Distribution network equipment health state evaluation method and system based on Gaussian mixture model

Abstract

The invention provides a distribution network equipment health state assessment method and system based on a Gaussian mixture model. According to the scheme, a multi-time-scale distribution network equipment health state assessment model with extremely high applicability and universality is constructed; a distribution network equipment abnormity monitoring and diagnosis process based on field measured data is formed; real-time anomaly evaluation of any distribution network equipment and any state parameter is realized without depending on data quality; the Gaussian mixture model constructed based on historical operation data can better represent operation characteristics of the distribution network equipment, and abnormal defects slowly occurring in the distribution network equipment are effectively captured; the recognition accuracy and the response speed of the abnormal operation state of the distribution network equipment with high concealment and strong randomness are improved; the problem that the health state of the equipment cannot be correctly evaluated under the condition that part of monitoring data is missing in the traditional technology is solved; a set of independent dynamic anomaly judgment standards is customized for each distribution network device, and the anomaly diagnosis conclusion has reference and analytic performance.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network, and in particular relates to a distribution network equipment health status assessment method and system based on a Gaussian mixture model. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] At present, the new distribution network equipment health status assessment algorithm based on machine learning and big data analysis has established a more accurate assessment model through in-depth mining and analysis of the equipment's historical operating data, and realized dynamic assessment and prediction of the equipment's health status. However, its core lies in judging the health status of the equipment by carefully analyzing the various operating data of the equipment, which relies heavily on accurate and comprehensive data support. However, the actual distribution network equipment is multi-faceted and widespread, and the monitoring devices cannot cover all of them. There are monitoring blind spots, which leads to the omission of key data and seriously affects the evaluation effect. At the same time, communication failures will cause data transmission interruption, resulting in temporary loss of key monitoring data, which will also affect the credibility of the evaluation results.

[0004] At the same time, the traditional distribution network equipment health status algorithm based on static thresholds does not fully consider the actual operating characteristics of the equipment and cannot distinguish whether the equipment status change is caused by an abnormality in the equipment itself or by the randomness of the distribution network. In other words, there is no fixed threshold to evaluate whether the equipment status is abnormal. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for evaluating the health status of distribution network equipment based on a Gaussian mixture model, so as to solve the problem that the status evaluation algorithm based on the evaluation model relies on accurate and comprehensive data support and cannot correctly evaluate the equipment status when some monitoring data of the equipment is missing; and the health status algorithm of distribution network equipment based on static thresholds does not fully consider the actual operating characteristics of the equipment and cannot distinguish whether the change of equipment status is caused by abnormality of the equipment itself or by randomness of the distribution network.

[0006] According to a first aspect of an embodiment of the present invention, a method for evaluating the health status of a distribution network device based on a Gaussian mixture model is provided, comprising:

[0007] Obtain the historical operation data of the distribution network equipment to be evaluated;

[0008] Fitting a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model;

[0009] Based on the real-time updated historical operation data, according to the preset strategy, the Gaussian mixture model is updated regularly or in real time to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timing moment, the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is not a preset timing moment, the Gaussian mixture model is updated in real time based on the real-time updated historical operation data;

[0010] The abnormality judgment of the equipment is performed based on the updated Gaussian mixture model to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model that is updated periodically, the divergence between the current Gaussian mixture model and the Gaussian mixture model that was updated last time is calculated, and horizontal and vertical comparisons are performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison results; for the Gaussian mixture model that is updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

[0011] Furthermore, the horizontal and vertical comparison specifically includes vertical analysis, iterative vertical analysis and horizontal analysis, wherein the vertical analysis is to update the Gaussian mixture model at a preset timing based on the historical operation data of a preset time period, and compare the current Gaussian mixture model with the Gaussian mixture model updated last; the iterative vertical analysis is to update the Gaussian mixture model at a preset timing based on all historical operation data, and compare the current Gaussian mixture model with the Gaussian mixture model updated last; the horizontal analysis is to construct a Gaussian mixture model for each device to be tested using the historical operation data of a preset time period, and compare the Gaussian mixture models of different devices.

[0012] Furthermore, the divergence between the current Gaussian mixture model and the previously updated Gaussian mixture model is calculated, wherein the divergence of the two models is calculated using JS divergence, which is specifically expressed as follows:

[0013]

[0014] Among them, P is the updated Gaussian mixture model, Q is the previously updated Gaussian mixture model, D(P||R) is the KL divergence of the two probability distributions P and R of the same random variable x, and D(Q||R) is the KL divergence of the two probability distributions Q and R of the same random variable x.

