Running day anomaly identification method based on similar scene analysis

Abstract

The invention discloses an operation day anomaly identification method based on similar scene analysis, and belongs to the technical field of power dispatching, and the method comprises the steps: S1, obtaining the deviation degree of a historical numerical value of an influence factor and a reference value of the influence factor based on the type of the influence factor; s2, acquiring a plurality of similar scenes corresponding to the influence factors based on the deviation degree, and acquiring reference values of the power operation key indexes in the similar scenes based on historical data rules of the power operation key indexes; and S3, according to the real-time data of the to-be-determined operation day, obtaining a similar scene to which the to-be-determined operation day belongs, and carrying out anomaly identification on the to-be-determined operation day based on absolute deviation between an actual value of a corresponding index in the real-time data of the to-be-determined operation day and a reference value of a power operation key index in the similar scene to which the to-be-determined operation day belongs. The method solves the problem of low anomaly discrimination accuracy caused by lack of analysis of the change rule of the power operation key index value in different operation scenes formed by different influence factors from the influence factors.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power dispatching, and in particular to an operating day anomaly identification method based on similar scenario analysis. Background Art

[0002] With the continuous progress and expansion of the power system, the abnormal identification of key indicators of power operation is an important technical content of power dispatching and operation management. Its role is to monitor the changes in the values ​​of key indicators of power operation, timely identify abnormal changes in indicator data, and judge the operating status based on abnormal changes. At present, the abnormal identification of key indicators of power operation mainly adopts expert threshold method and abnormal data mining method to achieve abnormal identification. For example, the patent number CN117743994A is a method and system for evaluating the abnormal identification of input data of a power prediction model, which includes: obtaining the historical measured data of meteorological elements of new energy stations, and for each meteorological element, The elements are classified; the probability density function and cumulative probability density function corresponding to the historical measured data of each meteorological element at any time under each type are obtained; the confidence interval of the cumulative probability density function of each meteorological element at any time under each type is set to obtain the upper and lower limits of the normal data range of each meteorological element at any time under each type; the envelope of the normal data corresponding to each meteorological element at different times under each type is determined to perform online anomaly identification of real-time input data; based on the results of anomaly identification, the evaluation indicators of the proportion of input data anomaly identification evaluation methods and the degree of anomaly are designed to complete the quantitative evaluation of the quality of real-time input data of the power prediction model. The above scheme analyzes the historical change rules of key indicators, improves the rationality of the threshold set by experts for anomaly identification, and thus improves the prediction accuracy. However, when identifying the abnormal changes of key indicators of power operation, the above scheme does not distinguish the operation scenarios. When identifying the abnormal changes of key indicators of power operation with the same threshold, there is an inaccurate anomaly identification. Summary of the invention

[0003] In view of the problem in the prior art that the accuracy of abnormality identification is low due to the lack of analysis of the changing rules of the values ​​of key indicators of power operation under different operating scenarios constituted by different influencing factors starting from the influencing factors, the present invention provides an operating day abnormality identification method based on similar scenario analysis, which obtains several similar scenarios corresponding to the influencing factors through skewness difference, and obtains reference values ​​of the key indicators of power operation in each similar scenario respectively; before judging the operating status of the operating day to be determined, the similar scenario to which the operating day to be determined belongs is first determined and then compared with the reference value under the scenario to which it belongs, and then targeted abnormality identification is performed, which solves the problem that the accuracy of abnormality identification is low due to the lack of analysis of the changing rules of the values ​​of key indicators of power operation under different operating scenarios constituted by different influencing factors starting from the influencing factors, and significantly improves the accuracy of abnormality identification.

[0004] In order to solve the above technical problems, the present invention provides an operation day anomaly identification method based on similar scenario analysis, comprising the following steps: S1: Obtain the influencing factors of the key indicators of power operation based on the correlation model of the key indicators of power operation, and obtain the deviation between the historical values ​​of the influencing factors and the benchmark values ​​of the influencing factors based on the types of the influencing factors; S2: Based on the deviation, several similar scenarios corresponding to the influencing factors are obtained, and based on the historical data rules of the key indicators of power operation, reference values ​​of the key indicators of power operation under similar scenarios are obtained; S3: Obtain similar scenarios to which the operation day to be determined belongs based on the real-time data of the operation day to be determined, and identify abnormalities on the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario.

