Power distribution terminal health state prediction method and device and electronic equipment

Abstract

The application provides a power distribution terminal health state prediction method and device and electronic equipment, and belongs to the technical field of power distribution terminal intelligent operation and maintenance. The method comprises the following steps: acquiring multi-source monitoring data to be tested of a power distribution terminal; inputting the multi-source monitoring data to be tested into a power distribution terminal health state prediction model to obtain a power distribution terminal health state output by the power distribution terminal health state prediction model; the power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm; the source domain data set comprises sample multi-source monitoring data under a single working condition, and the target domain data set comprises sample multi-source monitoring data under multiple working conditions. The application is used to solve the problems of single evaluation dimension and poor working condition adaptability when the existing method is used for power distribution terminal health state early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution terminal intelligent operation and maintenance, in particular to a power distribution terminal health state prediction method, a power distribution terminal health state prediction device, an electronic device, a machine readable storage medium and a computer program product. BACKGROUND

[0002] The sudden failure of key power distribution terminals such as power distribution transformers and circuit breakers in power distribution automation systems will lead to large-scale power outages, which will have a huge impact on people's production and life. Therefore, by conducting health state early warning on key power distribution terminals such as power distribution transformers and circuit breakers, the root cause of the failure can be located in advance, and the situation of sudden failure of power distribution terminals can be reduced.

[0003] The current health state early warning of power distribution terminals mainly relies on periodic maintenance and single-parameter threshold judgment. The existing method has multiple limitations: first, the evaluation dimension is single, only relying on isolated parameters such as temperature or current, ignoring the synergistic effect of multi-source data. Second, the working condition adaptability is poor, when the terminal encounters load mutation or extreme weather, the traditional single-parameter threshold judgment method is easy to produce serious deviation. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a power distribution terminal health state prediction method, device and electronic device, to solve the problem of single evaluation dimension and poor working condition adaptability when the existing method conducts power distribution terminal health state early warning.

[0005] In order to achieve the above-mentioned purpose, the embodiments of the present application provide a power distribution terminal health state prediction method, comprising:

[0006] Obtaining the to-be-tested multi-source monitoring data of the power distribution terminal;

[0007] Inputting the to-be-tested multi-source monitoring data into a power distribution terminal health state prediction model to obtain the power distribution terminal health state output by the power distribution terminal health state prediction model;

[0008] Wherein, the power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm; the source domain data set includes sample multi-source monitoring data under a single working condition, and the target domain data set includes sample multi-source monitoring data under multiple working conditions.

[0009] Optionally, the power distribution terminal health state prediction model is trained by the following steps:

[0010] In the case of working condition switching of the power distribution terminal, obtaining sample multi-source monitoring data of the power distribution terminal under different working conditions;

[0011] constructing a joint training set based on the sample multi-source monitoring data; the joint training set comprises the source domain data set and the target domain data set;

[0012] repeating the following steps until a set end condition is reached: calculating a sampling probability of each sample multi-source monitoring data in the joint training set; performing weighted random sampling on the joint training set based on the sampling probability to obtain a sub-sample data set; performing model training on the sub-sample data set by using a KD tree algorithm or a KNN algorithm to obtain a base classifier; calculating an error rate of the base classifier on the target domain data set; calculating an intermediate weight based on the error rate; adjusting sample weights of the source domain data set and sample weights of the target domain data set based on the intermediate weight;

[0013] constructing an optimized source domain data set based on the sample weights of the source domain data set after the set end condition is reached;

[0014] training the power distribution terminal health state prediction model based on the optimized source domain data set.

[0015] Optionally, the model training on the sub-sample data set by using the KD tree algorithm to obtain the base classifier comprises:

[0016] dividing the sub-sample data set into q sub-regions;

[0017] constructing a KD tree for each of the sub-regions in parallel;

[0018] searching, in parallel, k nearest neighbor samples of each of the sub-regions based on the KD tree of each of the sub-regions and the target domain data set based on the Euclidean distance to obtain q*k candidate samples;

[0019] constructing a target KD tree based on the q*k candidate samples;

[0020] searching, based on the target KD tree, k target nearest neighbor samples based on the Euclidean distance with the target domain data set that are globally minimum;

[0021] performing model training based on the k target nearest neighbor samples to obtain the base classifier.

[0022] Optionally, the model training on the sub-sample data set by using the KNN algorithm to obtain the base classifier comprises:

[0023] dividing the sub-sample data set into q sub-regions;

[0024] calculating the Euclidean distance between the target domain data set and sample multi-source monitoring data in each of the sub-regions;

[0025] determine k initial nearest neighbor samples of each of the sub-regions based on the Euclidean distance, to obtain q*k candidate samples;

[0026] select k target nearest neighbor samples with global minimum from the q*k candidate samples;

[0027] train a model based on the k target nearest neighbor samples, to obtain a base classifier.

[0028] Optionally, the adjusting the sample weight of the source domain data set and the sample weight of the target domain data set based on the intermediate weight comprises:

[0029] in the case that the sample multi-source monitoring data of the target domain data set is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the target domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the target domain data set is reduced; and in the case that the sample multi-source monitoring data of the target domain data set is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the target domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the target domain data set remains unchanged;

[0030] in the case that the sample multi-source monitoring data of the source domain data set is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the source domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain data set is increased; and in the case that the sample multi-source monitoring data of the source domain data set is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the source domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain data set is reduced.

[0031] Optionally, the to-be-tested multi-source monitoring data comprises at least two of electrical data, mechanical data and environmental data of the power distribution terminal; and the sample multi-source monitoring data comprises at least two of sample electrical data, sample mechanical data and sample environmental data of the power distribution terminal.

