Charging abnormity identification method and system for electric vehicle charging equipment, and medium

Abstract

The invention provides an electric vehicle charging equipment charging abnormity identification method and system and a medium. The method comprises the following steps: acquiring the operation state of charging equipment; reasoning a charging state and a defect reason probability based on the operation state of the charging equipment in combination with a pre-constructed Bayesian reasoning model to obtain a defect reason corresponding to the charging abnormality; wherein the Bayesian reasoning model is constructed by adopting a semantic recognition model to recognize text information on the basis of historical operation state data of the charging equipment, determining a reasoning model sample library through wavelet packet change and distance correlation analysis, and performing training on the basis of the reasoning model sample library. The semantic recognition model is adopted to recognize the text information, the text information is fully considered, and the charging abnormity reason in the charging process is accurately recognized.

Description

Technical Field

[0001] This application relates to the field of abnormal diagnosis technology for charging equipment, specifically to a method, system, and medium for identifying charging abnormalities in electric vehicle charging equipment. Background Technology

[0002] With the rapid development of the new energy vehicle industry, both its ownership and charging demand are experiencing sustained high-speed growth. As a core infrastructure for new energy vehicles, the large-scale deployment and high-frequency use of charging networks present multi-dimensional challenges: On the one hand, user charging needs exhibit significant differences, while charging equipment, due to intergenerational technological iterations, long-term high-frequency use, and widespread compatibility and adaptation failures, as well as hardware aging, ultimately manifests as abnormal charging processes on the user side. Typical abnormal phenomena include "unable to start charging" (such as the device not responding to the start command, or the vehicle-to-pile communication timeout), "abnormal charging shutdown" (such as unexplained interruption during charging, or the protection mechanism mistakenly triggering a shutdown), and "slow charging" (such as the actual output power being lower than the nominal value, or the charging time significantly exceeding expectations). These problems not only directly affect user charging efficiency but also exacerbate the "range anxiety" of new energy vehicle users due to "unreliable charging," becoming a key bottleneck restricting the improvement of the charging network experience.

[0003] On the other hand, the current operation and maintenance management of charging equipment anomalies still relies on the traditional model: the equipment edge only passively records fault information through operation parameter monitoring, and the anomaly diagnosis is mainly based on the preset fault code / shutdown code mapping table. However, this method has three limitations: (1) The information dimension is single, and it can only identify the explicit faults that have been preset and coded. It cannot fully enumerate the implicit anomalies reported by users subjectively; (2) The text information is fragmented. The charging evaluations submitted by users through APP / customer service hotline and other channels (such as "no response after charging for half an hour" or "sudden power outage in the middle") are not associated with the equipment operation logs and fault code records, and a large number of potential fault clues have not been discovered; (3) The platform is not intelligent enough. The existing system mostly uses rule engines to perform simple classification and statistics on fixed fault codes. It lacks the ability to understand the semantics and probabilistic association of multi-source heterogeneous data, making it difficult to accurately locate the root cause from complex phenomena. This results in low fault diagnosis accuracy and slow response time, which cannot meet the "efficient, accurate and proactive" operation and maintenance needs under the large-scale development of charging networks.

[0004] In summary, existing technologies cannot efficiently, accurately, and proactively identify and analyze anomalies such as charging failures and slow charging during the charging process. Summary of the Invention

[0005] To address the problem that existing technologies cannot efficiently, accurately, and proactively identify charging failures and slow charging anomalies during the charging process, this application proposes a method for identifying charging anomalies in electric vehicle charging equipment, including: Obtain the operating status of the charging equipment; Based on the operating status of the charging equipment and a pre-built Bayesian inference model, the charging status and the probability of the cause of the defect are inferred, and the cause of the defect corresponding to the charging anomaly is obtained. Among them, the Bayesian inference model is constructed based on the historical operating status data of charging equipment, by determining the sample library of the inference model through wavelet packet change and distance correlation analysis, and combined with prior probability.

