Method and device for detecting abnormity of multi-source heterogeneous operation and maintenance data of charging equipment

Abstract

The invention belongs to the technical field of intelligent operation and maintenance vehicle network interaction multi-source heterogeneous data anomaly detection, and particularly relates to a charging equipment multi-source heterogeneous operation and maintenance data anomaly detection method and device, and the method comprises the steps: collecting to-be-detected multi-source heterogeneous operation and maintenance data of charging equipment; based on the to-be-detected multi-source heterogeneous operation and maintenance data, determining abnormal data by using a pre-established anomaly detection model; based on a preset rule, determining fault information of the charging equipment according to the abnormal data; wherein the anomaly detection model is established by utilizing historical multi-source heterogeneous operation and maintenance data of the charging equipment. According to the technical scheme provided by the invention, the fault positioning efficiency and accuracy of the charging equipment are improved, the operation and maintenance cost of the electric vehicle is also reduced, and operation and maintenance personnel are assisted to quickly carry out fault processing.

Description

Technical Field

[0001] This invention belongs to the field of intelligent operation and maintenance vehicle-to-grid interactive multi-source heterogeneous data anomaly detection technology, specifically involving a method and device for detecting multi-source heterogeneous operation and maintenance data anomalies of charging equipment. Background Technology

[0002] Existing electric vehicle remote monitoring systems monitor and acquire real-time operational data from electric vehicles. This data, through acquisition, classification, processing, analysis, and application, can serve for safety monitoring and fault diagnosis at charging stations. Research also utilizes data mining algorithms to determine fuzzy control rules, ensuring optimal power, stability, and safety for pure electric vehicles on various road surfaces. Data mining techniques are used to collect and process data related to vehicle driving environment, driver operation, and vehicle motion state, and to adjust motor drive system control strategies in real-time based on driver intention prediction results, enabling vehicle control and decision-making. Furthermore, there are novel power system massive data analysis and fusion technologies characterized by high dimensionality, large scale, diverse types, multiple time scales, strong coupling, and high security requirements, addressing data needs, fusion goals, and common technologies for different application scenarios.

[0003] With the trend of globalization, domestic monitoring companies have successively established numerous safety monitoring and management systems. However, these systems are still unable to effectively utilize the massive amounts of data in vehicle monitoring systems to uncover potential patterns and provide more scientific decision-making support for road traffic regulatory departments and monitoring companies. Therefore, there is still significant room for improvement in using data mining technology for vehicle monitoring and management. Currently, research is limited to road traffic, and research on data mining technology based on electric vehicle charging and discharging monitoring is still rare.

[0004] Data from electric vehicle charging stations has four main characteristics: first, it is high-dimensional, large-scale, and growing rapidly; second, it is diverse in type, with wide-ranging sources, broad distribution, and varied structures; third, it has multiple time scales and high real-time requirements; and fourth, it is highly coupled. With the rapid development of new energy vehicles, the scale of charging equipment in operation is expanding daily, and the focus of work is gradually shifting from construction to operation and maintenance. As a major carrier of new infrastructure, the operation and maintenance of charging equipment has always faced challenges such as the difficulty of extracting fault characteristics from massive operational data. However, research on data anomaly analysis techniques for electric vehicle charging and discharging monitoring is scarce. Summary of the Invention

[0005] To overcome the problems existing in the above-mentioned related technologies, the present invention provides a method and device for detecting abnormal operation and maintenance data of multi-source heterogeneous charging equipment.

[0006] According to a first aspect of the present invention, a method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment is provided, comprising: Collect multi-source heterogeneous operation and maintenance data of charging equipment to be tested; Based on the multi-source heterogeneous operation and maintenance data to be detected, abnormal data is identified using a pre-established anomaly detection model. Based on preset rules, fault information of the charging equipment is determined according to the abnormal data; The anomaly detection model is established using historical multi-source heterogeneous operation and maintenance data of the charging equipment.

[0007] Preferably, the step of determining abnormal data based on the multi-source heterogeneous operation and maintenance data to be detected using a pre-established anomaly detection model includes: The multi-source heterogeneous operation and maintenance data to be detected is preprocessed to obtain the first dataset; The first dataset was clustered using the K-means clustering algorithm to obtain multiple clusters; Based on the sample type, feature extraction is performed on samples from each cluster to obtain key features; Using the key features as input to the anomaly detection model, the model outputs anomaly data.

[0008] Preferably, the preprocessing of the multi-source heterogeneous operation and maintenance data to obtain the first dataset includes: The multi-source heterogeneous operation and maintenance data to be detected is cleaned to obtain the cleaned multi-source heterogeneous operation and maintenance data to be detected. The cleaned multi-source heterogeneous operation and maintenance data to be tested is standardized to obtain standardized multi-source heterogeneous operation and maintenance data to be tested. The first dataset is constructed using the standardized, multi-source heterogeneous operation and maintenance data to be tested.

[0009] Preferably, the step of extracting features from samples in each cluster according to sample type to obtain key features includes: When the sample type in each cluster is data, principal component analysis is used to extract features from the samples to obtain key features. When the sample type in each cluster is an image, a pre-established feature extraction model is used to extract features from the sample to obtain key features.

[0010] Preferably, the principal component analysis method is used to extract features from the sample to obtain key features, including: Using samples from each cluster, construct a sample matrix for each cluster; Based on the sample matrix of each cluster, determine the covariance matrix of each cluster; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain the eigenvalues ​​of the covariance matrix and their corresponding principal component direction vectors. The feature values ​​are sorted in descending order, and a projection matrix is ​​constructed by selecting the top preset number of feature values. Using the projection matrix, the dimension-reduced feature vector is calculated, and the dimension-reduced feature vector is the key feature.