[0032] In another aspect, the embodiment of the present application also provides a power distribution terminal health state prediction device, comprising:

[0033] an acquisition module configured to acquire to-be-tested multi-source monitoring data of a power distribution terminal;

[0034] a prediction module configured to input the to-be-tested multi-source monitoring data into a power distribution terminal health state prediction model, to obtain a power distribution terminal health state output by the power distribution terminal health state prediction model.

[0035] The power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm; the source domain data set includes sample multi-source monitoring data under a single working condition, and the target domain data set includes sample multi-source monitoring data under multiple working conditions.

[0036] Optionally, the power distribution terminal health state prediction model is trained by the following steps:

[0037] In the case that the working condition of the power distribution terminal is switched, sample multi-source monitoring data of the power distribution terminal under different working conditions is acquired;

[0038] A joint training set is constructed based on the sample multi-source monitoring data; the joint training set includes the source domain data set and the target domain data set;

[0039] All the following steps are repeatedly executed until a set end condition is reached: a sampling probability of each sample multi-source monitoring data in the joint training set is calculated; the joint training set is weighted and randomly sampled based on the sampling probability to obtain a sub-sample data set; a base classifier is obtained by using a KD tree algorithm or a KNN algorithm to train the sub-sample data set; an error rate of the base classifier on the target domain data set is calculated; an intermediate weight is calculated based on the error rate; the sample weight of the source domain data set and the sample weight of the target domain data set are adjusted based on the intermediate weight;

[0040] An optimized source domain data set is constructed based on the sample weight of the source domain data set after the set end condition is reached;

[0041] The power distribution terminal health state prediction model is trained based on the optimized source domain data set.

[0042] Optionally, the KD tree algorithm is used to train the sub-sample data set to obtain a base classifier, including:

[0043] The sub-sample data set is divided into q sub-areas;

[0044] The KD tree of each sub-area is constructed in parallel;

[0045] Based on the KD tree of each sub-area, k nearest neighbor samples of each sub-area and the target domain data set based on the Euclidean distance are searched in parallel to obtain q*k candidate samples;

[0046] A target KD tree is constructed based on the q*k candidate samples;

[0047] Based on the target KD tree, k target nearest neighbor samples based on the Euclidean distance with the target domain data set are searched;

[0048] training a model based on the k target nearest neighbor samples to obtain a base classifier.

[0049] Optionally, the training of the model on the sub-sample dataset by using the KNN algorithm to obtain the base classifier comprises:

[0050] dividing the sub-sample dataset into q sub-regions;

[0051] calculating the Euclidean distance between the target domain dataset and the sample multi-source monitoring data in each of the sub-regions;

[0052] determining k initial nearest neighbor samples of each of the sub-regions based on the Euclidean distance to obtain q*k candidate samples;

[0053] selecting k target nearest neighbor samples with the smallest global value from the q*k candidate samples;

[0054] training a model based on the k target nearest neighbor samples to obtain a base classifier.

[0055] Optionally, the adjusting of the sample weights of the source domain dataset and the sample weights of the target domain dataset based on the intermediate weight comprises:

[0056] in the case that the sample multi-source monitoring data of the target domain dataset is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the target domain dataset based on the intermediate weight to reduce the sample weight of the sample multi-source monitoring data of the target domain dataset; and in the case that the sample multi-source monitoring data of the target domain dataset is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the target domain dataset based on the intermediate weight to keep the sample weight of the sample multi-source monitoring data of the target domain dataset unchanged;

[0057] in the case that the sample multi-source monitoring data of the source domain dataset is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the source domain dataset based on the intermediate weight to increase the sample weight of the sample multi-source monitoring data of the source domain dataset; and in the case that the sample multi-source monitoring data of the source domain dataset is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the source domain dataset based on the intermediate weight to reduce the sample weight of the sample multi-source monitoring data of the source domain dataset.

[0058] Optionally, the to-be-tested multi-source monitoring data comprises at least two of electrical data, mechanical data and environmental data of the power distribution terminal; and the sample multi-source monitoring data comprises at least two of sample electrical data, sample mechanical data and sample environmental data of the power distribution terminal.

[0059] In another aspect, the present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the power distribution terminal health state prediction method when executing the program.

[0060] In another aspect, the present application also provides a machine readable storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the power distribution terminal health state prediction method.

[0061] In another aspect, the present application also provides a computer program product comprising a computer program, wherein the computer program is executable on a processor to implement the power distribution terminal health state prediction method.

[0062] Through the above technical solution, the power distribution terminal health state prediction model of the present application is trained based on sample multi-source monitoring data under multiple working conditions composed of the source domain data set and the target domain data set, thereby avoiding the defect of single evaluation dimension; and the power distribution terminal health state prediction model of the present application is trained based on the source domain data set and the target domain data set using the transfer learning algorithm, which can filter the source domain data closer to the distribution of the target domain data set for training and suppress noise samples, thereby realizing the health state prediction of the power distribution terminal under multiple working conditions and solving the defect of poor working condition adaptability of the existing method.

[0063] Other features and advantages of the present application will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0064] The accompanying drawings are included to provide a further understanding of the present application and constitute a part of the specification, which together with the specific embodiments below, serve to explain the present application, but do not constitute a limitation on the present application. In the drawings:

[0065] Figure 1 is one of the flowcharts of the power distribution terminal health state prediction method provided by the present application;

[0066] Figure 2 is another flowchart of the power distribution terminal health state prediction method provided by the present application;

[0067] Figure 3 is a structural schematic diagram of the power distribution terminal health state prediction device provided by the present application;

[0068] Figure 4 is a structural schematic diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0069] The specific implementation of the embodiments of the present application is described in detail below with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiments of the present application, and is not used to limit the embodiments of the present application.