[0005] After adopting the above technical solution, the present invention has the following advantages: Considering that the historical values ​​of the influencing factors correspond to different scenarios, the historical values ​​of the influencing factors are divided by the deviation values ​​to obtain similar scenarios under the influencing factors, and the reference values ​​of the key indicators of power operation in each similar scenario are obtained respectively. Before judging the operating status of the operating day to be determined, the similar scenario to which the operating day to be determined belongs is first determined and then compared with the reference values ​​under the scenario to which it belongs, and then targeted abnormality analysis is performed, thereby significantly improving the accuracy of abnormality analysis; By calculating the degree of deviation, different influencing factors are divided according to the degree of deviation, and complex similar scenes are carefully combined to achieve the division of complex scenes. At the same time, it can flexibly adapt to various possible complex scenes, enhancing the flexibility and adaptability of identification; By pre-calculating and storing the reference value of each similar scene, the time for judging the status of the operating day to be determined is greatly shortened. When it is necessary to judge the status of the operating day to be determined, it is only necessary to quickly determine the similar scene to which it belongs and compare it with the pre-stored reference value, thereby improving the recognition efficiency and real-time performance; The problem of low accuracy in abnormality identification due to the lack of analysis of the changing patterns of key indicators of power operation under different operating scenarios composed of different influencing factors based on the influencing factors is solved.

[0006] Preferably, in S1, the expression for obtaining the influencing factors of the key power operation indicators based on the correlation model of the key power operation indicators is I=F(f1, f2, ...f N ), where I represents the key indicator of power operation, f1, f2, ... f N Both represent the influencing factors, N is the number of the influencing factors, and F is the correlation model.

[0007] Preferably, in S1, the deviation between the historical value of the influencing factor and the benchmark value of the influencing factor obtained based on the influencing factor type includes: Obtaining a benchmark value of the influencing factor based on the influencing factor type and the historical value of the influencing factor; If the influencing factor type is the first type, the deviation degree is obtained based on the Euclidean distance between the historical value and the reference value; if the influencing factor type is the second type, the deviation degree is obtained based on the difference between the historical value and the reference value.

[0008] In this scheme, the first type is a series of numerical types and there is a correlation between the series numerical types. The Euclidean distance is used as the deviation evaluation indicator, and the differences between all the numerical values ​​in the series are comprehensively considered, thereby improving the accuracy of the calculation. The second type is a data type that can directly perform four arithmetic operations. The deviation is obtained by directly obtaining the difference, which improves the efficiency of obtaining the deviation. Different calculation methods are used for different types of influencing factors, which can more accurately reflect the difference between its historical value and the benchmark value, and enhance the adaptability of the scheme.

[0009] Preferably, obtaining the benchmark value of the influencing factor based on the influencing factor type and the historical value of the influencing factor includes: if the influencing factor type is the third type, obtaining the benchmark value based on the direct statistical method; if the influencing factor type is the fourth type, obtaining the benchmark value based on the interval statistical method.

[0010] In this scheme, different statistical methods are used for different types of influencing factors, which can more accurately reflect the characteristics and laws of their historical values, help ensure the accuracy and reliability of the benchmark values, and at the same time enable the scheme to adapt to different types of influencing factors, thereby enhancing the adaptability of the scheme.

[0011] Preferably, the obtaining of the reference value based on interval statistics method includes: Based on the characteristics of the historical values, the historical values ​​are divided to obtain a number of first data intervals, and the number of times that data points in the data set of the key indicators of power operation fall in the first data intervals is counted, and the median of the first data interval with the highest number is the benchmark value.

[0012] Preferably, S2 includes: S21: constructing a combination rule according to the full coverage and uniqueness principle, dividing the historical values ​​based on the deviation to obtain a plurality of second data intervals, and combining the second data intervals according to the combination rule to obtain the plurality of similar scenarios; S22: Obtain reference values ​​of key power operation indicators under similar scenarios based on historical data patterns of key power operation indicators.

[0013] In this scheme, different influencing factors are divided by deviation, and simple and single scenarios are combined according to combination rules to form complex similar scenarios, thereby realizing the division of complex scenarios. At the same time, it can flexibly adapt to various possible complex scenarios, enhancing the flexibility and adaptability of identification. By obtaining the reference values ​​of key indicators of power operation in each similar scenario respectively, before judging the operating status of the operating day to be determined, the similar scenario to which the operating day to be determined belongs is first determined and then compared with the reference value under the scenario to which it belongs, and then targeted abnormal analysis is performed, thereby significantly improving the accuracy of abnormal analysis.

[0014] Preferably, the S22 includes: In the formula, Represents the reference value of key indicators of power operation under similar scenario H, Take the values ​​of the key power operation indicators of the historical operation day d belonging to the similar scenario H, is the mode function of the key indicators of power operation on the historical operation day d belonging to the similar scenario H.

