CN115617613A - Data early warning method, device, electronic equipment and storage medium - Google Patents

Abstract

The embodiment of the application relates to the technical field of monitoring, and provides a data early warning method, a data early warning device, electronic equipment and a storage medium, wherein a function list corresponding to an object to be monitored is obtained aiming at the object to be monitored, the function list comprises a plurality of functions to be fitted, and one function to be fitted is used for representing a variation trend of the object to be monitored; then, fitting the function to be fitted in the function list for multiple times based on the actual data value of the object to be monitored to obtain a target fitting function with the minimum deviation between the represented change trend of the object to be monitored and the actual change trend of the object to be monitored; finally, early warning is carried out on the object to be monitored by utilizing a target fitting function; therefore, accurate early warning can be carried out before a fault occurs.

Description

Data early warning method, device, electronic equipment and storage medium

technical field

The embodiments of the present application relate to the technical field of monitoring, and in particular, to a data early warning method, device, electronic equipment, and storage medium.

Background technique

During the operation of the information system, it has to face various risk factors such as hardware, network, and software itself. The above factors may cause the quality of system services to decline, or even the system to crash.

At present, in order to ensure the smooth operation of the system, system operation and maintenance personnel often add a series of monitoring objects after the system goes online, such as service delay, service success rate, etc., and set the monitoring threshold of each monitoring object. During the operation of the system, various monitoring objects are continuously sampled, and an alarm is issued when the data sampling value of a certain monitoring object exceeds its monitoring threshold.

However, this method can only give an alarm to the faults that have already occurred, and cannot give an early warning before the fault occurs.

Contents of the invention

The purpose of the embodiments of the present application is to provide a data early warning method, device, electronic equipment and storage medium, capable of performing accurate early warning before a fault occurs.

In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

In the first aspect, the embodiment of the present application provides a data early warning method, which is applied to electronic equipment, and the method includes:

Obtaining a function list corresponding to the object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to characterize a change trend of the object to be monitored;

Based on the actual data value of the object to be monitored, the functions to be fitted in the function list are fitted multiple times to obtain a target fitting function; wherein, the object to be monitored represented by the target fitting function The deviation between the change trend of and the actual change trend of the object to be monitored is the smallest;

Using the target fitting function, an early warning is given to the object to be monitored.

Optionally, the method also includes:

periodically sampling the actual data value of the object to be monitored according to a preset sampling interval;

For each sampling period, the sampling period and its corresponding actual data value are added as a sampling data to the pre-built sampling data set.

Optionally, the step of performing multiple fittings on the functions to be fitted in the function list based on the actual data value of the object to be monitored to obtain the target fitting function includes:

From the sampled data set, obtain the last sampled data and a set number of sampled data before it to obtain a plurality of candidate sampled data;

Fitting each of the functions to be fitted in the function list according to the plurality of candidate sampling data to obtain each fitting function;

Each of the fitting functions is verified, and a candidate fitting function is determined; wherein, the difference between the change trend of the object to be monitored represented by the candidate fitting function and the actual change trend of the object to be monitored the largest deviation;

removing the function to be fitted corresponding to the candidate fitting function from the function list;

The above steps are repeated until there is only one function to be fitted left in the function list, and the fitting function obtained in the last fitting of the function to be fitted is used as the target fitting function.

Optionally, the step of fitting each of the functions to be fitted in the function list according to the plurality of candidate sampling data to obtain each fitting function includes:

For each of the functions to be fitted, if the function to be fitted is a system time function, then according to the plurality of candidate sampling data, the least squares method is used to solve each coefficient in the function to be fitted to obtain the said fitting function;

If the function to be fitted is an extended time function, then according to the plurality of candidate sampling data, the coefficient partial derivative function group and the coefficient guess value of the extended time function, the gradient descent method is used to solve the function to be fitted Each coefficient in , get the fitting function.

Optionally, the candidate sampling data includes a candidate sampling period and its corresponding actual data value;

The step of verifying each of the fitting functions and determining the candidate fitting functions includes:

For each of the fitting functions, substituting the plurality of candidate sampling data into the fitting function to obtain predicted data values corresponding to each of the candidate sampling periods;

Calculate the variance sum corresponding to the fitting function according to the predicted data value and the actual data value corresponding to each of the candidate sampling periods, wherein the variance sum indicates the value of the object to be monitored represented by the fitting function The deviation between the change trend and the actual change trend of the object to be monitored;

According to the variance sum corresponding to each of the fitting functions, the fitting result weight of each of the fitting functions is calculated, wherein the fitting result weight represents the fitting effect of the fitting function;

The fitting function with the largest weight of the fitting result is used as the candidate fitting function.

Optionally, the step of calculating the fitting result weight of each fitting function according to the variance sum corresponding to each fitting function includes:

According to the sum of variances corresponding to each of the fitting functions, using a preset formula

calculating a fitting result weight of each of the fitting functions;

n表示参与第j次拟合的所述拟合函数的个数，k表示参与第j次拟合的所述拟合函数的编号。Among them, W _ij represents the fitting result weight of the i-th fitting function in the j-th fitting; W _i(j-1) represents the i-th fitting function in the j-1 fitting The weight of the fitting result in the fitting, and the initial default value of W _i(j-1) is 0; sum _ij represents the corresponding variance sum of the i-th fitting function in the j-th fitting; sum _total Indicates the sum of the variance sums corresponding to each of the fitting functions participating in the jth fitting,

n represents the number of the fitting functions participating in the j-th fitting, and k represents the number of the fitting functions participating in the j-th fitting.

Optionally, the step of using the target fitting function to give an early warning to the object to be monitored includes:

Obtain the sampling period of the last sampling data from the sampling data set;

Determine a sampling period to be predicted in the future according to the sampling period of the last sampling data, the sampling interval and the preset warning time offset;

Substituting the to-be-predicted sampling period into the target fitting function to obtain the predicted value of the to-be-monitored object in the to-be-predicted sampling period;

If the predicted value exceeds the early warning threshold, an early warning is given to the object to be monitored at the early warning time;

Wherein, the warning time is before the to-be-predicted sampling period, and the difference between the early-warning time and the to-be-predicted sampling period is the warning time offset.

