A fault diagnosis method, server and computer readable storage medium

Abstract

The present application relates to the technical field of fault diagnosis, and in particular to a fault diagnosis method, a server and a computer readable storage medium, the fault diagnosis method comprising: obtaining target collection data, fault feedback information and charging pile logs, predicting charging pile faults according to the target collection data to obtain fault prediction results, generating at least one fault reasoning link according to the fault feedback information, the charging pile logs and a preset fault reasoning model, and performing fault diagnosis on the charging pile according to the fault prediction results and the at least one fault reasoning link. The embodiment can respond to user feedback fault information in a timely manner and use the fault feedback information for fault reasoning, which is conducive to strengthening the generalization ability of fault diagnosis, thereby realizing more intelligent fault diagnosis. Furthermore, the embodiment verifies the fault reasoning link by using the fault prediction results, which can ensure the reliability of the fault reasoning link and thus the accuracy of the fault reasoning results, thereby improving the accuracy of fault diagnosis.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, specifically to a fault diagnosis method, a server, and a computer-readable storage medium. Background Technology

[0002] Charging piles are devices used to charge new energy vehicles. With the continuous promotion and application of new energy vehicles, the demand for charging piles is also growing rapidly. During use, charging piles often experience malfunctions such as communication abnormalities, current fluctuations, over-temperature protection, and hardware failures. When a charging pile malfunctions, fault diagnosis is necessary to find the root cause. Traditional fault diagnosis methods typically rely on rule-based judgments based on sensor data collected by the device or use traditional machine learning methods for fault classification. However, traditional fault diagnosis methods struggle to respond promptly to user feedback, have weak generalization capabilities, and are difficult to achieve intelligent fault diagnosis. Furthermore, when new or complex faults occur, traditional fault diagnosis methods often fail to provide accurate diagnoses. Summary of the Invention

[0003] One objective of this embodiment is to provide a fault diagnosis method, server, and computer-readable storage medium to address the technical problems of weak generalization ability or low diagnostic accuracy in existing charging pile fault diagnosis methods.

[0004] In a first aspect, embodiments of the present invention provide a fault diagnosis method, comprising:

[0005] Acquire target data, fault feedback information, and charging pile logs;

[0006] Based on the target data collected, charging pile faults are predicted, and fault prediction results are obtained.

[0007] At least one fault inference link is generated based on the fault feedback information, the charging pile log, and the preset fault inference model.

[0008] The charging pile is diagnosed based on the fault prediction results and at least one fault reasoning link.

[0009] Optionally, the step of predicting charging pile faults based on the target collected data and obtaining fault prediction results includes:

[0010] Extract the features from the target data to obtain target feature information;

[0011] Based on the target feature information and the preset fault prediction model, the fault of the charging pile is predicted, and the fault prediction result is obtained.

[0012] Optionally, the preset fault prediction model includes an anomaly detection sub-model and a fault type identification sub-model. The step of predicting charging pile faults based on the target feature information and the preset fault prediction model to obtain the fault prediction result includes:

[0013] The target feature information is input into the anomaly detection sub-model to obtain anomaly detection results and anomaly detection feature information, wherein the anomaly detection feature information includes anomaly detection labels.

[0014] The anomaly detection feature information is input into the fault type identification sub-model to obtain the fault type identification result;

[0015] The fault type identification result is verified based on the anomaly detection result to obtain the fault prediction result.

[0016] Optionally, the fault diagnosis of the charging pile based on the fault prediction result and at least one fault reasoning link includes:

[0017] Extract the fault information of each fault inference link to obtain the candidate fault information corresponding to each fault inference link;

[0018] The charging pile is diagnosed based on the fault prediction results and the candidate fault information.

[0019] Optionally, the step of diagnosing the charging pile based on the fault prediction result and the candidate fault information includes:

[0020] Based on the fault prediction results, target fault information is selected from the candidate fault information;

[0021] Determine the target fault inference link based on the target fault information;

[0022] Obtain the fault reasoning results of the target fault reasoning link;

[0023] The charging pile is diagnosed based on the fault reasoning results.

[0024] Optionally, the fault prediction result includes anomaly detection results and fault type identification results, each candidate fault information includes first type fault information and / or second type fault information, and the step of filtering target fault information from the candidate fault information according to the fault prediction result includes:

[0025] The anomaly detection results are matched with the first type of fault information and / or the second type of fault information to obtain a first matching degree;

[0026] The fault type identification result is matched with the first type of fault information and / or the second type of fault information to obtain the second matching degree;

[0027] Target fault information is selected from the candidate fault information based on the first matching degree and the second matching degree.