[0006] Optionally, the construction of the Bayesian inference model includes: Acquire user review texts, multi-source data from charging equipment operation monitoring, and waveform data to construct a charging equipment operation status dataset; Based on semantic model recognition of customer service record text information, by controlling generation parameters and prompt words, and by combining the entity relationship between the site and the complaint time with the charging order information of the corresponding time period, defect samples are generated. Wavelet packet transform is used to reconstruct the defect sample to obtain the reconstructed signal. Distance correlation analysis is then used to analyze the relationship between the reconstructed signal and the cause of the defect, and the defect features corresponding to the cause of the defect are extracted. Based on the defect characteristics and multi-source data corresponding to the defect causes, select state variables that characterize the operating status of the charging equipment. The conditional probability of the cause of the defect is determined based on the defect samples corresponding to the historical causes of defects and the charging equipment operating status dataset. Based on the state variables, defect causes, and conditional probabilities of defect causes, a sample library of Bayesian inference models is constructed. The Bayesian inference model is obtained by training the model based on the Bayesian inference model sample library.

[0007] Optionally, the step of reconstructing the defect sample using wavelet packet transform to obtain a reconstructed signal, and then using distance correlation analysis to analyze the correlation between the reconstructed signal and the defect cause, and extracting the defect features corresponding to the defect cause, includes: Wavelet packet decomposition is performed on the defect sample, and the decomposition coefficients are used to reconstruct the defect sample into a single branch to obtain the reconstructed signal. The reconstructed signal has the same length as the original signal and contains characteristic information of different frequency bands. The energy value of the node is obtained by performing a sum of squares processing on the signal of each node. A time lever is added to the energy value, the change of energy over time is observed, and the energy distance is calculated for the signal after the last layer of reconstruction, forming a feature vector of the energy distance. The feature vector of the energy distance is used as the factor index value, and the correlation coefficient between the factor index value and the failure rate is calculated using the distance correlation algorithm. Based on the correlation coefficient between factor index values ​​and failure rates, the defect characteristics corresponding to the causes of equipment failures are determined.

[0008] Optionally, the energy distance is calculated using the following formula:

[0009] In the formula, For energy distance, x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the middle, S ij To reconstruct the signal, i is the layer number obtained from the wavelet packet operation, j is the node number, t is the time of the signal, and k is the position of the signal in that node.

[0010] Optionally, determining the defect characteristics corresponding to the causes of equipment failure based on the correlation coefficient between factor index values ​​and failure rates includes: Select the correlation coefficients between factor index values ​​and failure rates that are greater than a set threshold, and determine the defect characteristics corresponding to the selected correlation coefficients as failure factors.

[0011] Optionally, the step of inferring the charging status and the probability of defect causes based on the charging equipment's operating status and a pre-built Bayesian inference model, and deriving the defect causes corresponding to the charging anomalies, includes: The charging device's operating status is matched with the inference model sample library in the pre-built Bayesian inference model, and the sample with the highest matching degree is used as the matching sample. Based on the conditional probability in the matched samples and the prior probability in the pre-built Bayesian inference model, calculate the probability of the cause of each defect category. The defect category corresponding to the maximum probability of the defect cause is taken as the defect cause corresponding to the charging anomaly.

[0012] Optionally, it also includes: generating a station equipment operation and maintenance strategy based on the defect cause corresponding to the charging anomaly.

[0013] Furthermore, the present invention also provides a charging anomaly identification system for electric vehicle charging equipment, comprising: The status acquisition module is used to acquire the operating status of the charging equipment; The defect identification module is used to infer the charging status and the probability of defect causes based on the charging equipment's operating status and a pre-built Bayesian inference model, and to determine the defect causes corresponding to charging anomalies. Among them, the Bayesian inference model is based on the historical operating status data of the charging equipment. It extracts the defect features corresponding to the defect causes through wavelet packet change and distance correlation analysis, and is constructed by the correlation between the defect causes and the operating status of the charging equipment.

[0014] Optionally, it further includes: a model building module; the model building module includes: The data collection submodule is used to acquire user review text, multi-source data on charging equipment operation monitoring, and waveform data to build a dataset of charging equipment operating status. The semantic transformation submodule is used to identify customer service record text information based on semantic models. By controlling the generation parameters and prompt words, it uses the entity relationship between the site and the complaint time and combines the charging order information of the corresponding time period to generate defect samples. The feature analysis submodule is used to reconstruct the defect sample using wavelet packet transform to obtain the reconstructed signal, and to use distance correlation analysis to analyze the relationship between the reconstructed signal and the defect cause to extract the defect features corresponding to the defect cause. The feature correspondence submodule is used to select state variables that characterize the operating status of the charging equipment based on the defect features corresponding to the defect causes and multi-source data. The conditional probability calculation submodule is used to determine the conditional probability of the defect cause based on the defect samples corresponding to historical defect causes and the charging equipment operating status dataset; The model learning submodule is used to construct an inference model sample library based on state variables, defect causes, and conditional probabilities of defect causes; and to learn a Bayesian inference model using data from the inference model sample library.