Optionally, the electronic device stores a function library, the function library includes multiple system time functions and multiple extended time functions, and the multiple extended time functions are customized by system operation and maintenance personnel;

The step of obtaining the function list corresponding to the object to be monitored includes:

Responding to the selection operation, obtaining a plurality of the functions to be fitted from the function library to obtain the function list; wherein the function to be fitted is at least one of the system time function and the extended time function kind.

In the second aspect, the embodiment of the present application also provides a data early warning device, which is applied to electronic equipment, and the device includes:

An acquisition module, configured to acquire a function list corresponding to the object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to represent a change trend of the object to be monitored;

A fitting module, configured to perform multiple fittings on the functions to be fitted in the function list based on the actual data values of the objects to be monitored to obtain a target fitting function; wherein, the target fitting function represented by The deviation between the change trend of the object to be monitored and the actual change trend of the object to be monitored is the smallest;

The early warning module is configured to use the target fitting function to give early warning to the object to be monitored.

In the third aspect, the embodiment of the present application also provides an electronic device, including a processor and a memory, the memory is used to store a program, and the processor is used to implement the data in the first aspect when executing the program. Early warning method.

In a fourth aspect, the embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the data early warning method in the above-mentioned first aspect is implemented.

Compared with the prior art, in the data early warning method, device, electronic equipment and storage medium provided by the embodiments of the present application, for the object to be monitored, the function list corresponding to the object to be monitored is first obtained, and the function list includes a plurality of functions to be fitted , a function to be fitted is used to represent a change trend of the object to be monitored; then, based on the actual data value of the object to be monitored, the function to be fitted in the function list is fitted multiple times to obtain the characterized The target fitting function with the smallest deviation between the change trend of the object and the actual change trend of the object to be monitored; finally, the target fitting function is used to give an early warning to the monitored object; thus, an accurate early warning can be given before the fault occurs.

Description of drawings

FIG. 1 shows a first schematic flowchart of a data early warning method provided by an embodiment of the present application.

FIG. 2 shows a second schematic flow diagram of a data early warning method provided by an embodiment of the present application.

FIG. 3 shows an example diagram of periodic sampling provided by the embodiment of the present application.

FIG. 4 is a schematic flowchart of step S103 in the data early warning method shown in FIG. 1 and FIG. 2 .

FIG. 5 is a schematic flowchart of step S105 in the data early warning method shown in FIG. 1 and FIG. 2 .

Fig. 6 shows a schematic block diagram of a data early warning device provided by an embodiment of the present application.

Fig. 7 shows a schematic block diagram of an electronic device provided by an embodiment of the present application.

Icons: 100-data warning device; 101-acquisition module; 103-fitting module; 105-warning module; 104-sampling module; 10-electronic equipment; 11-processor; 12-memory; 13-bus.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application.

Usually, in order to ensure the smooth operation of information systems (such as network monitoring systems, network management systems, IT operation and maintenance systems, etc.), system operation and maintenance personnel will add a series of monitoring objects to the system after it goes online, and set each item Monitor the monitoring threshold of the object to observe the operating status of the system in time and intervene in time for any abnormalities that may be triggered.

At present, the main monitoring method is: during the operation of the system, the monitoring objects are continuously sampled, and when the data sampling value of a certain monitoring object exceeds its monitoring threshold, an alarm is issued. However, since the data sampling values are samples of the existing state of the system, this method cannot predict the future through sampling data. That is to say, the alarm can only be given to the faults that have already occurred, and the alarm cannot be given in advance before the fault occurs.

At the same time, it is very difficult to give an early warning before a fault occurs by setting a monitoring threshold. The reason is that it is difficult to accurately determine the trend curve obtained by a certain monitoring object with the system running before the system runs, so when the system operation and maintenance personnel set the monitoring threshold, it is difficult to accurately set the monitoring threshold on the trend curve, and finally The result is: either the monitoring threshold has not reached the system abnormality, or a large number of false alarms and early warnings.

Therefore, it is necessary to achieve effective early warning and reduce unnecessary early warning misjudgments. It is very difficult to set the monitoring threshold, and it is impossible to provide accurate early warning before the fault occurs.

In order to solve the above technical problems, the embodiments of the present application perform multiple fittings on the functions to be fitted in the function list corresponding to the objects to be monitored based on the actual data values of the objects to be monitored, wherein the function list includes multiple functions to be fitted, A function to be fitted is used to represent a change trend of the object to be monitored, and the target fitting function with the smallest deviation between the change trend of the characterized change trend of the object to be monitored and the actual change trend of the object to be monitored is obtained, and then the target fitting function is used to The combination function can give early warning to the monitored object, so that the accurate warning can be given before the fault occurs. Details are given below.

The electronic device in the embodiment of the present application may be a server, for example, a single server, a server cluster, etc.; it may also be a terminal, for example, a desktop computer, a notebook computer, a smart phone, a tablet computer, and the like. The embodiments of this application do not impose any limitation on this.

In this embodiment, for information systems, such as network monitoring systems, network management systems, IT operation and maintenance systems, etc., before the system goes online, the system operation and maintenance personnel can configure the monitoring object library and function library for the system to facilitate subsequent The system is monitored.

That is, for the information system, the electronic device may pre-store the monitoring object library and the function library of the system. For ease of understanding, before introducing the specific implementation of the embodiment of the present application, the monitoring object library and the function library are introduced first.

The monitoring object library may include: system operation and maintenance personnel pre-configured for the system, various monitoring objects that need to be monitored during system operation, for example, service delay, service success rate, CPU usage rate, etc.

Each monitoring object can include: ID, name, warning threshold, warning time offset and sampling interval.

Wherein, the identifier is used to uniquely identify a monitoring object used for sampling. The name is used to describe the monitoring object. The early warning threshold refers to the data value of the monitoring object that may cause system exceptions. The sampling interval refers to the time interval between sampling the actual data value of the monitored object twice.

The early warning time offset refers to the early warning time when the monitored object deviates from the trend. For example, for a monitoring object, through the data early warning method provided by the embodiment of the present application, it is predicted that an abnormality will occur at 16:00 on the same day. If the early warning time offset is 30 minutes, an alarm will be issued at 15:30 on the same day, thereby Reserve controllable abnormal intervention time for system operation and maintenance personnel.