[0011] Preferably, determining the covariance matrix of each cluster based on the sample matrix of each cluster includes: Using the sample matrix of each cluster, the feature mean vector of each cluster is determined; Using the feature mean vector of each cluster, the samples in each cluster are centered to obtain the centered data of each cluster; The covariance matrix of each cluster is calculated using the centralized data of each cluster.

[0012] Preferably, the step of using a pre-established feature extraction model to extract features from the sample to obtain key features includes: The key features are output by inputting images as samples into the feature extraction model.

[0013] Preferably, the process of establishing the anomaly detection model includes: Collect historical, multi-source, heterogeneous operation and maintenance data of charging equipment; Based on the aforementioned historical multi-source heterogeneous operation and maintenance data, key historical features were determined; Anomalies in the historical key features are labeled to obtain historical anomaly data; A model dataset is constructed using the aforementioned key historical features and the aforementioned abnormal historical data. The dataset is divided into a training set and a test set; The long short-term memory network model is trained using the training set to obtain the trained long short-term memory network model. The trained Long Short-Term Memory (LSTM) network model is validated using the test set. If the validation is successful, the trained LTM network model is the anomaly detection model. If the validation fails, the hyperparameters of the LTM network model are adjusted, and the adjusted LTM network model is retrained until the validation is successful.

[0014] Preferably, the step of determining historical key features based on the historical multi-source heterogeneous operation and maintenance data includes: The historical multi-source heterogeneous operation and maintenance data is preprocessed to obtain the second dataset; The second dataset was clustered using the K-means clustering algorithm to obtain multiple historical clusters; Based on the sample type, features are extracted from samples of each historical cluster to obtain key historical features.

[0015] Preferably, the preprocessing of the historical multi-source heterogeneous operation and maintenance data to obtain the second dataset includes: The historical multi-source heterogeneous operation and maintenance data is cleaned to obtain cleaned historical multi-source heterogeneous operation and maintenance data. The historical multi-source heterogeneous operation and maintenance data after data cleaning is standardized to obtain standardized historical multi-source heterogeneous operation and maintenance data. The second dataset is constructed using the standardized historical multi-source heterogeneous operation and maintenance data.

[0016] Preferably, the step of extracting features from samples of each historical cluster according to sample type to obtain historical key features includes: When the sample type of each historical cluster is data, principal component analysis is used to extract features from the samples to obtain historical key features. When the sample type of each historical cluster is an image, a pre-established feature extraction model is used to extract features from the sample to obtain historical key features.

[0017] Preferably, the principal component analysis method is used to extract features from the sample to obtain historical key features, including: Using samples from each historical cluster, construct a sample matrix for each historical cluster; Based on the sample matrices of each historical cluster, determine the covariance matrix of each historical cluster. Eigenvalues ​​of the covariance matrices of each cluster in the history are obtained by performing eigenvalue decomposition on the covariance matrices of each cluster in the history and their corresponding principal component direction vectors. The feature values ​​are sorted in descending order, and a predetermined number of the top-ranked feature values ​​are selected to construct a historical projection matrix. Using the projection matrix of the history, the dimensionality-reduced feature vector of the history is calculated, and the dimensionality-reduced feature vector of the history is the key feature of the history.

[0018] Preferably, determining the covariance matrix of each historical cluster based on the sample matrix of each historical cluster includes: Using the sample matrix of each historical cluster, determine the feature mean vector of each historical cluster. Using the feature mean vectors of each historical cluster, the samples in each historical cluster are centered to obtain the centered data of each historical cluster. The covariance matrix of each historical cluster is calculated using the centralized data of each cluster.

[0019] Preferably, the step of using a pre-established feature extraction model to extract features from the sample to obtain historical key features includes: The image samples from each historical cluster are input into the feature extraction model, which outputs the historical key features.

[0020] Preferably, the process of establishing the feature extraction model includes: The sample type collected is historical images and their corresponding historical key features; The convolutional neural network is trained using the historical samples as input layer training samples and the historical key features as output layer training samples to obtain the feature extraction model.

[0021] Preferably, the formula for calculating the sample matrix of each cluster includes:

[0022] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The sample matrix of each cluster, For the first j The first sample in each cluster. For the first j The second sample in the cluster. For the first j The th cluster i One sample, For the first j The th cluster One sample, T This is a transpose.

[0023] Preferably, the formula for calculating the feature mean vector of each cluster includes:

[0024] In the above formula, , M The total number of clusters; , For the first jThe total number of samples in each cluster; For the first j The feature mean vector of each cluster For the first j The th cluster i One sample.

[0025] Preferably, the formula for calculating the centralized data of each cluster includes:

[0026] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j Centralized data of each cluster For the first j The th cluster i One sample, For the first j The feature mean vector of each cluster.

[0027] Preferably, the formula for calculating the covariance matrix of each cluster includes:

[0028] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The covariance matrix of each cluster, For the first j Centralized data of each cluster T This is a transpose.

[0029] Preferably, the formula for calculating the projection matrix includes:

[0030] The formula for calculating the dimensionality-reduced feature vector includes:

[0031] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The projection matrix of each cluster, PFor the preset quantity, For the first j The first eigenvalue in the projection matrix of each cluster. For the first j The first eigenvalue in the projection matrix of each cluster; For the first j The th cluster i The dimensionality-reduced feature vector of each sample. T For transpose, For the first j The th cluster i One sample.

[0032] According to a second aspect of the present invention, a device for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment is provided, comprising: The data acquisition unit is used to collect multi-source heterogeneous operation and maintenance data of the charging equipment to be tested. The detection unit is used to identify abnormal data based on the multi-source heterogeneous operation and maintenance data to be detected, using a pre-established anomaly detection model. The determining unit is used to determine the fault information of the charging device based on the abnormal data according to preset rules; The anomaly detection model is established using historical multi-source heterogeneous operation and maintenance data of the charging equipment.