[0028] Optionally, the step of filtering target fault information from the candidate fault information based on the first matching degree and the second matching degree includes:

[0029] Based on the first matching degree and the second matching degree, determine whether there is a completely matching and unique candidate fault information;

[0030] If so, then the candidate fault information is determined to be the target fault information;

[0031] If not, then calculate the comprehensive matching degree based on the first matching degree and the second matching degree, and filter out the target fault information from the candidate fault information based on the comprehensive matching degree.

[0032] Optionally, the fault diagnosis method further includes:

[0033] Acquire historical unstructured data, which includes a first part of data and a second part of data, wherein the first part of data and the second part of data include fault type labels;

[0034] Obtain structured sample data obtained by manually annotating the first part of the data;

[0035] The structured sample data and the second part of the data are input into the third model to obtain the generalized corpus;

[0036] The fourth model is fine-tuned based on the generalized corpus to obtain a preset fault reasoning model, wherein the fourth model is a sub-model of the third model.

[0037] In a second aspect, embodiments of the present invention provide a server, including a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, wherein when the processor executes the one or more computer programs, the server implements the fault diagnosis method as described above.

[0038] In a third aspect, embodiments of the present invention provide a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the fault diagnosis method as described above.

[0039] Compared with existing technologies, this invention provides a fault diagnosis method, a server, and a computer-readable storage medium. The fault diagnosis method includes: acquiring target data, fault feedback information, and charging pile logs; predicting charging pile faults based on the target data to obtain fault prediction results; generating at least one fault inference link based on the fault feedback information, charging pile logs, and a preset fault inference model; and performing fault diagnosis on the charging pile based on the fault prediction results and at least one fault inference link. This embodiment can respond promptly to fault information reported by users and use the fault feedback information for fault inference, which is beneficial to enhancing the generalization ability of fault diagnosis, thereby achieving more intelligent fault diagnosis. Furthermore, this embodiment can use the fault prediction results to verify the fault inference link with strong chain inference capabilities, ensuring the reliability of the fault inference link, thereby ensuring the accuracy of the fault inference results and improving the accuracy of fault diagnosis. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention 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.

[0041] Figure 1 This is a schematic diagram of a fault diagnosis scenario provided by an embodiment of the present invention;

[0042] Figure 2 This is a flowchart illustrating a fault diagnosis method provided in an embodiment of the present invention;

[0043] Figure 3 This is a flowchart illustrating step S22 of a fault diagnosis method provided in an embodiment of the present invention.

[0044] Figure 4 This is a flowchart illustrating step S222 of a fault diagnosis method provided in an embodiment of the present invention.

[0045] Figure 5 This is a flowchart illustrating step S24 of a fault diagnosis method provided in an embodiment of the present invention.

[0046] Figure 6 This is a schematic diagram of the structure of a fault diagnosis device provided in an embodiment of the present invention;

[0047] Figure 7 This is a schematic diagram of the structure of a prediction module in a fault diagnosis device provided in an embodiment of the present invention;

[0048] Figure 8This is a schematic diagram of the structure of a diagnostic module in a fault diagnosis device provided in an embodiment of the present invention;

[0049] Figure 9 This is a schematic diagram of the structure of a fault diagnosis device provided in another embodiment of the present invention;

[0050] Figure 10 This is a schematic diagram of the hardware structure of a server provided in an embodiment of the present invention. Detailed Implementation

[0051] To facilitate understanding of the present invention, a more detailed description is provided below with reference to the accompanying drawings and specific embodiments. It should be noted that when an element is described as "connected" to another element, it can be directly connected to the other element, or there may be one or more intervening elements. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0053] Please see Figure 1 This invention provides a fault diagnosis scenario. For example... Figure 1 As shown, the fault diagnosis scenarios include charging pile 100 and server 200.

[0054] A charging pile 100 is a device used to charge electric vehicles to replenish their power. Its working principle is to receive electrical energy from the power grid and then transmit that energy to the electric vehicle via a charging cable to charge it. In some embodiments, the charging pile 100 can be any type of charging pile, such as a DC charging pile, an AC charging pile, or an AC / DC integrated charging pile.

[0055] DC charging stations use direct current (DC) to charge the power batteries of electric vehicles; this charging method is also known as "fast charging." DC charging stations are connected to the power grid, receiving three-phase 380V AC power from the grid and converting it to DC. This DC power is then delivered to the electric vehicle's power battery through a standard DC charging plug and socket, thus achieving DC charging. The inherent power supply characteristics of DC charging stations allow them to output sufficient charging power, with a wide range of voltage and current adjustment, enabling rapid charging. DC charging stations also function as chargers, capable of monitoring and controlling the operating status of the battery being charged in real time, and can also measure the amount of electricity charged.

