Method and device for studying, judging and accurately tracking power distribution network line loss abnormity considering new energy access

Abstract

The invention discloses a new energy access-considered power distribution network line loss abnormity research and judgment and accurate tracking method and device, and the method comprises the steps: constructing a new energy line loss data generation model based on a conditional generative adversarial network, and taking a new energy output joint probability distribution function as a conditional variable; inputting the real line loss data of the regional power grid into a new energy line loss data generation model, and generating new energy access power distribution line loss data; performing power distribution network area scene division on the curve corresponding to the line loss data of the jurisdiction area of the power distribution network; constructing a power distribution network area line loss rate standard library, combining similar or similar scenes, and then carrying out anomaly study and judgment; and analyzing each factor of the line loss rate abnormity of the transformer area and an internal correlation degree between transformer area users, and realizing accurate positioning of the line loss abnormity associated users based on an improved Adtributtor algorithm. According to the invention, intelligent study and judgment of the line loss abnormity of the transformer area and accurate tracking of abnormal users are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of line loss fault determination, and specifically relates to a method for abnormal determination and precise tracking of line loss in a distribution network taking into account the access of new energy sources. Background Art

[0002] Lean management of line loss in substations has gradually become the mainstream, and energy conservation and loss reduction work has also achieved remarkable results. At present, problems such as abnormal line loss rate in substations and user electricity theft are still common, and their troubleshooting methods still rely mainly on the experience of operation and maintenance personnel for judgment. This not only greatly increases the workload of operation and maintenance personnel, but also makes some abnormalities difficult to find, resulting in power loss. Therefore, how to establish a complete intelligent analysis and judgment system for abnormal line loss rate in substations and use this as a basis to quickly and accurately find the related users who cause abnormal line loss rate in substations has become urgent.

[0003] At present, the research on line loss rate by researchers at home and abroad mainly focuses on the calculation and prediction of line loss rate, while the identification and tracking of abnormal line loss rate in the substation area are rarely involved. Some scholars have proposed a method for deep identification of abnormal line loss data in the same period based on the approximate equality characteristic of "rank sum". On the basis of the traditional α threshold calculation, an improved α∩β threshold calculation method is adopted to solve the problem of difficulty in identifying different types of abnormal data in the substation area; while other scholars have proposed a line loss anomaly identification method based on secondary K-means clustering based on the real data of low-voltage substation area to effectively identify line loss anomalies. The simulation results show that this method has largely solved the one-sidedness and limitations of traditional line loss anomaly identification methods. Although the above methods have analyzed the abnormal line loss rate problem in the substation area and achieved good results, they have not further explored the intrinsic relationship between the line loss rate and the users, and lack further research and analysis on the users who cause the abnormal line loss rate in the substation area. Summary of the invention

[0004] In view of the shortcomings in the prior art, the present invention provides a method for abnormal analysis and precise tracking of distribution network line losses taking into account the access of new energy sources. It takes into account the situation of new energy access to the distribution network, and is different from the method for abnormal analysis of substation line loss rate in traditional distribution network scenarios. It utilizes data between systems, divides typical scenarios, and builds a standard library to achieve intelligent analysis of substation line loss anomalies and precise tracking of abnormal users.

[0005] The present invention achieves the above technical objectives through the following technical means.

[0006] Abnormal analysis and accurate tracking method of distribution network line loss considering the access of new energy: Construct the joint probability distribution function of renewable energy output; Establish a new energy line loss data generation model based on a conditional generative adversarial network, and input the new energy output joint probability distribution function as a conditional variable and the real line loss data of the regional power grid into the new energy line loss data generation model to generate the line loss data of the distribution network connected to the new energy; For the curve corresponding to the line loss data of the distribution network, the ICEEMDAN algorithm is used to realize the distribution network substation scene division; The distribution network substation scene division results are divided once to obtain the distribution network substation line loss rate standard library; Based on the interval dynamic translation method, the line loss rate standard library of the distribution network substation is divided twice; Based on the interval overlap rate strategy, similar or close scenarios in the line loss rate standard library of the distribution network area after secondary division are merged; Conduct abnormal analysis based on the combined distribution network area line loss rate standard database; The factors causing abnormal line loss rate in the substation area and the intrinsic correlation between users in the substation area are analyzed, and the users associated with abnormal line loss are accurately located based on the improved Adtributor algorithm.