[0033] According to a third aspect of the present invention, an electronic device is provided, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment is implemented.

[0034] According to a fourth aspect of the present invention, a readable storage medium is provided having an executable program stored thereon, wherein when the executable program is executed, the method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment is implemented.

[0035] The technical solution provided by this invention has the following beneficial effects: This invention provides a method and apparatus for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment. By collecting multi-source heterogeneous operation and maintenance data of the charging equipment to be detected, and using a pre-established anomaly detection model based on the multi-source heterogeneous operation and maintenance data to be detected, abnormal data is identified. Based on preset rules, fault information of the charging equipment is determined according to the abnormal data. Furthermore, an anomaly detection model is established using historical multi-source heterogeneous operation and maintenance data of the charging equipment. This not only improves the efficiency and accuracy of fault location of charging equipment, but also reduces the operation and maintenance cost of electric vehicles and assists operation and maintenance personnel in quickly handling faults. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0037] Figure 1 This is a flowchart of a method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the principal component analysis method provided in an embodiment of the present invention; Figure 3 This is a structural block diagram of a multi-source heterogeneous operation and maintenance data anomaly detection device for charging equipment provided in an embodiment of the present invention; Figure 4 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the following embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0039] Example 1 This invention provides a method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, such as... Figure 1 As shown, it includes the following steps: Step 1: Collect multi-source heterogeneous operation and maintenance data of the charging equipment to be tested; Step 2: Based on the multi-source heterogeneous operation and maintenance data to be detected, use the pre-established anomaly detection model to identify abnormal data; Step 3: Based on preset rules, determine the fault information of the charging equipment according to the abnormal data; The anomaly detection model is built using historical multi-source heterogeneous operation and maintenance data of charging equipment.

[0040] In some embodiments, multi-source heterogeneous operation and maintenance data may include, but is not limited to, charging pile side operation data and vehicle and battery data; The charging pile side operation data may include, but is not limited to: charging voltage, charging current, charging power, state of charge (SOC), charging time, and alarm status, etc. Vehicle and battery data may include, but are not limited to: total battery pack voltage, individual battery cell voltage, individual cell temperature, state of health (SOH) of the battery, and alarm information reported by the BMS.

[0041] The present invention provides a method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, which can be applied to the analysis of operation and maintenance data of charging piles, and facilitates the determination of whether electric vehicles have faults.

[0042] Specifically, fault information includes at least the following three types: fault type, fault code, fault location, and fault level.

[0043] It should be noted that this invention does not limit the "preset rules," which can be set based on experimental data, expert experience, and actual engineering projects. For example, different abnormal data correspond to different fault types, fault locations, and fault levels. Furthermore, effective information such as the number of faults of the same type and the duration of faults can be determined based on the fault information, helping maintenance personnel to conduct further analysis.

[0044] Further, step 2 includes: Step 21: Preprocess the multi-source heterogeneous operation and maintenance data to be detected to obtain the first dataset; Step 22: Use the K-means clustering algorithm to cluster the first dataset to obtain multiple clusters; Step 23: Based on the sample type, extract features from the samples of each cluster to obtain key features; Step 24: Use key features as input to the anomaly detection model and output abnormal data.

[0045] It should be noted that the "K-means clustering algorithm" involved in the embodiments of the present invention is well known to those skilled in the art, therefore, its specific implementation will not be described in detail.

[0046] Understandably, using the K-means clustering algorithm to group operational data and identify data points with similar characteristics helps in understanding patterns in data from different sources.

[0047] Further, step 21 includes: Step 211: Clean the multi-source heterogeneous operation and maintenance data to be tested, and obtain the cleaned multi-source heterogeneous operation and maintenance data to be tested; Step 212: Standardize the multi-source heterogeneous operation and maintenance data to be tested after data cleaning to obtain standardized multi-source heterogeneous operation and maintenance data to be tested; Step 213: Construct the first dataset using the standardized multi-source heterogeneous operation and maintenance data to be tested.

[0048] In some embodiments, the multi-source heterogeneous operation and maintenance data to be detected is preprocessed, including: Data cleaning of multi-source heterogeneous operation and maintenance data to be tested: removing noise and duplicate data, eliminating fields unrelated to the operation of charging equipment (such as electricity price, electricity fee and other billing information), and imputing or removing missing values ​​and outliers to ensure the reliability of subsequent analysis. Standardization and data fusion are performed on the multi-source heterogeneous operation and maintenance data to be tested after data cleaning: unified encoding, unit conversion and field mapping are performed on heterogeneous data from different manufacturers and systems, timestamp format is unified and time series resampling is performed to eliminate differences in units and formats and ensure the semantic and scale comparability of each data source; data fusion: time series alignment and multi-source fusion strategies are adopted to merge and verify the consistency of data from charging piles and vehicle BMS to form a unified, complete and traceable dataset.

[0049] Further, step 23 includes: Step 231: When the sample type of each cluster is data, principal component analysis is used to extract features from the samples to obtain key features; Step 232: When the sample type of each cluster is an image, the pre-established feature extraction model is used to extract features from the sample to obtain key features.

