Power distribution network line fault prediction method, apparatus and device, and storage medium

Abstract

The invention discloses a power distribution network line fault prediction method and device, equipment and a storage medium, and the method comprises the steps: obtaining equipment in a power distribution network line and meteorological data corresponding to an area where the equipment is located, and obtaining n pieces of equipment and n pieces of meteorological data; the equipment comprises one of the following types: an overhead line, a cable line, pole-mounted equipment and a transformer; respectively inputting the n pieces of meteorological data into n equipment fault prediction models corresponding to the n pieces of equipment to obtain n equipment fault probabilities; determining a line fault index of the power distribution network line according to the n equipment fault probabilities; and determining a target risk level of the line fault according to the line fault index. Therefore, the corresponding fault prediction models are trained through the multiple devices in the power distribution network line, the fault index of the line is predicted according to the fault probability of the devices on the power distribution network line, and accurate prediction of the power distribution network line fault under the multi-source data is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power failure prediction, and in particular to a power distribution network line failure prediction method, device, equipment and storage medium. BACKGROUND

[0002] The power distribution network is located at the end of the power supply system, and its operation reliability directly affects the power supply quality of the end user. With the continuous expansion of the power distribution network, the aging of equipment and the increase of extreme weather events, the frequency of power distribution line failures is on the rise. The traditional failure prediction method focuses on a single meteorological factor, has insufficient model generalization ability, and is difficult to quantify the risk of different equipment categories, which makes it difficult to meet the needs of refined management.

[0003] Therefore, how to improve the accuracy of power distribution network line failure prediction has become a problem to be solved. SUMMARY

[0004] The embodiments of the present application provide a power distribution network line failure prediction method, device, equipment and storage medium, which trains a corresponding failure prediction model for each device in the power distribution network line, so as to predict the failure index of the line according to the failure probability of the device on the power distribution network line, and realizes accurate prediction of the power distribution network line failure under multi-source data.

[0005] In a first aspect, the embodiments of the present application provide a power distribution network line failure prediction method applied to a server, and the method comprises: Obtain the meteorological data corresponding to the devices in the power distribution network line and the regions where the devices are located, to obtain n devices and n meteorological data; the types of the devices include one of the following: overhead line, cable line, pole-mounted device, and transformer; each meteorological data includes daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, and maximum wind speed; n is a positive integer; Input the n meteorological data into n device failure prediction models corresponding to the n devices respectively, to obtain n device failure probabilities; each device failure prediction model is a pre-trained model for inputting meteorological data and outputting device failure probability; each device corresponds to a device failure prediction model; Determine the line failure index of the power distribution network line according to the n device failure probabilities; Determine the target risk level of line failure according to the line failure index.

[0006] In a second aspect, the embodiments of the present application provide a power distribution network line failure prediction device applied to a server, and the power distribution network line failure prediction device comprises: The data acquisition module is configured to acquire meteorological data corresponding to devices in a power distribution network line and areas where the devices are located, to obtain n devices and n meteorological data; the types of the devices include one of the following: overhead line, cable line, pole-mounted device, and transformer; each meteorological data includes the following: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, and maximum wind speed; n is a positive integer; The device fault probability prediction module is configured to input the n meteorological data into n device fault prediction models corresponding to the n devices respectively, to obtain n device fault probabilities; each device fault prediction model is a pre-trained model for inputting meteorological data and outputting device fault probability; each device corresponds to a device fault prediction model. The line fault index calculation module is configured to determine a line fault index of the power distribution network line according to the n device fault probabilities. The risk level determination module is configured to determine a target risk level of line fault according to the line fault index.

[0007] In a third aspect, an electronic device is provided, including: a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for performing the steps in the first aspect of the embodiments.

[0008] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program for electronic data exchange, wherein the computer program causes a computer to perform some or all of the steps described in the first aspect of the embodiments.

[0009] In a fifth aspect, a computer program product is provided, which includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform some or all of the steps described in the first aspect of the embodiments. The computer program product can be a software installation package.

[0010] It can be seen that, by using the embodiments of the present application, the following beneficial effects are achieved: By implementing the embodiment of the application, meteorological data corresponding to the devices in the power distribution network line and the area where the devices are located are obtained, n devices and n meteorological data are obtained; the type of the device includes one of the following: overhead line, cable line, pole-mounted device, transformer; each meteorological data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, maximum wind speed; the n meteorological data are respectively input into n device fault prediction models corresponding to the n devices, n device fault probabilities are obtained; each device fault prediction model is a model pre-trained for inputting meteorological data and outputting device fault probability; each device corresponds to a device fault prediction model; the line fault index of the power distribution network line is determined according to the n device fault probabilities; the target risk level of the line fault is determined according to the line fault index. As can be seen, by training corresponding fault prediction models for various devices in the power distribution network line, the fault index of the line is predicted according to the fault probability of the device on the power distribution network line, thereby realizing accurate prediction of the fault of the power distribution network line under multi-source data. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the application or the background art, the drawings needed to be used in the embodiments of the application or the background art will be described below.

[0012] Figure 1 is a flowchart of a power distribution network line fault prediction method provided by the embodiments of the application; Figure 2 is a system architecture diagram of a power distribution network line fault prediction system provided by the embodiments of the application; Figure 3 is a model application architecture diagram of a power distribution network line fault prediction method provided by the embodiments of the application; Figure 4 is a flowchart of training a device fault prediction model provided by the embodiments of the application; Figure 5 is a flowchart of decision tree construction provided by the embodiments of the application; Figure 6 is a structural diagram of a power distribution network line fault prediction device provided by the embodiments of the application; Figure 7 is a structural diagram of an electronic device provided by the embodiments of the application. DETAILED DESCRIPTION

[0013] In the following, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0014] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed or can optionally include other steps or units inherent to the process, method, product or device.

[0015] Reference to "embodiments" herein means that the specific features, structures or properties described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily independent or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0016] The related content, concepts, meanings, technical problems, technical solutions, beneficial effects, etc. involved in the embodiments of the present application are described below.