[0070] Method embodiments

[0071] Please refer to Figure 1 The embodiments of the present application provide a power distribution terminal health state prediction method, comprising:

[0072] Step 100, obtaining the to-be-tested multi-source monitoring data of the power distribution terminal.

[0073] The electronic device obtains the to-be-tested multi-source monitoring data of the power distribution terminal. The power distribution terminal can be various execution units in the power distribution automation system. For example, in the embodiments of the present application, the power distribution terminal can be a feeder terminal (FTU, Feeder Terminal Unit), a station terminal (DTU, Distribution Terminal Unit), a transformer terminal (TTU, Transformer Terminal Unit), or a low-voltage monitoring terminal (LTU, Low-voltage Terminal Unit), etc. The to-be-tested multi-source monitoring data can include at least two of the electrical data, mechanical data and environmental data of the power distribution terminal. That is, the to-be-tested multi-source monitoring data can include any two or three of the electrical data, mechanical data and environmental data of the power distribution terminal.

[0074] In order to expand the evaluation dimension, as much as possible to consider the synergistic effect of multi-source data, and in order to improve the dynamic feature fusion ability, effectively integrate the heterogeneous information generated in the operation of the power distribution terminal. In one embodiment, the to-be-tested multi-source monitoring data can include the electrical data, mechanical data and environmental data of the power distribution terminal. The electrical data of the power distribution terminal can include the voltage, current, power and harmonic data of the power distribution terminal. The mechanical data of the power distribution terminal can include the vibration data and structural protection level of the power distribution terminal. The environmental data of the power distribution terminal can include the temperature and humidity of the environment where the power distribution terminal is located.

[0075] Step 200, inputting the to-be-tested multi-source monitoring data into the power distribution terminal health state prediction model to obtain the power distribution terminal health state output by the power distribution terminal health state prediction model.

[0076] The electronic device inputs the to-be-tested multi-source monitoring data of the power distribution terminal into a power distribution terminal health state prediction model to obtain a power distribution terminal health state output by the power distribution terminal health state prediction model. The power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm. The source domain data set includes sample multi-source monitoring data under a single working condition, and the target domain data set includes sample multi-source monitoring data under multiple working conditions. In the embodiment of the present application, the source domain data set can be understood as sample multi-source monitoring data under a single working condition in old data of the power distribution terminal. The target domain data set can be understood as sample multi-source monitoring data under multiple working conditions in new data of the power distribution terminal.

[0077] In one embodiment, the multiple working conditions of the power distribution terminal can include a heavy load working condition, a light load working condition, and an impact load working condition of the power distribution terminal. The light load working condition can be that the actual load of the power distribution terminal is much lower than the rated capacity, for example, the load rate is usually ≤ 30%. The heavy load working condition can be that the actual load rate of the power distribution terminal is between 70% and 90%, close to the rated capacity. The impact load working condition can be a non-periodic or periodic sudden load, and the peak value can reach tens of times of the average load. The sample multi-source monitoring data includes at least two of the sample electrical data, the sample mechanical data, and the sample environmental data of the power distribution terminal. That is, the sample multi-source monitoring data can include any two or three of the sample electrical data, the sample mechanical data, and the sample environmental data of the power distribution terminal. In one embodiment, the sample multi-source monitoring data can include the sample electrical data, the sample mechanical data, and the sample environmental data of the power distribution terminal. The sample multi-source monitoring data can be obtained based on historical multi-source monitoring data of the power distribution terminal. The power distribution terminal health state prediction model can use various neural network models, such as convolutional neural networks, long short-term memory networks, etc.

[0078] Since the existing training data set under a single working condition is difficult to contain all potential working conditions when the working condition of the power distribution terminal changes, and the algorithm is prone to noise interference and poor effect when the distribution of the target domain and the source domain is greatly different, in order to make the data distribution of the source domain data set closer to the data distribution of the target domain data set, suppress noise samples, and realize health state prediction of the power distribution terminal under multiple working conditions, the power distribution terminal health state prediction model (or power distribution equipment health state early warning model) in the embodiment of the present application can be trained based on the source domain data set and the target domain data set by using a transfer learning algorithm. The goal of the transfer learning algorithm is to effectively extract the knowledge learned in one environment to help learning in a new environment. For example, the TrAdaBoost algorithm can be used for model fine-tuning in the embodiment of the present application. TrAdaBoost is an instance-based transfer learning algorithm with strong knowledge transfer capability. When the working condition of the power distribution terminal changes, the existing training data set is difficult to contain all potential working conditions, and the model fine-tuning method based on improved TrAdaBoost in the embodiment of the present application can filter and train the source domain data closer to the distribution of the target domain data set, suppress noise samples, and thus realize health state prediction of the power distribution terminal under multiple working conditions, solving the defect of poor working condition adaptability of the existing method.

[0079] Therefore, the power distribution terminal health state prediction model in the embodiment of the present application is trained based on sample multi-source monitoring data under multiple working conditions composed of the source domain data set and the target domain data set, avoiding the defect of single evaluation dimension; and the power distribution terminal health state prediction model in the embodiment of the present application is trained based on the source domain data set and the target domain data set by using a transfer learning algorithm, which can filter and train the source domain data closer to the distribution of the target domain data set, suppress noise samples, and thus realize health state prediction of the power distribution terminal under multiple working conditions, solving the defect of poor working condition adaptability of the existing method.

[0080] In other aspects of the embodiment of the present application, please refer to Figure 2 , the power distribution terminal health state prediction model is trained by the following steps:

[0081] Step 11, in the case of working condition switching of the power distribution terminal, sample multi-source monitoring data of the power distribution terminal under different working conditions is acquired.