[0015] Furthermore, the calculation of the divergence may also be: calculating the sum of JS divergences of each updated Gaussian mixture model and other updated Gaussian mixture models, and obtaining the Gaussian mixture model with the largest divergence as the Gaussian mixture model where the anomaly occurs.

[0016] Furthermore, the abnormality judgment of the device based on the calculated divergence and the comparison result is specifically as follows:

[0017] Build a unique Gaussian mixture model for each device based on historical operation data;

[0018] Update the Gaussian mixture model regularly;

[0019] Calculate the JS divergence between models and compare the JS divergence;

[0020] Based on the comparison result obtained, determine whether the device has an abnormality.

[0021] Furthermore, the Gaussian mixture model updated in real time calculates the theoretical probability estimate of the current operating condition of the equipment, and makes an abnormal judgment on the equipment based on the theoretical probability value, specifically: the theoretical probability estimate of the current operating condition of each device to be evaluated is calculated, and compared with the maximum value of the marginal distribution probability; if the probability of a certain parameter is lower than the minimum marginal probability of the parameter in the Gaussian mixture model, an abnormal state is recorded; if the number of consecutive abnormal states is greater than a preset threshold, the equipment is considered to be abnormal and an alarm is issued.

[0022] According to a second aspect of an embodiment of the present invention, a distribution network equipment health status assessment system based on a Gaussian mixture model is provided, comprising:

[0023] A data acquisition unit, which is used to acquire historical operation data of the distribution network equipment to be evaluated;

[0024] A GMM fitting unit, which is used to fit a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model;

[0025] A GMM updating unit, which is used to perform a timed update or a real-time update on the Gaussian mixture model based on the real-time updated historical operation data according to a preset strategy to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timed moment, the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is a non-preset timed moment, the Gaussian mixture model is updated in real time based on the real-time updated historical operation data;

[0026] An abnormality judgment unit is used to judge the abnormality of the equipment based on the updated Gaussian mixture model to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model updated at a regular interval, the divergence between the current Gaussian mixture model and the Gaussian mixture model updated last time is calculated, and horizontal and vertical comparisons are performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison results; for the Gaussian mixture model updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

[0027] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, comprising a memory, a processor, and a computer program stored and running on the memory, wherein when the processor executes the program, the method for assessing the health status of distribution network equipment based on a Gaussian mixture model is implemented.

[0028] According to a fourth aspect of an embodiment of the present invention, a non-transitory computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method for evaluating the health status of distribution network equipment based on a Gaussian mixture model is implemented.

[0029] According to the fifth invention of the embodiment of the present invention, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, it implements the method for evaluating the health status of distribution network equipment based on a Gaussian mixture model.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) The present invention provides a method and system for evaluating the health status of distribution network equipment based on a Gaussian mixture model. The scheme can better characterize the operating characteristics of distribution network equipment through a Gaussian mixture model constructed based on historical operating data, effectively capture abnormal defects that occur slowly inside the distribution network equipment; improve the recognition accuracy and response speed of abnormal operating states of distribution network equipment with high concealment and strong randomness; solve the problem that traditional technology cannot correctly evaluate the health status of equipment when some monitoring data is missing; and customize a set of independent dynamic abnormality judgment standards for each distribution network device, so that the abnormal diagnosis conclusion is more referenceable and analyzable.

[0032] (2) The scheme updates the Gaussian mixture model periodically or in real time based on real-time updated historical operation data and in accordance with a preset strategy to obtain an updated Gaussian mixture model; and based on the updated Gaussian mixture model, the scheme determines the abnormality of the equipment and implements the evaluation of the health status of the distribution network equipment. By constructing a multi-time scale distribution network equipment health status evaluation model with strong applicability and versatility, a distribution network equipment abnormality monitoring and diagnosis process based on field measured data is formed, and real-time abnormality evaluation of any distribution network equipment and any state parameter is realized without relying on data quality.