[0050] Further, step 231 includes: Step 2311: Construct a sample matrix for each cluster using samples from each cluster; Specifically, the formulas for calculating the sample matrix of each cluster include:

[0051] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The sample matrix of each cluster, For the first j The first sample in each cluster. For the first j The second sample in the cluster. For the first j The th cluster i One sample, For the first j The th cluster One sample, T For transpose; Step 2312: Determine the covariance matrix of each cluster based on the sample matrix of each cluster; Step 2313: Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues ​​of the covariance matrix and their corresponding principal component direction vectors; Step 2314: Sort the eigenvalues ​​in descending order and select the top preset number of eigenvalues ​​to construct the projection matrix; Step 2315: Calculate the dimension-reduced feature vector using the projection matrix. The dimension-reduced feature vector is the key feature. Specifically, the formula for calculating the projection matrix includes:

[0052] The formulas for calculating the dimensionality-reduced eigenvectors include:

[0053] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The projection matrix of each cluster, P For the preset quantity, For the first j The first eigenvalue in the projection matrix of each cluster. For the first j The first eigenvalue in the projection matrix of each cluster; For the first j The th cluster i The dimensionality-reduced feature vector of each sample. T For transpose, For the first j The th cluster i One sample.

[0054] Further, step 2312 includes: Step 2312a: Using the sample matrix of each cluster, determine the feature mean vector of each cluster; Specifically, the formulas for calculating the feature mean vector of each cluster include:

[0055] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The feature mean vector of each cluster For the first j The th cluster i One sample; Step 2312b: Using the feature mean vector of each cluster, the samples in each cluster are centered to obtain the centered data of each cluster; Specifically, the calculation formulas for the centralized data of each cluster include:

[0056] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j Centralized data of each cluster For the first j The th cluster i One sample, For the first j The feature mean vector of each cluster; Step 2312c: Calculate the covariance matrix of each cluster using the centered data of each cluster; Specifically, the formulas for calculating the covariance matrix of each cluster include:

[0057] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The covariance matrix of each cluster, For the first j Centralized data of each cluster T This is a transpose.

[0058] Further, step 232 includes: Step 2321: Input the sample of type image into the feature extraction model and output the key features.

[0059] In some embodiments, principal component analysis (PCA) is used to identify and select useful features, performing dimensionality reduction and feature extraction on data from different sources (such as state of charge (SOC), maximum single-cell voltage, maximum single-cell current, etc.) to analyze their impact on the operating status and abnormal behavior of the charging equipment. Specifically, the calculation is as follows: Let the clustering step yield the... j Each cluster contains There are 10 samples, each sample has 100 samples. M Numerical features constitute the sample matrix First, calculate the mean vector of features within the cluster. and perform centralized processing. Then construct the covariance matrix. Then, the principal component direction vectors and their eigenvalues ​​are obtained through eigenvalue decomposition. The eigenvalues ​​are then sorted in descending order, and the top [values] are selected. P The principal components form the projection matrix. The dimensionality-reduced feature vectors are obtained. This method can extract key features that reflect the main changing trends within a cluster, such as the rate of change of state of charge (SOC), the difference in individual cell voltage, and the amplitude of individual cell current fluctuations, thereby enabling quantitative analysis of the correlation between charging state parameters.

[0060] For image-based information, such as charging temperature distribution maps, fault code frequency distribution maps, or fire characteristic heatmaps, a Convolutional Neural Network (CNN) is used for deep feature extraction and analysis. The CNN model extracts spatial structure and texture features from the image layer by layer through multiple convolutional, pooling, and activation operations. Its typical computational form is as follows:

[0061] In the above formula, I Indicates the input image. K For convolution kernel, b For bias terms, It is a non-linear activation function. F The convolution outputs a feature map. After multiple convolutions and pooling, the feature map is flattened and input into a fully connected layer to obtain a high-dimensional feature vector f_img.

[0062] CNN networks can be trained using supervised learning, utilizing the cross-entropy loss function. Optimize network parameters, among which, y For real labels, For predicted values, N This represents the total number of samples. During training, learning rate decay, regularization, and data augmentation strategies are combined to improve the model's generalization ability. Finally, the numerical features extracted by PCA are... The image f_img extracted by CNN is fused to form a unified multimodal feature vector, which is used as input for subsequent time series analysis and anomaly detection models based on Long Short-Term Memory (LSTM) networks.

[0063] Furthermore, the process of establishing the anomaly detection model includes: Step 41: Collect historical multi-source heterogeneous operation and maintenance data of the charging equipment; Step 42: Determine key historical features based on historical multi-source heterogeneous operation and maintenance data; Step 43: Label the abnormal samples in the historical key features to obtain historical abnormal data; Step 44: Construct the model dataset using historical key features and historical anomaly data; Step 45: Divide the dataset into a training set and a test set; Step 46: Train the Long Short-Term Memory Network model using the training set to obtain the trained Long Short-Term Memory Network model; Step 47: Validate the trained Long Short-Term Memory (LSTM) network model using the test set. If the validation is successful, the trained LTM network model is an anomaly detection model. If the validation fails, adjust the hyperparameters of the LTM network model and retrain the adjusted LTM network model until the validation is successful.

[0064] Further, step 42 includes: Step 421: Preprocess the historical multi-source heterogeneous operation and maintenance data to obtain the second dataset; Step 422: Use the K-means clustering algorithm to cluster the second dataset to obtain multiple historical clusters; Step 423: Based on the sample type, extract features from the samples of each historical cluster to obtain key historical features.

[0065] Further, step 46 includes: The long short-term memory network model is trained by using historical key features from the training set as input layer training samples and historical outlier data from the training set as output layer training samples.

[0066] Further, step 47 includes: Using historical key features from the test set as input to the long short-term memory network model, the output is predicted outlier data; The prediction accuracy is determined by using the predicted outlier data and historical outlier data in the test set. If the prediction accuracy is greater than or equal to the accuracy threshold, the verification is successful; otherwise, the verification fails.