[0082] The electronic device acquires sample multi-source monitoring data of the power distribution terminal under different working conditions in the case of working condition switching of the power distribution terminal.

[0083] Step 12, a joint training set is constructed based on the sample multi-source monitoring data; the joint training set includes the source domain data set and the target domain data set.

[0084] The electronic device acquires sample multi-source monitoring data of the power distribution terminal under different working conditions, and constructs a joint training set , T a is a source domain dataset, T b is a target domain dataset. And the electronic device can initialize the sample weights in the source domain dataset and the target domain dataset: the sample weight of the source domain dataset is ; and the sample weight in the target domain dataset is . Wherein n, m are the sample quantity, i represents the i-th sample multi-source monitoring data in the source domain dataset, and j represents the j-th sample multi-source monitoring data in the target domain dataset.

[0085] Step 13, repeat the following all steps until the set end condition is reached: calculate the sampling probability of each sample multi-source monitoring data in the joint training set; based on the sampling probability, weight random sampling is performed on the joint training set to obtain a sub-sample dataset; using KD tree algorithm or KNN algorithm, the model is trained on the sub-sample dataset to obtain a base classifier; calculate the error rate of the base classifier on the target domain dataset; based on the error rate calculation, the intermediate weight is calculated; based on the intermediate weight, the sample weight of the source domain dataset and the sample weight of the target domain dataset are adjusted.

[0086] The electronic device can initialize the weight vector w 1 = (w 1 1, …, w 1 n+m ). Wherein, ; wherein n is the sample quantity of the source domain dataset, and m is the sample quantity of the target domain dataset. The electronic device calculates the sampling probability of each sample multi-source monitoring data in the joint training set as ; wherein t is the iteration number. The electronic device sampling probability p t is weighted randomly sampled on the joint training set to obtain a sub-sample dataset M. The electronic device can divide the sub-sample dataset M into q sub-areas, and independently call KD tree algorithm or KNN algorithm on each sub-area to train the model of the sub-sample dataset to obtain a base classifier h t . The embodiment of the application proposes an improved model fine-tuning method of TrAdaBoost, designs a parallel multi-channel mode, divides the sub-sample dataset M into multiple sub-areas (parallel channels), and each sub-area uses KD tree algorithm or KNN algorithm in parallel to speed up model training. And the embodiment of the application proposes to use a weighted random sampling method, dynamically adjusts the sampling probability according to the sample weight, selects the source domain data closer to the target domain distribution, and suppresses the noise samples. Then the electronic device calculates the error rate of the base classifier on the target domain dataset . Wherein ≥ 0.5, terminate the iteration. denotes the indicator function, which outputs 1 if the condition in the bracket is true, and 0 otherwise. For example, denotes 1 when the classification is wrong. x t j denotes the jth sample multi-source monitoring data in the target domain dataset, and t denotes the iteration number. t x t j denotes the base classifier h t for x t j denotes the health state prediction result of the power distribution terminal y t j denotes x t j denotes the true health state of the corresponding power distribution terminal. denotes the sample weight of the jth sample in the target domain dataset at the tth iteration. The electronic device further calculates an intermediate weight based on the error rate. In one embodiment, the electronic device updates the intermediate weight according to the error rate β t and γ t . The calculation formula of the intermediate weight β t and γ t is as follows:

[0087] ; The electronic device further adjusts the sample weights of the source domain dataset and the sample weights of the target domain dataset based on the intermediate weight. The above steps are repeatedly executed until the termination iteration condition ≥ 0.5 is reached.

[0088] Step 14, constructing an optimized source domain dataset based on the sample weights of the source domain dataset after the set end condition is reached.

[0089] Step 15, training the power distribution terminal health state prediction model based on the optimized source domain dataset.

[0090] ​The electronic device normalizes the sample weights of the source domain data set after the set end condition is reached to obtain an optimized source domain data set. The electronic device trains the power distribution terminal health state prediction model based on the optimized source domain data set. Specifically, the electronic device can train based on the optimized source domain data set in a supervised manner, calculate a loss function by calculating the deviation between the model prediction result and the true label, adjust the model parameters of the power distribution terminal health state prediction model based on the loss function, and obtain the trained power distribution terminal health state prediction model. The embodiment of the present application updates the power distribution terminal health state prediction model in real time in the case of working condition switching of the power distribution terminal, and improves the generalization ability of the model.

[0091] When the data distribution difference between the target domain data set and the source domain data set is large, the algorithm is susceptible to noise interference and the effect is poor. The model fine-tuning method based on the improved TrAdaBoost designed in the embodiment of the present application first uses KNN algorithm or KD tree as Learner, designs a parallel multi-channel mode, divides the sub-sample data set into multiple sub-regions (parallel channels), and accelerates model training. At the same time, a weighted random sampling method is proposed, which dynamically adjusts the sampling probability according to the sample weight, selects the source domain data closer to the target domain distribution, and suppresses the noise samples.

[0092] In other aspects of the embodiment of the present application, the model training of the sub-sample data set using the KD tree algorithm to obtain the base classifier comprises: dividing the sub-sample data set into q sub-regions; constructing the KD tree of each sub-region in parallel; based on the KD tree of each sub-region, searching out k nearest neighbor samples of each sub-region based on the Euclidean distance of the target domain data set to obtain q*k candidate samples; constructing a target KD tree based on the q*k candidate samples; searching out k target nearest neighbor samples based on the Euclidean distance of the target domain data set based on the target KD tree; and training the model based on the k target nearest neighbor samples to obtain the base classifier.

