A power distribution network fault prediction method and system

Abstract

The application discloses a power distribution network fault prediction method and system, and the method comprises the following steps: acquiring multi-dimensional feature data, performing first preprocessing on the multi-dimensional feature data to obtain first multi-dimensional feature data; extracting spatial features and time features in the multi-dimensional feature data by using a first neural network and a second neural network, and constructing a prediction model based on the spatial features and the time features; defining a loss function of the prediction model, training the prediction model by using an optimizer, obtaining a first prediction model, and evaluating the first prediction model by using an evaluation index, so that dynamic evaluation and prediction of power distribution network faults are realized.

Description

Technical Field

[0001] This invention relates to the fields of power system disaster prevention and mitigation and artificial intelligence technology, and in particular to a method and system for predicting distribution network faults. Background Technology

[0002] The frequent occurrence of external natural disasters such as ice storms, wildfires, cold waves, torrential rains and floods, extreme temperatures, and lightning has brought unprecedented challenges to the safe and stable operation of the power system. Furthermore, due to the diverse spatial and temporal distribution of the power system and the unique lifecycle of power equipment, equipment failures in different geographical regions exhibit diversified and specific characteristics. Extreme external disasters and inherent equipment defects can easily trigger cluster failures, potentially leading to various chain reactions that jeopardize power supply and energy availability, thereby impacting social stability. Currently, the Guizhou Power Grid Company generates a large amount of data during the operation and maintenance of distribution network lines, but this data is not being effectively utilized. Simultaneously, the lack of effective intelligent means to manage distribution network failures hinders the improvement of distribution network failure response strategies. In the daily operation and maintenance of the distribution network, predicting distribution network failures is a key path to achieving efficient disaster prevention and mitigation goals. Accurate research on distribution network failure prediction models can not only effectively guide the design and operation and maintenance of distribution network lines but also significantly improve the power grid's ability to withstand natural disasters, which is of great significance for ensuring the safe and stable operation of the power grid.

[0003] However, current distribution network fault models face numerous challenges. Traditional fault prediction methods based on experience and limited observational data often fail to comprehensively and accurately reflect the fault characteristics of different regions under complex weather conditions. Existing methods struggle to effectively integrate multiple influencing factors such as meteorological elements and environmental conditions, leading to a high reliance on human experience in the fault prediction process and a lack of automated and scientific prediction methods. These problems directly affect the accuracy and applicability of distribution network auxiliary decision-making, making it difficult to adapt to complex terrain and variable climate conditions, thus limiting its ability to provide accurate decision support for power grid planning and operation and maintenance. To overcome these challenges, combining multi-source data, applying advanced numerical simulation and data assimilation techniques, and introducing machine learning methods have become the main research trends. In particular, digital distribution network fault prediction methods based on big data and machine learning show great potential. However, research on intelligent systems for automatic and accurate distribution network fault prediction under complex modal data remains a technological gap in the industry, severely restricting the improvement of power grid disaster prevention and mitigation capabilities. Existing methods cannot fully utilize the increasingly abundant multi-source data resources and are ill-suited to rapidly changing climate conditions, resulting in distribution network fault prediction results failing to provide timely and accurate decision support for distribution network planning and operation and maintenance. Meanwhile, although machine learning methods have made significant progress in data analysis in recent years, effective solutions for complex modal data are still lacking. In particular, the potential of the 2DCNN-BLSTM algorithm in processing time-series, high-dimensional, and nonlinear relational data has not been fully explored and utilized. There is an urgent need for a distribution network fault prediction method that can effectively integrate multi-source data and automatically process complex modal data. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for predicting distribution network faults, which solves the problems of low accuracy in ultra-short-term fault prediction of distribution networks under complex modes, inability to effectively integrate multi-source data, and lack of automation and intelligent processing capabilities of traditional methods.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for predicting faults in a power distribution network, comprising:

[0008] Obtain multidimensional feature data, and perform a first preprocessing on the multidimensional feature data to obtain first multidimensional feature data;

[0009] The spatial and temporal features in the multidimensional feature data are extracted using a first neural network and a second neural network, respectively, and a prediction model is constructed based on the spatial and temporal features.