[0017] First, some terms involved in the present application are explained and described: XGBoost model: XGBoost model is an ensemble learning model based on gradient boosting tree improvement, which realizes prediction through weighted combination of multiple decision trees, and is widely used in classification, regression and other tasks, which not only retains the stability of ensemble model, but also improves the training speed and practicability through engineering optimization.

[0018] Please refer to Figure 1 , Figure 1 is a flowchart of a power distribution network line fault prediction method provided by the embodiments of the present application, the method is applied to a server, and the method includes but is not limited to the following steps: S101, obtain meteorological data corresponding to the equipment in the power distribution network line and the area where the equipment is located, obtain n equipment and n meteorological data; the type of the equipment includes one of the following: overhead line, cable line, pole-mounted equipment, transformer; each meteorological data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, maximum wind speed.

[0019] In the embodiment of the application, the equipment in the power distribution network line refers to various power equipment that constitutes or is attached to the power distribution network line, which is used to ensure the normal operation of the line and realize the function of power transmission and distribution. These equipment is the main carrier of the fault of the power distribution network line. The type of the equipment includes one of the following: overhead line, cable line, pole-mounted equipment, transformer. It should be noted that the unit of measurement of the overhead line and the cable line is length, for example, the length units such as kilometers (km) and meters (m) can be used to quantify the actual laying scale of the overhead line and the cable line in the power distribution network line, so as to facilitate subsequent calculation of the line fault index by reasonably allocating weights in combination with the length scale and the fault probability, and accurately reflecting the contribution degree of different types of equipment to the overall fault risk of the line.

[0020] Please refer to Figure 2 , Figure 2 is a system architecture diagram of a power distribution network line fault prediction system provided by the embodiment of the application, as shown in the figure, the power distribution network line fault prediction system includes: a meteorological station, a server, a monitoring platform. The meteorological station can collect real-time meteorological data of the power distribution network area and transmit the data to the server through data interaction. The server as the core processing carrier receives the meteorological data, calls the equipment fault prediction model respectively constructed for the overhead line, the cable line, the pole-mounted equipment and the transformer, outputs the equipment fault probability, and then calculates the line fault index and determines the risk level. The monitoring platform is the interactive interface of the user (power distribution network operation and maintenance personnel) and the system, which displays the line fault index, risk level and other information pushed by the server, and supports the user to interactively adjust the system parameters and model configuration. The power distribution network line fault prediction system ensures the real-time and accuracy of fault prediction, and provides data support for the operation and maintenance personnel to develop differentiated inspection strategies, and finally improves the practicality of power distribution network fault early warning and operation and maintenance efficiency.

[0021] In specific embodiments, the equipment management system of the power distribution network includes the equipment name, equipment type and other equipment information of all equipment on the line of the power distribution network. Therefore, the equipment in the power distribution network line can be obtained from the equipment management system of the power distribution network, and n equipment is obtained. Then, the meteorological data of the same period of the installation area of each of the above n equipment is called from the meteorological database, and n meteorological data is obtained. The meteorological data corresponding to each equipment includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, maximum wind speed, etc. without limitation. Wherein, n is a positive integer.

[0022] Since different types of equipment have different sensitivities to meteorological factors, for example, overhead lines are more sensitive to wind speed, and transformers are more sensitive to temperature, it is necessary to customize the corresponding fault prediction model for different types of equipment.

[0023] It can be seen that by effectively integrating power distribution network equipment information and regional meteorological data, multi-source and high-quality data are provided to ensure the accuracy of the power distribution network line fault prediction method.

[0024] S102, input the n meteorological data into the n device fault prediction models corresponding to the n devices respectively to obtain n device fault probabilities; each device fault prediction model is a pre-trained model for inputting meteorological data and outputting device fault probability; each device corresponds to a device fault prediction model.

[0025] In the embodiments of the present application, the device fault prediction model refers to a binary classification model that is constructed for different types of equipment in the power distribution network, pre-trained, and can realize input of meteorological data and output of device fault probability. The model can be an XGBoost model. The device fault prediction model can learn the device fault law in the coupling data of historical meteorological and historical faults to quantitatively predict the device fault probability under specific meteorological conditions.

[0026] In specific embodiments, first, the type of each device in the n devices is determined, and then the pre-trained device fault prediction model of the device is matched according to the device type, that is, it is ensured that each device corresponds to a prediction model that adapts to its own fault law. Then, the meteorological data corresponding to each device is input into the corresponding device fault prediction model, and the device fault probability of the device under the current meteorological condition can be obtained, and finally n device fault probabilities corresponding to the n devices are obtained.

[0027] Please refer to Figure 3 , Figure 3 is a model application architecture diagram of a power distribution network line fault prediction method provided by the embodiments of the present application, as shown in the figure, the power distribution network line is composed of device 1, device 2, …, device n, and these devices belong to four types of overhead lines, cable lines, pole-mounted devices, and transformers. For different types of equipment, corresponding fault prediction model sets are constructed, including device fault prediction models corresponding to overhead lines, cable lines, pole-mounted devices, and transformers. When the device data in the power distribution network line is input into the corresponding fault prediction model, the fault probability of each of device 1, device 2, …, device n can be output, thereby realizing accurate prediction of the fault probability of different types of equipment in the power distribution network line.

[0028] It can be seen that the device fault prediction model is matched based on the device type difference, and the accurate and efficient prediction of the device fault probability is realized.

[0029] Optionally, before the n device fault probabilities are obtained by inputting the n meteorological data into the n device fault prediction models corresponding to the n devices respectively, the method further includes the following steps: A201, obtaining historical fault record data and corresponding historical meteorological data of a target device; the target device is any one of the n devices; A202, processing the historical fault record data and the historical meteorological data to obtain label data and training sample data; A203, training a preset model according to the label data and the training sample data to obtain a device fault prediction model corresponding to the target device.

[0030] In the embodiment of the application, the target device refers to any one of the n devices, and its type belongs to one of overhead lines, cable lines, pole-mounted devices, and transformers. The embodiment of the application is based on the target device for interpretation.