[0093] The electronic device divides the sub-sample data set M into q sub-regions, independently constructs a KD tree for the data of each sub-region, searches all KD trees in parallel during query, and takes the k nearest neighbor samples of the target domain data set based on the Euclidean distance after merging the results. Specifically, the electronic device constructs the KD tree of each sub-region in parallel during the construction phase. During the query phase, the electronic device searches the k nearest neighbor samples of each sub-region and the target domain data set based on the Euclidean distance in parallel to obtain q*k candidate samples. Finally, the electronic device merges the q*k candidate samples into a large set, constructs a target KD tree, and searches the globally minimum k target nearest neighbor samples of the target domain data set based on the Euclidean distance based on the target KD tree. Finally, the k target nearest neighbor samples are used for model training to obtain a base classifier.

[0094] Therefore, the embodiment of the application solves the problem of low calculation efficiency of the power distribution terminal when the data distribution difference between the source domain data set and the target domain data set is large under multiple working conditions by using weighted random sampling and parallel calculation based on the KD tree algorithm.

[0095] In other aspects of the embodiment of the application, the model training of the sub-sample data set by using the KNN algorithm to obtain a base classifier includes: dividing the sub-sample data set into q sub-regions; calculating the Euclidean distance between the target domain data set and the sample multi-source monitoring data in each sub-region; determining the k initial nearest neighbor samples of each sub-region based on the Euclidean distance to obtain q*k candidate samples; screening the globally minimum k target nearest neighbor samples from the q*k candidate samples; and performing model training based on the k target nearest neighbor samples to obtain a base classifier.

[0096] The electronic device first uses the KNN algorithm as a Learner, designs a parallel multi-channel mode, divides the sub-sample data set into multiple sub-regions (parallel channels), calculates the Euclidean distance between the sub-regions and the target domain data set after combining the results, and accelerates model training. Specifically, the sub-sample data set is divided into q sub-regions; the Euclidean distance between the target domain data set and the sample multi-source monitoring data in each sub-region is calculated; the k initial nearest neighbor samples of each sub-region are determined based on the Euclidean distance to obtain q*k candidate samples; the globally minimum k target nearest neighbor samples are screened from the q*k candidate samples; and model training is performed based on the k target nearest neighbor samples to obtain a base classifier.

[0097] Therefore, the embodiment of the application solves the problem of low calculation efficiency of the power distribution terminal when the data distribution difference between the source domain data set and the target domain data set is large under multiple working conditions by using weighted random sampling and parallel calculation based on the KNN algorithm.

[0098] In other aspects of embodiments of the present invention, adjusting the sample weights of the source domain dataset and the target domain dataset based on the intermediate weights includes:

[0099] If the multi-source monitoring data of the target domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to reduce the sample weights of the multi-source monitoring data of the target domain dataset; and if the multi-source monitoring data of the target domain dataset is incorrectly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to keep the sample weights of the multi-source monitoring data of the target domain dataset unchanged.

[0100] In this embodiment of the invention, the intermediate weight β t The calculation formula is as follows:

[0101] ;in, The error rate of the base classifier on the target domain dataset. β t The intermediate weights are used as follows: If the multi-source monitoring data of the target domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to reduce their overall weight; conversely, if the multi-source monitoring data of the target domain dataset is incorrectly classified, the sample weights are adjusted based on the intermediate weights to keep them unchanged. Specifically, the formula for updating the sample weights of the multi-source monitoring data of the target domain dataset is as follows:

[0102] ;

[0103] in, w b j ( t +1) is the first t The sample weights of the j-th sample from the multi-source monitoring data of the +1th iteration. w b j ( t ) is the first t The sample weight of the j-th sample from the multi-source monitoring data; β tis an intermediate weight; t represents the number of iterations. The embodiment of the present application reduces the weight of the correctly classified samples and keeps the weight of the misclassified samples unchanged for the target domain dataset. Thus, the embodiment of the present application prevents the misclassified samples from being ignored due to the weight reduction, forces the subsequent base classifier to continuously optimize these difficult samples, and improves the generalization ability by retaining the difficult samples. The embodiment of the present application reduces the weight of the correctly classified samples, reduces the redundant calculation of the model on the simple samples, that is, reduces the redundant learning, and thus focuses on the difficult samples. Thus, the embodiment of the present application accelerates the learning of the difficult samples of the target domain and avoids underfitting.

[0104] In the case that the sample multi-source monitoring data of the source domain dataset is correctly classified, the sample weight of the sample multi-source monitoring data of the source domain dataset is adjusted based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain dataset is increased; and in the case that the sample multi-source monitoring data of the source domain dataset is misclassified, the sample weight of the sample multi-source monitoring data of the source domain dataset is adjusted based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain dataset is reduced.

[0105] In the embodiment of the present application, the intermediate weight β t The calculation formula is as follows:

[0106] ; wherein, is the error rate of the base classifier on the target domain dataset, β t is an intermediate weight. In the case that the sample multi-source monitoring data of the source domain dataset is correctly classified, the sample weight of the sample multi-source monitoring data of the source domain dataset is adjusted based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain dataset is increased; and in the case that the sample multi-source monitoring data of the source domain dataset is misclassified, the sample weight of the sample multi-source monitoring data of the source domain dataset is adjusted based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain dataset is reduced. Specifically, the sample weight updating formula of the sample multi-source monitoring data of the source domain dataset is as follows:

[0107] ;