[0010] Define the loss function of the prediction model, train the prediction model using an optimizer to obtain the first prediction model, and evaluate the first prediction model using evaluation metrics.

[0011] As a preferred embodiment of the power distribution network fault prediction method of the present invention, the first preprocessing of the multidimensional feature data includes:

[0012] Calculate the local outlier value for each multidimensional feature data, and calculate the mean and standard deviation of the local outlier values ​​for all multidimensional feature data;

[0013] If the local outlier value of the multidimensional feature data is greater than the sum of the mean of the local outlier values ​​of the multidimensional feature data and three times the standard deviation of the local outlier values ​​of the multidimensional feature data, then the multidimensional feature data is an outlier.

[0014] As a preferred embodiment of the power distribution network fault prediction method of the present invention, it further includes:

[0015] Calculate the mean and standard deviation of each multidimensional feature data, and calculate the standardized value of the mean of each multidimensional feature data;

[0016] Multiply the standardized value by the negative exponent of the nonlinear factor, and add 1 to the exponent term to get the denominator;

[0017] Divide 2 by the denominator and subtract 1 to obtain the normalized value.

[0018] As a preferred embodiment of the power distribution network fault prediction method of the present invention, it further includes:

[0019] Missing values ​​are filled using spatiotemporal data interpolation based on tensor decomposition. The optimal tensor rank is obtained by dividing multidimensional feature data into K parts, using one part as the test set and the remaining K-1 parts as the training set, repeating this process K times, selecting a different part as the test set each time, and calculating the average performance index of the K evaluations.

[0020] As a preferred embodiment of the power distribution network fault prediction method of the present invention, the prediction model constructed based on the spatial and temporal characteristics includes:

[0021] For the input layer, the shape of the input data is expanded, the spatial features of the expanded input data are extracted using the first neural network, the number of input channels is increased while the dimension remains unchanged, and weights are learned at the same time.

[0022] Pooling operations reduce spatial dimensionality and lower computational complexity in subsequent layers. During training, a portion of the input units are randomly set to zero to reshape the data and obtain sequential data.

[0023] The second neural network is used to process the sequence data and output the final prediction result.

[0024] As a preferred embodiment of the power distribution network fault prediction method of the present invention, it further includes:

[0025] For the l-th layer in the distribution network, a convolution operation is performed on the selected input mapping using a convolution kernel matrix, and a nonlinear activation function is used to introduce nonlinearity. The feature map is then operated on through a subsampling function to output the mapping.

[0026] By introducing a gate control mechanism, we can capture the temporal correlation in time series data.

[0027] As a preferred embodiment of the power distribution network fault prediction method described in this invention, the training prediction model includes:

[0028] The learning rate for the current time step is half the difference between the maximum and minimum learning rates. This is used to obtain the adjustment factor. The result of the cosine function is multiplied by 1 and then multiplied by the adjustment factor. Finally, the result is added to the minimum learning rate.

[0029] Secondly, the present invention provides a power distribution network fault prediction system, comprising:

[0030] The preprocessing module is used to acquire multidimensional feature data, perform a first preprocessing on the multidimensional feature data, and obtain first multidimensional feature data.

[0031] A model building module is used to extract spatial and temporal features from the multidimensional feature data using a first neural network and a second neural network, respectively, and to build a prediction model based on the spatial and temporal features.

[0032] The evaluation module is used to define the loss function of the prediction model, train the prediction model using an optimizer to obtain a first prediction model, and evaluate the first prediction model using evaluation metrics.

[0033] Thirdly, the present invention provides a computing device, comprising:

[0034] Memory and processor;

[0035] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the power distribution network fault prediction method.

[0036] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the power distribution network fault prediction method.