[0031] The preset model refers to a model of a pre-set basic framework. The model can be a binary classification XGBoost model, which includes an optimizable hyperparameter, such as tree depth max_depth, learning rate learning_rate, sample sampling ratio subsample, and has the ability of regularization, missing value processing, and parallel pruning.

[0032] In specific embodiments, the historical fault record data of the target device is first extracted from the device management system of the power distribution network. The data includes the fault occurrence time of the target device, the fault device type and other key information. At the same time, the historical meteorological data of the region where the target device is located is obtained from the meteorological database at the same period as the historical fault record data.

[0033] Then, the obtained historical fault record data and historical meteorological data are processed to generate label data and training sample data. The label data refers to data used to identify whether the historical operation state of the target device has failed, and is the basis for judging the accuracy of the prediction result in the model training process. It is usually presented in binary form, that is, the corresponding label is 1 when the device fails, and the corresponding label is 0 when the device does not fail. The training sample data includes meteorological feature vectors and other data.

[0034] The generated training sample data is taken as input, and the label data is taken as a true result reference to train the preset model. In the training process, a grid search or a Bayesian optimization method is used to determine model hyperparameters. The generalization ability of the model is evaluated through K-fold cross-validation. The precision, recall, and F1 score are used to verify the performance of the model. Finally, a device fault prediction model that adapts to the fault rules of the target device is obtained.

[0035] Optionally, the step of processing the historical fault record data and the historical meteorological data to obtain label data and training sample data specifically includes the following steps: B201, performing a first processing operation on the historical fault record data and the historical meteorological data respectively to obtain corrected fault record data and corrected meteorological data. The first processing operation includes: deduplication, outlier correction, and missing value marking. B202, aligning the corrected fault record data and the corrected meteorological data based on a date scale to obtain first sample data. B203, expanding the first sample data through a preset oversampling method to obtain second sample data. B204, constructing the label data and the training sample data according to the second sample data.

[0036] In the embodiments of the application, the first processing operation refers to a set of basic data cleaning operations performed on the historical fault record data and the historical meteorological data, specifically including: deduplication, outlier correction, and missing value marking. Among them, the missing value marking specifically marks the missing items in the historical meteorological data, so that the XGBoost model automatically learns the default split direction based on its missing value processing mechanism during training.

[0037] The preset oversampling method refers to a technical means for expanding the number of fault class samples in advance. For example, the oversampling method can be a synthetic minority over-sampling technique (Synthetic Minority Over-sampling Technique, SMOTE), which can alleviate the imbalance between fault samples and non-fault samples in the data set by synthesizing new fault class samples.

[0038] In specific embodiments, the first processing operation is first performed on the historical fault record data to delete duplicate fault records and invalid data entered incorrectly, and to correct abnormal information in the record that does not conform to the actual logic. At the same time, the first processing operation is performed on the historical meteorological data to correct abnormal meteorological indicators that are outside the reasonable range, such as abnormally high rainfall and wind speed values, and to mark the missing indicator items in the meteorological data as a specific identifier (such as NaN). Finally, the corrected fault record data and the corrected meteorological data are obtained.

[0039] Then, with date as the unified time scale, the date corresponding to each fault record in the corrected fault record data is associated and matched with the meteorological indicators of the same date in the corrected meteorological data, ensuring that the fault state of the same equipment on the same date corresponds to the meteorological conditions of the same date, forming first sample data containing date, equipment fault state, and contemporaneous meteorological indicators.

[0040] The first sample data is processed by using a preset oversampling method. For the fault class samples with a small number in the data set, new samples similar to the existing fault class sample features are synthesized to supplement the number of fault class samples, so that the number of fault class samples and non-fault class samples tends to be balanced, and second sample data is obtained.

[0041] Finally, the meteorological data is extracted from the second sample data and integrated into a structured meteorological feature vector as training sample data, and the equipment fault state information is extracted from the second sample data to generate corresponding label data in a binary form, and the label data and the training sample data are constructed.

[0042] It can be seen that by preprocessing the data, high-quality, class-balanced, and format-adapted label data and training sample data can be obtained, effectively improving the model training effect and the accuracy of subsequent fault prediction.

[0043] Please refer to Figure 4 , Figure 4 is a flowchart of training an equipment fault prediction model provided by an embodiment of the present application. Optionally, the preset model is a binary classification XGBoost model. The above step of training the preset model according to the label data and the training sample data to obtain an equipment fault prediction model corresponding to the target equipment specifically includes the following steps: C201, determining a training set and a test set according to the label data and the training sample data; C202, determining a target function; the target function includes a differentiable loss function and a regularization term; C203, constructing a plurality of decision trees based on the preset model according to the target function and the training set; C204, determining a logic function; C205, determining an initial equipment fault prediction model according to the logic function and the plurality of decision trees; C206, evaluating the initial equipment fault prediction model according to the test set to obtain a target evaluation result; C207, if the target evaluation result meets a preset evaluation condition, determining that the initial equipment fault prediction model is the equipment fault prediction model corresponding to the target equipment.

[0044] In this embodiment of the application, the preset model is a binary classification XGBoost model.

[0045] In this embodiment, the objective function includes a differentiable loss function and a regularization term. The differentiable loss function measures the error between the model's predicted values ​​and the true values ​​of the labeled data. Due to its differentiability, it supports optimization of model parameters through methods such as gradient descent. In this embodiment, a logarithmic loss function can be used to adapt to binary classification tasks. The regularization term controls the preset model complexity, constraining the depth of the decision tree, the number of leaf nodes, etc., to prevent the model from overfitting the training set data and causing a decrease in generalization ability.

[0046] Preset evaluation conditions refer to the pre-defined standards for judging whether the model performance meets the requirements. They are usually presented in the form of thresholds for indicators such as precision, recall, and F1 score. The model only meets the usage requirements when the target evaluation result exceeds the corresponding threshold.

[0047] In a specific embodiment, the label data and training sample data are divided into a training set and a test set according to a preset ratio. This preset ratio can be 8:2, that is, the training set and the test set are divided at a ratio of 80% and 20%.