[0007] Furthermore, the ICEEMDAN algorithm is used to realize the distribution network area scenario division, specifically: The curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network is decomposed and reconstructed to obtain the preprocessed line loss data curve of the substation under the jurisdiction of the distribution network, and the "burr" data is removed to obtain the line loss change trend of the substation. The line loss change trend of the substation is clustered and analyzed, and then the distribution network substation scene division result is obtained.

[0008] Furthermore, the distribution network area scene division result is divided once, specifically: According to the distribution network area scenario division results, the line loss rates under the same scenario are stored in the same set , for the set The line loss rate data in the network is mined and analyzed, the number of clusters is dynamically adjusted, and the silhouette coefficient corresponding to each cluster number is calculated to determine the optimal number of clusters. ; Use K-means++ to perform cluster analysis and extract the cluster with the largest number of cases ; Will The minimum value of the elements in is used as the lower limit of the interval, and the maximum value is used as the upper limit of the interval, and finally the line loss rate standard library after one division is obtained.

[0009] Furthermore, the line loss rate standard library of the distribution network area is divided twice by the interval dynamic translation, specifically: The pre-processed distribution network substation line loss data is generated by K-means++ clustering clusters, keep the cluster with the largest number of cases ; Assumed Cluster The number of cases is less than , select cluster and , calculated relative to The offset is calculated as:

[0010] Where: , Represents clusters and Relative to cluster The offset of and Cluster and The length of the interval, and Cluster and The number of cases is the total sample size; Combined with the relative offset results, the standard library interval of the line loss rate in the substation area is solved and updated as follows:

[0011] Where: , They represent the lower and upper limits of the standard library interval of the area line loss rate, and Respectively represent clusters The minimum and maximum values ​​of the inner elements.

[0012] Furthermore, similar or close scenes are merged, specifically: Construct the overlap rate of two scenes: Assume A and B are two fixed intervals, then the interval overlap rate of A and B is expressed as:

[0013] Where: L represents the length of the corresponding interval, (A∩B) represents the overlapping interval of A and B; Set an adaptive threshold for interval overlap α ,pass α The inter-class variance obtained by segmentation is used as the evaluation index. The specific formula is as follows:

[0014] Where: is the between-class variance, and Respectively represent [0 ,α )and[ α, 1] The proportion of the two parts, u 1 and u 2 are the average values ​​of the overlap rate of each part; Find The threshold corresponding to the maximum is taken as the optimal threshold α best , and the overlap ratio is greater than or equal to α best The standard library intervals are merged according to the union principle, otherwise they are not merged.

[0015] Furthermore, the grey correlation analysis method is used to analyze the internal correlation between the various factors of abnormal line loss rate in the substation area and the users in the substation area. For the three factors with the largest grey correlation, the most suspicious users are excavated through the improved Adtributor algorithm, so as to achieve the accurate positioning of users associated with line loss anomalies.

[0016] Furthermore, the accurate positioning of users associated with abnormal line loss is achieved based on the improved Adtributor algorithm, specifically: For the three factors with the largest grey correlation, the most suspicious users are mined through the Adtributor algorithm. The algorithm involves the calculation of two parameters: the unexpectedness value and the explanatory power value EP. The explanatory power value EP satisfies:

[0017] Where: A ij Indicates the true value of the fault, F ij represents the normal predicted value, the subscript i and j Represents the date and user number respectively, m is the abnormal index, A( m )、F( m ) are A ij 、F ij The sum of The F( m ) is realized by the following improved method: extract the line loss rate data in the normal time period of the substation area, use the Lagrange interpolation method to solve the estimated value of the abnormal time period, and use it as F( m ) to calculate the explanatory power value.

[0018] A device for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy sources, comprising: A new energy line loss data generation model construction module is established, which builds a new energy line loss data generation model based on a conditional generative adversarial network, and uses the new energy output joint probability distribution function as a conditional variable. The real line loss data of the regional power grid is input into the new energy line loss data generation model to generate the line loss data of the distribution network connected to the new energy; The multi-dimensional scenario division module of the substation line loss rate uses the ICEEMDAN algorithm to realize the distribution network substation scenario division for the curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network; The abnormal intelligent judgment module based on the line loss rate standard library of the distribution network is used to build the line loss rate standard library of the distribution network and merge similar or close scenes, and then conduct abnormal analysis and judgment; The abnormal correlation of line loss rate in the substation area is used for precise positioning module, which analyzes the internal correlation between various factors of abnormal line loss rate in the substation area and users in the substation area, and realizes precise positioning of users associated with abnormal line loss rate based on the improved Adtributor algorithm.