[0033] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0035] Figure 1 Schematic diagram of the Gaussian mixture model described in an embodiment of the present invention;

[0036] Figure 2 1 is a relative comparison of the JS divergences of the two probability distributions described in the embodiments of the present invention, wherein (a), (b), (c), (d), (e) and (f) are schematic diagrams of JS divergences under different probability distributions, respectively;

[0037] Figure 3 It is an overall framework diagram of the distribution network equipment health status assessment method based on Gaussian mixture model described in an embodiment of the present invention;

[0038] Figure 4 It is a flow chart of a method for evaluating the health status of distribution network equipment based on a Gaussian mixture model described in an embodiment of the present invention;

[0039] Figure 5 It is a schematic diagram of different horizontal and vertical comparison characteristics described in an embodiment of the present invention;

[0040] Figure 6 A schematic diagram of health status evaluation of distribution network equipment under different update strategies described in an embodiment of the present invention;

[0041] Figure 7 It is a schematic diagram of on-site measured data of the cable equipment described in an embodiment of the present invention;

[0042] Figure 8 is the marginal probability distribution of the Gaussian mixture model under the specific working conditions described in the embodiment of the present invention;

[0043] Fig. 9 The real-time measurement data after injecting different contents and different types of random noises as described in the embodiment of the present invention;

[0044] Fig.10 It is a schematic diagram of the comparison results of the GMM models at different time points described in the embodiments of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0046] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0047] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0048] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0049] like Figure 4 As shown, a method for evaluating the health status of distribution network equipment based on a Gaussian mixture model is provided, including the following processing steps:

[0050] Step 1: Obtain the historical operation data of the distribution network equipment to be evaluated;

[0051] Step 2: Fitting a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model;

[0052] In a specific implementation, the Gaussian mixture model is defined as follows:

[0053] The Gaussian mixture model is famous for its high accuracy in describing random variables. It is a model that uses the Gaussian probability density function (normal distribution curve) to accurately quantify things. It is a model that decomposes things into several parts based on the Gaussian probability density function (normal distribution curve).

[0054] GMM is used to describe the joint probability distribution of a random vector X, which is defined as a convex combination of multiple Gaussian distribution functions. m ,μ m ,σ m; m = 1, ..., M} is the adjustable parameter set of GMM. The mathematical expression of GMM is as follows:

[0055]

[0056] Among them, f X (x) represents the joint probability density function of X; ω m is the weight coefficient; W in the denominator of equation (1-3) represents the dimension of X; N m (·) represents the multidimensional normal distribution, which is called the mth Gaussian component of GMM; det(·) represents the matrix determinant; M represents the total number of Gaussian components; μ m and σ m Respectively represent the mean vector and covariance matrix of the mth Gaussian component.

[0057] Figure 1 A schematic diagram of a GMM with four Gaussian components is given. In this figure, the probability density function of the GMM (black line in the figure) is a convex combination of four Gaussian distribution probability density functions (red, blue, green, and purple lines in the figure). By adjusting the parameter set Ω, the GMM can characterize the probability density function of any random variable. For ease of understanding, we can also regard it as the "Fourier decomposition" of the probability density function.

[0058] exist Figure 1 In the equation, M=1, and the weight coefficient ωm is 0.4, 0.1, 0.3, and 0.2 respectively.

[0059] The key to the GMM model is to determine the model's weight coefficient ω m , mean vector μ m and the covariance matrix σ m These parameter sets Ω, determining the parameter set Ω of GMM is a typical parameter estimation problem. Based on the historical data of X, the maximum likelihood estimation technique and Bayesian estimation technique can be used to obtain the parameter set of GMM.

[0060] Step 3: Based on the real-time updated historical operation data, according to the preset strategy, the Gaussian mixture model is updated regularly or in real time to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timing moment (i.e., the slow dynamics hereinafter), the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is a non-preset timing moment (i.e., the fast dynamics hereinafter), the Gaussian mixture model is updated in real time based on the real-time updated historical operation data;

[0061] Step 4: Based on the updated Gaussian mixture model, the abnormality judgment of the equipment is performed to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model updated at a regular interval, the divergence between the current Gaussian mixture model and the Gaussian mixture model updated last time is calculated, and a horizontal and vertical comparison is performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison result; for the Gaussian mixture model updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

[0062] In a specific implementation, the horizontal and vertical comparisons specifically include vertical analysis, iterative vertical analysis and horizontal analysis, wherein the vertical analysis is to update the Gaussian mixture model at a preset timing based on the historical operation data of a preset time period, and compare the current Gaussian mixture model with the previously updated Gaussian mixture model; the iterative vertical analysis is to update the Gaussian mixture model at a preset timing based on all historical operation data, and compare the current Gaussian mixture model with the previously updated Gaussian mixture model; the horizontal analysis is to construct a Gaussian mixture model for each device to be tested using the historical operation data of a preset time period, and compare the Gaussian mixture models of different devices.