[0048] Next, construct the objective function, which can be as follows:

[0049] In the above formula, This represents the set of all optimizable parameters of the preset model, such as structural parameters like split nodes and leaf node weights of each decision tree, loss function parameters, regularization parameters, and model hyperparameters. The model is optimized through... , so that the objective function Minimize the value; This represents the total number of samples in the training set; Indicates the first The true label (fault status) of each sample; The model represents the first The predicted output for each sample is the probability of equipment failure. This represents a differentiable loss function used to calculate the true label of a single sample. With predicted probability The error between; This indicates the total number of decision trees integrated in the model; Indicates the first A decision tree; Represents a single decision tree Complexity penalty term.

[0050] Next, based on the pre-defined binary classification XGBoost model framework, multiple decision trees are constructed iteratively with the training set as input and the objective function as the optimization objective. In each iteration, the residual of the new decision tree is fitted to the current model, and the second-order Taylor expansion is used to approximate the increment of the objective function to select the optimal split node, ultimately forming multiple collaborative decision trees.

[0051] Determine the logistic function, which uses the Sigmoid function to map the predictions from multiple decision trees to failure probability values ​​between 0 and 1. The logistic function can be as follows:

[0052] In the above formula, This represents the device failure probability output by the model, and its value ranges from [0,1]. The model represents the input The original prediction output, whose value is Decision tree prediction output The sum of .

[0053] Among them, the model is for the input The original prediction output It can be as follows:

[0054] In the above formula, The model represents the input The original prediction output, This represents the input feature vector of the model, i.e., the meteorological feature vector; This indicates the total number of decision trees integrated in the model; Indicates the first Decision trees for input The single-tree prediction output, Essentially, it involves inputting meteorological feature vectors. The final output is the weight of the leaf node to which the decision tree ultimately belongs in the k-th decision tree, i.e., when Enter the first After constructing the decision tree, it will be progressively divided into leaf nodes according to the tree's splitting rules (based on meteorological feature thresholds). The weight of that leaf node is... The value of reflects the first decision trees The predictive contribution of the corresponding equipment failure status.

[0055] An initial equipment fault prediction model can be determined based on a logic function and multiple decision trees. Specifically, the integrated output of multiple decision trees... The input logic function, through the mapping effect of the Sigmoid function, will... compressing to the [0, 1] interval to obtain a device failure probability The probability value directly corresponds to the quantitative prediction result of whether the device fails under the input meteorological condition. Through the series connection of the multi-decision tree ensemble calculation and the logical function probability mapping, a model structure capable of receiving the meteorological feature vector input and outputting the device failure probability, i.e., the initial device failure prediction model, is formed.

[0056] According to the test set, the initial device failure prediction model is evaluated to obtain a target evaluation result. The target evaluation result is compared with a preset evaluation condition. If the target evaluation result meets the preset evaluation condition, it is determined that the initial device failure prediction model is a device failure prediction model adapted to the target device.

[0057] As can be seen, through the above steps, a device failure prediction model adapted to the target device and having high prediction accuracy and strong generalization ability can be constructed.

[0058] See Figure 5 , Figure 5 is a flowchart of a decision tree construction provided by an embodiment of the present application. Optionally, the above step of constructing a plurality of decision trees based on the preset model according to the target function and the training set specifically includes the following steps: D201, initializing an initial prediction value of the preset model; D202, determining a residual error of each sample in the training set and the initial prediction value to obtain a plurality of residual errors; D203, generating a plurality of candidate split nodes according to the plurality of residual errors; D204, determining a plurality of first target function values and a plurality of second target function values according to the target function and the plurality of candidate split nodes; the first target function value is the target function value before the candidate split node is split; and the second target function value is the target function value after the candidate split node is split; D205, determining a plurality of gains according to the plurality of first target function values and the plurality of second target function values; D206, determining a maximum gain in the plurality of gains and a target candidate split node corresponding to the maximum gain; D207, constructing a node of a target decision tree according to the target candidate split node, and determining the weight of each leaf node after the node is split; D208, if the target decision tree does not reach a preset complexity, determining a target prediction value according to the weight of each leaf node of the target decision tree, and updating the initial prediction value according to the target prediction value; repeating the step of determining a residual error of each sample in the training set and the initial prediction value to obtain a plurality of residual errors; D209、if the target decision tree reaches the preset complexity, re-performing the step of initializing the initial prediction value of the preset model, and iteratively generating the plurality of decision trees.

[0059] In the embodiments of the present application, the initial prediction value refers to the initial estimation value of the preset model for all sample fault states in the training set at the initial stage of training, which is usually set based on the statistical characteristics of the training set label data, for example, the initial prediction value can be the mean value of all sample labels.

[0060] The preset complexity refers to a standard preset for judging whether the decision tree stops building, which is usually in the form of maximum depth of the decision tree, maximum number of samples of the leaf node, or minimum gain threshold, etc., for controlling the complexity of the decision tree to avoid overfitting.

[0061] In specific embodiments, the initial prediction value of the preset model is first initialized. Then the residual of each sample in the training set and the initial prediction value is determined to obtain a plurality of residuals, wherein the residual reflects the deviation of the current prediction result of the model from the true situation. A plurality of candidate split nodes are generated according to the plurality of residuals, i.e. all weather features in the training set and their possible split thresholds are traversed to generate a plurality of candidate split nodes, each node corresponding to a combination of a feature and a threshold, for dividing samples.

[0062] Then, a plurality of first objective function values and a plurality of second objective function values can be determined according to the objective function and the plurality of candidate split nodes, wherein the first objective function value is the objective function value before the candidate split node is split, i.e. the objective function value calculated based on all samples contained in the node when a certain candidate split node does not split the samples. The second objective function value is the objective function value after the candidate split node is split, i.e. after a certain candidate split node is split into left and right child nodes according to the corresponding threshold, the objective function values of the samples contained in the two child nodes are calculated respectively and summed.