[0108] wherein, is the sample weight of the i th sample multi-source monitoring data in the t th iteration, t is the sample weight of the i th sample multi-source monitoring data in the t+1 th iteration, is the sample weight of the i th sample multi-source monitoring data in the t th iteration, t is the sample weight of the i th sample multi-source monitoring data in the t+1 th iteration; β t and gammat is the intermediate weight; t represents the iteration number. Thus, the embodiment of the present application increases the weight of the correctly classified samples in the source domain dataset and reduces the weight of the incorrectly classified samples. Thus, the embodiment of the present application reduces the weight of the incorrectly classified samples in the source domain dataset, gradually eliminates the source domain samples with large distribution difference from the target domain dataset, filters the interference samples with distribution drift, and realizes the suppression of noise with inconsistent distribution. The embodiment of the present application increases the weight of the correctly classified samples in the source domain dataset, retains the transferable common knowledge, and improves the utilization rate of the source domain dataset. That is, the effective knowledge is retained, and the knowledge related to the target domain dataset is strengthened. Thus, the embodiment of the present application realizes the precise screening of the knowledge of the source domain dataset.

[0109] In other aspects of the embodiment of the present application, after step 200, the method further includes: sending an alarm signal to an operation and maintenance system based on the abnormal health state of the power distribution terminal. Thus, the health state early warning of the power distribution terminal under multiple working conditions is realized, and the fault root cause can be located in time.

[0110] To sum up, the embodiment of the present application realizes the health state prediction of the power distribution terminal under multiple working conditions based on the model fine-tuning method of the improved transfer learning algorithm, innovatively introduces the parallel multi-channel weighted sampling mechanism and the dynamic distribution adaptation strategy, can quickly predict the health state of the power distribution terminal based on the cross-domain knowledge transfer and the historical multi-source monitoring data of the power distribution terminal, and solves the problems of calculation efficiency and model generalization.

[0111] Device embodiment

[0112] Please refer to Figure 3 On the other hand, the embodiment of the present application also provides a power distribution terminal health state prediction device, which comprises:

[0113] The acquisition module 301 is configured to acquire the to-be-tested multi-source monitoring data of the power distribution terminal.

[0114] The prediction module 302 is configured to input the to-be-tested multi-source monitoring data into a power distribution terminal health state prediction model to obtain the power distribution terminal health state output by the power distribution terminal health state prediction model.

[0115] The power distribution terminal health state prediction model is trained based on a source domain dataset and a target domain dataset by using a transfer learning algorithm; the source domain dataset comprises sample multi-source monitoring data under a single working condition, and the target domain dataset comprises sample multi-source monitoring data under multiple working conditions.

[0116] Optionally, the power distribution terminal health state prediction model is trained by the following steps:

[0117] In the case where the power distribution terminal switches the working condition, the sample multi-source monitoring data of the power distribution terminal under different working conditions is acquired.

[0118] constructing a joint training set based on the sample multi-source monitoring data; the joint training set comprises the source domain data set and the target domain data set;

[0119] repeating the following steps until a set end condition is reached: calculating a sampling probability of each sample multi-source monitoring data in the joint training set; performing weighted random sampling on the joint training set based on the sampling probability to obtain a sub-sample data set; performing model training on the sub-sample data set by using a KD tree algorithm or a KNN algorithm to obtain a base classifier; calculating an error rate of the base classifier on the target domain data set; calculating an intermediate weight based on the error rate; adjusting sample weights of the source domain data set and sample weights of the target domain data set based on the intermediate weight;

[0120] constructing an optimized source domain data set based on the sample weights of the source domain data set after the set end condition is reached;

[0121] training the power distribution terminal health state prediction model based on the optimized source domain data set.

[0122] Optionally, the model training on the sub-sample data set by using the KD tree algorithm to obtain the base classifier comprises:

[0123] dividing the sub-sample data set into q sub-regions;

[0124] constructing KD trees of each of the sub-regions in parallel;

[0125] searching, in parallel, k nearest neighbor samples of each of the sub-regions based on the Euclidean distance from the target domain data set to obtain q*k candidate samples;

[0126] constructing a target KD tree based on the q*k candidate samples;

[0127] searching, based on the target KD tree, k target nearest neighbor samples with the smallest global Euclidean distance from the target domain data set;

[0128] performing model training based on the k target nearest neighbor samples to obtain the base classifier.

[0129] Optionally, the model training on the sub-sample data set by using the KNN algorithm to obtain the base classifier comprises:

[0130] dividing the sub-sample data set into q sub-regions;

[0131] calculating the Euclidean distance between the target domain data set and sample multi-source monitoring data in each of the sub-regions;

[0132] determine k initial nearest neighbor samples of each of the sub-regions based on the Euclidean distance, to obtain q*k candidate samples;

[0133] select k target nearest neighbor samples with global minimum from the q*k candidate samples;

[0134] train a model based on the k target nearest neighbor samples, to obtain a base classifier.

[0135] Optionally, the adjusting the sample weight of the source domain data set and the sample weight of the target domain data set based on the intermediate weight comprises:

[0136] in the case that the sample multi-source monitoring data of the target domain data set is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the target domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the target domain data set is reduced; and in the case that the sample multi-source monitoring data of the target domain data set is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the target domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the target domain data set remains unchanged;

[0137] in the case that the sample multi-source monitoring data of the source domain data set is classified correctly, adjusting the sample weight of the sample multi-source monitoring data of the source domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain data set is increased; and in the case that the sample multi-source monitoring data of the source domain data set is classified incorrectly, adjusting the sample weight of the sample multi-source monitoring data of the source domain data set based on the intermediate weight, so that the sample weight of the sample multi-source monitoring data of the source domain data set is reduced.

[0138] Optionally, the to-be-tested multi-source monitoring data comprises at least two of electrical data, mechanical data and environmental data of the power distribution terminal; and the sample multi-source monitoring data comprises at least two of sample electrical data, sample mechanical data and sample environmental data of the power distribution terminal.