[0015] Optionally, the feature analysis submodule is specifically used for: Wavelet packet decomposition is performed on the defect sample, and the decomposition coefficients are used to reconstruct the defect sample into a single branch to obtain the reconstructed signal. The reconstructed signal has the same length as the original signal and contains characteristic information of different frequency bands. The energy value of the node is obtained by performing a sum of squares processing on the signal of each node. By adding a time lever to the energy value, the change of energy over time is observed, and the energy distance is used as a characteristic value to represent different analyzed signals. The energy distance is calculated from the reconstructed signal of the last layer, forming the eigenvector of the energy distance. The feature vector of the energy distance is used as the factor index value, and the correlation coefficient between the factor index value and the failure rate is calculated using the distance correlation algorithm. Based on the correlation coefficient between factor index values ​​and failure rates, the defect characteristics corresponding to the causes of equipment failures are determined.

[0016] Optionally, the energy distance is calculated using the following formula:

[0017] In the formula, For energy distance, x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the middle, S ij To reconstruct the signal, i is the layer number obtained from the wavelet packet operation, j is the node number, t is the time of the signal, and k is the position of the signal in that node.

[0018] In another aspect, this application also provides an electronic device, 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, a method for identifying charging anomalies in electric vehicle charging equipment as described above is implemented.

[0019] Furthermore, this application also provides a readable storage medium having an executable program stored thereon, which, when executed, implements the above-described method for identifying charging anomalies in electric vehicle charging equipment.

[0020] Compared with the prior art, the beneficial effects of this application are as follows: This application provides a method for identifying charging anomalies in electric vehicle charging equipment, comprising: acquiring the operating status of the charging equipment; based on the operating status of the charging equipment and a pre-constructed Bayesian inference model, inferring the charging status and the probability of defect causes, and deriving the defect causes corresponding to the charging anomalies; wherein, the Bayesian inference model is constructed based on historical operating status data of the charging equipment, using a semantic recognition model to identify text information, and determining the inference model sample library through wavelet packet change and distance correlation analysis, combined with prior probabilities. This invention uses a semantic recognition model to identify text information, fully considering text information, and accurately identifying the causes of charging anomalies during the charging process. Attached Figure Description

[0021] Figure 1 This is a flowchart of a charging anomaly identification method for an electric vehicle charging device according to this application. Figure 2 This is a schematic diagram of the three-layer wavelet packet decomposition in this application; Figure 3 This is a schematic diagram of an electronic device structure according to this application. Detailed Implementation

[0022] This application proposes a method for identifying charging anomalies in electric vehicle charging equipment. It uses a semantic model to identify user review text information and generate sample data for an inference model. Based on a hierarchical logic of "charging state - feature variables - defect cause," a Bayesian inference model is constructed to accurately determine different charging states and locate defect causes, providing technical support for intelligent operation and maintenance.

[0023] To better understand this application, the content of this application will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] Example 1: A method for identifying charging anomalies in electric vehicle charging equipment, such as Figure 1 As shown, it includes: Step S1: Obtain the operating status of the charging device; Step S2: Based on the operating status of the charging equipment and the pre-built Bayesian inference model, infer the charging status and the probability of the cause of the defect, and obtain the cause of the defect corresponding to the charging anomaly. Among them, the Bayesian inference model is constructed based on the historical operating status data of charging equipment, by determining the sample library of the inference model through wavelet packet change and distance correlation analysis, and combined with prior probability.