[0063] A plurality of gains can be determined according to the plurality of first objective function values and the plurality of second objective function values, the gain representing the difference between the objective function values before and after a certain candidate split node is split, i.e. the first objective function value minus the second objective function value, the greater the gain, the more significant the contribution of the split node to reducing model error and improving prediction accuracy.

[0064] The maximum gain in the plurality of gains and the target candidate split node corresponding to the maximum gain are determined, which is the optimal sample division node in the current decision tree construction. The node of the target decision tree is constructed according to the target candidate split node, and the weights of the leaf nodes after the node is split are determined. The leaf node weight refers to the weight value determined by solving the optimization problem of minimizing the objective function based on the residual of the sample contained in the leaf node and the objective function after the decision tree is split to the leaf node, which directly affects the contribution of the leaf node to the prediction result of the model.

[0065] If the target decision tree does not reach the preset complexity, a target prediction value is determined according to the weight of each leaf node of the target decision tree, and the initial prediction value is updated according to the target prediction value, that is, the leaf node weight is superimposed on the original prediction value, and the updated prediction value is the new initial prediction value. The step of determining the residual error of each sample in the training set and the initial prediction value is repeatedly executed to obtain a plurality of residual errors. If the target decision tree reaches the preset complexity, the construction of the decision tree is stopped, the decision tree is taken as a base learner of the XGBoost model, and the step of initializing the initial prediction value of the preset model is re-executed to construct the next decision tree, so as to iteratively generate a plurality of decision trees.

[0066] Optionally, the step of evaluating the initial equipment fault prediction model according to the test set to obtain a target evaluation result comprises the following steps: E201, inputting the test set into the initial equipment fault prediction model to obtain a predicted fault probability; E202, calculating a target evaluation index according to the predicted fault probability; the target evaluation index comprises: precision, recall, and F1 score; E203, determining the target evaluation result according to the target evaluation index.

[0067] In the embodiments of the present application, the target evaluation index comprises: precision, recall, and F1 score.

[0068] In specific embodiments, the test set divided in advance is input into the initial equipment fault prediction model. For each test sample, the model first calculates the integrated output through the internal multiple decision trees, and then maps the integrated output through a logical function to obtain the predicted fault probability corresponding to each test sample.

[0069] Based on the predicted fault probability of the test set and the real fault label, the precision, recall, and F1 score are calculated, specifically, a preset probability threshold (such as 0.5) can be set, samples with a predicted fault probability greater than or equal to the threshold are determined as fault samples predicted by the model, and samples with a predicted fault probability less than the threshold are determined as non-fault samples predicted by the model, then by constructing a confusion matrix (statistical true positive, false positive, true negative, and false negative sample number), according to the calculation formulas of precision (true positive / (true positive+false positive)), recall (true positive / (true positive+false negative)), and F1 score (2*precision*recall / (precision+recall)), the specific values of the three target evaluation indexes are obtained.

[0070] Then, the precision, recall and F1 score calculated are integrated to form a target evaluation result that can comprehensively reflect the performance of the initial equipment fault prediction model, and the result needs to include specific values of the three indicators and the balance relationship among the three.

[0071] It can be seen that the performance of the initial equipment fault prediction model is quantified objectively and comprehensively to obtain an equipment fault prediction model that meets the fault prediction requirements of the power distribution network, thereby improving the accuracy of subsequent power distribution network line fault prediction and risk classification.

[0072] S103, determining a line fault index of the power distribution network line according to the n equipment fault probabilities.

[0073] In the embodiments of the present application, the line fault index refers to a comprehensive quantitative index calculated based on the fault probabilities of all the equipment included in the power distribution network line and the influence weight of the equipment on the line power supply reliability, and the numerical value directly reflects the overall fault risk level of the line.

[0074] In specific embodiments, the line fault index of the power distribution network line can be determined according to the n equipment fault probabilities. The fault risk of the power distribution network line is not determined by a single equipment, but is jointly affected by the fault probabilities of all the equipment included in the line. Therefore, the fault probabilities of the equipment in the line can be converted into the line fault index reflecting the overall risk of the line.

[0075] Optionally, the step of determining the line fault index of the power distribution network line according to the n equipment fault probabilities specifically includes the following steps: A301, determining a weight corresponding to each of the n equipment to obtain n weights; each weight is determined by the equipment type corresponding to the equipment and the influence degree on the line; A302, determining the line fault index according to the n equipment fault probabilities and the n weights.

[0076] In the embodiments of the present application, the weight corresponding to the equipment refers to a numerical value given to each equipment for quantifying the contribution proportion of the equipment to the line fault risk according to the equipment type of each equipment in the power distribution network line and the influence degree of the equipment on the line operation. The numerical value is not a fixed constant and needs to be dynamically adjusted in combination with the core degree of the equipment. The setting of the weight focuses more on the absolute influence degree of the equipment fault, for example, the weight of a longer overhead line is higher than that of a shorter overhead line.

[0077] In specific embodiments, the weight corresponding to each of the n devices is determined first, and n weights can be obtained. Specifically, according to the device type of the n devices and the influence degree indicators such as the line outage duration caused by device failure in the historical failure data, a corresponding weight is assigned to each device. Among them, the greater the influence on the line power supply reliability of the device (such as a transformer), the higher the weight value; the smaller the influence of the device (such as some auxiliary overhead devices), the lower the weight value.

[0078] Then, the line failure index can be determined according to the n device failure probabilities and the n weights, and the line failure index calculation formula can be as follows:

[0079] In the above formula, represents the line failure index; represents the device failure probability of the first device; represents the weight of the first device; represents the device failure probability of the nth device; represents the weight of the nth device.

[0080] Specifically, the n device failure probabilities and the n weights are multiplied one by one to obtain the weighted contribution value of each device to the line failure risk, and then the weighted contribution values of all devices are summed to obtain the line failure index of the distribution network line.

[0081] As can be seen, by determining the device weight and accurately calculating the line failure index, the quantification from the device level failure probability to the line level risk is realized, which provides accurate data support for subsequent line risk classification and differentiated operation strategy formulation.