[0063] In a specific implementation, the divergence between the current Gaussian mixture model and the previously updated Gaussian mixture model is calculated, wherein the divergence of the two models is calculated using JS divergence.

[0064] In a specific implementation, the divergence may also be calculated by calculating the sum of JS divergences of each updated Gaussian mixture model and other updated Gaussian mixture models, and obtaining the Gaussian mixture model with the largest divergence as the Gaussian mixture model where the anomaly occurs.

[0065] Specifically, the reason why this embodiment adopts JS divergence is that:

[0066] Kullback-Leibler divergence (KL divergence) is also called relative entropy. For two probability distributions P(x) and Q(x) of the same random variable x, KL divergence can be used to measure the difference between the two probability distributions. The calculation method is as follows:

[0067]

[0068] Where X is all possible values ​​of the random variable x. Since the logarithmic function is a convex function, the value of KL divergence is non-negative. KL divergence is sometimes called KL distance, but it does not satisfy the properties of distance: First, KL divergence is not symmetric, that is,

[0069] D KL (P||Q)≠D KL (Q||P) (1-5)

[0070] Secondly, KL divergence does not satisfy the triangle inequality, that is, for the probability distribution H(x) of the random variable x, the following equation holds:

[0071] D KL (P||Q)>D KL (P||H)+D KL (H||Q) (1-6)

[0072] Since the numerical variation of KL divergence is large, in order to intuitively reflect the difference between two probability distributions, many scholars further calculated Jensen-Shannon divergence (JS divergence) based on KL divergence. JS divergence not only solves the problem of asymmetric KL divergence, but also has a value range between 0 and 1. JS divergence takes 0 to indicate that the two probability distributions are completely consistent, and takes 1 to indicate that the two probability distributions are extremely different. The calculation method of JS divergence is as follows:

[0073]

[0074] in

[0075]

[0076] like Figure 2 As shown in (a) to (f), the JS divergence of two probability distributions compared with each other in various cases is shown. It can be seen that when the overlap between the two probability distributions is small, the JS value is large, which is 0.7. As the overlap between the two probability distributions gradually increases, the JS value gradually decreases until the two probability distributions are almost identical, and the JS value is close to 0.

[0077] However, in the comparison of GMM models, the similarity of two models cannot be judged only by the absolute value of JS divergence, because in the process of GMM fitting, due to insufficient samples or deviations in the selection of Gaussian components, the GMM model cannot be guaranteed to be completely accurate. If the fitting degree is good, the absolute value of JS divergence between GMM models will be small. If there is a small anomaly in the data, the JS divergence between the abnormal GMM model and other models will increase, but the value itself is still small. If the JS divergence value is used to judge only, it will cause the anomaly to be missed.

[0078] Therefore, we need to use a relative comparison method to judge the similarity of the two models. Based on the equipment historical operating condition data set, a set of GMM models are constructed, denoted as {GMM1, GMM2, …, GMMn}; For each pair of GMM models (GMM i ,GMM j ), i≠j, calculate the JS divergence between them according to formula (1-7), denoted as JS ij ; For each GMM i , calculate the sum of JS divergences between it and all other GMM models according to formula (1-9), recorded as TotalJS i ; According to formula (1-10), we find the GMM model with the largest sum that is multiple of the others. In the figure, it is intuitively shown that the JS divergences of the i-th row and j-th column are all high. At this time, we have reason to believe that the i-th GMM model has an abnormality.

[0079]

[0080] TotalJS k >α·max(TotalJS i ) (1-10)

[0081] Among them, k≠i; α>1, is a given multiple threshold, used to determine whether the "multiple" condition is met.

[0082] In the specific implementation, the Gaussian mixture model updated in real time calculates the theoretical probability estimate of the current operating condition of the equipment, and makes an abnormal judgment of the equipment based on the theoretical probability value. Specifically: construct a separate Gaussian mixture model for each device based on historical operating data, and calculate the probability estimate of the historical operating condition of the equipment; update the GMM model in real time, calculate the theoretical probability estimate of the current operating condition of the equipment, and compare it with the probability estimate of the historical operating condition to determine whether the equipment is abnormal; if the equipment is abnormal, further determine the abnormal parameters: extract the marginal distribution of each variable through GMM and estimate its probability range. If the probability of a certain parameter is lower than the minimum marginal probability density of the parameter in the Gaussian mixture model, record an abnormal state; if the number of consecutive abnormal states is greater than the preset threshold, it is considered that the equipment is abnormal and an alarm is issued.