The function library may include multiple system time functions, and one system time function is used to represent a change trend of the monitored object.

It should be pointed out that each system time function in the function library is not a definite function, but a definite function template. That is, the expression of the function is definite, but each coefficient in the function is unknown, and specific coefficients need to be solved according to actual monitoring requirements.

Optionally, the system time functions in the function library may include, but not limited to, standard linear functions, standard polynomial functions, and the like. Among them, the standard linear function is used to represent the steady change trend of the monitored object, and the standard polynomial function is used to represent the drastic change trend of the monitored object. The function expressions for standard linear functions and standard polynomial functions are as follows:

Standard linear function: f(t)=a ₀ +a ₁ *t

Standard polynomial function: f(t)=a ₀ +a ₁ *t+a ₂ *t ² +…+a _n *t ⁿ⁺¹ (n>1)

In a possible implementation manner, each system time function in the function library may include: a function identifier, a function expression, a default parameter, and a function fitting method.

Wherein, the function identifier is a character string used to identify a unique system time function in the function library.

The function expression is used by the system operation and maintenance personnel to select the corresponding system time function, for example, the function expression of the standard linear function and the function expression of the standard polynomial function shown above.

The default parameters refer to the initial configuration parameters of the system time function, for example, the value of the number of items n in the above-mentioned standard polynomial function, etc.

The function fitting method refers to the fitting method configured for the system time function in advance, that is, the method of solving specific coefficients in the system time function, for example, the least square method. That is, the coefficient calculation of the system time function can be realized by the least square method, and each coefficient value in the system time function can be obtained.

In a possible situation, in actual applications, the function library provided by the system may not meet the specific monitoring requirements. In this case, system operation and maintenance personnel can also customize the extended time function according to the actual monitoring requirements, and automatically Each defined extended time function is added to the function library, which is convenient for selection from the function library in subsequent monitoring. An extended time function is used to characterize a change trend of the monitored object.

Like the system time function, the various extended time functions customized by the system operation and maintenance personnel here are not definite functions, but definite function templates. That is, the expression of the function is definite, but each coefficient in the function is unknown, and specific coefficients need to be solved according to actual monitoring requirements.

Optionally, the extended time function may include: a function identifier, a function expression, a coefficient partial derivative function group, a coefficient guess value, and a function fitting method.

Wherein, the function identifier is a character string used to identify a unique extended time function in the function library.

The function expression is customized by the system operation and maintenance personnel, and it is also a string, which is used to describe the evaluation calculation formula of the extended time function.

The coefficient partial derivative function group is an array of strings describing the partial derivative calculation formula for each coefficient in the extended time function.

The coefficient guess value is an array of floating point values, which is used for the initial value of the coefficient when the gradient descent method calculates the coefficient of the extended time function.

The function fitting method refers to the fitting method of the extended time function, that is, the method of solving the specific coefficients in the extended time function, for example, the gradient descent method. That is, the coefficient approximation calculation can be realized by the gradient descent method, and each coefficient value in the extended time function can be obtained.

It should be pointed out that the above-mentioned function expression of the extended time function and the partial derivative calculation formula of each coefficient in the extended time function are both an arithmetic expression, and the expression may include coefficients, function arguments, operators, etc.

Wherein, the coefficients may be represented by a _i (i≥0). For example, the extended time function may include multiple coefficients, and each coefficient may be represented by a ₀ , a ₁ , a ₂ . . . respectively.

Meanwhile, since the system time function and the extended time function involved in the embodiment of the present application both correspond to time functions, the function argument is fixed as t.

Operators may include common operators, such as +, -, *, /, parentheses, etc., where parentheses allow nesting in expressions. Operators can also include functional operators, for example, Math.abs(x), Math.pow(x,y), Math.log(x), Math.sin(x), Math.cos(x), Math. tan(x), among them, abs is the absolute value function of x, pow is the y power function of x, log is the logarithmic function of x, sin is the sine function of x, cos is the cosine of x function, tan is the tangent function of x.

Optionally, the extended time function defined by the system operation and maintenance personnel may include, but not limited to, an exponential function, a logarithmic function, a mixed function, and the like. Exponential function, logarithmic function, and mixed function are all used to characterize the drastic change trend of the monitored object. The function expression of exponential function, logarithmic function, mixed function, coefficient partial derivative function group and coefficient guess value are as follows:

1. Exponential function

Function expression: f(t)＝a ₀ *Math.pow(a ₁ ,t)

Coefficient partial derivative function set:

Fa ₀ ＝Math.pow(a ₁ ,t)

Fa ₁ ＝a ₀ *Math.pow(a ₁ ,(t-1))*t

Coefficient guess: {5.0, 1.0}. That is, the initial values of a ₀ and a ₁ are 5.0 and 1.0, respectively.

2. Logarithmic function

Function expression: f(t)=a ₀ *Math.log(a ₁ *t)+a ₂

Coefficient partial derivative function set:

Fa ₀ =Math.log(t)+Math.log(a ₁ )

Fa ₁ =a ₀ /a ₁

Fa ₂ =1

Coefficient guess value: {1.0, 1.0, 0.8}. That is, the initial values of a ₀ , a ₁ and a ₂ are 1.0, 1.0 and 0.8, respectively.

3. Mixed function

Function expression: f(t)＝a ₀ +a ₁ /(a ₂ +Math.pow(t,a ₃ ))

Coefficient partial derivative function set:

Fa ₀ =1

Fa ₁ =1/(Math.pow(t,a ₃ )+a ₂ )

Fa ₂ ＝-a ₁ /(Math.pow(t,(2*a ₃ ))+(2*a ₂ *Math.pow(t,a ₃ )+Math.pow(a ₂ ,2))

Fa ₃ ＝-(a ₁ *Math.pow(t,a ₃ )*Math.log(t))/(Math.pow(t,(2*a ₃ ))+2*a ₂ *Math.pow( t,a ₃ ))+Math.pow(a ₂ ,2))

Coefficient guesses: {5.0, 2.0, 1.0, 1.0}. That is, the initial values of a ₀ , a ₁ , a ₂ and a ₃ are 5.0, 2.0, 1.0 and 1.0, respectively.

The specific implementation of the embodiment of the present application will be introduced in detail below in combination with the monitoring object library and function library introduced above.