[0067] In some embodiments, during the training phase of the Long Short-Term Memory (LSTM) network model, historical key features are first obtained based on historical multi-source heterogeneous operation and maintenance data. These historical key features are then labeled to obtain anomalous data. A training dataset is constructed using the historical key features and the anomalous data. The historical key features are then sequentially input into the LSTM network model. Iterative optimization of the model parameters is achieved by minimizing the loss function, often using binary cross-entropy or weighted cross-entropy to measure the difference between the model's output probability and the true label. During training, adaptive gradient algorithms or stochastic gradient descent (SGD) with momentum are used for optimization, combined with learning rate decay, L2 regularization, and early stopping strategies to suppress overfitting. After model training, accuracy, recall, and ROC-AUC curves are calculated to evaluate the model's discriminative performance, and the optimal anomaly detection threshold is determined based on the curve shape. The resulting LSTM model, as an anomaly detection model, can predict the input feature sequence and output anomaly probabilities in real-time, enabling anomaly detection, early warning, and status monitoring during the operation and maintenance of charging equipment.

[0068] Further, step 421 includes: Step 4211: Clean the historical multi-source heterogeneous operation and maintenance data to obtain the cleaned historical multi-source heterogeneous operation and maintenance data; Step 4212: Standardize the historical multi-source heterogeneous operation and maintenance data after data cleaning to obtain standardized historical multi-source heterogeneous operation and maintenance data; Step 4213: Construct a second dataset using standardized historical multi-source heterogeneous operation and maintenance data.

[0069] Further, step 423 includes: Step 4231: When the sample type of each historical cluster is data, principal component analysis is used to extract features from the samples to obtain historical key features. Step 4232: When the sample type of each historical cluster is an image, a pre-established feature extraction model is used to extract features from the sample to obtain historical key features.

[0070] Further, step 4231 includes: Step 4231a: Construct a sample matrix for each historical cluster using samples from each historical cluster; Step 4231b: Determine the covariance matrix of each historical cluster based on the sample matrix of each historical cluster. Step 4231c: Perform eigenvalue decomposition on the covariance matrix of each historical cluster to obtain the eigenvalues ​​of the covariance matrix of each historical cluster and their corresponding principal component direction vectors. Step 4231d: Sort the eigenvalues ​​in descending order and select the top 100 eigenvalues ​​to construct the historical projection matrix; Step 4231e: Using the projection matrix of the history, calculate the dimensionality-reduced feature vector of the history. The dimensionality-reduced feature vector of the history is the key feature of the history.

[0071] It is understandable that the calculation formula for the sample matrix of each cluster in history is the same as in step 2311, the calculation formula for the projection matrix of history is the same as in step 2314, and the calculation formula for the dimensionality-reduced eigenvector of history is the same as in step 2315.

[0072] Further, step 4231b includes: Step 4231b1: Using the sample matrices of each historical cluster, determine the feature mean vector of each historical cluster. Step 4231b2: Using the feature mean vectors of each historical cluster, the samples in each historical cluster are centered to obtain the centered data of each historical cluster. Step 4231b3: Calculate the covariance matrix of each historical cluster using the centralized data of each cluster.

[0073] It is understandable that the formula for calculating the characteristic mean vector of each historical cluster is the same as that in step 2312a, the formula for calculating the centered data of each historical cluster is the same as that in step 2312b, and the formula for calculating the covariance matrix of each historical cluster is the same as that in step 2312c.

[0074] Further, step 4232 includes: Step 42321: Input the image samples from each historical cluster into the feature extraction model and output the key historical features.

[0075] Specifically, the process of building a feature extraction model includes: The sample type collected is historical images and their corresponding historical key features; The convolutional neural network is trained by using historical samples as input layer training samples and historical key features as output layer training samples, thus obtaining a feature extraction model.

[0076] In some embodiments, historical samples of image type and their corresponding historical key features can also be used to verify the trained convolutional neural network, and the verification method is the same as the training and verification process of the anomaly detection model.

[0077] To further illustrate the above-mentioned method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, this invention provides a specific example. First, it aggregates multi-source heterogeneous operation and maintenance data such as voltage, current, and SOC collected from various companies for different vehicles during each charging session. Data preprocessing is required: removing information useless for judging charging facility malfunctions, such as electricity prices, i.e., data cleaning; standardizing the data to ensure format consistency for subsequent research, given the different file formats from various companies; and integrating the multi-source heterogeneous operation and maintenance data from different companies into a unified dataset using data fusion techniques such as time series alignment.

[0078] Figure 2 Principal component analysis was used to extract features from preprocessed multi-source heterogeneous operation and maintenance data, thereby filtering out useful data related to operation and maintenance status.

[0079] Finally, by combining convolutional neural networks and long short-term memory neural networks (CNN-LSTM), and using CNN convolutional layers and pooling layers as feature extraction modules for image type information, data features were obtained. The reconstructed information was then injected into the LSTM network to mine the temporal features of the data, analyze fault data, and perform anomaly detection of operation and maintenance data.

[0080] The present invention provides a method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment. By analyzing anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, the method can significantly reduce the operation and maintenance costs of electric vehicles. By adopting appropriate intelligent data processing technology, the method can improve the data processing speed, realize the analysis and processing of real-time status information, and assist operation and maintenance personnel in quickly determining the fault type, fault location and fault level.

[0081] The method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment provided by this invention has the following advantages: 1. Effective data from the operation monitoring of multi-source charging equipment were extracted and feature analysis was performed, revealing the differential influence patterns of the knowledge base and database; 2. Data preprocessing technology based on big data analysis enables the accurate reflection and efficient utilization of data under multi-source heterogeneous data conditions, and further classifies risk levels based on the frequency and impact of typical faults. 3. The addition of convolutional neural networks makes the dimensionality reduction types more comprehensive. It can not only process charging pile message data, but also extract features from image-type fire information, which is in line with the multi-source heterogeneous operation and maintenance data processing method. 4. Compared with using LSTM or CNN alone for prediction, the CNN-LSTM model has higher prediction accuracy.