[0019] An electronic device comprising a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the above-mentioned distribution network line loss abnormality analysis and precise tracking method taking into account the access of new energy when executing the computer program.

[0020] A storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the above-mentioned distribution network line loss abnormality analysis and precise tracking method taking into account the access of new energy.

[0021] The beneficial effects of the present invention are: (1) The present invention takes into account the situation of renewable energy access to the distribution network, constructs a new energy output joint probability distribution function, and uses the new energy output joint probability distribution function as a conditional variable of a new energy line loss data generation model, thereby improving the data generation quality of the new energy line loss data generation model, and the generated distribution network line loss data is more in line with the regional renewable energy output law; (2) The present invention uses the ICEEMDAN algorithm to implement distribution network substation scene division for the curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network, divides the distribution network substation scene division result once, and obtains the distribution network substation line loss rate standard library. The distribution network substation line loss rate standard library is divided twice by interval dynamic translation, thereby improving the interval accuracy of the line loss rate standard library; (3) Based on the interval overlap rate strategy, the present invention merges similar or close scenes in the line loss rate standard library of the distribution network area after secondary division, reduces the amount of calculation for abnormal analysis and precise positioning, and improves calculation efficiency; (4) The present invention uses the grey correlation analysis method to analyze the internal correlation between the various factors of abnormal line loss rate in the substation area and the users in the substation area, and then uses the improved Adtributor algorithm to accurately locate the users associated with abnormal line loss, thereby reducing the probability of misjudgment of abnormal associated users. Compared with the traditional correlation algorithm, it can more accurately discover the users associated with abnormal line loss rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of the method for abnormal analysis and accurate tracking of line losses in a distribution network taking into account the access of new energy sources in the present invention; Figure 2 This is a conceptual diagram of a new energy line loss data generation model based on a conditional adversarial generation network of the present invention; Figure 3 It is a modal decomposition effect diagram of the line loss rate curve under the ICEEMDAN algorithm of the present invention; Figure 4 It is a scene division result diagram under the K-means clustering strategy of the present invention; Figure 5 It is a scene division result diagram under the fusion clustering strategy of the present invention; Figure 6 It is a schematic diagram of the standard library interval under different partitioning strategies of the present invention; Figure 7 It is a schematic diagram of line loss rate anomaly prediction based on Lagrange interpolation of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0024] like Figure 1 As shown, the present invention provides a method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy sources, comprising the following steps: S1. Construct a joint probability distribution function of renewable energy output based on Copula theory. The joint probability distribution function can help understand the mutual influence and correlation between the output types of renewable energy. Since renewable energy output is often affected by many factors, such as wind speed, light intensity, etc., these factors may lead to a certain degree of dependence between the outputs. Through the joint probability distribution function, this dependence can be more accurately portrayed, thereby providing theoretical support for subsequent data generation.

[0025] S11. Establish marginal distribution function.

[0026] The marginal distribution function is the basis for constructing the joint probability distribution function. The non-parametric kernel density estimation method is used to construct the marginal distribution function of renewable energy output, which can capture different characteristics in the data and obtain accurate data analysis results.x Marginal distribution function The calculation formula is:

[0027] In the formula, n represents the number of samples; h Indicates bandwidth; K represents the kernel density function; X i Indicates i Sample wind power output data.

[0028] Similarly, we can get the photovoltaic y Marginal distribution function , and its calculation formula is:

[0029] In the formula, Y Indicates i Samples of photovoltaic output data.

[0030] S12. Select an appropriate kernel density function.

[0031] The kernel density function is an effective nonparametric estimation method used to estimate the probability density function of data. The Gaussian kernel density function is used here because it has good universality and low space complexity compared to other kernel functions. The expression of the Gaussian kernel density function is:

[0032] Where exp represents the exponential function.

[0033] S13. Obtain the probability distribution functions of wind power and photovoltaic power.