Please refer to FIG. 1 , which shows a schematic flowchart of a data early warning method provided by an embodiment of the present application. The data early warning method is applied to electronic equipment, and may include the following steps:

S101. Obtain a function list corresponding to an object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to represent a change trend of the object to be monitored.

In this embodiment, the object to be monitored may be any monitored object in the monitored object library, for example, CPU usage. It should be pointed out that, in the embodiment of the present application, any object to be monitored is taken as an example for illustration, and it does not mean that only the object to be monitored is monitored. It can be understood that during the running of the system, each monitoring object in the monitoring object library will be monitored, and the object to be monitored can be any one of these monitoring objects.

From the foregoing, it can be seen that the electronic device stores a function library, and the function library includes multiple system time functions and multiple extended time functions. The extended time function is customized by the system operation and maintenance personnel. The process of function list can include:

In response to the selection operation, multiple functions to be fitted are acquired from the function library to obtain a function list; wherein, the functions to be fitted are at least one of a system time function and an extended time function.

In this embodiment, a time function in the function library is used to represent a change trend of the monitored object. However, in practice, for the object to be monitored, the system operation and maintenance personnel may not be able to predict the change trend of the object to be monitored in advance . Therefore, system operation and maintenance personnel can select multiple time functions from the function library, and use each selected time function as a function to be fitted.

Correspondingly, the selection operation refers to the operation that the system operation and maintenance personnel select the time function from the function library, and each time function selected by the system operation and maintenance personnel is used as a function to be fitted, and finally a function list of objects to be monitored is obtained.

Understandably, the functions to be fitted in the function list may all be system time functions, all may be extended time functions, or may be a combination of system time functions and extended time functions. The system operation and maintenance personnel can flexibly select the function to be fitted according to the possible change trend of the object to be monitored, which is not limited in this embodiment of the present application.

S103, based on the actual data value of the object to be monitored, perform multiple fittings on the function to be fitted in the function list to obtain the target fitting function; wherein, the change trend of the object to be monitored represented by the target fitting function is related to the The deviation between the actual change trends of the objects is minimal.

In this embodiment, after the function list of the object to be monitored is determined, the functions to be fitted in the function list corresponding to the object to be monitored are fitted multiple times based on the actual data value of the object to be monitored. That is, since the function to be fitted is a function template, the coefficients in it are unknown, so it is necessary to solve the coefficients of each function to be fitted in the function list based on the actual data value of the object to be monitored. That is, to fit the function to be fitted.

Since there are multiple functions to be fitted in the function list, one function to be fitted is used to represent a change trend of the object to be monitored. Therefore, it is necessary to perform multiple fittings on the functions to be fitted in the function list until one of them is selected. The target fitting function with the smallest deviation between the characteristic change trend of the object to be monitored and the actual change trend of the object to be monitored. It should be pointed out that the target fitting function here is different from the function to be fitted, and it is a definite function, that is, each coefficient in the target fitting function is known.

S105, using the target fitting function to give an early warning to the object to be monitored.

In this embodiment, after the target fitting function is determined through step S102, the target fitting function can be used to give an early warning to the object to be monitored. That is, use the target fitting function to predict the time when the object to be monitored may appear abnormal, and give an early warning.

At the same time, since the deviation between the change trend of the object to be monitored represented by the target fitting function and the actual change trend of the object to be monitored is the smallest, the accuracy of the early warning can be ensured, so that an accurate early warning can be carried out before the fault occurs.

It can be seen from the foregoing that, in order to fit the functions to be fitted in the function list according to the actual data values of the objects to be monitored, it is necessary to sample the actual data values of the objects to be monitored. That is, the data early warning method provided by the embodiment of the present application, in addition to the function fitting process in step S103 and the early warning process in step S105, also includes a sampling process of actual data values.

Therefore, on the basis of FIG. 1 , please refer to FIG. 2 , the data early warning method provided by the embodiment of the present application further includes steps S102-S104.

S102. Periodically sample the actual data value of the object to be monitored according to a preset sampling interval.

S104. For each sampling period, add the sampling period and its corresponding actual data value as a sampling data to the pre-built sampling data set.

In this embodiment, the actual data value of the object to be monitored can be periodically sampled according to the sampling interval defined in the monitored object library. Each sampling period may be the time of each sampling, and the sampling period is t in the system time function and extended time function introduced above.

As shown in Figure 3, t1, t2, and t3 respectively represent the first sampling period, the second sampling period, and the third sampling period, that is, the first sampling and the second sampling of the actual data value of the object to be monitored Sample, 3rd sample. Each sampling period can sample the actual data value of the object to be monitored. After each sampling, the current sampling period and the actual data value sampled at that time are taken as a sampling data and added to the pre-built sampling data set. Each sampling data in the sampling data set can be represented by (t _i , y _i ), where t _i represents the sampling period, and y _i represents the actual data value.

Optionally, for the object to be monitored, the current sampling period can be determined by the following formula:

Among them, t _i represents the current sampling period, T _m represents the current system time, T ₀ represents the first sampling time of the object to be monitored, and Δt represents the sampling interval.

It should be pointed out that the sampling process of actual data values in steps S102-S104 and the function fitting process in step S103 are two independent processes, and the two processes are performed in parallel.

Step S103 will be described in detail below.

On the basis of FIG. 2 , please refer to FIG. 4 , step S103 may include S1031-S1035.

S1031. Obtain the last sampled data and a set number of sampled data before it from the sampled data set to obtain a plurality of candidate sampled data.

In this embodiment, since the sampling of the actual data values is independent of the function fitting process, and the actual data values are periodically sampled according to the sampling interval, that is, the sampled data in the sampled data set is constantly increasing. Therefore, in order to ensure the accuracy of subsequent early warnings, each time the function to be fitted in the function list is fitted, the last sampled data and the m sampled data before it must be obtained from the sampled data set. The specific value of m The value can be flexibly set by system operation and maintenance personnel according to actual needs, and this embodiment of the present application does not impose any limitation on this.

Optionally, a data queue with a set length can also be pre-configured, and multiple candidate sampling data for fitting each time can be added to the data queue, and then the data to be fitted in the function list can be directly adjusted according to the data queue function to fit.