[0083] In the specific implementation, Figure 3 As shown in the figure, the scheme described in this embodiment includes two dynamic evaluation algorithms, fast and slow. The principle is to transform the random changes in the time domain of the device status into an invariant joint probability distribution, and combine the JS divergence and probability estimation value to determine whether the device is abnormal. The technical framework and algorithm flow chart of the health status evaluation of distribution network equipment based on Gaussian mixture model are shown in the figure. Figure 3 , 4 shown.

[0084] For slow dynamics, the GMM model is updated regularly, the JS divergence between models is calculated, and horizontal and vertical comparisons are performed to find obviously abnormal operating states and identify some slowly occurring abnormal defects inside the equipment. Slow dynamics include vertical analysis, iterative vertical analysis, and horizontal analysis. The characteristics of the analysis are as follows: Figure 5 The specific process of the slow dynamic algorithm is as follows:

[0085] 1) Build a unique Gaussian mixture model for each device based on historical operation data (essentially a probability distribution function that can characterize the device's own operating characteristics);

[0086] 2) Update the GMM model regularly;

[0087] 3) Calculate the JS divergence between models according to formula (1-7) and compare the JS divergence according to formula (1-10);

[0088] 4) Based on the comparison results of step 3, determine whether the equipment is abnormal, and identify some slowly occurring abnormal defects inside the equipment, such as insulation moisture, reduced insulation strength, etc.

[0089] In response to fast dynamics, the GMM model is updated in real time, and the real-time measurement data of the distribution network equipment is brought into the latest GMM model to calculate the theoretical probability estimate of the current operating condition of the equipment, and compare it with the corresponding minimum edge probability density in the Gaussian mixture model. If the index is low, an abnormal state is recorded. If the abnormality occurs multiple times in succession, an alarm is issued to achieve real-time alarm of abnormal probability operating state of the equipment. The specific process is:

[0090] 1) Build a unique Gaussian mixture model for each device based on historical operating data and calculate the probability estimate of the device's historical operating conditions;

[0091] 2) Update the GMM model in real time;

[0092] 3) Calculate the theoretical probability estimate of the current operating condition of the equipment, compare it with the probability estimate of the historical operating condition, and determine whether the equipment is abnormal;

[0093] 4) If the equipment is abnormal, further determine the abnormal parameters: extract the marginal distribution of each variable through GMM and estimate its probability range. If the probability of a parameter is lower than the minimum marginal probability density of the parameter in the Gaussian mixture model, record an abnormal state;

[0094] 5) If the probability density is lower than the minimum edge probability for a period of time, the number of recorded abnormalities exceeds the threshold, and an alarm message is issued. The frequency of the alarm can be determined based on the degree of lower than edge probability.

[0095] Comparison of slow and fast dynamic methods Figure 6 Compared with slow dynamics, fast dynamics have higher real-time performance and are suitable for rapid detection and real-time alarm of equipment status, which can effectively prevent the development of equipment failures.

[0096] In this way, the scheme described in this embodiment can access any multi-dimensional data of the power distribution equipment, and can characterize variables that obey any distribution without preprocessing the original data, solving the problems of many types of distribution network business data, strong randomness and complex probability distribution. The joint probability distribution does not depend on the quality of single-dimensional data, and can effectively reduce the impact of missing data sources on the evaluation results. The scenarios generated by the GMM model can basically reflect the operating characteristics of the equipment. Compared with other models, it has the advantages of high accuracy, strong generalization ability, strong real-time performance and high visualization.

[0097] The scheme described in this embodiment realizes a comprehensive assessment of the health status of all types of distribution network equipment, can effectively capture abnormal defects that occur slowly inside the distribution network equipment, improves the recognition accuracy and response speed of abnormal operating conditions that are highly concealed and random, can promptly discover potential fault hazards, provide early warnings, and ensure the stable operation of the distribution network equipment.

