A method and system for predicting power grid equipment failure under extreme weather conditions

Abstract

The present invention discloses a method and system for predicting power grid equipment failures under extreme weather conditions, including: collecting first object data and preprocessing to obtain a data set. According to the test set situation, further sampling is performed to obtain a balanced data set. A prediction fault model is established, and a prediction result is output. The method and system for predicting power grid equipment failures under extreme weather conditions provided by the present invention can well solve the problems of low probability, small number of samples, and nonlinearity of extreme disasters in the power grid. For the multi-temporal, multi-dimensional, multi-scale, and uncertain characteristics of meteorological data, the prediction effect is very good. The GRNN neural network has a strong nonlinear mapping ability and learning speed. The method is simple and practical, the calculation is reliable, it is easy to implement and deploy on-site, and it has strong operability and promotion and application value.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault prediction, and in particular to a method and system for predicting power grid equipment faults under extreme weather conditions. Background Art

[0002] The global greenhouse effect and the continuous deterioration of the ecological environment have led to an increasing incidence of extreme natural disasters such as typhoons and rainstorms. The tripping failure of transmission lines caused by natural disasters is the main cause of power grid failures in my country. The failure and shutdown model of natural disasters can be expressed by the failure rate of a single device. The current power outage defense system only collects power information and does not consider external factors such as the weather environment. The safety and stability of the power grid are analyzed with a constant failure rate and failure set. Existing research considers external information to establish a failure rate model on the original basis. The current failure rate evaluation model is mainly divided into three categories, namely, mechanism model, historical data statistical model, and hybrid model.

[0003] At present, there are not many failure rate models based on disaster mechanisms, most of which are for disasters such as wildfires, heavy rains, and ice cover. Although the failure rate research based on the mechanism can clearly reflect the causal relationship between disasters and failures, it is difficult to ensure the accuracy of predictions due to the lack and roughness of external information, especially for natural disasters with complex mechanisms, and it is difficult to achieve in practice. Traditional statistical theories require large-capacity sample support and are not very applicable to the assessment of failure rates in extreme weather with low probability. As "big data" technology has become a hot topic in current research, machine learning algorithms such as neural networks and support vector machines have become more and more perfect. Power grid equipment failure warning is also a pattern recognition and classification problem, which is also applicable to data mining models, and artificial neural networks have very good performance in pattern recognition, optimization calculation, and nonlinear mapping.

[0004] During the operation of the power grid, some historical meteorological station monitoring information and historical power grid fault information have been accumulated. These historical information contain the correlation between external meteorological information and historical faults, but they have not been fully explored in power grid fault prediction. This scheme proposes a method for predicting the failure rate of power grid equipment under extreme weather conditions. According to the characteristics of extreme weather, it can mine the correlation model between historical disaster meteorological information and historical power grid line tower fault information, thereby guiding the prediction of power grid failure rate under real-time meteorological data conditions, and providing reliable forecast support and decision-making basis for the safe operation of the power grid. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is: the existing technology has the problems of low probability of extreme disasters, small samples and nonlinearity.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: A method for predicting power grid equipment failures under extreme weather conditions, comprising: collecting first object data and preprocessing it to obtain a data set.

[0008] According to the test set situation, further sampling is performed to obtain a balanced data set.

[0009] Establish a predictive fault model and output the prediction results.

[0010] As a preferred solution of the method for predicting power grid equipment failures under extreme weather conditions described in the present invention, the preprocessing includes performing correlation analysis on the first object data, obtaining input data at each first object by weighted calculation of the meteorological information closest to the first object, and matching it with historical data to form a sample.

[0011] As a preferred solution of the method for predicting power grid equipment failure under extreme weather conditions described in the present invention, the data set includes a test set and a failure set, both of which are composed of the samples.

[0012] As a preferred solution of the method for predicting power grid equipment failures under extreme weather conditions described in the present invention, further sampling to obtain a balanced data set based on the test set situation includes resampling to obtain a balanced data set in response to a situation in which non-fault samples in the data set are much larger than fault samples.

