 In formula 1, X _j represents the unit load at time j, and d _j represents the difference between unit load at time j and time j-1; Step 2: Suppose the time window size is k,

express yes

Rounding, for k≤j≤(n-k+1), on the unit load difference sequence data {d ₁ , d ₂ , d ₃ ,..., d _n }, with a step size of 1, from the left Move the time window to the right; Step 3: For each time window, using the interval estimation method, under a given significance level α, estimate the confidence interval of the overall mean m of the unit load difference within the time window

and

satisfy

Under a given significance level α, first calculate the mean value of the unit load difference samples in the time window:

In formula 2,

Represents the mean value of the unit load difference samples in the time window, d _j represents the unit load difference between time j and time j-1; Compute the standard deviation of unit load differences over a time window:

Calculate the upper and lower confidence limits of the confidence interval of the overall mean m of the unit load difference sample within the time window:

In formula 4 and formula 5,

Represents the t(n-1) distribution

quantile; Step 4: Judgment of unit operating conditions: when j=k or j=n-k+1, if

and

Then it is determined that the unit is in a stable working condition within this time window, otherwise it is determined that the unit is in an unstable working condition within this time window; when k<j<(n-k+1), if

and

Then it is determined that the unit is in a stable working condition at time j, otherwise, it is determined that the unit is in an unsteady working condition at time j. 2. A big data analysis method based on interval estimation for all working conditions of thermal power units according to claim 1, characterized in that: the significance level α is 0.05 or 0.01.

Images