[0037] Compared with existing technologies, the advantages of this invention are as follows: By integrating two-dimensional convolutional neural networks (2DCNN) and bidirectional long short-term memory (BLSTM) methods, this invention significantly improves the accuracy of distribution network fault prediction under complex modalities; by employing multi-source distribution network data, meteorological elements, and holiday data, combined with the feature extraction capabilities of deep learning models, the temporal resolution of distribution network fault prediction is improved from the original ultra-short-term to the short-term, and even medium-term, level; an adaptive feature weight adjustment mechanism is introduced, enabling the model to automatically adjust the weights of various environmental factors according to the data characteristics of different regions. This adaptability allows the model to maintain high accuracy under different meteorological environments, greatly expanding the application scope of the method; by integrating time-series distribution network data and meteorological forecast data, dynamic assessment and prediction of distribution network faults are achieved. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the overall process logic of the power distribution network fault prediction method according to an embodiment of the present invention;

[0040] Figure 2 This is a CNN-BLSTM structure diagram of the power distribution network fault prediction method according to an embodiment of the present invention;

[0041] Figure 3 This is the LSTM layer of the power distribution network fault prediction method described in one embodiment of the present invention. Detailed Implementation

[0042] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0043] Example 1

[0044] Reference Figures 1-3 As an embodiment of the present invention, a method for predicting faults in a power distribution network is provided, comprising:

[0045] S100: Obtain multidimensional feature data, perform first preprocessing on the multidimensional feature data, and obtain first multidimensional feature data;

[0046] S200: The first and second neural networks are used to extract spatial and temporal features from multidimensional feature data, and a prediction model is built based on the spatial and temporal features.

[0047] S300: Define the loss function of the prediction model, train the prediction model using the optimizer to obtain the first prediction model, and evaluate the first prediction model using evaluation metrics.

[0048] It should be noted that the fusion of two-dimensional convolutional neural networks (2D CNN) and bidirectional long short-term memory (BLSTM) significantly improves the accuracy of distribution network fault prediction under complex modalities. Compared with traditional single-factor algorithms, the error rate is significantly reduced, and the results are closer to the actual situation. The 2DCNN-BLSTM ensemble learning algorithm architecture significantly improves computational efficiency. Compared with traditional manual experience-based classification methods, this method significantly reduces computation and investigation time while ensuring the accuracy of the results. This effect makes large-scale, high-frequency distribution network fault monitoring possible.

[0049] In this embodiment of the application, step S100 includes the following sub-steps A1-A5;

[0050] In A1: Calculate the local outlier value for each multidimensional feature data, and calculate the mean and standard deviation of the local outlier values ​​for all multidimensional feature data;

[0051] In A2: If the local outlier value of the multidimensional feature data is greater than the sum of the mean of the local outlier values ​​of the multidimensional feature data and three times the standard deviation of the local outlier values ​​of the multidimensional feature data, then the multidimensional feature data is an outlier.

[0052] In A3: Calculate the mean and standard deviation of each multidimensional feature data, and calculate the standardized value of the mean of each multidimensional feature data;

[0053] In A4: Multiply the standardized value by the negative exponent of the nonlinear factor, add 1 to the exponent term, and get the denominator;

[0054] In A5: Divide 2 by the denominator and subtract 1 to get the normalized value.

[0055] In this embodiment of the application, an advanced multi-source data acquisition system is used to collect multi-dimensional feature data, including historical power distribution network faults, meteorological elements, and holidays;

[0056] Specifically, historical distribution network fault information includes the time, location, and affected area of ​​the fault, as well as fault handling information. Data types include line information, time information, and string information. Meteorological factors include considerations such as rising temperatures leading to decreased equipment withstand voltage and insulation characteristics; high humidity easily reducing equipment insulation performance, leading to current leakage, flashover, etc.; and lightning strikes potentially causing equipment to experience lightning overvoltage. Data includes time-based classifications, regional classifications, average temperature, average high temperature, extreme high temperature, average low temperature, extreme low temperature, number of rainy days, number of thunderstorm days, snowy days, hazy days, foggy days, number of days with winds of level 3-5, number of days with winds above level 5, relative humidity, average wind speed, monthly average maximum wind speed, and maximum sustained wind speed. Holidays include those indirectly affecting distribution network operation and causing faults.