[0082] S104, determining the target risk level of the line failure according to the line failure index.

[0083] In the embodiments of the present application, the target risk level refers to setting a classification threshold based on the line failure index of the distribution network line, combining the overall distribution characteristics of all distribution network line failure indexes, and classifying the specific level of the current line failure risk, which is used to intuitively distinguish the failure risk differences of different lines and provide clear basis for formulating differentiated inspection and maintenance strategies for operation and maintenance personnel. In the embodiments of the present application, the level is specifically divided into five levels, including a complete risk interval of low risk, lower risk, medium risk, higher risk, and extremely high risk.

[0084] In specific embodiments, the target risk level of the line failure can be determined according to the line failure index, that is, the line failure index is compared with the threshold corresponding to different risk intervals to determine the risk interval and the corresponding risk level of the line failure index.

[0085] The monitoring platform or the mobile terminal can show the fault probability and risk level of each line to the operation and maintenance personnel, and provide a high-risk line list and key weather indicators to assist in formulating inspection and maintenance plans. Based on the target risk level, the maintenance personnel can perform differentiated strategies according to the risk level, for example, high-risk lines can be arranged for key inspection, medium-risk lines can be strengthened for real-time monitoring, and low-risk lines can be normally operated.

[0086] Optionally, new fault records and new weather data can also be collected periodically to retrain the model or update the parameters, so as to realize self-iteration of the model and continuously improve the prediction accuracy.

[0087] In summary, by implementing the embodiments of the present application, the weather data corresponding to the devices in the power distribution network line and the areas where the devices are located are obtained, and n devices and n weather data are obtained. The types of devices include one of the following: overhead lines, cable lines, pole-mounted devices, and transformers. Each weather data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, and maximum wind speed. The n weather data are input into n device fault prediction models corresponding to the n devices, respectively, to obtain n device fault probabilities. Each device fault prediction model is a pre-trained model for inputting weather data and outputting device fault probability. Each device corresponds to a device fault prediction model. The line fault index of the power distribution network line is determined according to the n device fault probabilities. The target risk level of the line fault is determined according to the line fault index. As can be seen, by training corresponding fault prediction models for various devices in the power distribution network line, the fault index of the line is predicted according to the fault probability of the device on the power distribution network line, thereby realizing accurate prediction of the power distribution network line fault under multi-source data.

[0088] Please refer to Figure 6 , Figure 6 is a structural schematic diagram of a power distribution network line fault prediction device provided by the embodiments of the present application. The power distribution network line fault prediction device 600 is applied to a server, and the power distribution network line fault prediction device 600 includes: The data acquisition module 601 is configured to obtain weather data corresponding to devices in a power distribution network line and areas where the devices are located, and obtain n devices and n weather data. The types of devices include one of the following: overhead lines, cable lines, pole-mounted devices, and transformers. Each weather data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, and maximum wind speed. n is a positive integer. The device failure probability prediction module 602 is configured to input the n pieces of meteorological data into n device failure prediction models corresponding to the n devices respectively to obtain n device failure probabilities; each device failure prediction model is a pre-trained model configured to input meteorological data and output a device failure probability; and each device corresponds to a device failure prediction model. The line fault index calculation module 603 is configured to determine a line fault index of the power distribution network line according to the n device failure probabilities. The risk level determination module 604 is configured to determine a target risk level of a line fault according to the line fault index.

[0089] Optionally, before the n pieces of meteorological data are input into the n device failure prediction models corresponding to the n devices respectively to obtain the n device failure probabilities, the power distribution network line fault prediction apparatus 600 is further configured to: obtain historical failure record data and corresponding historical meteorological data of a target device; the target device is any one of the n devices; process the historical failure record data and the historical meteorological data to obtain label data and training sample data; train a preset model according to the label data and the training sample data to obtain a device failure prediction model corresponding to the target device.

[0090] Optionally, in the processing of the historical failure record data and the historical meteorological data to obtain the label data and the training sample data, the power distribution network line fault prediction apparatus 600 is further configured to: perform a first processing operation on the historical failure record data and the historical meteorological data respectively to obtain corrected failure record data and corrected meteorological data; the first processing operation includes: deduplication, outlier correction, and missing value marking; align the corrected failure record data and the corrected meteorological data based on a date scale to obtain first sample data; expand the first sample data by a preset oversampling method to obtain second sample data; construct the label data and the training sample data according to the second sample data.

[0091] Optionally, the preset model is a binary classification XGBoost model; and in the training of the preset model according to the label data and the training sample data to obtain the device failure prediction model corresponding to the target device, the power distribution network line fault prediction apparatus 600 is further configured to: determine a training set and a test set according to the label data and the training sample data. determining a target function; the target function comprising a differentiable loss function and a regularization term; constructing a plurality of decision trees based on the preset model according to the target function and the training set; determining a logical function; determining an initial device fault prediction model according to the logical function and the plurality of decision trees; evaluating the initial device fault prediction model according to the test set to obtain a target evaluation result; if the target evaluation result meets a preset evaluation condition, determining that the initial device fault prediction model is a device fault prediction model corresponding to the target device.

[0092] Optionally, in the aspect of constructing a plurality of decision trees based on the preset model according to the target function and the training set, the power distribution network line fault prediction device 600 is further specifically configured to: initializing an initial prediction value of the preset model; determining a residual error of each sample in the training set and the initial prediction value to obtain a plurality of residual errors; generating a plurality of candidate split nodes according to the plurality of residual errors; determining a plurality of first target function values and a plurality of second target function values according to the target function and the plurality of candidate split nodes; the first target function value is the target function value before the candidate split node is split; the second target function value is the target function value after the candidate split node is split; determining a plurality of gains according to the plurality of first target function values and the plurality of second target function values; determining a maximum gain in the plurality of gains and a target candidate split node corresponding to the maximum gain; constructing a node of a target decision tree according to the target candidate split node, and determining the weight of each leaf node after the node is split; if the target decision tree does not reach a preset complexity, determining a target prediction value according to the weight of each leaf node of the target decision tree, and updating the initial prediction value according to the target prediction value; repeating the step of determining a residual error of each sample in the training set and the initial prediction value to obtain a plurality of residual errors; if the target decision tree reaches the preset complexity, re-executing the step of initializing the initial prediction value of the preset model, and iteratively generating the plurality of decision trees.