[0139] The power distribution terminal health state prediction device comprises a processor and a memory, the acquisition module 301, the prediction module 302 and the like are stored in the memory as program units, and the corresponding functions are realized by the processor executing the program units stored in the memory.

[0140] The processor comprises a core, and the core retrieves the corresponding program units from the memory. The core can be provided with one or more than one.

[0141] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0142] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a power distribution terminal health status prediction method. The method includes: acquiring multi-source monitoring data to be tested from the power distribution terminal; inputting the multi-source monitoring data to be tested into a power distribution terminal health status prediction model to obtain the power distribution terminal health status output by the power distribution terminal health status prediction model; wherein the power distribution terminal health status prediction model is trained using a transfer learning algorithm based on a source domain dataset and a target domain dataset; the source domain dataset includes sample multi-source monitoring data under a single operating condition, and the target domain dataset includes sample multi-source monitoring data under multiple operating conditions.

[0143] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0144] In another aspect, the present application also provides a computer program product comprising a computer program, which can be stored on a machine readable storage medium, and the computer program can be executed by a processor to enable a computer to perform a power distribution terminal health state prediction method, which comprises: obtaining to-be-tested multi-source monitoring data of a power distribution terminal; inputting the to-be-tested multi-source monitoring data into a power distribution terminal health state prediction model to obtain a power distribution terminal health state output by the power distribution terminal health state prediction model; wherein the power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm; the source domain data set comprises sample multi-source monitoring data under a single working condition, and the target domain data set comprises sample multi-source monitoring data under multiple working conditions.

[0145] In another aspect, the present application also provides a machine readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement a power distribution terminal health state prediction method, which comprises: obtaining to-be-tested multi-source monitoring data of a power distribution terminal; inputting the to-be-tested multi-source monitoring data into a power distribution terminal health state prediction model to obtain a power distribution terminal health state output by the power distribution terminal health state prediction model; wherein the power distribution terminal health state prediction model is trained based on a source domain data set and a target domain data set by using a transfer learning algorithm; the source domain data set comprises sample multi-source monitoring data under a single working condition, and the target domain data set comprises sample multi-source monitoring data under multiple working conditions.

[0146] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement without creative labor.

[0147] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary general hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0148] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the health status of a power distribution terminal, characterized in that, include: Acquire the multi-source monitoring data to be tested from the power distribution terminal; The multi-source monitoring data to be tested is input into the power distribution terminal health status prediction model to obtain the power distribution terminal health status output by the power distribution terminal health status prediction model. The power distribution terminal health status prediction model is trained using a transfer learning algorithm based on a source domain dataset and a target domain dataset. The source domain dataset includes multi-source monitoring data of samples under a single operating condition, and the target domain dataset includes multi-source monitoring data of samples under multiple operating conditions. The power distribution terminal health status prediction model is trained through the following steps: When the power distribution terminal switches operating conditions, sample multi-source monitoring data of the power distribution terminal under different operating conditions are acquired. A joint training set is constructed based on the multi-source monitoring data of the samples; the joint training set includes the source domain dataset and the target domain dataset. Repeat all the following steps until the set termination condition is met: Calculate the sampling probability of each sample from the multi-source monitoring data in the joint training set; perform weighted random sampling on the joint training set based on the sampling probability to obtain a subsample dataset; train the model on the subsample dataset using the KD-tree algorithm or the KNN algorithm to obtain a base classifier; calculate the error rate of the base classifier on the target domain dataset; calculate intermediate weights based on the error rate; adjust the sample weights of the source domain dataset and the target domain dataset based on the intermediate weights. An optimized source domain dataset is constructed based on the sample weights of the source domain dataset after the set termination conditions are met. The power distribution terminal health status prediction model is trained based on the optimized source domain dataset. Using the KNN algorithm or KD tree as the learner, a parallel multi-channel approach was designed to divide the subsample dataset into multiple parallel channels. At the same time, a weighted random sampling method was proposed to dynamically adjust the sampling probability according to the sample weight and select source domain data that are closer to the distribution of the target domain.

2. The method for predicting the health status of a power distribution terminal according to claim 1, characterized in that, The step of training a model on the subsample dataset using the KD-tree algorithm to obtain a base classifier includes: The subsample dataset is divided into q sub-regions; Construct the KD tree for each of the sub-regions in parallel; Based on the KD tree of each sub-region, the k nearest neighbor samples of each sub-region and the target domain dataset based on Euclidean distance are searched in parallel to obtain q*k candidate samples; Construct a target KD tree based on the q*k candidate samples; Based on the target KD tree, search for the k target nearest neighbor samples that are globally minimum based on Euclidean distance to the target domain dataset; The base classifier is obtained by training the model based on the k nearest neighbor samples of the target.

3. The method for predicting the health status of a power distribution terminal according to claim 1, characterized in that, The step of training a model on the subsample dataset using the KNN algorithm to obtain a base classifier includes: The subsample dataset is divided into q sub-regions; Calculate the Euclidean distance between the target domain dataset and the multi-source monitoring data of samples in each of the sub-regions; Based on the Euclidean distance, k initial nearest neighbor samples are determined for each sub-region, resulting in q*k candidate samples; Select the k smallest target nearest neighbor samples from the q*k candidate samples; The base classifier is obtained by training the model based on the k nearest neighbor samples of the target.