Understandably, for the object to be monitored, when the function to be fitted in the function list is fitted for the first time, the data queue may be empty, so it is necessary to obtain the last sampled data and its previous ones from the sampled data set m sampling data are added to the data queue; and from the second function fitting, in order to avoid data duplication and improve fitting efficiency, only the last sampling data needs to be obtained from the sampling data set and added to the data queue.

Optionally, the process of adding the last sampled data in the sampled data set to the data queue may include:

Determine whether the data queue reaches the set length;

If not, add the last sample data in the sample data set to the end of the data queue;

If so, after dequeuing the head data in the data queue, add the last sampled data in the sampled data set to the tail of the data queue.

S1032. Fit each function to be fitted in the function list according to the plurality of candidate sampling data to obtain each fitted function.

In this embodiment, each time the function to be fitted in the function list is fitted, the last sampled data and the m sampled data before it can be obtained directly from the sampled data set for fitting, or it can be fitted according to the aforementioned data set for fitting.

In the fitting process, multiple candidate sampling data can be converted into a binomial (t, y) array, where t is the sampling period, y is the actual data value, and then according to the binomial (t, y) array fits each function to be fitted in the function list.

It can be seen from the foregoing that the multiple functions to be fitted in the function list may be system time functions, or extended time functions, or a combination of system time functions and extended time functions. For the system time function, each coefficient of the system time function can be solved by the least square method. For the extended time function, each coefficient of the extended time function can be approximated and solved by the gradient descent method according to the coefficient partial derivative function group of the extended time function and the guessed value of the coefficient.

Therefore, in S1032, according to multiple candidate sampling data, each function to be fitted in the function list is fitted, and the process of obtaining each fitted function may include:

For each function to be fitted, if the function to be fitted is a system time function, then according to multiple candidate sampling data, the least square method is used to solve each coefficient in the function to be fitted to obtain the fitted function;

If the function to be fitted is an extended time function, then use the gradient descent method to solve each coefficient in the function to be fitted according to multiple candidate sampling data, the coefficient partial derivative function group and the coefficient guess value of the extended time function, and obtain the fitting function.

That is, if the function to be fitted is a system time function, then according to the least square method, the binomial (t, y) array is substituted into the coefficient equation system, and each coefficient value [a ₀ ,a ₁ ,…, a _k ]. If the function to be fitted is an extended time function, an iterative operation is performed according to the gradient descent method to approximate and solve each coefficient value [a ₀ , a ₁ ,…, a _k ] in the extended time function.

It should be pointed out that the fitting function here is the function expression obtained by solving each coefficient in the function to be fitted, and then substituting each coefficient into the function to be fitted.

For example, the function to be fitted is: f(t)=a ₀ +a ₁ *t, and a ₀ =1 and a ₁ =2 are obtained through solution, then the obtained fitting function is: f(t)=1+2t.

S1033, verify each fitting function, and determine a candidate fitting function; wherein, the deviation between the change trend of the object to be monitored represented by the candidate fitting function and the actual change trend of the object to be monitored is the largest.

In this embodiment, the candidate sampling data may include candidate sampling periods and their corresponding actual data values.

Optionally, the process of verifying each fitting function and determining a candidate fitting function may include S10331-S10334.

S10331. For each fitting function, substitute a plurality of candidate sampling data into the fitting function to obtain a predicted data value corresponding to each candidate sampling period.

That is, after each fitting function is obtained through S1032, taking any fitting function as an example, each candidate sampling period t is substituted into the fitting function, and the predicted data value corresponding to each candidate sampling period is calculated, namely f(t ) value.

S10332. Calculate the variance sum corresponding to the fitting function according to the predicted data value and the actual data value corresponding to each candidate sampling period, where the variance sum indicates the variation trend of the object to be monitored represented by the fitting function and the actual value of the object to be monitored. The magnitude of the deviation between trends.

计算拟合函数对应的方差和，其中，f(t_i)表示第i个候选采样周期对应的预测数据值，y_i表示第i个候选采样周期对应的实际数据值。Taking any fitting function as an example, after using the fitting function to calculate the predicted data value corresponding to each candidate sampling period, use the formula

Calculate the variance sum corresponding to the fitting function, where f(t _i ) represents the predicted data value corresponding to the i-th candidate sampling period, and y _i represents the actual data value corresponding to the i-th candidate sampling period.

Since the fitting function represents the change trend of the object to be monitored, the variance value represents the deviation between the change trend of the object to be monitored represented by the fitting function and the actual change trend of the object to be monitored. The larger the variance value, the greater the deviation between the change trend of the object to be monitored represented by the fitting function and the actual change trend of the object to be monitored; the smaller the variance value, the greater the deviation between the change trend of the object to be monitored represented by the fitting function. The smaller the deviation between the trend and the actual change trend of the object to be monitored.

S10333. Calculate the fitting result weight of each fitting function according to the variance sum corresponding to each fitting function, wherein the fitting result weight represents the fitting effect of the fitting function.

In this embodiment, after the variance sum corresponding to each fitting function is calculated, the fitting result weight of each fitting function may be calculated according to the variance sum corresponding to each fitting function. The weight of the fitting result represents the fitting effect of the fitting function. The larger the weight of the fitting result, the worse the fitting effect of the fitting function; the smaller the weight of the fitting result, the better the fitting effect of the fitting function.

Optionally, the process of calculating the fitting result weight of each fitting function according to the variance sum corresponding to each fitting function includes:

According to the variance sum corresponding to each fitting function, use the preset formula

Calculate the fitting result weight of each fitting function;

n表示参与第j次拟合的拟合函数的个数，k表示参与第j次拟合的拟合函数的编号。Among them, W _ij represents the fitting result weight of the i-th fitting function in the j-th fitting; W _i(j-1) represents the weight of the i-th fitting function in the j-1 fitting The weight of the fitting result, and the initial default value of W _i(j-1) is 0; sum _ij represents the variance sum corresponding to the i-th fitting function in the j-th fitting; sum _total represents the participation in the j-th fitting The sum of variances corresponding to each fitting function of the fit,

n represents the number of fitting functions participating in the j-th fitting, and k represents the number of the fitting function participating in the j-th fitting.

S10334. Use the fitting function with the largest fitting result weight as a candidate fitting function.