[0034] Substituting the Gaussian kernel density function in S12 into the marginal distribution function of S11, the probability distribution functions of wind power and photovoltaic power can be obtained:

[0035] In the formula, represents the probability distribution function of wind power output; represents the probability distribution function of photovoltaic output; h (·) represents the adaptive bandwidth.

[0036] S14. Construct the joint probability distribution function of new energy output based on Copula theory.

[0037] The Copula theory is introduced, and based on the probability distribution functions of S13 wind power and photovoltaic power, the joint probability distribution function of new energy output is obtained:

[0038] In the formula, represents the cumulative probability distribution function of wind power; represents the cumulative probability distribution function of photovoltaic; represents the random variable correlation coefficient.

[0039] S2. Establish a new energy line loss data generation model based on conditional generative adversarial network (CGAN) (including conditional variables c , generator G and discriminator D), and input the joint probability distribution function of renewable energy output in S1 as the conditional variable and the real line loss data of the regional power grid into the data generation model to generate a large amount of distribution network line loss data of renewable energy access. Figure 2 The DCNN structure represents the Deep Convolution Neural Network (DCNN) model.

[0040] S21. Set condition variables c .

[0041] Take the joint probability distribution function of renewable energy output in S1 as the conditional variable c Input into the generator G to improve its data generation quality.

[0042] S22. Set the generator G and the discriminator D.

[0043] The real line loss data (real data) of the regional power grid is input into the generator G to obtain a large amount of generated data. The discriminator D not only needs to judge the similarity between the generated data and the real data, but also needs to verify whether the generated data meets the conditional variables. c The structure and parameters of the generator G and the discriminator D are shown in Table 1.

[0044] The loss function expressions of the generator and discriminator are:

[0045] In the formula, is the loss function of the generator, is the loss function of the discriminator, E is the expected value of the distribution, is the discriminator output function, is the generated sample output by the generator, For real samples, and is the distribution of generated samples and the distribution of real samples.

[0046] Table 1 Structure and parameters of the generator and discriminator

[0047] S23. Data generation model game training.

[0048] During the training process, the generator G and the discriminator D compete with each other by updating parameters alternately. The generator G tries to deceive the discriminator so that it cannot correctly distinguish between real samples and generated samples, while the discriminator D strives to improve its discrimination ability. The final game training objective function of the data generation model is:

[0049] S24. Evaluate the validity of the data generated by new energy line losses.

[0050] In order to further evaluate whether the data generated by the data generation model fits the real data, the root mean square error is selected RMSE To quantitatively analyze the difference between the generated samples and the real samples, the expression is:

[0051] In the formula, m Indicates the output dimension; represents the true value; Indicates that the data generation model generates values.

[0052] S3. Construct a multi-dimensional scenario division strategy for substation line loss rate based on historical data mining.

[0053] ICEEMDAN is an improved denoising algorithm proposed by Colominas et al. It overcomes the drawback of the traditional EMD method that directly discards the first few order IMF components, can effectively weaken the residual noise in the IMF components, improve the pseudo-mode and modal aliasing problems in the traditional denoising algorithm, is not easy to distort the reconstructed signal, and can better smooth the initial data, laying a solid foundation for subsequent research. The noise of the data will affect the trend analysis of the line loss rate curve, so step S3 considers denoising the distribution network line loss data generated by S2, including the following steps: S31. Decompose the curve corresponding to the line loss data of the distribution network using the ICEEMDAN algorithm to generate k Intrinsic Mode Function (IMF); S32, for the IMF components obtained by decomposition, the Normalized Autocorrelation Function (NAF) is introduced as the signal correlation evaluation index, and the correlation coefficient with the distribution network line loss data curve generated by S2 is calculated:

[0054] in,r 1(·) represents the first residual component generated by EMD, M (·) indicates modality, To calculate the first residual component of the noisy signal generated by EMD, r k (·) indicates the k The residual component, For the k The standard deviation of the special white noise is added to the EMD. E k represents the empirical function, ω (i) ( t ) represents the added Gaussian white noise, IMF 1(·) is the first-order IMF component, IMF k (·) indicates the k The order IMF component, X (·) represents the noise signal, N represents the total order of the IMF, It represents the correlation function between the generated distribution network line loss data curve and each order residual component, It represents the correlation coefficient between the generated distribution network line loss data curve and each order IMF component; S33, setting a correlation coefficient threshold, filtering out a highly correlated IMF component set, and superimposing all components in the set, thereby completing the reconstruction process, and then removing the "burr" data in the curve to obtain the line loss change trend of the substation area; S34. Perform K-means++ cluster analysis based on the line loss change trend of the substation area to determine the clustering results of historical data and obtain the clustering on the trend, such as Figure 4 As shown (Scenario 1 and Scenario 2); S35. Based on S34, the support vector machine (SVC) algorithm is used to perform secondary clustering on the given trend clustering results. Figure 5 As shown (scenarios 3 to 6), the distribution network area scenario division results are obtained.