[0015] Preferably, in S3, the abnormality identification of the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario includes: If the absolute deviation is greater than the preset absolute deviation, the operation day to be determined is judged to be abnormal; if it is less than or equal to the preset absolute deviation, the operation day to be determined is judged to be normal.

[0016] Preferably, the second type includes a continuous numerical type, a discrete numerical type and a character numerical type.

[0017] Preferably, the third type includes the discrete numerical type and the character numerical type, and the fourth type includes the continuous numerical type and the series numerical type.

[0018] The beneficial effects of this program: Considering that the historical values ​​of the influencing factors correspond to different scenarios, the historical values ​​of the influencing factors are divided by the deviation values ​​to obtain similar scenarios under the influencing factors, and the reference values ​​of the key indicators of power operation in each similar scenario are obtained respectively. Before judging the operating status of the operating day to be determined, the similar scenario to which the operating day to be determined belongs is first determined and then compared with the reference values ​​under the scenario to which it belongs, and then targeted abnormality analysis is performed, thereby significantly improving the accuracy of abnormality analysis; By calculating the degree of deviation, different influencing factors are divided according to the degree of deviation, and simple and single scenes are combined according to the combination rules to form complex similar scenes, the division of complex scenes is realized, and at the same time, it can flexibly adapt to various possible complex scenes, thereby enhancing the flexibility and adaptability of identification; By pre-calculating and storing the reference value of each similar scene, the time for judging the status of the operating day to be determined is greatly shortened. When it is necessary to judge the status of the operating day to be determined, it is only necessary to quickly determine the similar scene to which it belongs and compare it with the pre-stored reference value, thereby improving the recognition efficiency and real-time performance; The problem of low accuracy in abnormality identification due to the lack of analysis of the changing patterns of key indicators of power operation under different operating scenarios composed of different influencing factors based on the influencing factors is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0020] Figure 1 The present invention is a flow chart of the method for identifying anomalies on operating days based on similar scenario analysis. DETAILED DESCRIPTION

[0021] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0022] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0023] Embodiment 1: like Figure 1 As shown, the method for identifying abnormalities on operating days based on similar scenario analysis includes the following steps: S1: Obtain influencing factors of the key indicators of power operation based on the correlation model of the key indicators of power operation, and obtain the deviation between the historical values ​​of the influencing factors and the benchmark values ​​of the influencing factors based on the types of the influencing factors.

[0024] In S1, the expression for obtaining the influencing factors of the key power operation indicators based on the correlation model of the key power operation indicators is I=F(f1, f2, ...f N ), where I represents the key indicator of power operation, f1, f2, ... f N Both represent the influencing factors, N is the number of the influencing factors, and F is the correlation model.

[0025] In S1, the deviation between the historical value of the influencing factor obtained based on the influencing factor type and the benchmark value of the influencing factor includes: Obtaining a benchmark value of the influencing factor based on the influencing factor type and the historical value of the influencing factor; If the influencing factor type is the first type, the deviation degree is obtained based on the Euclidean distance between the historical value and the reference value; if the influencing factor type is the second type, the deviation degree is obtained based on the difference between the historical value and the reference value.

[0026] The base value of the influencing factor obtained based on the influencing factor type and the historical value of the influencing factor includes: If the influencing factor type is the third type, the reference value is obtained based on a direct statistical method; if the influencing factor type is the fourth type, the reference value is obtained based on an interval statistical method.

[0027] The obtaining of the reference value based on the interval statistics method comprises: Based on the characteristics of the historical values, the historical values ​​are divided to obtain a number of first data intervals, and the number of times that data points in the data set of the key indicators of power operation fall in the first data intervals is counted, and the median of the first data interval with the highest number is the benchmark value.

[0028] The second type includes a continuous numeric type, a discrete numeric type, and a character numeric type.

[0029] The third type includes the discrete numeric type and the character numeric type, and the fourth type includes the continuous numeric type and the series numeric type.

[0030] In this embodiment, the influencing factors of the key indicators of power operation refer to a set of related factors that have a great influence on the value of the key indicators of power operation. For example, the influencing factors of the load forecast accuracy index generally include: date, sunshine, temperature, wind power, production plan of large industrial users, etc. The correlation model can accurately identify the factors that have a significant correlation with the key indicators of power operation based on a large amount of basic information through statistical analysis and calculation. The influencing factors are extracted by corresponding related factors of the basic information selected in the correlation model, which avoids the subjectivity and uncertainty of human judgment and improves the accuracy of analysis.