[0098] The scheme described in this embodiment constructs a dynamic Gaussian mixture model, which greatly enhances the generalization ability and stability of the health status assessment algorithm, can process monitoring data from different sources and distributions, can adapt to various complex operating environments, and can give accurate assessment results under any equipment type, operating state, and environmental factors. It effectively solves the problem that the health status assessment algorithm of distribution network equipment is heavily dependent on the quality of monitoring data, and significantly reduces the impact of missing data sources on the assessment results.

[0099] In order to verify the effectiveness of the solution described in this embodiment, the effectiveness of the method described in this embodiment was tested based on field measured data of a certain cable equipment.

[0100] Accuracy test: First, the equipment status assessment GMM model was constructed using the measurement data on Monday, and then the real-time data of the next day (Tuesday) was brought into the model to calculate the probability distribution, such as Figure 7 ,It was found that all points were within the normal probability interval, which was consistent with the actual situation on site.

[0101] Taking the current of 400A and the voltage of 10336V as the boundary, draw the edge distribution diagram of the constructed GMM model, as shown in Figure 8 As shown, it can be seen that except for temperatures around 32 degrees and 41 degrees, the probability of other temperatures occurring is much lower than the general value, which is considered to be an abnormal situation, which is consistent with the actual situation on site and effectively verifies the accuracy.

[0102] Furthermore, a sensitivity test was conducted to manually inject noise data of different nature and content into the original data to construct abnormal situations. Fig. 9 The figure shows the maximum amplitude of three different types of noise injected into the temperature while keeping the voltage and current data unchanged, simulating the cable temperature change under high-resistance fault. The three types of noise are random errors of ±0.1 to ±2 (interval of 0.1) (without changing the mean of the original data), random errors of 0.2 to 4 (interval of 0.2), and constant errors of 0.1 to 2 (interval of 0.1).

[0103] from Fig. 9 It can be found that it is difficult for field personnel to judge whether the data is abnormal by relying only on time series data. Using the GMM model, we can compare it with the GMM model of the measurement point the day before. Fig.10 As shown, abnormal data can be effectively identified. The vertical axis in the figure represents different amounts of noise, and the color represents the similarity between the measured data and the existing GMM model. The following conclusions can be drawn from the figure:

[0104] 1) Constant error has the highest resolution, and can be effectively distinguished after adding 0.8 degree error (JS divergence is greater than 0.2), that is, if the fault does not change the voltage and current, and only the energy changes, it can be effectively identified after the temperature changes by more than 0.8 degree;

[0105] 2) The random error resolution that does not change the original data mean is the lowest;

[0106] 3) All methods can effectively distinguish when the error exceeds 1.6 degrees.

[0107] The sensitivity of this method was effectively verified.

[0108] In one or more embodiments, corresponding to the above-mentioned method for evaluating the health status of distribution network equipment based on a Gaussian mixture model, a distribution network equipment health status evaluation system based on a Gaussian mixture model is provided, including:

[0109] A data acquisition unit, which is used to acquire historical operation data of the distribution network equipment to be evaluated;

[0110] A GMM fitting unit, which is used to fit a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model;

[0111] A GMM updating unit, which is used to perform a timed update or a real-time update on the Gaussian mixture model based on the real-time updated historical operation data according to a preset strategy to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timed moment, the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is a non-preset timed moment, the Gaussian mixture model is updated in real time based on the real-time updated historical operation data;

[0112] An abnormality judgment unit is used to judge the abnormality of the equipment based on the updated Gaussian mixture model to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model updated at a regular interval, the divergence between the current Gaussian mixture model and the Gaussian mixture model updated last time is calculated, and horizontal and vertical comparisons are performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison results; for the Gaussian mixture model updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

[0113] Furthermore, the relevant technical details of the system described in this embodiment are described in detail in Embodiment 1, and thus will not be repeated here.

[0114] In further embodiments, there is also provided:

[0115] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, no further description is given here.

[0116] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0117] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0118] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the method described in embodiment 1 is completed.

[0119] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0120] A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for evaluating the health status of distribution network equipment based on a Gaussian mixture model is implemented.