[0055] Figure 3 Although the local extreme values ​​of ICEEMDAN still have short-term jumps, the change frequency is low and the fluctuation amplitude is not large, which better suppresses modal aliasing.

[0056] S4. A one-time partitioning strategy for the line loss rate standard library based on cluster analysis is proposed.

[0057] According to the distribution network area scenario division results, the line loss rates under the same scenario are stored in the same set Wi In this paper, the distribution of line loss rate in different scenarios is studied. The traditional division method divides W i The line loss rate data in the set are sorted, and the minimum value is taken as the lower limit of the interval and the maximum value is taken as the upper limit of the interval, such as Figure 6 As shown. The traditional method considers all the set data and ignores the selection of outliers in different clusters, which leads to insufficient division accuracy. Therefore, the present invention provides a one-time division strategy for the line loss rate standard library based on cluster analysis, and the specific steps are as follows: S41, for the collection W i The line loss rate data in the network is mined and analyzed, the number of clusters is dynamically adjusted, and the silhouette coefficient (SC) corresponding to each cluster number is calculated to determine the optimal number of clusters. K ; S42. Based on the calculation results of S41, K-means++ is used to perform cluster analysis and extract the cluster with the largest number of cases. C i ( Figure 6 pentagonal point group in ); S43, Select C i The minimum value of the elements in is used as the lower limit of the interval, and the maximum value is used as the upper limit of the interval, and finally the line loss rate standard library after one division is obtained.

[0058] S5. Establish a dynamic interval translation strategy based on the number of cluster cases to determine the retention or removal of data points near the upper and lower limits of the line loss rate in the substation area.

[0059] When using the interval dynamic translation method, the three clusters with the largest number of cases need to be analyzed. For the sake of convenience, the spatial distribution order of the clusters is specified from top to bottom as { C k , C i , C j The principle steps of interval dynamic translation method are as follows: S51, the pre-processed distribution network substation line loss data is generated by K-means++ clustering k clusters, keep the cluster with the largest number of cases C i ; S52, Assumed Cluster C k The number of cases is less than C j , select cluster C j and C k, calculated relative to C i The offset is calculated as:

[0060] Where: , Represents clusters C j and C k Relative to cluster C i The offset of L Cj and L Ck Cluster C j and C k The length of the interval; m j and m k Cluster C j and C k the number of cases; t is the total sample size; S53, combined with the relative offset result in S52, solve and update the standard library interval of the area line loss rate as follows:

[0061] Where: , They represent the lower and upper limits of the standard library interval of the line loss rate of the substation area respectively; C imin and C imax Respectively represent clusters C i The minimum and maximum values ​​of the inner elements.

[0062] S6. Based on the interval overlap rate strategy, similar or close scenes are merged to reduce the amount of calculation for subsequent intelligent judgment and precise tracking, and improve calculation efficiency. Since some substation scenes are divided into many dimensions, scene overlap is inevitable. Overlap refers to a large number of intersections between two or more substation line loss rate standard library intervals. The main impact is that a large number of overlaps cause the standard library intervals to be not distinguishable, increasing the system judgment time and affecting work efficiency. The steps for scene merging are as follows: S61, constructing the overlap ratio of two scenes: Assuming A and B are two determined intervals, the interval overlap ratio of A and B is expressed as:

[0063] Where: L represents the length of the corresponding interval; (A∩B) represents the overlapping interval of A and B; where, sim (A,B)∈[0,1].