[0082] Example 2 This invention also provides a device for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, such as... Figure 3 As shown, it includes: The data acquisition unit is used to collect multi-source heterogeneous operation and maintenance data of the charging equipment to be tested. The detection unit is used to identify abnormal data based on the multi-source heterogeneous operation and maintenance data to be detected, using a pre-established anomaly detection model. The determination unit is used to determine the fault information of the charging equipment based on preset rules and abnormal data; The anomaly detection model is built using historical multi-source heterogeneous operation and maintenance data of charging equipment.

[0083] Furthermore, the detection unit includes: The first acquisition subunit is used to preprocess the multi-source heterogeneous operation and maintenance data to be detected to obtain the first dataset. The clustering subunit is used to cluster the first dataset using the K-means clustering algorithm to obtain multiple clusters; Extraction sub-units are used to extract features from samples in each cluster based on sample type to obtain key features; The second acquisition subunit is used to take key features as input to the anomaly detection model and output abnormal data.

[0084] Furthermore, the first acquisition subunit includes: The first acquisition module is used to clean the multi-source heterogeneous operation and maintenance data to be tested, and to obtain the cleaned multi-source heterogeneous operation and maintenance data to be tested. The second acquisition module is used to standardize the multi-source heterogeneous operation and maintenance data to be tested after data cleaning, and obtain the standardized multi-source heterogeneous operation and maintenance data to be tested. The third acquisition module is used to construct the first dataset using the standardized, multi-source heterogeneous operation and maintenance data to be detected.

[0085] Furthermore, extracting sub-units includes: The first extraction module is used to extract key features from samples when the sample type in each cluster is data, by using principal component analysis. The second extraction module is used to extract key features from samples when the sample type in each cluster is an image, using a pre-established feature extraction model.

[0086] Furthermore, the first extraction module includes: The construction submodule is used to construct the sample matrix of each cluster using samples from each cluster; The determination submodule is used to determine the covariance matrix of each cluster based on the sample matrix of each cluster; The decomposition submodule is used to perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues ​​of the covariance matrix and their corresponding principal component direction vectors. The selection submodule is used to sort the feature values ​​in descending order and select the top preset number of feature values ​​to construct the projection matrix; The computation submodule is used to calculate the dimension-reduced feature vector using the projection matrix. The dimension-reduced feature vector is the key feature.

[0087] Furthermore, the first defined submodule is specifically used for: Using the sample matrix of each cluster, determine the feature mean vector of each cluster; By using the feature mean vectors of each cluster, the samples in each cluster are centered to obtain the center of each cluster. The covariance matrix of each cluster is calculated using the centralized data of each cluster.

[0088] Furthermore, a pre-established feature extraction model is used to extract features from the samples, obtaining key features, including: Input images into the feature extraction model and output key features.

[0089] Furthermore, the device also includes: a modeling unit for establishing an anomaly detection model; the modeling unit includes: The data acquisition subunit is used to collect historical multi-source heterogeneous operation and maintenance data of the charging equipment. Determine sub-units to identify key historical features based on historical multi-source heterogeneous operation and maintenance data; The annotation sub-unit is used to annotate abnormal samples in historical key features to obtain historical abnormal data; Construct sub-units to build model datasets using historical key features and historical anomaly data; Divide the dataset into sub-units to separate the dataset into training and test sets; The training subunit is used to train the long short-term memory network model using the training set to obtain the trained long short-term memory network model. The validation subunit is used to validate the trained Long Short-Term Memory (LSTM) network model using a test set. If the validation is successful, the trained LTM network model is an anomaly detection model. If the validation fails, the hyperparameters of the LTM network model are adjusted, and the adjusted LTM network model is retrained until the validation is successful.

[0090] Furthermore, the sub-units are determined, including: The fourth acquisition module is used to preprocess historical multi-source heterogeneous operation and maintenance data to obtain the second dataset; The clustering module is used to cluster the second dataset using the K-means clustering algorithm to obtain multiple historical clusters; The extraction module is used to extract features from samples of each historical cluster based on the sample type, thereby obtaining key historical features.

[0091] Furthermore, the fourth acquisition module includes: The first acquisition submodule is used to clean historical multi-source heterogeneous operation and maintenance data to obtain cleaned historical multi-source heterogeneous operation and maintenance data. The second acquisition submodule is used to standardize the historical multi-source heterogeneous operation and maintenance data after data cleaning, and obtain standardized historical multi-source heterogeneous operation and maintenance data. The third acquisition submodule is used to construct a second dataset using standardized historical multi-source heterogeneous operation and maintenance data.

[0092] Furthermore, the extraction module includes: The first extraction submodule is used to extract features from samples when the sample type in each historical cluster is data, using principal component analysis to obtain historical key features. The second extraction submodule is used to extract features from samples when the sample type in each historical cluster is an image, using a pre-established feature extraction model to obtain historical key features.

[0093] Furthermore, the first extraction submodule includes: Using samples from each historical cluster, construct a sample matrix for each historical cluster; Based on the sample matrices of each historical cluster, determine the covariance matrix of each historical cluster. Eigenvalues ​​of the covariance matrices of each historical cluster are obtained by performing eigenvalue decomposition on the covariance matrices of each historical cluster and their corresponding principal component direction vectors. Sort the eigenvalues ​​in descending order and select the top 100 eigenvalues ​​to construct a historical projection matrix. By using the projection matrix of history, the dimensionality-reduced eigenvectors of history are calculated, and these dimensionality-reduced eigenvectors represent the key features of history.