It should be pointed out that the process of fitting each function to be fitted in S1032 and the process of verifying each fitting function in S1033 can be executed concurrently, thereby shortening the time for multidimensional fitting calculation and ensuring Timeliness of early warning.

S1034. Remove the function to be fitted corresponding to the candidate fitting function from the function list.

In this embodiment, since the greater the weight of the fitting result, the worse the fitting effect of the fitting function is; the smaller the weight of the fitting result is, the better the fitting effect of the fitting function is. Therefore, in the current fitting, it is necessary to find out the candidate fitting function with the largest fitting result weight, and delete the function to be fitted corresponding to the candidate fitting function from the function list. In this case, the function to be fitted Will not participate in the next fitting.

Optionally, each monitoring object in the monitoring object library may also include a fitting interval. The fitting interval refers to the time interval between two fittings of the function to be fitted in the function list corresponding to the monitoring object. That is, after one fitting is completed, wait for the fitting interval before proceeding to the next fitting.

It should be pointed out that the above steps S1031 to S1034 are a process of fitting the functions to be fitted in the function list once. After one fitting (that is, after executing S1034), the function to be fitted in the function list can be detected Whether the number of is 1, if it is not 1, you need to continue to perform subsequent re-fitting, that is, repeat S1031~S1034 until there is only one function to be fitted in the function list; if it is 1, stop subsequent re-fitting and use the fitting function obtained in the last fitting of the only remaining function to be fitted in the function list as the target fitting function.

Optionally, the participating fitting flag can be set for each function to be fitted in the function list, and the participating fitting flag is "true" by default. After completing a fitting, change the participating fitting flag of the function to be fitted corresponding to the candidate fitting function to "false", indicating that it will not participate in the subsequent re-fitting; after that, the participating fitting flag in the statistical function list is "true "The number of functions to be fitted, if the number is greater than 1, use each function to be fitted with the fitting flag "true" for subsequent re-fitting; if the number is equal to 1, the fitting flag will be involved The fitting function obtained in the last fitting of the function to be fitted that is "true" is used as the target fitting function.

S1035, there is only one function to be fitted left in the function list, and the fitting function obtained in the last fitting of the function to be fitted is used as the target fitting function.

Step S105 will be described in detail below.

On the basis of FIG. 2 , please refer to FIG. 5 , step S105 may include S1051-S1054.

S1051. Obtain a sampling period of the last sampling data from the sampling data set.

It can be seen from FIG. 2 that for the object to be monitored, the actual data value of the object to be monitored will be periodically sampled according to the set sampling interval. The sampling process of actual data values is continuous and independent from the function fitting process and the early warning process. Therefore, in order to ensure the accuracy of the early warning, during the early warning process, the early warning trial calculation should be carried out based on the sampling period of the last sampled data in the sampled data set.

S1052. Determine a sampling period to be predicted in the future according to the sampling period, sampling interval and preset warning time offset of the last sampling data.

In this embodiment, according to the preset formula:

Determine a sampling period to be predicted in the future; among them, t _Δ represents the sampling period to be predicted, T _p represents the early warning time offset, Δt represents the sampling interval, and t _max represents the sampling period of the last sampled data.

S1053. Substitute the to-be-predicted sampling period into the target fitting function to obtain the predicted value of the to-be-monitored object in the to-be-predicted sampling period.

In this embodiment, assuming that the target fitting function is f(t)=1+2t, then _t =tΔ is substituted into the target fitting function to obtain the predicted value y _estimate of the object to be monitored in the sampling period to be predicted.

S1054. If the predicted value exceeds the warning threshold, give a warning to the object to be monitored at the warning time; wherein, the warning time is before the sampling period to be predicted, and the warning time is different from the sampling period to be predicted by the warning time offset.

In this embodiment, after obtaining the predicted value y _{estimate of the object to be monitored in the predicted sampling period, the predicted value y estimate is} compared with the warning threshold of the object to be monitored.

If the predicted value does not exceed the warning threshold, it means that the object to be monitored will not be abnormal in the sampling period to be predicted, and the next warning detection is continued, that is, return to S1051.

If the predicted value exceeds the warning threshold, it means that the object to be monitored will be abnormal in the sampling period to be predicted, and an alarm will be given to the object to be monitored at the time of early warning. The warning time is at the warning time offset before the sampling period to be predicted. For example, if the sampling period to be predicted is 16:00 and the warning time offset is 30 minutes, the warning time is 15:30. Obviously, alarming the monitored objects at the early warning moment can reserve a controllable abnormal intervention time for the system operation and maintenance personnel.

Compared with the prior art, the embodiment of the present application has the following beneficial effects:

First, use the time function to predict the future operation of the object to be monitored, and the system operation and maintenance personnel can customize and expand the time function to obtain a time function that is more in line with the monitoring requirements, thus overcoming the way of setting the monitoring threshold in the prior art Difficult problems;

Secondly, multiple time functions are selected from the function library as the functions to be fitted. One function to be fitted is used to represent a change trend of the object to be monitored. Multiple functions to be fitted can be system time functions or extended time functions. , or both the system time function and the extended time function, so as to avoid the problem of lack of a pre-determined corresponding fitting function for the monitoring object in practice, and the actual operability is high;

Third, the least squares method is used to fit the system time function, and the gradient descent method is used to fit the extended time function, which increases the support type for the fitting function and improves the time efficiency of the fitting calculation;

Fourth, in the function fitting calculation, the processing method of parallel execution of multi-fitting calculation is adopted, which greatly shortens the time required for multi-dimensional function fitting calculation and ensures the timeliness of early warning;

Fifth, introduce the fitting result feedback. In the multi-function fitting, through multiple fitting calculations, the fitting variance and the larger function to be fitted are eliminated, and finally the change trend of the characterized to-be-monitored object and the to-be-fit function are selected. The target fitting function with the smallest deviation between the actual change trends of the monitored objects can reduce the error range between the actual data value and the predicted data value when using the target fitting function for subsequent early warning, and improve the accuracy of the early warning;

Sixth, in the early warning stage, if it is predicted that the object to be monitored may be abnormal in a certain sampling period in the future, an alarm will be issued in advance, so that controllable abnormal intervention time can be reserved for system operation and maintenance personnel.