[0064] S62: To ensure that the standard library is sufficiently streamlined, continue to use the OTSU algorithm to set an adaptive threshold for the interval overlap rate. α ,pass α The inter-class variance obtained by segmentation is used as the evaluation index. The specific formula is as follows:

[0065] Where: δ 2 is the between-class variance; w 1 and w 2 respectively represent [0 ,α )and[ α, 1] The proportion of the two parts; u 1 and u 2 are the average values ​​of the overlap rate of each part; S63, find out δ 2 The threshold corresponding to the maximum is taken as the optimal threshold α best , and the overlap ratio is greater than or equal to α best The standard library intervals are merged according to the union principle, otherwise they are not merged.

[0066] S7. Automatic analysis method of line loss rate anomaly based on substation standard library.

[0067] S71. Import the line loss rate sequence of the substation area to be studied ; S72, get sequence The area number and time date in the , determine the corresponding scene according to the scene interval merging result, and extract the corresponding standard library W; S73, judge the set Is there a value in W that is not in W? If so, proceed to S74; otherwise, proceed to S75. S74, the line loss rate of the substation area is abnormal within a given period, and the abnormal point is stored in the set T; S75. The line loss rate of the substation area is normal within the given time period, and the process ends.

[0068] S8. Use the grey correlation analysis method to analyze the internal correlation between the various factors of abnormal line loss rate in the substation area (power sales, power supply, average temperature, number of public transformers and total capacity of public transformers) and the users in the substation area, and accurately locate the users associated with abnormal line loss based on the improved Adtributor algorithm. The specific steps are as follows: S81. Grey correlation analysis is an important part of the grey system. It determines whether the connection between different data labels is close according to the similarity of the geometric shapes of the data curves. If the correlation between two factors is stronger, the grey correlation degree is greater, and vice versa. The expressions for calculating the correlation coefficient and correlation degree are:

[0069] Where: For the i Factors j The correlation coefficient of the moment; For the i The relevance of the factors; For reference sequence j The value of the moment, For the i Factors j The value of the moment, and is the normalized value; is the resolution coefficient, which is generally taken as 0.5.

[0070] S82. For the three factors with the largest grey correlation, the most suspicious users are mined through the Adtributor algorithm. When using the algorithm, the calculation of two parameters, Surprise (S) and Explanatory power (EP), is involved. The calculation formula of S is as follows:

[0071] Where: p is the predicted probability; q is the true probability; The calculation formula of EP value is as follows:

[0072] Where: A ij Indicates the true value of the fault; F ij denotes normal predicted value; subscript i and j Represents date and user number respectively; m is an abnormal indicator; A( m )、F( m ) are A ij 、F ij The sum of .

[0073] S83. In actual power grids, small disturbances caused by user load fluctuations are common, and the Adtributor algorithm is highly sensitive to disturbance factors. At the same time, the ARMA time series model used in the algorithm has weak processing capabilities for nonlinear data. If the traditional Adtributor algorithm is continued to be used, the EP value calculation formula will be The solution efficiency is low and the prediction accuracy is difficult to guarantee, which seriously affects the efficiency and accuracy of abnormal user tracking.

[0074] To overcome the above problems, the present invention improves the Adtributor algorithm and adopts the Lagrange interpolation method to obtain the estimated value of the outlier point, which can achieve the fitting of nonlinear functions to a certain extent and greatly reduce the amount of calculation; see Figure 7 . Specifically: First, based on S74, the abnormal line loss rate in the platform area was identified ( Figure 7 middle A ( m )area).

[0075] Next, the line loss rate data of the substation area during the normal period of line loss rate is extracted, and the estimated value of the abnormal period of time is solved by Lagrange interpolation method ( Figure 7 middle F ( m ) area), assuming that the line loss rate curve is in the interval The value on Known, the abnormal line loss rate period of the substation area The estimated value of a position in Should meet:

[0076] Where: represents the interpolation polynomial (corresponding to the area F ( m ) values ​​of each point on the inner curve); s The value range of represents the normal time period.

[0077] Finally, combined with the above obtained A( m ) and F( m ) value, and calculate the improved EP value according to the EP value formula in step S82.