[0094] Furthermore, based on the sample matrices of each historical cluster, the covariance matrix of each historical cluster is determined, including: Using the sample matrices of each historical cluster, determine the feature mean vector of each historical cluster. By using the feature mean vectors of each historical cluster, the samples in each historical cluster are centered to obtain the centered data of each historical cluster. By using the centralized data of each historical cluster, the covariance matrix of each historical cluster is calculated.

[0095] Furthermore, the second extraction submodule includes: Input the image samples from each historical cluster into the feature extraction model, and output the key historical features.

[0096] Furthermore, the process of establishing the feature extraction model includes: The sample type collected is historical images and their corresponding historical key features; The convolutional neural network is trained by using historical samples as input layer training samples and historical key features as output layer training samples, thus obtaining a feature extraction model.

[0097] Furthermore, the formulas for calculating the sample matrix of each cluster include:

[0098] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The sample matrix of each cluster, For the first j The first sample in each cluster. For the first j The second sample in the cluster. For the first j The th cluster i One sample, For the first j The th cluster One sample, T This is a transpose.

[0099] Furthermore, the formulas for calculating the feature mean vectors of each cluster include:

[0100] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The feature mean vector of each cluster For the first j The th cluster i One sample.

[0101] Furthermore, the calculation formulas for the centralized data of each cluster include:

[0102] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j Centralized data of each cluster For the first j The th cluster i One sample, For the first j The feature mean vector of each cluster.

[0103] Furthermore, the formulas for calculating the covariance matrix of each cluster include:

[0104] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The covariance matrix of each cluster, For the first j Centralized data of each cluster T This is a transpose.

[0105] Furthermore, the formula for calculating the projection matrix includes:

[0106] The formulas for calculating the dimensionality-reduced eigenvectors include:

[0107] In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The projection matrix of each cluster, P For the preset quantity, For the first j The first eigenvalue in the projection matrix of each cluster. For the first j The first eigenvalue in the projection matrix of each cluster; For the first jThe th cluster i The dimensionality-reduced feature vector of each sample. T For transpose, For the first j The th cluster i One sample.

[0108] It is understood that the device embodiments provided above correspond to the method embodiments described above, and the specific details can be referred to each other, which will not be repeated here.

[0109] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0110] Example 3 like Figure 4 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0111] The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to realize the steps of the multi-source heterogeneous operation and maintenance data anomaly detection method for charging equipment in the above embodiments.

[0112] Example 4 Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of the multi-source heterogeneous operation and maintenance data anomaly detection method for charging devices described in the above embodiments.

[0113] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0115] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0116] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0117] Finally, 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 the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment, characterized in that, include: Collect multi-source heterogeneous operation and maintenance data of charging equipment to be tested; Based on the multi-source heterogeneous operation and maintenance data to be detected, abnormal data is identified using a pre-established anomaly detection model. Based on preset rules, fault information of the charging equipment is determined according to the abnormal data; The anomaly detection model is established using historical multi-source heterogeneous operation and maintenance data of the charging equipment.

2. The method according to claim 1, characterized in that, Based on the multi-source heterogeneous operation and maintenance data to be detected, anomaly data is identified using a pre-established anomaly detection model, including: The multi-source heterogeneous operation and maintenance data to be detected is preprocessed to obtain the first dataset; The first dataset was clustered using the K-means clustering algorithm to obtain multiple clusters; Based on the sample type, feature extraction is performed on samples from each cluster to obtain key features; Using the key features as input to the anomaly detection model, the model outputs anomaly data.

3. The method according to claim 2, characterized in that, The preprocessing of the multi-source heterogeneous operation and maintenance data to obtain the first dataset includes: The multi-source heterogeneous operation and maintenance data to be detected is cleaned to obtain the cleaned multi-source heterogeneous operation and maintenance data to be detected. The cleaned multi-source heterogeneous operation and maintenance data to be tested is standardized to obtain standardized multi-source heterogeneous operation and maintenance data to be tested. The first dataset is constructed using the standardized, multi-source heterogeneous operation and maintenance data to be tested.

4. The method according to claim 2, characterized in that, Based on the sample type, feature extraction is performed on samples from each cluster to obtain key features, including: When the sample type in each cluster is data, principal component analysis is used to extract features from the samples to obtain key features. When the sample type in each cluster is an image, a pre-established feature extraction model is used to extract features from the sample to obtain key features.

5. The method according to claim 4, characterized in that, The principal component analysis method is used to extract features from the sample to obtain key features, including: Construct a sample matrix for each cluster using samples from each cluster; Based on the sample matrix of each cluster, determine the covariance matrix of each cluster; The covariance matrix is ​​subjected to eigenvalue decomposition to obtain the eigenvalues ​​of the covariance matrix and their corresponding principal component direction vectors. The feature values ​​are sorted in descending order, and a projection matrix is ​​constructed by selecting the top preset number of feature values. Using the projection matrix, the dimension-reduced feature vector is calculated, and the dimension-reduced feature vector is the key feature.

6. The method according to claim 5, characterized in that, The step of determining the covariance matrix of each cluster based on the sample matrix of each cluster includes: Using the sample matrix of each cluster, the feature mean vector of each cluster is determined; Using the feature mean vector of each cluster, the samples in each cluster are centered to obtain the centered data of each cluster; The covariance matrix of each cluster is calculated using the centralized data of each cluster.

7. The method according to claim 4, characterized in that, The process employs a pre-established feature extraction model to extract features from the samples, obtaining key features, including: The key features are output by inputting images as samples into the feature extraction model.