[0121] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in the present embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model, characterized in that: include: Obtain the historical operation data of the distribution network equipment to be evaluated; Fitting a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model; Based on the real-time updated historical operation data, according to the preset strategy, the Gaussian mixture model is updated regularly or in real time to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timing moment, the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is not a preset timing moment, the Gaussian mixture model is updated in real time based on the real-time updated historical operation data; The abnormality judgment of the equipment is performed based on the updated Gaussian mixture model to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model that is updated periodically, the divergence between the current Gaussian mixture model and the Gaussian mixture model that was updated last time is calculated, and horizontal and vertical comparisons are performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison results; for the Gaussian mixture model that is updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

2. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as claimed in claim 1, characterized in that: The horizontal and vertical comparison specifically includes vertical analysis, iterative vertical analysis and horizontal analysis, wherein the vertical analysis is to update the Gaussian mixture model at a preset timing based on the historical operation data of a preset time period, and compare the current Gaussian mixture model with the Gaussian mixture model updated last; the iterative vertical analysis is to update the Gaussian mixture model at a preset timing based on all the historical operation data, and compare the current Gaussian mixture model with the Gaussian mixture model updated last; the horizontal analysis is to construct a Gaussian mixture model for each device to be tested using the historical operation data of a preset time period, and compare the Gaussian mixture models of different devices.

3. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as claimed in claim 1, characterized in that: The divergence between the current Gaussian mixture model and the previously updated Gaussian mixture model is calculated, wherein the divergence of the two models is calculated using JS divergence, which is specifically expressed as follows: Among them, P is the updated Gaussian mixture model, Q is the previously updated Gaussian mixture model, D(P||R) is the KL divergence of the two probability distributions P and R of the same random variable x, and D(Q||R) is the KL divergence of the two probability distributions Q and R of the same random variable x.

4. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as claimed in claim 3, characterized in that: The calculation of the divergence may also be: calculating the sum of JS divergences of each updated Gaussian mixture model and other updated Gaussian mixture models, and obtaining the Gaussian mixture model with the largest divergence as the Gaussian mixture model where the anomaly occurs.

5. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as claimed in claim 1, characterized in that: The abnormality judgment of the device based on the calculated divergence and the comparison result is specifically as follows: Build a unique Gaussian mixture model for each device based on historical operation data; Update the Gaussian mixture model regularly; Calculate the JS divergence between models and compare the JS divergence; Based on the comparison result obtained, determine whether the device has an abnormality.

6. A method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as claimed in claim 1, characterized in that: For the Gaussian mixture model updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the equipment abnormality judgment is performed based on the theoretical probability value, specifically: the theoretical probability estimate of the current operating condition of each device to be evaluated is calculated, and compared with the maximum value of the marginal distribution probability. If the probability of a certain parameter is lower than the minimum marginal probability of the parameter in the Gaussian mixture model, an abnormal state is recorded; if the number of consecutive abnormal states is greater than a preset threshold, it is considered that the equipment is abnormal and an alarm is issued.

7. A distribution network equipment health status assessment system based on Gaussian mixture model, characterized in that: include: A data acquisition unit, which is used to acquire historical operation data of the distribution network equipment to be evaluated; A GMM fitting unit, which is used to fit a Gaussian mixture model based on the historical operation data to obtain a fitted Gaussian mixture model; A GMM updating unit, which is used to perform a timed update or a real-time update on the Gaussian mixture model based on the real-time updated historical operation data according to a preset strategy to obtain an updated Gaussian mixture model; wherein the preset strategy is specifically: when the current moment is a preset timed moment, the Gaussian mixture model is updated regularly based on the real-time updated historical operation data; when the current moment is a non-preset timed moment, the Gaussian mixture model is updated in real time based on the real-time updated historical operation data; An abnormality judgment unit is used to judge the abnormality of the equipment based on the updated Gaussian mixture model to realize the evaluation of the health status of the distribution network equipment; wherein, the abnormality judgment of the equipment based on the updated Gaussian mixture model is specifically as follows: for the Gaussian mixture model updated at a regular interval, the divergence between the current Gaussian mixture model and the Gaussian mixture model updated last time is calculated, and horizontal and vertical comparisons are performed, and the abnormality judgment of the equipment is performed based on the calculated divergence and the comparison results; for the Gaussian mixture model updated in real time, the theoretical probability estimate of the current operating condition of the equipment is calculated, and the abnormality judgment of the equipment is performed based on the theoretical probability value.

8. An electronic device comprising a memory, a processor and a computer program stored and running on the memory, characterized in that: When the processor executes the program, it implements a distribution network equipment health status assessment method based on a Gaussian mixture model as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, a method for evaluating the health status of distribution network equipment based on a Gaussian mixture model as described in any one of claims 1 to 6 is implemented.

Images