[0013] As a preferred solution of the method for predicting power grid equipment failure under extreme weather conditions described in the present invention, the resampling includes using an oversampling method for fault samples to generate artificial samples to expand the sampling of minority class samples.

[0014] For non-fault samples, an under-sampling algorithm is used to distinguish the distance between samples and divide the majority class samples into noise samples, boundary samples and safety samples, thereby reducing noise samples and boundary samples and streamlining safety samples.

[0015] As a preferred solution of the method for predicting power grid equipment failure under extreme weather conditions described in the present invention, the first object includes but is not limited to transmission towers, and the first object data includes but is not limited to the longitude and latitude information of transmission towers, transmission tower failure statistics data, longitude and latitude information of meteorological stations in the transmission tower area, and meteorological information in the past three years.

[0016] As a preferred solution of the method for predicting power grid equipment failures under extreme weather conditions described in the present invention, the establishment of a prediction fault model and outputting prediction results include establishing a neural network prediction model, including an input layer, a pattern layer, a summation layer and an output layer, calculating the optimal smoothing factor, optimizing the neural network prediction model, and outputting a fault prediction result.

[0017] A power grid equipment failure prediction system under extreme weather conditions, characterized in that:

[0018] The preprocessing module collects and preprocesses the first object data to obtain a data set.

[0019] The resampling module further samples the test set to obtain a balanced data set.

[0020] Prediction module, establishes a prediction fault model and outputs the prediction results.

[0021] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0022] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.

[0023] The beneficial effects of the invention are as follows: the problems of low probability, small sample, nonlinearity, etc. of extreme disasters in power grids are well solved. The prediction effect is very good for the characteristics of multi-time, multi-dimensional, multi-scale, and uncertain meteorological data. The GRNN neural network has strong nonlinear mapping ability and learning speed. The method is simple and practical, the calculation is reliable, easy to implement and deploy on the engineering site, and has strong operability and promotion and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0025] Figure 1 An overall flow chart of a method and system for predicting power grid equipment failures under extreme weather conditions provided in the first embodiment of the present invention.

[0026] Figure 2 A schematic diagram of the SMOTE principle of a method and system for predicting power grid equipment failure under extreme weather conditions provided in the first embodiment of the present invention.

[0027] Figure 3 A generalized regression neural network topology diagram of a method and system for predicting power grid equipment failures under extreme weather conditions provided in the first embodiment of the present invention.

[0028] Figure 4 This is a GRNN prediction result diagram of a method and system for predicting power grid equipment failure under extreme weather conditions provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0030] Example 1, reference Figure 1 to Figure 3 , as an embodiment of the present invention, provides a method for predicting power grid equipment failures under extreme weather conditions, comprising:

[0031] S1: Collect and preprocess the first object data to obtain a data set.

[0032] In the present invention, the first object is a transmission tower, and the first object data includes the longitude and latitude information of the transmission tower, transmission tower fault statistics, longitude and latitude information of the meteorological station in the transmission tower area, and meteorological information in the past three years, including wind speed, wind direction, and rainfall.

[0033] Furthermore, the first object may also be substation equipment, distribution lines and wind turbine generator sets.

[0034] In an optional embodiment of the present invention, the first object is a substation device, which is an important part of the power system and faces a higher risk of failure under extreme weather conditions. There are many types of substation equipment, including transformers, circuit breakers, busbars, disconnectors, etc. The stable operation of these devices is crucial to the reliability of the power system. In extreme weather conditions, such as heavy rain, thunderstorms, high temperatures or cold waves, substation equipment may face an increased risk of failure due to factors such as temperature changes, voltage fluctuations, and increased humidity.