[0057] In this embodiment of the application, the first preprocessing includes outlier detection and removal, data normalization, and missing value processing;

[0058] Outlier detection and removal includes using an improved Local Outlier Factor (LOF) algorithm to identify and remove outlier data points;

[0059] Specifically, for feature j, the criteria for determining outliers are expressed as follows:

[0060]

[0061] in, Let be the mean of the LOF values ​​of feature j. Let be the standard deviation of the LOF values ​​of feature j;

[0062] The kd-tree algorithm is used to accelerate the calculation process of LOF in order to handle large-scale datasets;

[0063] It should be noted that the improved Local Outlier Factor (LOF) algorithm is suitable for anomaly detection in high-dimensional data, and can effectively identify local outliers, thereby improving the reliability of the data.

[0064] In an alternative embodiment, data normalization may further include a linear transformation of the original data using a traditional Min-Max normalization method, mapping the resulting values ​​to the range [0, –1]. The transformation function is expressed as:

[0065]

[0066] Where, x max x is the maximum value of the sample data. min The minimum value of the sample data;

[0067] In an optional embodiment, data normalization may further include Z-score normalization, which standardizes the data by giving it the mean and standard deviation. The processed data conforms to a standard normal distribution, i.e., the mean is 0 and the standard deviation is 1. The transformation function is expressed as:

[0068]

[0069] Where, x mean Let σ be the mean of all sample data, and σ be the standard deviation of all sample data.

[0070] In this embodiment of the application, data normalization includes:

[0071] Using an improved Min-Max normalization method, a nonlinear transformation is introduced to enhance the discriminative power of the features, represented as follows:

[0072]

[0073] Where α is an adjustable nonlinear factor, typically set between 0.1 and 10, and μ j and σ j Let be the mean and standard deviation of feature j, respectively;

[0074] It should be noted that the preferred nonlinear normalization method in this paper can better handle features of different scales and distributions, especially those with long-tailed distributions.

[0075] In this embodiment of the application, after completing steps A1-A5, step S100 also includes step A6;

[0076] In A6: missing values ​​are filled using spatiotemporal data interpolation based on tensor decomposition. The optimal tensor rank is obtained by dividing the multidimensional feature data into K parts, using one part as the test set and the remaining K-1 parts as the training set, repeating this process K times, each time selecting a different part as the test set, and calculating the average performance index of the K evaluations to obtain the optimal tensor rank.

[0077] In one alternative embodiment, missing value handling may include a moving average window method. If the data at the i-th position in a list a is missing, the average of the preceding and following window data is taken as the imputed data.

[0078] For example, given temperature data: a = [13, 10, 11, 5, None, 4, 9, 7, 8] and window = 3, the data at the None position is: (10 + 11 + 5 + 4 + 9 + 7) / 6 = 7.67;

[0079] In an alternative embodiment, missing value handling may further include regression imputation, which establishes a regression equation based on the complete dataset, and for objects containing missing values, substitutes known attribute values ​​into the equation to estimate unknown attribute values, and uses these estimates to fill in the missing values. This can lead to biased estimations when the variables are not linearly correlated.

[0080] For example, for temperature data that needs to be imputed: a = [13, 10, 11, 5, None, 4, 9, 7, 8], construct a time x-temperature y regression model y = -0.774x + 11.856, substitute x = 5, and get a missing temperature value of 7.99;

[0081] In this embodiment of the application, a spatiotemporal data interpolation method based on tensor decomposition is used to fill missing values. For the missing value x... ij The fill value is represented as:

[0082]

[0083] in, For the core tensor, a i b j c and c are spatial, temporal, and feature pattern factors, respectively;

[0084] The tensor decomposition problem is solved using alternating least squares (ALS), and the optimal tensor rank is selected through cross-validation.