[0025] The following is a further description of a method for identifying charging anomalies in electric vehicle charging equipment provided by the present invention, specifically including: Before step S1, the process of constructing a Bayesian inference model is also included. The construction process of the Bayesian inference model is further described below: The construction of a Bayesian inference model includes: Acquire user review texts, multi-source data from charging equipment operation monitoring, and waveform data to construct a charging equipment operation status dataset; Based on semantic model recognition of customer service record text information, by controlling generation parameters and prompt words, and by combining the entity relationship between the site and the complaint time with the charging order information of the corresponding time period, defect samples are generated. Wavelet packet transform is used to reconstruct the defect sample to obtain the reconstructed signal. Distance correlation analysis is then used to analyze the relationship between the reconstructed signal and the cause of the defect, and the defect features corresponding to the cause of the defect are extracted. Based on the defect characteristics and multi-source data corresponding to the defect causes, select state variables that characterize the operating status of the charging equipment. The conditional probability of the cause of the defect is determined based on the defect samples corresponding to the historical causes of defects and the charging equipment operating status dataset. Based on the state variables, defect causes, and conditional probabilities of defect causes, a sample library for inference models is constructed. The Bayesian inference model is obtained by training the model based on the Bayesian inference model sample library.

[0026] The construction of a Bayesian inference model specifically includes: Step 1: Obtain user review text, multi-source data from charging equipment operation monitoring, and waveform recording data through standard interfaces to construct a charging equipment operation status dataset. The multi-source data includes charging order information, voltage, current, and power during charging, as well as millisecond-level waveform recording data; user review text; ambient temperature and humidity; season; charging equipment operating years; manufacturer information; and operating location. The charging order information includes transaction serial number, charging equipment number, vehicle VIN code, reason for transaction termination, charging start and end times, and charging amount.

[0027] Charging status is divided into three types: normal charging, abnormal charging termination, and slow charging. Abnormal charging termination (i.e., defect classification) includes BRM message reception timeout, insulation defect, and abnormal power metering.

[0028] By combining actual charging phenomena, causes of defects, and operation and maintenance guidelines, a knowledge base for handling typical defects is compiled.

[0029] Step 2: Based on the semantic model, identify customer service record text information, control the generation parameters and prompt words, and use the entity relationship between the site and the complaint time to combine the charging order information of the corresponding time period to generate defect samples including charging device number, charging start time, end time, and voltage and current of 0, and store them in the sample database of the probability model.

[0030] Step 3: For charging anomalies (i.e., defect samples) occurring during the charging process of the charging equipment, the fundamental wave and harmonic signals are reconstructed using wavelet packet transform. Distance correlation analysis is then used to analyze the relationship between the harmonics and the defect type, extracting defect features. These features include the recorded fundamental wave and recorded harmonics, specifically the fundamental current wave and the amplitude of the 5th harmonic.

[0031] Step 4: Based on the historical defects of the charging equipment, charging orders, and waveform recording data, select feature quantities that characterize the operating status of the charging equipment and construct a feature set. A collection of actual charging phenomena Defect Cause Classification Set Based on historical defect data statistics, determine the conditional probability of the corresponding defect cause classification under the given characteristics. ;in, This represents 1, 2, ..., n state categories. Representing 1, 2, ..., k state variables. This represents 1, 2, ..., g categories of defect causes. This represents the category of the i-th defect cause; Based on the dataset (i.e., feature set Z, actual charging phenomenon classification set Y, defect cause classification set C, and conditional probability P) obtained from the aforementioned steps and the extracted state variables, a sample library for the inference model is constructed. State variables include charging voltage, charging current, charging power, fundamental frequency of recorded waveform, harmonic frequency of recorded waveform, ambient temperature and humidity, seasonal factors, charging amount per unit time, number of charging interruptions, rated power of charging pile, maximum current of charging pile, operating years of charging pile, vehicle manufacturer, charging pile manufacturer, and operating location. For continuous variables (such as voltage and current), discretization is performed based on business thresholds, transforming them into finite-state categorical variables, i.e., normal or abnormal states. For operating years, discretization is based on whether it exceeds a set threshold. Vehicle manufacturers are divided into A (mainstream manufacturers and other manufacturers), and charging pile manufacturers are divided into B (mainstream manufacturers and other manufacturers). Seasonal factors are discretized based on whether it occurs during a high-temperature season, and operating locations are discretized based on whether they occur in a high-temperature region.

[0032] The Bayesian inference model categorizes actual charging phenomena as Y = {normal charging, BRM message reception timeout, insulation defect, abnormal power metering, and slow charging}. The defect cause classifications are shown in Table 1. The conditional probabilities of the corresponding defect cause classifications under discrete features are also presented. The value is calculated by statistically analyzing the frequency of the feature (i.e., the state variable) in the sample dataset.