[0093] Optionally, in the aspect of evaluating the initial device fault prediction model according to the test set to obtain a target evaluation result, the power distribution network line fault prediction device 600 is further specifically configured to: inputting the test set into the initial device failure prediction model to obtain a predicted failure probability; calculating a target evaluation index according to the predicted failure probability; the target evaluation index comprises: precision, recall, and F1 score; determining the target evaluation result according to the target evaluation index.

[0094] Optionally, in the aspect of determining the line failure index of the power distribution network line according to the n device failure probabilities, the risk level determination module 604 is further specifically configured to: determining a weight corresponding to each of the n devices to obtain n weights; each weight is determined by a device type corresponding to the device and an influence degree on the line; determining the line failure index according to the n device failure probabilities and the n weights.

[0095] The power distribution network line failure prediction apparatus 600 described in the present application can obtain meteorological data corresponding to devices in a power distribution network line and regions where the devices are located to obtain n devices and n meteorological data; the types of the devices include one of the following: overhead line, cable line, pole-mounted device, and transformer; each meteorological data includes the following: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, and maximum wind speed; the n meteorological data are respectively input into n device failure prediction models corresponding to the n devices to obtain n device failure probabilities; each device failure prediction model is a pre-trained model for inputting meteorological data and outputting device failure probability; each device corresponds to a device failure prediction model; the line failure index of the power distribution network line is determined according to the n device failure probabilities; and the target risk level of line failure is determined according to the line failure index. As can be seen, the corresponding failure prediction models are trained by various devices in the power distribution network line, so as to predict the failure index of the line according to the failure probability of the devices on the power distribution network line, thereby realizing accurate prediction of the failure of the power distribution network line under multi-source data.

[0096] Please refer to Figure 7 , Figure 7 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. The electronic device can include a processor, a memory, a communication interface, and one or more programs. The processor, the memory, and the communication interface can be connected to each other through a bus. The one or more programs are stored in the memory and are configured to be executed by the processor. In the present application, the program includes instructions for performing the following steps: Obtain meteorological data corresponding to devices in a power distribution network line and areas where the devices are located, to obtain n devices and n meteorological data; the types of devices include one of the following: overhead line, cable line, pole-mounted device, transformer; each meteorological data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, maximum wind speed; n is a positive integer; Input the n meteorological data into n device fault prediction models corresponding to the n devices respectively, to obtain n device fault probabilities; each device fault prediction model is a pre-trained model for inputting meteorological data and outputting device fault probability; each device corresponds to a device fault prediction model; Determine a line fault index of the power distribution network line according to the n device fault probabilities; Determine a target risk level of line fault according to the line fault index.

[0097] The electronic device described in the present application can obtain meteorological data corresponding to devices in a power distribution network line and areas where the devices are located, to obtain n devices and n meteorological data; the types of devices include one of the following: overhead line, cable line, pole-mounted device, transformer; each meteorological data includes: daily cumulative rainfall, sliding rainfall, maximum temperature, minimum temperature, average temperature, relative humidity, average wind speed, maximum wind speed; input the n meteorological data into n device fault prediction models corresponding to the n devices respectively, to obtain n device fault probabilities; each device fault prediction model is a pre-trained model for inputting meteorological data and outputting device fault probability; each device corresponds to a device fault prediction model; determine a line fault index of the power distribution network line according to the n device fault probabilities; determine a target risk level of line fault according to the line fault index. It can be seen that by training corresponding fault prediction models for various devices in the power distribution network line, the fault index of the line can be predicted according to the fault probability of the devices on the power distribution network line, thereby realizing accurate prediction of the fault of the power distribution network line under multi-source data.

[0098] The embodiment of the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to execute part or all steps of any method described in the above method embodiment, and the computer includes the electronic device.

[0099] The embodiment of the present application further provides a computer program product, and the computer program product includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all steps of any method described in the above method embodiment. The computer program product can be a software installation package, and the computer includes the electronic device.

[0100] Those skilled in the art can understand that all or part of the processes in the above-mentioned method embodiments can be implemented by a computer program instructing relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, the processes of the above-mentioned method embodiments can be included. The storage medium includes a ROM or a random access memory (RAM), a magnetic disk or an optical disk, and various storage media that can store program codes.

[0101] The steps of the methods or algorithms described in the embodiments of the present application can be implemented in the form of hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a RAM, a flash memory, a ROM, an electrically EPROM (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in a terminal device or a management device. Of course, the processor and the storage medium can also exist as discrete components in the terminal device or the management device.

[0102] Those skilled in the art should be aware that, in one or more examples described above, the functions described in the embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, the computer program instructions generate, in whole or in part, the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital video disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0103] The various modules / units included in the various devices and products described in the above embodiments can be software modules / units or hardware modules / units, or partially software modules / units and partially hardware modules / units. For example, for the various devices and products applied to or integrated into a chip, the various modules / units included therein can all be implemented in the form of hardware such as circuitry, or at least some of the modules / units can be implemented in the form of software programs running on a processor integrated in the chip, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuitry; for the various devices and products applied to or integrated into a chip module, the various modules / units included therein can all be implemented in the form of hardware such as circuitry, and different modules / units can be located in the same component (e.g., a chip, a circuit module, etc.) or different components of the chip module, or at least some of the modules / units can be implemented in the form of software programs running on a processor integrated in the chip module, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuitry; for the various devices and products applied to or integrated into a terminal device, the various modules / units included therein can all be implemented in the form of hardware such as circuitry, and different modules / units can be located in the same component (e.g., a chip, a circuit module, etc.) or different components of the terminal device, or at least some of the modules / units can be implemented in the form of software programs running on a processor integrated in the terminal device, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuitry.

[0104] The above detailed description of the specific embodiments of the present application has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the application to the precise form described, and many modifications, equivalents and alternatives shown thereby will be apparent to those skilled in the art.