[0078] The distribution network line loss abnormality analysis and accurate tracking device taking into account the access of new energy sources includes: A new energy line loss data generation model construction module is used to construct a new energy line loss data generation model based on a conditional generative adversarial network, and the new energy output joint probability distribution function is used as a conditional variable. The real line loss data of the regional power grid is input into the new energy line loss data generation model to generate the line loss data of the distribution network connected to the new energy; The multi-dimensional scenario division module of the substation line loss rate uses the ICEEMDAN algorithm to realize the distribution network substation scenario division for the curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network; The abnormal intelligent judgment module based on the line loss rate standard library of the distribution network is used to build the line loss rate standard library of the distribution network and merge similar or close scenes, and then conduct abnormal analysis and judgment; The abnormal correlation of line loss rate in the substation area is used for precise positioning module, which analyzes the internal correlation between various factors of abnormal line loss rate in the substation area and users in the substation area, and realizes precise positioning of users associated with abnormal line loss rate based on the improved Adtributor algorithm.

[0079] Based on the same inventive concept as a method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy, the present application also provides an electronic device, which includes one or more processors and one or more memories, wherein a computer-readable code is stored in the memory, wherein the computer-readable code, when executed by one or more processors, implements a method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy. Among them, the memory may include a non-volatile storage medium and an internal memory; the non-volatile storage medium may store an operating system and a computer-readable code. The computer-readable code includes program instructions, which, when executed, may enable the processor to execute any method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy. The processor is used to provide computing and control capabilities to support the operation of the entire electronic device. The memory provides an environment for the operation of the computer-readable code in the non-volatile storage medium, and when the computer-readable code is executed by the processor, the processor may execute any method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy.

[0080] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0081] The computer-readable storage medium may be an internal storage unit of the electronic device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card, FC), etc. equipped on the electronic device.

[0082] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. Abnormal analysis and accurate tracking method of distribution network line loss considering the access of new energy, Features: Construct the joint probability distribution function of renewable energy output; Establish a new energy line loss data generation model based on a conditional generative adversarial network, and input the new energy output joint probability distribution function as a conditional variable and the real line loss data of the regional power grid into the new energy line loss data generation model to generate the line loss data of the distribution network connected to the new energy; For the curve corresponding to the line loss data of the distribution network, the ICEEMDAN algorithm is used to realize the distribution network substation scene division; The distribution network substation scene division results are divided once to obtain the distribution network substation line loss rate standard library; Based on the interval dynamic translation method, the line loss rate standard library of the distribution network substation is divided twice; Based on the interval overlap rate strategy, similar or close scenarios in the line loss rate standard library of the distribution network area after secondary division are merged; Conduct abnormal analysis based on the combined distribution network area line loss rate standard database; The factors causing abnormal line loss rate in the substation area and the intrinsic correlation between users in the substation area are analyzed, and the users associated with abnormal line loss are accurately located based on the improved Adtributor algorithm.

2. According to claim 1, the method for abnormal analysis and accurate tracking of line loss in the distribution network taking into account the access of new energy sources, It is characterized in that The distribution network area scene division results are divided once, specifically: According to the distribution network area scenario division results, the line loss rates under the same scenario are stored in the same set W i , for the set W i The line loss rate data in the network is mined and analyzed, the number of clusters is dynamically adjusted, and the silhouette coefficient corresponding to each cluster number is calculated to determine the optimal number of clusters. K ; Use K-means++ to perform cluster analysis and extract the cluster with the largest number of cases C i ; Will C i The minimum value of the elements in is used as the lower limit of the interval, and the maximum value is used as the upper limit of the interval, and finally the line loss rate standard library after one division is obtained.

3. According to claim 2, the method for abnormal analysis and accurate tracking of line loss in the distribution network taking into account the access of new energy sources, It is characterized in that The line loss rate standard library of the distribution network substation area is divided into two parts based on the interval dynamic translation method, specifically: The pre-processed distribution network substation line loss data is generated by K-means++ clustering k clusters, keep the cluster with the largest number of cases C i ; Assumed Cluster C k The number of cases is less than C j , select cluster C j and C k , calculated relative to C i The offset is calculated as: ； ； Where: , Represents clusters C j and C k Relative to cluster C i The offset of L Cj and L Ck Cluster C j and C k The length of the interval, m j and m k Cluster C j and C k The number of cases t is the total sample size; Combined with the relative offset results, the standard library interval of the line loss rate in the substation area is solved and updated as follows: ； Where: , They represent the lower and upper limits of the standard library interval of the line loss rate in the substation area, C imin and C imax Respectively represent clusters C i The minimum and maximum values ​​of the inner elements.