[0033] In the formula, Let be the conditional probability of classifying the corresponding defect cause under two state variable features. The defect is caused by c. i The feature quantity under the classification is the number of samples Z1.

[0034] Table 1. Typical Defect Classifications and Causes

[0035] Step 5: Construct a Bayesian inference model. Given the feature quantities (i.e., state quantities), infer the charging state and the probability of the cause of the defect, and derive the cause of the defect corresponding to the charging anomaly.

[0036] In examining evidence The calculation method is as follows:

[0037] In the formula, α is the regularization factor. This represents the prior probability of the charging phenomenon occurring. For class The posterior probability, i.e., the class probability revised after obtaining certain information. The probability of occurrence. According to the Bayesian maximum a posteriori criterion, the Bayesian network classifier selects the classifier that maximizes the posterior probability. Largest class For class tags.

[0038] Specifically, the reasoning process of this Bayesian inference model includes: When multiple state variables are input, these state variables are matched with samples in the inference model's sample library. The sample with the highest matching degree is taken as the matched sample. Then, the conditional probability of the corresponding defect cause classification (the defect cause determined in the matched sample) under discrete features is calculated. The posterior probability is calculated from the prior probability and the conditional probability. The class with the highest posterior probability is then selected. For class labels (i.e. defect reasons).

[0039] Step 6: Generate a maintenance and repair strategy for the station equipment based on the cause of the charging anomaly, and output the strategy. This maintenance and repair strategy is a fixed measure formed manually according to experience and procedures, as shown in Table 2.

[0040] Table 2

[0041] Wavelet analysis principle of recording.

[0042] When analyzing defect features using wavelet packet operations, wavelet packet decomposition and reconstruction are required. The signal S is decomposed into i-th layer wavelet packets, and the decomposition coefficients X of the j-th node in the i-th layer are used. ij k S is obtained by performing single-branch reconstruction on the signal. ij After reconstruction, the signals of each node in the same layer have the same length. The reconstruction algorithm follows the principles of initialization, iteration, and termination. The signal of the previous layer is reconstructed from the filtered signal of the next layer, as shown in equation (1): (1) In the formula, h 0k and h 1k These are the low-pass and high-pass filter coefficients for wavelet packet reconstruction, respectively, and k is the signal position at that node. This refers to the signal component at position k in the (i+1)th layer and the 2jth node. The signal component at position k is located at the (i+1)th layer and the (2j+1)th node; the analyzed signal S is the reconstructed signal S of all nodes in this layer. ij Summing yields the result shown in equation (2): (2) The reconstructed signal has the same length as the original signal and contains characteristic information from different frequency bands. By performing a sum-of-squares operation on the signal at each node, the energy level of that node can be obtained. A time lever is then applied to the energy value to observe its change over time. This energy distance is used as a characteristic value to represent different analyzed signals. The formula is shown in equation (3): (3) In the formula x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the signal are calculated. Multi-level wavelet packet decomposition is performed on the current and voltage signals. The energy distance is calculated for the reconstructed signal from the last layer, forming the eigenvector of the energy distance. This embodiment uses three-level wavelet packet decomposition as an example. Figure 2 As shown, S is the reconstructed signal, A1 is the low-frequency component, D1 is the high-frequency component, AA2 is the coarser-grained low-frequency signal; DA2 is the mid-frequency detail signal, AAA3 is the lowest frequency backbone signal; DAA3 is the lower-frequency signal; ADA3 is the mid-frequency signal after the first approximation, second detail, and third approximation; DDA3 is the high-frequency and mid-frequency mixed signal obtained from the first two detail levels and the third approximation; AD2 is the mid-high frequency signal; DD2 is the ultra-high frequency signal; AAD3 is the high-frequency signal in the lowest frequency band; DAD3 is the specific transition frequency band signal obtained from the first detail level, second approximation, and third detail level; ADD3 is the mid-high frequency band signal obtained from the first approximation and the last two detail levels; DDD3 is the highest frequency component signal obtained from all three levels of detail.

[0043] The principle of distance correlation: The eigenvector of energy distance is used as the factor index value. A distance correlation algorithm is used to calculate the correlation coefficient between the factor index value and the failure rate. This correlation characterizes the relationship between the factor and the failure. A correlation ranking algorithm is then used to rank the correlation coefficients between multiple factors and the failure, ultimately determining the factors contributing to equipment failure.