8. The method according to claim 1, characterized in that, The process of establishing the anomaly detection model includes: Collect historical, multi-source, heterogeneous operation and maintenance data of charging equipment; Based on the aforementioned historical multi-source heterogeneous operation and maintenance data, key historical features were determined; Anomalies in the historical key features are labeled to obtain historical anomaly data; A model dataset is constructed using the aforementioned key historical features and the aforementioned abnormal historical data. The dataset is divided into a training set and a test set; The long short-term memory network model is trained using the training set to obtain the trained long short-term memory network model. The trained Long Short-Term Memory (LSTM) network model is validated using the test set. If the validation is successful, the trained LTM network model is the anomaly detection model. If the validation fails, the hyperparameters of the LTM network model are adjusted, and the adjusted LTM network model is retrained until the validation is successful.

9. The method according to claim 8, characterized in that, The determination of historical key features based on the historical multi-source heterogeneous operation and maintenance data includes: The historical multi-source heterogeneous operation and maintenance data is preprocessed to obtain the second dataset; The second dataset was clustered using the K-means clustering algorithm to obtain multiple historical clusters; Based on the sample type, features are extracted from samples of each historical cluster to obtain key historical features.

10. The method according to claim 9, characterized in that, The preprocessing of the historical multi-source heterogeneous operation and maintenance data to obtain the second dataset includes: The historical multi-source heterogeneous operation and maintenance data is cleaned to obtain cleaned historical multi-source heterogeneous operation and maintenance data. The historical multi-source heterogeneous operation and maintenance data after data cleaning is standardized to obtain standardized historical multi-source heterogeneous operation and maintenance data. The second dataset is constructed using the standardized historical multi-source heterogeneous operation and maintenance data.

11. The method according to claim 9, characterized in that, Based on the sample type, feature extraction is performed on samples from each historical cluster to obtain key historical features, including: When the sample type of each historical cluster is data, principal component analysis is used to extract features from the samples to obtain historical key features. When the sample type of each historical cluster is an image, a pre-established feature extraction model is used to extract features from the sample to obtain historical key features.

12. The method according to claim 11, characterized in that, The principal component analysis method is used to extract features from the sample to obtain historical key features, including: Using samples from each historical cluster, construct a sample matrix for each historical cluster; Based on the sample matrices of each historical cluster, determine the covariance matrix of each historical cluster. Eigenvalues ​​of the covariance matrices of each cluster in the history are obtained by performing eigenvalue decomposition on the covariance matrices of each cluster in the history and their corresponding principal component direction vectors. The feature values ​​are sorted in descending order, and a predetermined number of the top-ranked feature values ​​are selected to construct a historical projection matrix. Using the projection matrix of the history, the dimensionality-reduced feature vector of the history is calculated, and the dimensionality-reduced feature vector of the history is the key feature of the history.

13. The method according to claim 12, characterized in that, The step of determining the covariance matrix of each historical cluster based on the sample matrix of each historical cluster includes: Using the sample matrix of each historical cluster, determine the feature mean vector of each historical cluster. Using the feature mean vectors of each historical cluster, the samples in each historical cluster are centered to obtain the centered data of each historical cluster. The covariance matrix of each historical cluster is calculated using the centralized data of each cluster.

14. The method according to claim 11, characterized in that, The process employs a pre-established feature extraction model to extract features from the samples, obtaining historical key features, including: The image samples from each historical cluster are input into the feature extraction model, which outputs the historical key features.

15. The method according to claim 4 or 11, characterized in that, The process of establishing the feature extraction model includes: The sample type collected is historical images and their corresponding historical key features; The convolutional neural network is trained using the historical samples as input layer training samples and the historical key features as output layer training samples to obtain the feature extraction model.

16. The method according to claim 5, characterized in that, The formulas for calculating the sample matrix of each cluster include: In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The sample matrix of each cluster, For the first j The first sample in each cluster. For the first j The second sample in the cluster. For the first j The th cluster i One sample, For the first j The th cluster One sample, T This is a transpose.

17. The method according to claim 6, characterized in that, The formula for calculating the feature mean vector of each cluster includes: In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The feature mean vector of each cluster For the first j The th cluster i One sample.

18. The method according to claim 6, characterized in that, The formulas for calculating the centralized data of each cluster include: In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j Centralized data of each cluster For the first j The th cluster i One sample, For the first j The feature mean vector of each cluster.

19. The method according to claim 6, characterized in that, The formulas for calculating the covariance matrix of each cluster include: In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The covariance matrix of each cluster, For the first j Centralized data of each cluster T This is a transpose.

20. The method according to claim 5, characterized in that, The formula for calculating the projection matrix includes: The formula for calculating the dimensionality-reduced feature vector includes: In the above formula, , M The total number of clusters; , For the first j The total number of samples in each cluster; For the first j The projection matrix of each cluster, P For the preset quantity, For the first j The first eigenvalue in the projection matrix of each cluster. For the first j The first eigenvalue in the projection matrix of each cluster; For the first j The th cluster i The dimensionality-reduced feature vector of each sample. T For transpose, For the first j The th cluster i One sample.

21. A device for detecting abnormal operation and maintenance data of multi-source heterogeneous charging equipment, characterized in that, include: The data acquisition unit is used to collect multi-source heterogeneous operation and maintenance data of the charging equipment to be tested. The detection unit is used to identify abnormal data based on the multi-source heterogeneous operation and maintenance data to be detected, using a pre-established anomaly detection model. The determining unit is used to determine the fault information of the charging device based on the abnormal data according to preset rules; The anomaly detection model is established using historical multi-source heterogeneous operation and maintenance data of the charging equipment.

22. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment as described in any one of claims 1 to 20 is implemented.

23. A readable storage medium, characterized in that, It contains an execution program, which, when executed, implements the method for detecting anomalies in multi-source heterogeneous operation and maintenance data of charging equipment as described in any one of claims 1 to 20.

Images