4. According to claim 1, the method for abnormal analysis and accurate tracking of line loss in the distribution network taking into account the access of new energy sources, It is characterized in that Merge similar or close scenes, specifically: Construct the overlap rate of two scenes: Assume A and B are two fixed intervals, then the interval overlap rate of A and B is expressed as: ； Where: L represents the length of the corresponding interval, (A∩B) represents the overlapping interval of A and B; Set an adaptive threshold for interval overlap α ,pass α The inter-class variance obtained by segmentation is used as the evaluation index. The specific formula is as follows: ； Where: δ 2 is the between-class variance, w 1 and w 2 Respectively represent [0 ,α )and[ α, 1] The proportion of the two parts, u 1 and u 2 are the average values ​​of the overlap rate of each part respectively; Find δ 2 The threshold corresponding to the maximum is taken as the optimal threshold α best , and the overlap ratio is greater than or equal to α best The standard library intervals are merged according to the union principle, otherwise they are not merged.

5. According to claim 1, the method for abnormal analysis and accurate tracking of line loss in distribution network considering the access of new energy, It is characterized in that The grey correlation analysis method is used to analyze the internal correlation between the various factors of abnormal line loss rate in the substation area and the users in the substation area. For the three factors with the largest grey correlation, the most suspicious users are mined through the improved Adtributor algorithm, so as to achieve accurate positioning of users associated with line loss anomalies.

6. According to claim 1, the method for abnormal analysis and accurate tracking of line loss in the distribution network taking into account the access of new energy sources, It is characterized in that The ICEEMDAN algorithm is used to realize the distribution network substation scenario division. Specifically, the curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network is decomposed and reconstructed to obtain the preprocessed line loss data curve of the substation under the jurisdiction of the distribution network, and the "burr" data is removed to obtain the line loss change trend of the substation. The line loss change trend of the substation is clustered and analyzed, and then the distribution network substation scenario division result is obtained.

7. According to claim 1, the method for abnormal analysis and accurate tracking of line loss in distribution network considering the access of new energy sources, It is characterized in that The precise positioning of users associated with abnormal line loss based on the improved Adtributor algorithm is specifically as follows: For the three factors with the largest grey correlation, the most suspicious users are mined through the Adtributor algorithm. The algorithm involves the calculation of two parameters: the unexpectedness value and the explanatory power value EP. The explanatory power value EP satisfies: ； Where: A ij Indicates the true value of the fault, F ij represents the normal predicted value, the subscript i and j Represents the date and user number respectively, m is the abnormal index, A( m )、F( m ) are A ij 、F ij The sum of The F( m ) is realized by the following improved method: extract the line loss rate data in the normal time period of the substation area, use the Lagrange interpolation method to solve the estimated value of the abnormal time period, and use it as F( m ) to calculate the explanatory power value.

8. A device for implementing the method for abnormal analysis and accurate tracking of line loss in a distribution network taking into account the access of new energy sources as described in any one of claims 1 to 7, It is characterized in that include: A new energy line loss data generation model construction module is established, which builds a new energy line loss data generation model based on a conditional generative adversarial network, and uses the new energy output joint probability distribution function as a conditional variable. The real line loss data of the regional power grid is input into the new energy line loss data generation model to generate the line loss data of the distribution network connected to the new energy; The multi-dimensional scenario division module of the substation line loss rate uses the ICEEMDAN algorithm to realize the distribution network substation scenario division for the curve corresponding to the line loss data of the substation under the jurisdiction of the distribution network; The abnormal intelligent judgment module based on the line loss rate standard library of the distribution network is used to build the line loss rate standard library of the distribution network and merge similar or close scenes, and then conduct abnormal analysis and judgment; The abnormal correlation of line loss rate in the substation area is used for precise positioning module, which analyzes the internal correlation between various factors of abnormal line loss rate in the substation area and users in the substation area, and realizes precise positioning of users associated with abnormal line loss rate based on the improved Adtributor algorithm.

9. An electronic device, It is characterized in that including memory and processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the method for abnormal analysis and precise tracking of distribution network line losses taking into account the access of new energy as described in any one of claims 1-7 when executing the computer program.

10. A storage medium, It is characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the processor executes the method for abnormal analysis and precise tracking of distribution network line losses taking into account the access of new energy as described in any one of claims 1-7.

Images