[0044] Distance correlation is one method for calculating the correlation coefficient. The basic formula is: Where X is the sequence of the mean values ​​of each sub-interval of the factor index, and Y is the sequence of the failure rate. This represents the distance covariance between the factor index and the failure rate. ; This represents the covariance between the factor index and the failure rate. ,in, The square of the sample distance covariance. Let n be the square of the sample distance variance, and n be the sample size. For the distance matrix elements related to Y, This is the square of the sample distance variance. For distance matrix elements related to X, , This represents the mean of the samples in the j-th row. This represents the mean of the samples in the k-th row. The distance matrix is ​​the sample mean. Let dCor(X,Y) represent the Euclidean distance between the factor indicators, and k represent the position of the signal at that node. Distance correlation has the following properties: 0 ≤ dCor(X,Y) ≤ 1. When dCor(X,Y) = 0, X and Y are independent; when dCor(X,Y) = 1, X and Y are perfectly correlated.

[0045] The correlation ranking algorithm arranges the coefficients of variation obtained from the correlation between quantitative factors and faults in descending order, transforming a set of disordered sequences into an ordered sequence, and thus obtaining a sequence of the strength of correlation between factors and faults. Defect features with a correlation coefficient greater than 0.8 are identified as fault factors.

[0046] The beneficial effects of this invention are: This invention identifies textual data from user feedback of unsuccessful charging through semantic models and combines it with current structured data to provide more comprehensive data for evaluating the charging operation status.

[0047] This invention adds millisecond-level data sources to the existing technology, introduces waveform data of the vehicle-to-charging process, and uses wavelet analysis to identify abnormal situations in the transient charging process.

[0048] Taking into account charging anomalies, voltage, current, harmonics, season, environment, and regional data, a Bayesian inference algorithm is used to calculate the probability of fault categories under different feature quantities. The probabilities of fault categories under different feature quantities are updated based on new operational evidence, thereby identifying the causes of defects and meeting the diagnostic needs of multiple overlapping problems in actual operation and maintenance. Simultaneously, the algorithm automatically adjusts the prior probability indicators of equipment faults in the model based on new detection data and annotation conclusions, ensuring the accuracy of fault classification.

[0049] Example 2: Based on the same inventive concept, this application also provides a charging anomaly identification system for electric vehicle charging equipment, comprising: The status acquisition module is used to acquire the operating status of the charging equipment; The defect identification module is used to infer the charging status and the probability of defect causes based on the charging equipment's operating status and a pre-built Bayesian inference model, and to determine the defect causes corresponding to charging anomalies. Among them, the Bayesian inference model is based on the historical operating status data of the charging equipment. It extracts the defect features corresponding to the defect causes through wavelet packet change and distance correlation analysis, and is constructed by the correlation between the defect causes and the operating status of the charging equipment.

[0050] Furthermore, it also includes: a model building module; the model building module includes: The data collection submodule is used to acquire user review text, multi-source data on charging equipment operation monitoring, and waveform data to build a dataset of charging equipment operating status. The semantic transformation submodule is used to identify customer service record text information based on semantic models. By controlling the generation parameters and prompt words, it uses the entity relationship between the site and the complaint time and combines the charging order information of the corresponding time period to generate defect samples. The feature analysis submodule is used to reconstruct the defect sample using wavelet packet transform to obtain the reconstructed signal, and to use distance correlation analysis to analyze the relationship between the reconstructed signal and the defect cause to extract the defect features corresponding to the defect cause. The feature correspondence submodule is used to select state variables that characterize the operating status of the charging equipment based on the defect features corresponding to the defect causes and multi-source data. The conditional probability calculation submodule is used to determine the conditional probability of the defect cause based on the defect samples corresponding to historical defect causes and the charging equipment operating status dataset; The model learning submodule is used to construct an inference model sample library based on state variables, defect causes, and conditional probabilities of defect causes; and to learn a Bayesian inference model using data from the inference model sample library.