In order to execute the corresponding steps in the foregoing method embodiments and various possible implementation manners, an implementation manner of a data early warning device is given below.

Please refer to FIG. 6 , which shows a schematic block diagram of a data warning device 100 provided by an embodiment of the present application. The data warning device 100 is applied to electronic equipment, and includes: an acquisition module 101 , a fitting module 103 and a warning module 105 .

The obtaining module 101 is configured to obtain a function list corresponding to the object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to represent a change trend of the object to be monitored.

The fitting module 103 is used to perform multiple fittings on the function to be fitted in the function list based on the actual data value of the object to be monitored to obtain the target fitting function; wherein, the target fitting function represents the target to be monitored. The deviation between the change trend and the actual change trend of the object to be monitored is the smallest.

The early warning module 105 is configured to use the target fitting function to give early warning to the object to be monitored.

Optionally, the electronic device stores a function library, the function library includes multiple system time functions and multiple extended time functions, and the multiple extended time functions are customized by system operation and maintenance personnel;

The acquisition module 101 is specifically configured to: in response to a selection operation, acquire a plurality of functions to be fitted from a function library to obtain a function list; wherein, the function to be fitted is at least one of a system time function and an extended time function.

Optionally, the data early warning device 100 also includes a sampling module 104, and the sampling module 104 is used for:

According to the preset sampling interval, the actual data value of the object to be monitored is periodically sampled;

For each sampling period, add the sampling period and its corresponding actual data value as a sampling data to the pre-built sampling data set.

Optionally, the fitting module 103 is specifically used for:

From the sampling data set, obtain the last sampling data and a set number of sampling data before it, and obtain multiple candidate sampling data;

Fitting each function to be fitted in the function list according to multiple candidate sampling data to obtain each fitting function;

Verifying each fitting function, and determining a candidate fitting function; wherein, the deviation between the change trend of the object to be monitored represented by the candidate fitting function and the actual change trend of the object to be monitored is the largest;

Remove the function to be fitted corresponding to the candidate fitting function from the function list;

Repeat the above steps until there is only one function to be fitted left in the function list, and use the fitting function obtained in the last fitting of the function to be fitted as the target fitting function.

Optionally, the fitting module 103 executes fitting each function to be fitted in the function list according to a plurality of candidate sampling data to obtain each fitting function, which may include:

For each function to be fitted, if the function to be fitted is a system time function, then according to multiple candidate sampling data, the least square method is used to solve each coefficient in the function to be fitted to obtain the fitted function;

If the function to be fitted is an extended time function, then use the gradient descent method to solve each coefficient in the function to be fitted according to multiple candidate sampling data, the coefficient partial derivative function group and the coefficient guess value of the extended time function, and obtain the fitting function.

Optionally, the candidate sampling data includes the candidate sampling period and its corresponding actual data value, and the fitting module 103 executes the verification of each fitting function and determines the way of the candidate fitting function, which may include:

For each fitting function, multiple candidate sampling data are substituted into the fitting function to obtain the predicted data value corresponding to each candidate sampling period;

Calculate the variance sum corresponding to the fitting function according to the predicted data value and the actual data value corresponding to each candidate sampling period, where the variance sum indicates the change trend of the object to be monitored represented by the fitting function and the actual change trend of the object to be monitored The size of the deviation between;

Calculate the fitting result weight of each fitting function according to the variance sum corresponding to each fitting function, where the fitting result weight represents the fitting effect of the fitting function;

The fitting function with the largest weight of the fitting result is used as a candidate fitting function.

Optionally, the fitting module 103 executes the method of calculating the fitting result weight of each fitting function according to the variance sum corresponding to each fitting function, which may include:

According to the variance sum corresponding to each fitting function, use the preset formula

Calculate the fitting result weight of each fitting function;

n表示参与第j次拟合的拟合函数的个数，k表示参与第j次拟合的拟合函数的编号。Among them, W _ij represents the fitting result weight of the i-th fitting function in the j-th fitting; W _i(j-1) represents the weight of the i-th fitting function in the j-1 fitting The weight of the fitting result, and the initial default value of W _i(j-1) is 0; sum _ij represents the variance sum corresponding to the i-th fitting function in the j-th fitting; sum _total represents the participation in the j-th fitting The sum of variances corresponding to each fitting function of the fit,

n represents the number of fitting functions participating in the j-th fitting, and k represents the number of the fitting function participating in the j-th fitting.

Optionally, the early warning module 105 is specifically used for:

Obtain the sampling period of the last sampled data from the sampled data set;

Determine a sampling period to be predicted in the future according to the sampling period, sampling interval and preset warning time offset of the last sampling data;

Substitute the sampling period to be predicted into the target fitting function to obtain the predicted value of the object to be monitored in the sampling period to be predicted;

If the predicted value exceeds the early warning threshold, an early warning is given to the monitored object at the early warning time; wherein, the early warning time is before the sampling period to be predicted, and the warning time is different from the sampling period to be predicted by the warning time offset.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the data warning device 100 described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

Please refer to FIG. 7 , which shows a schematic block diagram of an electronic device 10 provided by an embodiment of the present application. The electronic device 10 includes a processor 11 , a memory 12 and a bus 13 , and the processor 11 is connected to the memory 12 through the bus 13 .

The memory 12 is used to store programs, such as the data early warning device 100 shown in Figure 6, the data early warning device 100 includes at least one software function module that can be stored in the memory 12 in the form of software or firmware (firmware), and the processor 11 receives After the instruction is executed, the program is executed to realize the data early warning method disclosed in the foregoing embodiments.

The memory 12 may include a high-speed random access memory (Random Access Memory, RAM), and may also include a non-volatile memory (non-volatile memory, NVM).

The processor 11 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 11 or instructions in the form of software. The above-mentioned processor 11 can be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a micro control unit (Microcontroller Unit, MCU), a complex programmable logic device (Complex Programmable LogicDevice, CPLD), a field programmable gate Array (Field Programmable Gate Array, FPGA), embedded ARM and other chips.