[0035] In this case, the first object data includes equipment location and type information: including the geographical location of the substation, the type of equipment (such as main transformers, disconnectors, etc.), the time when the equipment was installed, etc. This information helps to analyze the specific weather adaptability of the equipment. Historical fault statistics: records historical failure events of various types of equipment in the substation, including failure time, failure cause and repair time. This data can be used to analyze the aging of equipment and potential failure trends. Environmental monitoring data: including meteorological information in the area where the substation is located (such as temperature, humidity, wind speed, precipitation, etc.), as well as environmental variables such as temperature and humidity inside the substation. This information can help identify environmental factors that may cause failures under specific extreme weather conditions. Electrical operating parameters: such as voltage, current, load rate, etc., these parameters fluctuate greatly under extreme weather conditions, can reflect the stress of substation equipment, and serve as an important input for fault prediction.

[0036] During data preprocessing, substation equipment data usually needs to be standardized, especially the time alignment of fault statistics and meteorological data of different equipment, so as to analyze the failure risk of different equipment at the same time point. At the same time, environmental data and operating parameters can be calculated through weighted association to generate a comprehensive failure impact factor for each device.

[0037] In an optional embodiment of the present invention, the first object is a distribution line, which is a key channel connecting the substation and the power consumption end. The distribution line is particularly prone to failure in extreme weather. For example, in strong winds or thunderstorms, problems such as disconnection and short circuit of the distribution line occur frequently and have a wide impact. Distribution lines are usually laid in an open outdoor environment, and the threat posed to them by extreme weather mainly comes from interference from the physical environment.

[0038] In this case, the first object data includes line latitude and longitude and path information: including detailed geographic coordinates, installation path and laying height of the distribution line. This information can be used to identify the degree of weather impact on the distribution line in a certain area. Line material and age: including line material type (such as aluminum core, copper core, steel core aluminum stranded wire, etc.), insulation type and its laying years. This information helps to analyze the tolerance of different lines under extreme weather. Historical fault records: recording the time and cause of line disconnection, short circuit, insulation damage and other faults in the past, which helps to identify patterns of potential faults under similar extreme weather conditions. Line load information: recording the load of the line in each period. The load may change dramatically in extreme weather, thereby increasing the risk of failure. The data preprocessing of the distribution line mainly includes matching the historical fault records with meteorological data, and using weighted distance calculation to extract the failure probability of each line section under extreme weather conditions. For example, meteorological data such as wind speed, humidity, and temperature can be used to weightedly predict the failure risk of distribution lines in severe weather.

[0039] In an optional embodiment of the present invention, the first object is a wind turbine generator set. In modern power grid systems, wind turbine generator sets are an important renewable energy device and also have a high risk of failure under extreme weather conditions. Wind turbine generator sets are generally operated at a high altitude and exposed outdoors, and are easily affected by meteorological factors such as strong winds, heavy rain, ice and snow. In addition, the wind turbine generator set itself has a complex structure (such as blades, transmission systems, and power generation systems, etc.), and these components have a high failure rate in bad weather.

[0040] In this case, the first object data includes the location and height of the generator set: recording the geographical coordinates, altitude and total height of the tower of the wind turbine. This information helps to analyze the possible weather impacts on the location, such as changes in wind speed and wind direction. Equipment parameters and operating status: including operating parameters such as power output, speed, vibration frequency, bearing temperature, etc. of the generator set. Abnormal changes in these parameters under severe weather conditions can serve as early warning signals of faults. Environmental meteorological information: recording parameters such as wind speed, wind direction, temperature, humidity and precipitation, which directly affect the operational stability of the wind turbine. For example, when the wind speed is too high or the wind direction changes suddenly, the blades and transmission system are susceptible to additional physical pressure, leading to failure. Historical maintenance and fault records: recording the maintenance cycle and fault history of the unit, and analyzing the loss of the unit under different weather conditions. In the data preprocessing process, the data of the wind turbine generator set needs to be processed in layers. First, the environmental meteorological information and equipment parameters are correlated and calculated, and the comprehensive influence coefficient of the wind turbine generator set under different weather conditions is obtained by the distance weighting method. Secondly, the historical fault records are statistically analyzed to identify high-risk meteorological factor combinations. Ultimately, this data is used as model input to help improve the accuracy of predictions.