[0051] Furthermore, the feature analysis submodule is specifically used for: Wavelet packet decomposition is performed on the defect sample, and the decomposition coefficients are used to reconstruct the defect sample into a single branch to obtain the reconstructed signal. The reconstructed signal has the same length as the original signal and contains characteristic information of different frequency bands. The energy value of the node is obtained by performing a sum of squares processing on the signal of each node. By adding a time lever to the energy value, the change of energy over time is observed, and the energy distance is used as a characteristic value to represent different analyzed signals. The energy distance is calculated from the reconstructed signal of the last layer, forming the eigenvector of the energy distance. The feature vector of the energy distance is used as the factor index value, and the correlation coefficient between the factor index value and the failure rate is calculated using the distance correlation algorithm. Based on the correlation coefficient between factor index values ​​and failure rates, the defect characteristics corresponding to the causes of equipment failures are determined.

[0052] Furthermore, the energy distance is calculated using the following formula:

[0053] In the formula, For energy distance, x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the middle,S ij To reconstruct the signal, i is the layer number obtained from the wavelet packet operation, j is the node number, t is the time of the signal, and k is the position of the signal in that node.

[0054] Example 3 like Figure 3 As shown, this application 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.

[0055] The processor may be a Central Processing Unit (CPU), or it may be 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 core and control core of the terminal, and it 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 implement the steps of the electric vehicle charging equipment charging anomaly identification method in the above embodiments.

[0056] Example 4 Based on the same inventive concept, this application 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 electric vehicle charging device charging anomaly identification method described in the above embodiments.

[0057] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application 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.

[0058] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] 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 1 The function specified in one or more boxes.

[0060] 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.

[0061] The above are merely embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application shall be included within the scope of the claims of this application pending approval.

Claims

1. A method for identifying charging anomalies in electric vehicle charging equipment, characterized in that, include: Obtain the operating status of the charging equipment; Based on the operating status of the charging equipment and a pre-built Bayesian inference model, the charging status and the probability of the cause of the defect are inferred, and the cause of the defect corresponding to the charging anomaly is obtained. Among them, the Bayesian inference model is based on the historical operating status data of charging equipment, uses a semantic recognition model to identify text information, and determines the inference model sample library through wavelet packet change and distance correlation analysis, and is trained and constructed based on the inference model sample library.

2. The method as described in claim 1, characterized in that, The construction of the Bayesian inference model includes: Acquire user review texts, multi-source data from charging equipment operation monitoring, and waveform data to construct a charging equipment operation status dataset; Based on semantic model recognition of customer service record text information, by controlling generation parameters and prompt words, and by combining the entity relationship between the site and the complaint time with the charging order information of the corresponding time period, defect samples are generated. Wavelet packet transform is used to reconstruct the defect sample to obtain the reconstructed signal. The correlation between the reconstructed signal and the defect cause is analyzed by distance correlation analysis to extract the defect features corresponding to the defect cause. Based on the defect characteristics and multi-source data corresponding to the defect causes, select state variables that characterize the operating status of the charging equipment. The conditional probability of the cause of the defect is determined based on the defect samples corresponding to the historical causes of defects and the charging equipment operating status dataset. Based on the state variables, defect causes, and conditional probabilities of defect causes, a sample library of Bayesian inference models is constructed. The Bayesian inference model is obtained by training the model based on the Bayesian inference model sample library.

3. The method as described in claim 2, characterized in that, The defect sample is reconstructed using wavelet packet transform to obtain a reconstructed signal. Distance correlation analysis is then used to analyze the relationship between the reconstructed signal and the cause of the defect, and defect features corresponding to the cause of the defect are extracted, including: Wavelet packet decomposition is performed on the defect sample, and the decomposition coefficients are used to reconstruct the defect sample into a single branch to obtain the reconstructed signal. The reconstructed signal has the same length as the original signal and contains characteristic information of different frequency bands. The energy value of the node is obtained by performing a sum of squares processing on the signal of each node. A time lever is added to the energy value, the change of energy over time is observed, and the energy distance is calculated for the signal after the last layer of reconstruction, forming a feature vector of the energy distance. The feature vector of the energy distance is used as the factor index value, and the correlation coefficient between the factor index value and the failure rate is calculated using the distance correlation algorithm. Based on the correlation coefficient between factor index values ​​and failure rates, the defect characteristics corresponding to the causes of equipment failures are determined.

4. The method as described in claim 3, characterized in that, The energy distance is calculated using the following formula: In the formula, For energy distance, x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the middle, S ij To reconstruct the signal, i is the layer number obtained from the wavelet packet operation, j is the node number, t is the time of the signal, and k is the position of the signal in that node.