Claims

1. A method for predicting faults in power distribution network lines, characterized in that, Applied to a server, the method includes: Obtain meteorological data corresponding to the equipment in the power distribution network and the area where the equipment is located, resulting in n equipment and n meteorological data. The equipment types include one of the following: overhead lines, cable lines, pole-mounted equipment, and transformers. Each meteorological data includes: daily cumulative rainfall, sliding rainfall, highest temperature, lowest temperature, average temperature, relative humidity, average wind speed, and maximum wind speed; n is a positive integer. The n meteorological data are respectively input into the n equipment fault prediction models corresponding to the n equipment to obtain the n equipment fault probabilities; each equipment fault prediction model is a pre-trained model that takes meteorological data as input and outputs the equipment fault probability; each equipment corresponds to one equipment fault prediction model; The line fault index of the distribution network line is determined based on the failure probabilities of the n devices. The target risk level of the line fault is determined based on the line fault index.

2. The method as described in claim 1, characterized in that, Before inputting the n meteorological data into the n equipment failure prediction models corresponding to the n equipment to obtain the failure probabilities of the n equipment, the method further includes: Acquire historical fault record data and corresponding historical meteorological data of the target device; the target device is any one of the n devices. The historical fault record data and the historical meteorological data are processed to obtain label data and training sample data; The preset model is trained based on the label data and the training sample data to obtain a device fault prediction model corresponding to the target device.

3. The method as described in claim 2, characterized in that, The process of processing the historical fault record data and the historical meteorological data to obtain label data and training sample data includes: The historical fault record data and the historical meteorological data are respectively subjected to a first processing operation to obtain corrected fault record data and corrected meteorological data; the first processing operation includes: deduplication, outlier correction, and missing value marking. The corrected fault record data and the corrected meteorological data are aligned based on the date scale to obtain the first sample data; The first sample data is augmented using a preset oversampling method to obtain the second sample data; The label data and the training sample data are constructed based on the second sample data.

4. The method as described in claim 3, characterized in that, The preset model is a binary classification XGBoost model; the step of training the preset model based on the label data and the training sample data to obtain the equipment fault prediction model corresponding to the target device includes: The training set and the test set are determined based on the label data and the training sample data; Determine the objective function; the objective function includes a differentiable loss function and a regularization term; Based on the objective function and the training set, construct multiple decision trees according to the preset model; Determine the logical function; The initial equipment fault prediction model is determined based on the logical function and the multiple decision trees; The initial equipment fault prediction model is evaluated based on the test set to obtain the target evaluation result; If the target evaluation result meets the preset evaluation conditions, then the initial equipment fault prediction model is determined to be the equipment fault prediction model corresponding to the target equipment.

5. The method as described in claim 4, characterized in that, The step of constructing multiple decision trees based on the objective function and the training set using the preset model includes: Initialize the initial prediction values ​​of the preset model; Determine the residual between each sample in the training set and the initial predicted value to obtain multiple residuals; Multiple candidate split nodes are generated based on the multiple residuals; Based on the objective function and the plurality of candidate splitting nodes, a plurality of first objective function values ​​and a plurality of second objective function values ​​are determined; the first objective function values ​​are the objective function values ​​of the candidate splitting nodes before splitting; the second objective function values ​​are the objective function values ​​of the candidate splitting nodes after splitting. Multiple gains are determined based on the multiple first objective function values ​​and the multiple second objective function values; Determine the maximum gain among the plurality of gains and the target candidate split node corresponding to the maximum gain; Based on the target candidate splitting node, construct the nodes of the target decision tree and determine the weights of each leaf node after the node splits. If the target decision tree does not reach the preset complexity, then the target prediction value is determined according to the weight of each leaf node of the target decision tree, and the initial prediction value is updated according to the target prediction value; the step of determining the residual between each sample in the training set and the initial prediction value is repeated to obtain multiple residuals. If the target decision tree reaches the preset complexity, then the step of initializing the initial prediction value of the preset model is executed again to iteratively generate the multiple decision trees.

6. The method as described in claim 5, characterized in that, The step of evaluating the initial equipment fault prediction model based on the test set to obtain the target evaluation result includes: The test set is input into the initial equipment fault prediction model to obtain the predicted fault probability; The target evaluation metrics are calculated based on the predicted failure probability; the target evaluation metrics include: precision, recall, and F1 score. The target evaluation result is determined based on the target evaluation indicators.

7. The method according to any one of claims 1-6, characterized in that, The step of determining the line fault index of the distribution network line based on the failure probabilities of the n devices includes: The weight corresponding to each of the n devices is determined, resulting in n weights; each weight is determined by the device type and the degree of impact on the line. The line fault index is determined based on the failure probabilities of the n devices and the n weights.

8. A fault prediction device for power distribution lines, characterized in that, The power distribution network line fault prediction device, applied to servers, includes: The data acquisition module is used to acquire meteorological data corresponding to the equipment in the power distribution network and the area where the equipment is located, resulting in n equipment and n meteorological data. The types of equipment include one of the following: overhead lines, cable lines, pole-mounted equipment, and transformers. Each meteorological data includes: daily cumulative rainfall, sliding rainfall, highest temperature, lowest temperature, average temperature, relative humidity, average wind speed, and maximum wind speed; n is a positive integer. The equipment failure probability prediction module is used to input the n meteorological data into the n equipment failure prediction models corresponding to the n equipment respectively, so as to obtain the n equipment failure probabilities; each equipment failure prediction model is a pre-trained model that takes meteorological data as input and outputs the equipment failure probability; each equipment corresponds to one equipment failure prediction model; The line fault index calculation module is used to determine the line fault index of the distribution network line based on the fault probabilities of the n devices. The risk level determination module is used to determine the target risk level of a line fault based on the line fault index.

9. An electronic device, characterized in that, include: Processor, memory, communication interface, and one or more programs; The one or more programs are stored in the memory and configured to be executed by the processor, the programs including instructions for performing the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1-7.

Images