[0085] It should be noted that the spatiotemporal data interpolation method based on tensor decomposition is suitable for data with spatiotemporal correlation. It can simultaneously consider the spatial and temporal correlation of the data and provide more accurate missing value estimation. Compared with other interpolation methods, the tensor decomposition method can better handle nonlinear relationships, provide more accurate missing value imputation, and can adapt to various complex data patterns, including nonlinear and high-dimensional data, which increases the flexibility and robustness of the method.

[0086] In this embodiment of the application, step S200 includes the following sub-steps B1-B3;

[0087] In B1: For the input layer, the shape of the input data is expanded, the spatial features of the expanded input data are extracted using the first neural network, the number of input channels is increased while keeping the dimension unchanged, and weights are learned at the same time.

[0088] In B2: Pooling operations are used to reduce spatial dimensionality and lower the computational complexity of subsequent layers. During training, a portion of the input units are randomly set to zero to reshape the data and obtain sequential data.

[0089] In B3: The second neural network is used to process the sequence data and output the final prediction result.

[0090] Specifically, the first neural network is a two-dimensional convolutional neural network, and the second neural network is a bidirectional long short-term memory. The two-dimensional convolutional neural network is used to learn and extract the spatial features of the dataset, and the bidirectional long short-term memory is used to learn and extract the temporal features of the dataset.

[0091] like Figure 2 As shown, the repeating vector layer expands the input shape from (16,6) to (16,6,1) to accommodate the two-dimensional convolutional layer. The two-dimensional convolutional layer increases the number of channels from 1 to 64 while maintaining the dimension of (16,6) and learns 640 weights. The two-dimensional max pooling layer reduces the spatial dimension, preserves important features, and reduces computational complexity. The regularization layer mitigates overfitting by randomly setting a portion of the input units to zero during training. The repeating vector layer reshapes the data to (16,384) to be compatible with the subsequent bidirectional long short-term memory layer. The bidirectional long short-term memory layer processes the sequence data in both forward and backward directions, capturing the temporal dependencies with 31,600 parameters. Another regularization layer is applied to the output of the bidirectional long short-term memory layer to further prevent overfitting. The fully connected layer transforms the output of the bidirectional long short-term memory layer into the final prediction result, reducing the output shape from 20 to 1 and outputting a predicted distribution network fault with 21 parameters.

[0092] It should be noted that two-dimensional convolutional neural networks effectively extract spatial features from input data, which is crucial for understanding the geometric and structural information of the data. Reducing spatial dimensions through pooling operations helps to reduce the computational complexity of the model while retaining important features, making the model more efficient. The use of regularization layers helps to mitigate overfitting, enhance the model's generalization ability, and ensure that the model performs well on unseen data. Bidirectional long short-term memory networks process sequential data and capture temporal dependencies. By combining two-dimensional convolutional neural networks and bidirectional long short-term memory networks, high-precision prediction of distribution network faults is achieved while ensuring computational efficiency and the model's generalization ability.

[0093] In this embodiment of the application, after completing steps B1-B3, step S200 above also includes the following steps B4-B5;

[0094] In B4: For the l-th layer in the distribution network, the selected input mapping is convolved using a convolution kernel matrix, and nonlinearity is introduced using a nonlinear activation function. The feature map is then manipulated through a subsampling function to output the mapping.

[0095] In B5: a gate control mechanism is introduced to capture the temporal correlation in time series data.

[0096] Specifically, two-dimensional CNN layers possess excellent spatial feature extraction capabilities, greatly simplifying the prediction process. Convolutional kernels can be used to extract spatial features from data. A typical convolutional neural network (CNN) consists of convolutional layers (CL), pooling layers (PL), and fully connected layers (FL). Features of the input samples are extracted through convolutional and pooling layers. The equations for convolutional and pooling layers are expressed as follows:

[0097]

[0098] in, For convolutional layers, l represents the l-th layer in the network, k represents the convolutional kernel matrix of size S×S, and M... j For the selection of the input mapping, f is a nonlinear activation function, and the modified linear unit (ReLU) is used as the nonlinear activation function; This represents a pooling layer. down() represents a subsampling function that sums all distinct n×n blocks in the feature map, thereby reducing the output feature map by a factor of n in both spatial dimensions. Each output map has a corresponding additive bias b and multiplicative bias β.