[0041] In the present invention, when the first object is a transmission tower, the tripping of the transmission tower is closely related to the wind speed borne by the tower, the angle between the line direction and the wind direction. This solution divides the time of extreme weather into discrete time periods at intervals of 1 hour, and counts the meteorological information in each time period: the maximum wind speed of each tower X1, the maximum angle between the wind direction and the line direction X2, the two-minute average wind speed X3, the two-minute average angle between the wind direction and the line X4, and the rainfall X5.

[0042] Since the meteorological information obtained by the power grid comes from the meteorological station of the meteorological department, the meteorological information at the location of the power grid tower needs to be correlated and analyzed based on the meteorological station data, and the meteorological parameters at the tower are calculated using the weighted data of the three meteorological stations closest to the tower. Considering the error between the measurement data of the distant meteorological station and the meteorological parameters at the tower, the weight of the distant meteorological station data is lower. According to the distance attenuation law, when other conditions are the same, the effect between geographical elements is inversely proportional to the square of the distance. The calculation method of the weight of different meteorological station data is:

[0043]

[0044] Where k i (i=1,2,3) represents the data weight of different weather stations, d i (i=1,2,3) indicates the distance between weather station i and the tower.

[0045] The calculation method of wind speed data at the tower is:

[0046]

[0047] Where v i Represents wind speed data of weather station i.

[0048] The calculation method of wind direction data at the tower is:

[0049]

[0050] Where a i Represents the wind direction data of weather station i, a line It indicates the angle between the line where the tower is located and the earth's true north direction.

[0051] The above formula can be used to obtain the meteorological input data at each tower, which can be matched with the historical data of tower failures to form a sample.

[0052] S2: According to the test set situation, further sampling is performed to obtain a balanced data set.

[0053] Since most of the towers do not fail during extreme weather, the number of non-fault samples is much larger than the number of fault samples, i.e., an unbalanced sample set. If an unbalanced dataset is used directly as input, the prediction results will be biased towards the majority of samples. Therefore, this scheme uses a resampling method to balance the dataset.

[0054] The SMOTE oversampling method is used for fault samples. By generating artificial samples, the minority class samples are sampled. It can significantly improve the classification overfitting phenomenon that occurs in the random oversampling method and is currently the most widely used oversampling method. Figure 2, artificial fault class samples are inserted between fault class samples with close distance, thereby expanding the number of fault samples and improving the imbalance condition.

[0055] The undersampling algorithm OSS is used for non-fault samples, and the majority class samples are divided into noise samples, boundary samples and safety samples by judging the distance between samples. The main idea of ​​this method is to reduce noise samples and boundary samples, streamline safety samples, and achieve the purpose of reducing majority class samples. It selectively retains useful samples, deletes redundant noise samples and boundary samples, and improves classification results.

[0056] S3: Establish a prediction fault model and output the prediction results.

[0057] A generalized regression neural network (GRNN) prediction model is established. The network structure of GRNN consists of an input layer, a pattern layer, a summation layer, and an output layer. The topology is as follows: Figure 3 As shown. The input is X = [x 1 ,x 2 ,…,x n ] T , the output is Y=[y 1 ,y 2 ,…,y k ] T The input layer has a total of p neurons, corresponding to the p-dimensional input features of the sample. Let the input vector be X 0 =[x 01 ,x 02 ,…,x 0p ] T .

[0058] The pattern layer has a total of n neurons, the number of which is equal to the number of training samples. The i-th neuron corresponds to the i-th sample. The neuron activation function uses the Gaussian radial basis function to calculate the square of the Euclidean distance (X 0 -X i ) T (X 0 -X i ), the output of the i-th neuron is:

[0059]

[0060] Where σ is the smoothing factor of the radial basis function.

[0061] The summation layer has two neurons, which perform weighted summation on the output of the pattern layer:

[0062]

[0063] In the formula, the weight of the sum of SD is 1, and the weight of the sum of SN is the output vector y of the corresponding i-th training sample i .

[0064] The output layer divides the two outputs of the sum layer to get the final output of the network:

[0065]

[0066] The calculation of the smoothing factor includes the smoothing factor of the radial basis function, which will affect the shape of the radial basis function. If the smoothing factor is too large, the predicted output will be biased towards the average value of the training sample output. If the value is too small, the output result will be biased towards the training sample with the closest covariate.