[0031] In this embodiment, the first type is a numerical series type, and the degree of deviation is obtained based on the Euclidean distance between the historical value and the reference value as follows: In the formula, is the deviation between the historical value of the numerical influencing factor f and the benchmark value of the influencing factor f, that is, the deviation degree of the influencing factor f, N f is the number of values ​​in the sequence of the influencing factor, The historical value, i.e., the historical value of the nth value of the influencing factor, and the benchmark value are used respectively. By using the Euclidean distance as the deviation evaluation index, the differences between all the values ​​in the series are comprehensively considered, thereby improving the accuracy of the calculation. The deviation degree is obtained based on the difference between the historical value and the benchmark value as follows: In the formula, Represent the historical value and benchmark value of the influencing factor f, respectively. The deviation between the historical value of the continuous numerical type, discrete numerical type or character numerical type influencing factor f and the benchmark value of the influencing factor f is the influencing factor f deviation. The deviation is obtained by directly obtaining the difference, which improves the efficiency of obtaining the deviation. Different calculation methods are used for different types of influencing factors, which can more accurately reflect the difference between their historical values ​​and benchmark values, and enhance the adaptability of the solution.

[0032] In this embodiment, when the third type is a discrete numerical type, that is, the influencing factor takes a numerical value, but it is a discrete numerical value, such as the "date" in the load forecast accuracy influencing factor. The direct statistical method is used to calculate its benchmark value, that is, the number of occurrences of historical data with different values ​​is directly counted, and the one with the largest number of occurrences is the benchmark value of the discrete numerical influencing factor; when the third type is a character numerical type, that is, the influencing factor takes a character, and is generally a finite type, such as the "date" in the load forecast accuracy influencing factor, the conversion method is used to convert it into a discrete numerical type, and the direct statistical method is used to calculate its benchmark value; when the fourth type is a continuous numerical type, that is, the influencing factor takes a numerical value, and it takes continuous values, such as the "production plan of large industrial users" in the load forecast accuracy influencing factor, the method is used. The interval statistics method is used to calculate its benchmark value, and the number of times the data points in the data set of the key indicators of power operation fall into the first data interval is used as the estimated value of the probability of occurrence. The median of the first data interval with the largest probability of occurrence is the benchmark value of the continuous numerical influencing factor; when the fourth type is a series numerical type, that is, the influencing factor value is a series, such as the "temperature" in the load forecast accuracy influencing factor, the series type value can be regarded as a sequential combination of multiple continuous numerical types, and the interval statistics method can be used to calculate the benchmark value of each sequential position, and then arrange them in order to obtain the benchmark value of the series type value. Using different statistical methods for different types of influencing factors can more accurately reflect the characteristics and laws of their historical values, help ensure the accuracy and reliability of the benchmark value, and at the same time enable the scheme to adapt to different types of influencing factors, thereby enhancing the adaptability of the scheme.

[0033] S2: Based on the deviation degree, several similar scenarios corresponding to the influencing factors are obtained, and based on the historical data rules of the key indicators of power operation, reference values ​​of the key indicators of power operation under similar scenarios are obtained.

[0034] The S2 includes: S21: constructing a combination rule according to the full coverage and uniqueness principle, dividing the historical values ​​based on the deviation to obtain a plurality of second data intervals, and combining the second data intervals according to the combination rule to obtain the plurality of similar scenarios; S22: Obtain reference values ​​of key power operation indicators under similar scenarios based on historical data patterns of key power operation indicators.

[0035] The S22 includes: In the formula, Represents the reference value of key indicators of power operation under similar scenario H, Take the values ​​of the key power operation indicators of the historical operation day d belonging to the similar scenario H, is the mode function of the key indicators of power operation on the historical operation day d belonging to the similar scenario H.

[0036] In this embodiment, the full coverage and unique principle is specifically: if the second data interval corresponding to the historical value of the first influencing factor is A and B, the second data interval corresponding to the historical value of the second influencing factor is D and E, and the second data interval corresponding to the historical value of the third influencing factor is H and G, then several similar scenes are ADH, ADG, AEH, AEG, BDH, BDG, BEH, BEG. Several similar scenes are obtained by combining according to the full coverage and unique principle, which avoids the loss of some scene features, and also avoids overlap and redundancy between scenes, so that each scene represents a different situation and feature combination, thereby improving the comprehensiveness and accuracy of the obtained similar scenes.

[0037] S3: Obtain similar scenarios to which the operation day to be determined belongs based on the real-time data of the operation day to be determined, and identify abnormalities on the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario.