[0099] Specifically, such as Figure 3 As shown, the LSTM layer used to capture temporal correlations is represented as follows:

[0100] f t =σ(W f [h t-1 ,x t ]+b f )

[0101] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0102]

[0103] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0104] h t =o t *tanh(C t )

[0105] Among them, f t i t o t These are the forget gate, input gate, and output gate, respectively. (C) tIt is the unit state, h t It is in a hidden state, W f W i W o W C Let b be the weight matrix for the forget gate, input gate, output gate, and candidate cell states. f b i b o b C For bias terms, Let h be a candidate cell state, σ be the activation function, and h be the activation state. t-1 The hidden state of the previous time step, x t For the input of the current time step, C t-1 f represents the cell state at the previous time step. t and i t The update of the previous unit state and the unit state candidate are controlled separately.

[0106] It should be noted that the two-dimensional CNN layer can extract rich features from spatial data, and the introduction of nonlinear activation functions enhances the model's ability to handle nonlinear problems. The pooling layer reduces the spatial dimension of the feature map through the subsampling function, effectively reducing the computational complexity of the model while retaining important features. The introduction of the LSTM layer enables the model to capture the temporal correlation in time series data. The comprehensive analysis of spatial and temporal features makes the final prediction results more accurate and improves the reliability of distribution network fault prediction.

[0107] In this embodiment of the application, step S300 includes the following sub-step C1;

[0108] In C1: The learning rate for the current time step is half the difference between the maximum and minimum learning rates. This is the adjustment factor. The result of the cosine function is multiplied by 1 and then multiplied by the adjustment factor. Finally, the result is added to the minimum learning rate.

[0109] Specifically, the loss function is expressed as follows:

[0110]

[0111] Among them, y i Let i be the true value of the i-th sample. Let n be the model's predicted value for the i-th sample, and n be the total number of samples.

[0112] The model is trained using the Adam optimizer, with the learning rate expressed using a cosine annealing strategy:

[0113]

[0114] Where, η tLet η be the learning rate at time t. max η min These are the maximum and minimum learning rates, respectively, and T is the total training step size;

[0115] The model's performance is validated using a pre-defined prediction test set. The performance is evaluated using the ROC curve. The area under the ROC curve, AUC, measures the quality of the algorithm. The closer the curve is to the top left corner, the larger the AUC, and the better the classification performance.

[0116] It should be noted that the cosine annealing strategy allows the learning rate to change dynamically during training, which helps the model converge quickly in the early stages and stabilize and refine in the later stages, avoiding getting trapped in local minima. By combining the adjustment factor and the cosine function, it can gradually approach the optimal solution while maintaining a certain degree of randomness, balancing exploration and utilization. The cosine annealing strategy helps to avoid training instability caused by an excessively high learning rate or training stagnation caused by an excessively low learning rate.

[0117] The above is an illustrative scheme of a distribution network fault prediction method according to this embodiment. It should be noted that the technical solution of this distribution network fault prediction system and the technical solution of the above-described distribution network fault prediction method belong to the same concept. For details not described in detail in the technical solution of the distribution network fault prediction system in this embodiment, please refer to the description of the technical solution of the above-described distribution network fault prediction method.

[0118] The power distribution network fault prediction system in this embodiment includes:

[0119] The preprocessing module is used to acquire multidimensional feature data, perform a first preprocessing on the multidimensional feature data, and obtain first multidimensional feature data.

[0120] A model building module is used to extract spatial and temporal features from the multidimensional feature data using a first neural network and a second neural network, respectively, and to build a prediction model based on the spatial and temporal features.