To sum up, in the data early warning method, device, electronic equipment and storage medium provided by the embodiments of the present application, for the object to be monitored, first obtain the function list corresponding to the object to be monitored, the function list includes a plurality of functions to be fitted , a function to be fitted is used to represent a change trend of the object to be monitored; then, based on the actual data value of the object to be monitored, the function to be fitted in the function list is fitted multiple times to obtain the characterized The target fitting function with the smallest deviation between the change trend of the object and the actual change trend of the object to be monitored; finally, the target fitting function is used to give an early warning to the monitored object; thus, an accurate early warning can be given before the fault occurs.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, various modifications and changes may be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included within the protection scope of this application.

Claims (11)

1. A data early warning method, characterized in that being applied to electronic equipment, said method comprising: Obtaining a function list corresponding to the object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to characterize a change trend of the object to be monitored; Based on the actual data value of the object to be monitored, the functions to be fitted in the function list are fitted multiple times to obtain a target fitting function; wherein, the object to be monitored represented by the target fitting function The deviation between the change trend of and the actual change trend of the object to be monitored is the smallest; Using the target fitting function, an early warning is given to the object to be monitored. 2. The method of claim 1, further comprising: periodically sampling the actual data value of the object to be monitored according to a preset sampling interval; For each sampling period, the sampling period and its corresponding actual data value are added as a sampling data to the pre-built sampling data set. 3. the method for claim 2, is characterized in that, described based on the actual data value of described to-be-monitored object, the function to be fitted in the described function list is carried out multiple fitting, obtains target fitting function steps, including: From the sampled data set, obtain the last sampled data and a set number of sampled data before it to obtain a plurality of candidate sampled data; Fitting each of the functions to be fitted in the function list according to the plurality of candidate sampling data to obtain each fitting function; Each of the fitting functions is verified, and a candidate fitting function is determined; wherein, the difference between the change trend of the object to be monitored represented by the candidate fitting function and the actual change trend of the object to be monitored the largest deviation; removing the function to be fitted corresponding to the candidate fitting function from the function list; The above steps are repeated until there is only one function to be fitted left in the function list, and the fitting function obtained in the last fitting of the function to be fitted is used as the target fitting function. 4. The method according to claim 3, wherein, according to the plurality of candidate sampling data, each function to be fitted in the function list is fitted to obtain each fitting The steps of the function, including: For each of the functions to be fitted, if the function to be fitted is a system time function, then according to the plurality of candidate sampling data, the least squares method is used to solve each coefficient in the function to be fitted to obtain the said fitting function; If the function to be fitted is an extended time function, then according to the plurality of candidate sampling data, the coefficient partial derivative function group and the coefficient guess value of the extended time function, the gradient descent method is used to solve the function to be fitted Each coefficient in , get the fitting function. 5. The method according to claim 3, wherein the candidate sampling data comprises a candidate sampling period and its corresponding actual data value; The step of verifying each of the fitting functions and determining the candidate fitting functions includes: For each of the fitting functions, substituting the plurality of candidate sampling data into the fitting function to obtain predicted data values corresponding to each of the candidate sampling periods; Calculate the variance sum corresponding to the fitting function according to the predicted data value and the actual data value corresponding to each of the candidate sampling periods, wherein the variance sum indicates the value of the object to be monitored represented by the fitting function The deviation between the change trend and the actual change trend of the object to be monitored; According to the variance sum corresponding to each of the fitting functions, the fitting result weight of each of the fitting functions is calculated, wherein the fitting result weight represents the fitting effect of the fitting function; The fitting function with the largest weight of the fitting result is used as the candidate fitting function. 6. The method according to claim 5, wherein the step of calculating the fitting result weight of each of the fitting functions according to the corresponding variance of each of the fitting functions includes: According to the sum of variances corresponding to each of the fitting functions, using a preset formula

calculating a fitting result weight of each of the fitting functions; Among them, W _ij represents the fitting result weight of the i-th fitting function in the j-th fitting; W _i(j-1) represents the i-th fitting function in the j-1 fitting The weight of the fitting result in the fitting, and the initial default value of W _i(j-1) is 0; sum _ij represents the corresponding variance sum of the i-th fitting function in the j-th fitting; sum _total Indicates the sum of the variance sums corresponding to each of the fitting functions participating in the jth fitting,

n represents the number of the fitting functions participating in the j-th fitting, and k represents the number of the fitting functions participating in the j-th fitting. 7. The method according to claim 2, wherein the step of using the target fitting function to warn the object to be monitored includes: Obtain the sampling period of the last sampling data from the sampling data set; Determine a sampling period to be predicted in the future according to the sampling period of the last sampling data, the sampling interval and the preset warning time offset; Substituting the to-be-predicted sampling period into the target fitting function to obtain the predicted value of the to-be-monitored object in the to-be-predicted sampling period; If the predicted value exceeds the early warning threshold, an early warning is given to the object to be monitored at the early warning time; Wherein, the warning time is before the to-be-predicted sampling period, and the difference between the early-warning time and the to-be-predicted sampling period is the warning time offset. 8. The method according to claim 1, wherein the electronic device stores a function library, the function library includes a plurality of system time functions and a plurality of extended time functions, and the plurality of extended time functions are system Customized by operation and maintenance personnel; The step of obtaining the function list corresponding to the object to be monitored includes: Responding to the selection operation, obtaining a plurality of the functions to be fitted from the function library to obtain the function list; wherein the function to be fitted is at least one of the system time function and the extended time function kind. 9. A data early warning device, characterized in that it is applied to electronic equipment, and the device includes: An acquisition module, configured to acquire a function list corresponding to the object to be monitored, wherein the function list includes a plurality of functions to be fitted, and one function to be fitted is used to represent a change trend of the object to be monitored; A fitting module, configured to perform multiple fittings on the functions to be fitted in the function list based on the actual data values of the objects to be monitored to obtain a target fitting function; wherein, the target fitting function represented by The deviation between the change trend of the object to be monitored and the actual change trend of the object to be monitored is the smallest; The early warning module is configured to use the target fitting function to give early warning to the object to be monitored. 10. An electronic device, characterized in that it comprises a processor and a memory, the memory is used to store a program, and the processor is used to implement the method described in any one of claims 1-8 when executing the program. Data alert method. 11. A computer-readable storage medium, wherein a computer program is stored thereon, and when the computer program is executed by a processor, the data early warning method according to any one of claims 1-8 is implemented.

Images