[0067] Considering the wide application of mean absolute error (MAE) in error analysis, this scheme uses MAE to select the optimal smoothing factor σ.

[0068]

[0069] The neural network training based on historical data first associates the weather station information during extreme weather with the towers based on historical weather station monitoring data, tower longitude and latitude data, historical tower fault data, etc., to form the neural network input data. Then, new fault samples are generated, non-fault samples are extracted, and a balanced sample set is formed. The sample set is used as the training data of the neural network to train the GRNN. Through the training process of the neural network, the parameters of the neural network can be optimized and the correlation between the input data and the power grid fault can be identified.

[0070] The prediction model is used to verify the online prediction of power grid faults. The test samples are used to verify the fault prediction effect online, and the real-time collected weather station data is input into the trained GRNN model to obtain the power grid fault output results and predict whether the power grid equipment will fail.

[0071] The computer device may be a server. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data cluster data of the power monitoring system. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting power grid equipment failures under extreme weather conditions is implemented.

[0072] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0073] Example 2, reference Figure 4 , which is an embodiment of the present invention, provides a method and system for predicting power grid equipment failures under extreme weather conditions. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0074] The following is a specific implementation example using extreme windy weather as an example, which shows in detail how to establish a sample data set and build a power grid failure model for "a method for predicting power grid equipment failures under extreme weather conditions".

[0075] Step 1: Process historical meteorological information and historical fault information to form a test set and a fault set. Extract the information collected by the meteorological station, the longitude and latitude of the tower and the fault information, and the meteorological information around the tower in the time period before the accident, including the maximum wind speed X1, the maximum angle between the wind direction and the line direction X2, the two-minute average wind speed X3, the two-minute average angle between the wind direction and the line X4, and the rainfall X5. In addition, collect tower cases that did not fail during the period of strong winds, and the two types of samples form the original samples.

[0076] Step 2: For the case where the number of non-fault samples is much larger than that of fault samples, a resampling method is used to obtain a balanced data set. In this case, the original sample includes 140,000 non-fault samples and 38 fault samples. Due to the large difference in the number of fault samples and non-fault samples, the 140,000 non-fault samples are first randomly undersampled to reduce the number of non-fault samples to 300. Secondly, the 38 fault samples are oversampled by SMOTE to expand the number of minority class samples to 240. The non-fault samples are denoised to further reduce the number of non-fault samples to 240. The training sample set obtained based on the resampling algorithm has a total of 480 samples.

[0077] Step 3: Establish a generalized regression neural network (GRNN) prediction model.

[0078] According to the GRNN neural network model, the first 360 of the 480 experimental samples (240 fault samples and 240 normal samples) are used as training samples, and the last 120 are used as prediction samples. The sample attributes X1 (two-minute average wind speed), X2 (maximum wind speed), X3 (rainfall), X4 (maximum wind direction), and X5 (two-minute average wind direction) are normalized as network input. Whether the line tower has a power outage during this period (if the fault output is 1, if there is no fault, it is 0) is used as the output to construct the GRNN model.

[0079] After the sample is input into GRNN and the GRNN neural network output is obtained, this scheme takes 0.5 as the judgment threshold. i '≥0.5, the sample is classified as a fault and a fault warning is issued. i When '<0.5, no fault warning signal is issued. The fault warning results of the training samples obtained after using 0.5 as the threshold to classify the predicted output are shown in Table 1:

[0080] Table 1. Fault warning results of GRNN training samples

[0081] type Classification as Fault Classified as normal Fault sample 174 6 Normal sample 12 168

[0082] This scheme uses the forecast accuracy rate FDR, false alarm rate FAR and total forecast accuracy PA to evaluate the accuracy of the model:

[0083]

[0084] Where FP and FN are the number of faulty samples classified as non-faulty and faulty, respectively; TP and TN are the number of non-faulty samples classified as non-faulty and faulty, respectively.