[0121] The evaluation module is used to define the loss function of the prediction model, train the prediction model using an optimizer to obtain a first prediction model, and evaluate the first prediction model using evaluation metrics.

[0122] This embodiment also provides a computing device suitable for power distribution network fault prediction, including:

[0123] The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement the power distribution network fault prediction method proposed in the above embodiments.

[0124] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the method for predicting power distribution network faults as proposed in the above embodiments.

[0125] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for predicting distribution network faults proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0126] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0127] Example 2

[0128] Unlike the first embodiment, a verification test of the power distribution network fault prediction method is provided to verify and explain the technical effects of the method.

[0129] If there are 7 types of independent variable data for the period from January 1, 2021 to September 1, 2024, including cable coverage rate, commissioning date, line type, line area characteristics, weather, temperature, and wind force, and 1 objective variable data for whether there is a fault.

[0130] Outlier detection, data normalization, and missing value processing were performed to obtain nine fault factors, including cable coverage rate, commissioning date, line type, line area characteristics, weather, highest temperature, lowest temperature, average temperature, and wind force.

[0131] Correlation analysis was performed on the nine fault factors and whether the distribution network was faulty. The three fault factors with a correlation of less than 0.3 were processed by PCA and combined with the five fault factors with a correlation of greater than 0.3 to form the six final distribution network fault factors.

[0132] The dataset for constructing and predicting distribution network fault models consists of six final distribution network fault factors and one objective variable indicating whether a fault occurs, covering the period from January 1, 2021 to September 1, 2024.

[0133] The dataset was divided into 16 time steps as samples. The samples from January 1, 2021 to July 31, 2024 were used as samples for the construction and evaluation of the distribution network fault prediction model (the ratio of training set to validation set was 8:2). The samples from August 1, 2024 to September 1, 2024 were used as the prediction sample set.

[0134] The distribution network fault prediction model and evaluation training set are used as input (input layer (16, 6)). A RepeatVector layer expands the input shape from (16, 6) to (16, 6, 1) to fit the Conv2D layer. A Conv2D layer increases the number of channels from 1 to 64 while maintaining the (16, 6) dimension and learning 640 weights. A MaxPooling2D layer reduces the spatial dimension, preserving important features and reducing computational complexity. A Dropout layer randomly sets a portion of the input cells to zero during training. To mitigate overfitting, the data is reshaped into (16,384) values ​​using another RepeatVector layer for compatibility with subsequent BLSTM layers. A bidirectional Long Short-Term Memory (BLSTM) layer processes the sequence data in both forward and backward directions, capturing the temporal dependencies with 31,600 parameters. A Dropout layer is then applied to the BLSTM output to further prevent overfitting. Finally, a fully connected Dense layer transforms the BLSTM output into the final prediction, reducing the output shape from 20 to 1 and outputting a predicted distribution network fault with 21 parameters. The loss function is set to Mean Squared Error (MSE).

[0135] The trained model is used to predict the validation set, and finally, the accuracy of the validation set is evaluated using AUC, thus completing the model construction. The trained model is used to predict the validation set, and the accuracy of the validation set is evaluated using AUC, ultimately completing the entire process of the power distribution network fault prediction method and system.