4. The method for predicting the health status of a power distribution terminal according to claim 1, characterized in that, The step of adjusting the sample weights of the source domain dataset and the target domain dataset based on the intermediate weights includes: If the multi-source monitoring data of the target domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to reduce the sample weights of the multi-source monitoring data of the target domain dataset; and if the multi-source monitoring data of the target domain dataset is incorrectly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to keep the sample weights of the multi-source monitoring data of the target domain dataset unchanged. If the multi-source monitoring data of the source domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the source domain dataset are adjusted based on the intermediate weights to increase the sample weights of the multi-source monitoring data of the source domain dataset; and if the multi-source monitoring data of the source domain dataset is incorrectly classified, the sample weights of the multi-source monitoring data of the source domain dataset are adjusted based on the intermediate weights to decrease the sample weights of the multi-source monitoring data of the source domain dataset.

5. The method for predicting the health status of a power distribution terminal according to any one of claims 1 to 4, characterized in that, The multi-source monitoring data to be tested includes at least two of the electrical data, mechanical data, and environmental data of the power distribution terminal; the sample multi-source monitoring data includes at least two of the sample electrical data, sample mechanical data, and sample environmental data of the power distribution terminal.

6. A power distribution terminal health status prediction device, characterized in that, include: The acquisition module is used to acquire the multi-source monitoring data to be tested from the power distribution terminal. The prediction module is used to input the multi-source monitoring data to be tested into the power distribution terminal health status prediction model to obtain the power distribution terminal health status output by the power distribution terminal health status prediction model. The power distribution terminal health status prediction model is trained using a transfer learning algorithm based on a source domain dataset and a target domain dataset. The source domain dataset includes multi-source monitoring data of samples under a single operating condition, and the target domain dataset includes multi-source monitoring data of samples under multiple operating conditions. The power distribution terminal health status prediction model is trained through the following steps: When the power distribution terminal switches operating conditions, sample multi-source monitoring data of the power distribution terminal under different operating conditions are acquired. A joint training set is constructed based on the multi-source monitoring data of the samples; the joint training set includes the source domain dataset and the target domain dataset. Repeat all the following steps until the set termination condition is met: Calculate the sampling probability of each sample from the multi-source monitoring data in the joint training set; perform weighted random sampling on the joint training set based on the sampling probability to obtain a subsample dataset; train the model on the subsample dataset using the KD-tree algorithm or the KNN algorithm to obtain a base classifier; calculate the error rate of the base classifier on the target domain dataset; calculate intermediate weights based on the error rate; adjust the sample weights of the source domain dataset and the target domain dataset based on the intermediate weights. An optimized source domain dataset is constructed based on the sample weights of the source domain dataset after the set termination conditions are met. The power distribution terminal health status prediction model is trained based on the optimized source domain dataset. Using the KNN algorithm or KD tree as the learner, a parallel multi-channel approach was designed to divide the subsample dataset into multiple parallel channels. At the same time, a weighted random sampling method was proposed to dynamically adjust the sampling probability according to the sample weight and select source domain data that are closer to the distribution of the target domain.

7. The power distribution terminal health status prediction device according to claim 6, characterized in that, The step of training a model on the subsample dataset using the KD-tree algorithm to obtain a base classifier includes: The subsample dataset is divided into q sub-regions; Construct the KD tree for each of the sub-regions in parallel; Based on the KD tree of each sub-region, the k nearest neighbor samples of each sub-region and the target domain dataset based on Euclidean distance are searched in parallel to obtain q*k candidate samples; Construct a target KD tree based on the q*k candidate samples; Based on the target KD tree, search for the k target nearest neighbor samples that are globally minimum based on Euclidean distance to the target domain dataset; The base classifier is obtained by training the model based on the k nearest neighbor samples of the target.

8. The power distribution terminal health status prediction device according to claim 6, characterized in that, The step of training a model on the subsample dataset using the KNN algorithm to obtain a base classifier includes: The subsample dataset is divided into q sub-regions; Calculate the Euclidean distance between the target domain dataset and the multi-source monitoring data of samples in each of the sub-regions; Based on the Euclidean distance, k initial nearest neighbor samples are determined for each sub-region, resulting in q*k candidate samples; Select the k smallest target nearest neighbor samples from the q*k candidate samples; The base classifier is obtained by training the model based on the k nearest neighbor samples of the target.

9. The power distribution terminal health status prediction device according to claim 6, characterized in that, The step of adjusting the sample weights of the source domain dataset and the target domain dataset based on the intermediate weights includes: If the multi-source monitoring data of the target domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to reduce the sample weights of the multi-source monitoring data of the target domain dataset; and if the multi-source monitoring data of the target domain dataset is incorrectly classified, the sample weights of the multi-source monitoring data of the target domain dataset are adjusted based on the intermediate weights to keep the sample weights of the multi-source monitoring data of the target domain dataset unchanged. If the multi-source monitoring data of the source domain dataset is correctly classified, the sample weights of the multi-source monitoring data of the source domain dataset are adjusted based on the intermediate weights to increase the sample weights of the multi-source monitoring data of the source domain dataset; and if the multi-source monitoring data of the source domain dataset is incorrectly classified, the sample weights of the multi-source monitoring data of the source domain dataset are adjusted based on the intermediate weights to decrease the sample weights of the multi-source monitoring data of the source domain dataset.

10. The distribution terminal health status prediction device according to any one of claims 6 to 9, characterized in that, The multi-source monitoring data to be tested includes at least two of the electrical data, mechanical data, and environmental data of the power distribution terminal; the sample multi-source monitoring data includes at least two of the sample electrical data, sample mechanical data, and sample environmental data of the power distribution terminal.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the power distribution terminal health status prediction method according to any one of claims 1 to 5.

12. A machine-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the power distribution terminal health status prediction method according to any one of claims 1 to 5.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the power distribution terminal health status prediction method according to any one of claims 1 to 5.

Images