[0085] The prediction results of the training samples are shown in Table 2, where FDR is 96.7%, FAR is 6.67%, and PA is 95%. Further analysis shows that most of the unpredictable faults occur in areas where the wind speed is not very high and the strong wind does not last long. At the same time, the lines in the samples of false tripping alarms are all located in dangerous areas and are in a dangerous state where failures may occur.

[0086] Table 2 Warning accuracy of GRNN training samples

[0087] type FDR FAR PA Prediction Sample 96.7％ 6.67％ 95％

[0088] Step 4: Use the prediction model to perform real-time online prediction of power grid faults.

[0089] Based on the prediction model established in step 3, 120 prediction samples (the first 60 of which are faulty samples and the last 60 are non-faulty samples) are used as the prediction output y i The actual output y i 'The difference |y i -y i '|As Figure 4 shown.

[0090] The predicted sample fault warning results and warning accuracy obtained after classifying the predicted output using a threshold of 0.5 are shown in Tables 3 and 4. It can be seen that the prediction success rate of the fault warning model based on GRNN reaches 90%, and the false alarm rate is lower.

[0091] Table 3 GRNN prediction sample fault warning results

[0092] type Classification as Fault Classified as normal Fault sample 54 6 Normal sample 1 59

[0093] Table 4 Warning accuracy of GRNN training samples

[0094]

[0095] Embodiment 3 is an embodiment of the present invention, including a power grid equipment failure prediction system under extreme weather conditions, specifically:

[0096] The preprocessing module collects and preprocesses the first object data to obtain a data set.

[0097] The resampling module further samples the test set to obtain a balanced data set.

[0098] Prediction module, establishes a prediction fault model and outputs the prediction results.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting power grid equipment failures under extreme weather conditions, characterized in that: include: Collecting and preprocessing first object data to obtain a data set; According to the test set situation, further sampling is performed to obtain a balanced data set; Establish a predictive fault model and output the prediction results.

2. The method for predicting power grid equipment failures under extreme weather conditions according to claim 1, characterized in that: The preprocessing includes performing association analysis on the first object data, obtaining input data at each first object by weighted calculation of the meteorological information closest to the first object, and matching with historical data to form a sample.

3. The method for predicting power grid equipment failure under extreme weather conditions according to claim 2, characterized in that: The data set includes a test set and a fault set, both of which are composed of the samples.

4. The method for predicting power grid equipment failures under extreme weather conditions according to claim 3, characterized in that: The further sampling to obtain a balanced data set according to the test set situation includes, for a situation in which non-fault samples in the data set are much larger than fault samples, resampling to obtain a balanced data set.

5. The method for predicting power grid equipment failures under extreme weather conditions according to claim 4, characterized in that: The resampling includes using an oversampling method on the fault samples to generate artificial samples to expand the sampling of minority class samples; For non-fault samples, an under-sampling algorithm is used to distinguish the distance between samples and divide the majority class samples into noise samples, boundary samples and safety samples, thereby reducing noise samples and boundary samples and streamlining safety samples.

6. The method for predicting power grid equipment failures under extreme weather conditions according to claim 1, characterized in that: The first object includes but is not limited to transmission towers, and the first object data includes but is not limited to the longitude and latitude information of the transmission towers, transmission tower fault statistics, longitude and latitude information of weather stations in the transmission tower area, and weather information in the past three years.

7. The method for predicting power grid equipment failures under extreme weather conditions according to claim 1, characterized in that: The method of establishing a prediction fault model and outputting a prediction result includes establishing a neural network prediction model, including an input layer, a pattern layer, a summation layer and an output layer, calculating an optimal smoothing factor, optimizing the neural network prediction model, and outputting a fault prediction result.

8. A power grid equipment failure prediction system under extreme weather conditions using the method according to any one of claims 1 to 7, characterized in that: A preprocessing module collects and preprocesses the first object data to obtain a data set; A resampling module, further sampling to obtain a balanced data set according to the test set; Prediction module, establishes a prediction fault model and outputs the prediction results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images