[0136] By integrating two-dimensional convolutional neural networks (2D CNN) and bidirectional long short-term memory (BLSTM) methods, the accuracy of distribution network fault prediction under complex modes is significantly improved. The 2DCNN-BLSTM ensemble learning algorithm architecture significantly improves computational efficiency. An adaptive feature weight adjustment mechanism is introduced, enabling the model to automatically adjust the weights of various environmental factors according to the data characteristics of different regions. By integrating time-series distribution network data and meteorological forecast data, dynamic assessment and prediction of distribution network faults can be achieved.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting faults in a power distribution network, characterized in that, include: Obtain multidimensional feature data, and perform a first preprocessing on the multidimensional feature data to obtain first multidimensional feature data; The spatial and temporal features in the multidimensional feature data are extracted using a first neural network and a second neural network, respectively, and a prediction model is constructed based on the spatial and temporal features. Define the loss function of the prediction model, train the prediction model using an optimizer to obtain the first prediction model, and evaluate the first prediction model using evaluation metrics. The first preprocessing of the multidimensional feature data includes: Missing values ​​are filled using spatiotemporal data interpolation based on tensor decomposition. The optimal tensor rank is obtained by dividing multidimensional feature data into K parts, using one part as the test set and the remaining K-1 parts as the training set, repeating this process K times, each time selecting a different part as the test set, and calculating the average performance index of the K evaluations. The prediction model constructed based on the aforementioned spatial and temporal features includes: For the input layer, the shape of the input data is expanded, the spatial features of the expanded input data are extracted using the first neural network, the number of input channels is increased while the dimension remains unchanged, and weights are learned at the same time. Pooling operations reduce spatial dimensionality and lower computational complexity in subsequent layers. During training, a portion of the input units are randomly set to zero to reshape the data and obtain sequential data. The second neural network is used to process the sequence data and output the final prediction result. The first neural network is a two-dimensional convolutional neural network, and the second neural network is a bidirectional long short-term memory. The two-dimensional convolutional neural network is used to learn and extract the spatial features of the dataset, and the bidirectional long short-term memory is used to learn and extract the temporal features of the dataset. The repeating vector layer expands the input shape from (16,6) to (16,6,1), the 2D convolutional layer increases the number of channels from 1 to 64 while maintaining the dimension of (16,6) and learns 640 weights, the regularization layer mitigates overfitting by randomly setting a portion of the input cells to zero during training, and the repeating vector layer reshapes the data to (16,384).

2. The distribution network fault prediction method as described in claim 1, characterized in that, The first preprocessing of the multidimensional feature data includes: Calculate the local outlier value for each multidimensional feature data, and calculate the mean and standard deviation of the local outlier values ​​for all multidimensional feature data; If the local outlier value of the multidimensional feature data is greater than the sum of the mean of the local outlier values ​​of the multidimensional feature data and three times the standard deviation of the local outlier values ​​of the multidimensional feature data, then the multidimensional feature data is an outlier.

3. The distribution network fault prediction method as described in claim 2, characterized in that, Also includes: Calculate the mean and standard deviation of each multidimensional feature data, and calculate the standardized value of the mean of each multidimensional feature data; Multiply the standardized value by the negative exponent of the nonlinear factor, and add 1 to the exponent term to get the denominator; Divide 2 by the denominator and subtract 1 to obtain the normalized value.

4. The distribution network fault prediction method as described in claim 3, characterized in that, Also includes: For the first in the distribution network The layer performs convolution operations on the selected input mapping using a convolution kernel matrix, introduces nonlinearity using a nonlinear activation function, operates on the feature map through a subsampling function, and outputs the mapping. By introducing a gate control mechanism, we can capture the temporal correlation in time series data.

5. The distribution network fault prediction method as described in claim 4, characterized in that, The training prediction model includes: The learning rate for the current time step is half the difference between the maximum and minimum learning rates. This is used to obtain the adjustment factor. The result of the cosine function is multiplied by 1 and then multiplied by the adjustment factor. Finally, the result is added to the minimum learning rate.

6. A system applying the power distribution network fault prediction method as described in any one of claims 1-5, characterized in that, include: The preprocessing module is used to acquire multidimensional feature data, perform a first preprocessing on the multidimensional feature data, and obtain first multidimensional feature data. A model building module is used to extract spatial and temporal features from the multidimensional feature data using a first neural network and a second neural network, respectively, and to build a prediction model based on the spatial and temporal features. The evaluation module is used to define the loss function of the prediction model, train the prediction model using an optimizer to obtain a first prediction model, and evaluate the first prediction model using evaluation metrics.

7. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the power distribution network fault prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the power distribution network fault prediction method according to any one of claims 1 to 5.

Images