5. The method as described in claim 3, characterized in that, The method of determining the defect characteristics corresponding to the causes of equipment failure based on the correlation coefficient between factor index values ​​and failure rates includes: Select the correlation coefficients between factor index values ​​and failure rates that are greater than a set threshold, and determine the defect characteristics corresponding to the selected correlation coefficients as failure factors.

6. The method as described in claim 2, characterized in that, The method, based on the charging equipment's operating status and a pre-built Bayesian inference model, infers the charging status and the probability of defect causes, deriving the defect causes corresponding to charging anomalies, including: The charging device's operating status is matched with the inference model sample library in the pre-built Bayesian inference model, and the sample with the highest matching degree is used as the matching sample. Based on the conditional probabilities in the matched samples and the prior probabilities obtained from the pre-built Bayesian inference model, the probability of the cause of each defect category is calculated. The defect category corresponding to the maximum probability of the defect cause is taken as the defect cause corresponding to the charging anomaly.

7. The method as described in claim 1, characterized in that, Also includes: Based on the cause of the defect corresponding to the charging anomaly, generate the operation and maintenance strategy for the station equipment.

8. A charging anomaly detection system for electric vehicle charging equipment, characterized in that, include: The status acquisition module is used to acquire the operating status of the charging equipment; The defect identification module is used to infer the charging status and the probability of defect causes based on the charging equipment's operating status and a pre-built Bayesian inference model, and to determine the defect causes corresponding to charging anomalies. Among them, the Bayesian inference model is based on the historical operating status data of the charging equipment. It extracts the defect features corresponding to the defect causes through wavelet packet change and distance correlation analysis, and is constructed by the correlation between the defect causes and the operating status of the charging equipment.

9. The system as described in claim 8, characterized in that, Also includes: Model building module; The model building module includes: The data collection submodule is used to acquire user review text, multi-source data on charging equipment operation monitoring, and waveform data to build a dataset of charging equipment operating status. The semantic transformation submodule is used to identify customer service record text information based on semantic models. By controlling the generation parameters and prompt words, it uses the entity relationship between the site and the complaint time and combines the charging order information of the corresponding time period to generate defect samples. The feature analysis submodule is used to reconstruct the defect sample using wavelet packet transform to obtain the reconstructed signal, and to use distance correlation analysis to analyze the relationship between the reconstructed signal and the defect cause to extract the defect features corresponding to the defect cause. The feature correspondence submodule is used to select state variables that characterize the operating status of the charging equipment based on the defect features corresponding to the defect causes and multi-source data. The conditional probability calculation submodule is used to determine the conditional probability of the defect cause based on the defect samples corresponding to historical defect causes and the charging equipment operating status dataset; The model learning submodule is used to construct an inference model sample library based on state variables, defect causes, and conditional probabilities of defect causes; and to learn a Bayesian inference model using data from the inference model sample library.

10. The system as described in claim 8, characterized in that, The feature analysis submodule is specifically used for: Wavelet packet decomposition is performed on the defect sample, and the decomposition coefficients are used to reconstruct the defect sample into a single branch to obtain the reconstructed signal. The reconstructed signal has the same length as the original signal and contains characteristic information of different frequency bands. The energy value of the node is obtained by performing a sum of squares processing on the signal of each node. By adding a time lever to the energy value, the change of energy over time is observed, and the energy distance is used as a characteristic value to represent different analyzed signals. The energy distance is calculated from the reconstructed signal of the last layer, forming the eigenvector of the energy distance. The feature vector of the energy distance is used as the factor index value, and the correlation coefficient between the factor index value and the failure rate is calculated using the distance correlation algorithm. Based on the correlation coefficient between factor index values ​​and failure rates, the defect characteristics corresponding to the causes of equipment failures are determined.

11. The system as described in claim 8, characterized in that, The energy distance is calculated using the following formula: In the formula, For energy distance, x ij k Represents the reconstructed signal S ij The values ​​at each discrete point in the middle, S ij To reconstruct the signal, i is the layer number obtained from the wavelet packet operation, j is the node number, t is the time of the signal, and k is the position of the signal in that node.

12. 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, a method for identifying charging anomalies in an electric vehicle charging device as described in any one of claims 1 to 7 is implemented.

13. A readable storage medium, characterized in that, It contains an execution program, which, when executed, implements a method for identifying charging anomalies in electric vehicle charging equipment as described in any one of claims 1 to 7.

Images