[0038] In S3, the abnormality identification of the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario includes: If the absolute deviation is greater than the preset absolute deviation, the operation day to be determined is judged to be abnormal; if it is less than or equal to the preset absolute deviation, the operation day to be determined is judged to be normal.

[0039] In this embodiment, the deviation between the real-time data of the operating day to be determined and the benchmark value is calculated, and the similar scene to which the operating day to be determined belongs is determined by the deviation. Then, the actual value of the corresponding indicator in the real-time data of the operating day to be determined is compared with the reference value of the scene to which it belongs to determine whether the operating day is abnormal. By pre-calculating and storing the reference value of each similar scene, the time for judging the status of the operating day to be determined is greatly shortened. When it is necessary to judge the status of the operating day to be determined, it is only necessary to quickly determine the similar scene to which it belongs and compare it with the pre-stored reference value, thereby improving the recognition efficiency and real-time performance.

[0040] The specific implementation described above is a preferred implementation of the operating day anomaly identification method based on similar scene analysis of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. The method for identifying abnormalities on operating days based on similar scenario analysis is characterized by: The following steps are involved: S1: Obtain the influencing factors of the key indicators of power operation based on the correlation model of the key indicators of power operation, and obtain the deviation between the historical values ​​of the influencing factors and the benchmark values ​​of the influencing factors based on the types of the influencing factors; S2: Based on the deviation, several similar scenarios corresponding to the influencing factors are obtained, and based on the historical data rules of the key indicators of power operation, reference values ​​of the key indicators of power operation under similar scenarios are obtained; S3: Obtain similar scenarios to which the operation day to be determined belongs based on the real-time data of the operation day to be determined, and identify abnormalities on the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario.

2. The method for identifying anomalies on operating days based on similar scenario analysis according to claim 1, characterized in that: In S1, the expression for obtaining the influencing factors of the key power operation indicators based on the correlation model of the key power operation indicators is I=F ( f1,f2,...f N) , where I represents the key indicators of power operation, f1,f2,...f N Both represent the influencing factors, N is the number of the influencing factors, and F is the correlation model.

3. The method for identifying anomalies on operating days based on similar scenario analysis according to claim 1, characterized in that: In S1, the deviation between the historical value of the influencing factor obtained based on the influencing factor type and the benchmark value of the influencing factor includes: Obtaining a benchmark value of the influencing factor based on the influencing factor type and the historical value of the influencing factor; If the influencing factor type is the first type, the deviation degree is obtained based on the Euclidean distance between the historical value and the reference value; if the influencing factor type is the second type, the deviation degree is obtained based on the difference between the historical value and the reference value.

4. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 3 is characterized in that: The base value of the influencing factor obtained based on the influencing factor type and the historical value of the influencing factor includes: If the influencing factor type is the third type, the reference value is obtained based on a direct statistical method; if the influencing factor type is the fourth type, the reference value is obtained based on an interval statistical method.

5. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 4 is characterized in that: The obtaining of the reference value based on the interval statistics method comprises: Based on the characteristics of the historical values, the historical values ​​are divided to obtain a number of first data intervals, and the number of times that data points in the data set of the key indicators of power operation fall in the first data intervals is counted, and the median of the first data interval with the highest number is the benchmark value.

6. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 1, characterized in that: The S2 includes: S21: constructing a combination rule according to the full coverage and uniqueness principle, dividing the historical values ​​based on the deviation to obtain a plurality of second data intervals, and combining the second data intervals according to the combination rule to obtain the plurality of similar scenarios; S22: Obtain reference values ​​of key power operation indicators under similar scenarios based on historical data patterns of key power operation indicators.

7. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 6 is characterized in that: The S22 includes: In the formula, represents the reference value of key indicators of power operation under similar scenario H, Take the values ​​of the key power operation indicators of the historical operation day d belonging to the similar scenario H, is the mode function of the key indicators of power operation on the historical operation day d belonging to the similar scenario H.

8. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 1, characterized in that: In S3, the abnormality identification of the operation day to be determined based on the absolute deviation between the actual value of the corresponding indicator in the real-time data of the operation day to be determined and the reference value of the key indicator of power operation in the similar scenario includes: If the absolute deviation is greater than the preset absolute deviation, the operation day to be determined is judged to be abnormal; if it is less than or equal to the preset absolute deviation, the operation day to be determined is judged to be normal.

9. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 4, characterized in that: The second type includes a continuous numeric type, a discrete numeric type, and a character numeric type.

10. The method for identifying abnormalities on operating days based on similar scenario analysis according to claim 9, characterized in that: The third type includes the discrete numeric type and the character numeric type, and the fourth type includes the continuous numeric type and the series numeric type.

Images