[0056] AC charging stations can provide power to electric vehicles using single or dual 220VAC / 380VAC AC output interfaces, allowing them to use onboard chargers to charge their batteries. This charging method is also known as "slow charging." The output power of an AC charging station is typically 5kW (220VAC) / 20kW (380VAC), but the actual charging power is limited by the onboard charger; generally, the onboard charging power for small electric vehicles is between 2 and 3kW. The onboard charger for electric vehicles converts AC power to DC power through filtering and rectification, then stores the DC power in the electric vehicle's battery, thus charging the vehicle. This charging method is mainly used in small pure electric vehicles.

[0057] The input voltage of an AC / DC integrated charging pile is generally three-phase four-wire 380VAC±15% at a frequency of 50Hz. It includes a DC output port and an AC output port. The DC output port outputs adjustable DC power to charge the electric vehicle's battery, with a charging power typically ranging from 10 to 40kW. The AC output port outputs 220VAC (5kW) / 380VAC (20kW) AC power to provide charging power to the electric vehicle's onboard charger. The AC / DC integrated charging pile can provide conventional charging via the AC output port and fast charging via the DC output port. During peak charging hours in the daytime, fast charging is used; at night, when there are fewer users, conventional charging is used for slower charging. The AC / DC integrated charging pile can achieve simultaneous AC and DC charging and interlocked charging. Its modular design facilitates maintenance.

[0058] In some embodiments, the charging pile 100 is configured with one or more charging guns. The charging gun is an interface device connecting the charging pile and the electric vehicle, primarily used for charging the electric vehicle by transmitting electrical energy. The charging gun typically has a plug and a connecting cable, one end of which connects to the charging pile, and the other end is inserted into the charging port of the electric vehicle. Depending on different charging requirements and technical standards, charging guns can be of different types; in some embodiments, charging guns include fast-charging guns and slow-charging guns.

[0059] Fast charging guns, also known as DC fast charging guns, are typically used at fast charging stations. They have a large power output and can quickly charge the vehicle's power battery.

[0060] Slow charging guns, also known as AC charging guns, are typically used in home charging stations, commercial charging stations, and public charging stations. They have lower power and are suitable for charging with ordinary household power supplies, resulting in a relatively slow charging speed.

[0061] Server 200 is communicatively connected to charging pile 100 and is used to monitor and manage the charging pile network, process data from charging equipment, and ensure the stable and efficient operation of the charging service. Specifically, the charging pile server achieves real-time monitoring, data analysis, and fault handling of the charging pile's operating status through remote connection, ensuring the continuity and efficiency of the charging service. In some embodiments, server 200 can be a single physical server or a logical server virtualized from multiple physical servers. In some embodiments, server 200 can also be a server cluster composed of multiple interconnected servers, with each functional module distributed across different servers within the server cluster.

[0062] In some embodiments, server 200 may include an edge server or a cloud server. An edge server is a hardware or software system deployed close to a data source (such as a charging station) and possesses data storage, computing, and processing capabilities. The edge server processes data and provides services at the network edge without relying entirely on a remote cloud server. The cloud server is used to centrally manage the charging of multiple charging piles 100, including charging control, order data, and billing services. Server 200 can also function as an intelligent charging pile fault diagnosis system, performing the fault diagnosis methods described below.

[0063] This invention provides a fault diagnosis method; please refer to [link / reference]. Figure 2 Fault diagnosis methods include:

[0064] S21. Obtain target data, fault feedback information, and charging pile logs.

[0065] In this step, the target data collected includes any data used for fault diagnosis. The target data collected can include charging pile-side data and server-side data. Charging pile-side data refers to data collected by the charging pile through IoT (Internet of Things) devices, sensors, smart meters, and other devices or equipment. Server-side data can include logs stored on the server. In some embodiments, the target data collected can include equipment operation data, communication and protocol data, hardware status data, user interaction behavior data, environmental and geographical data, historical fault data, and maintenance records. Equipment operation data can include multi-dimensional operation data such as power parameters, environmental parameters, equipment status data, user operation logs, and charging transaction logs. Power parameters can include parameters such as voltage, current, or power; environmental parameters can include parameters such as temperature and humidity; and equipment status data can include charging gun connection status data, module temperature data, or communication status data.

[0066] Communication and protocol data may include BMS (Battery Management System) communication logs, message content, communication interruption logs, protocol versions, and other data.

[0067] Hardware status data may include interface status data (obtained through plug-in detection), module health status, meter readings, fan / heatsink operating status data, and switch status data, where switches include, but are not limited to, contactors and relays.

[0068] User interaction behavior data can include startup method, startup / stop time, user operation logs, abnormal click data, etc.

[0069] Environmental and geographic data can include ambient temperature, ambient humidity, etc.

[0070] Historical fault data and maintenance records can include historical fault records, historical repair records, mean time between failures (MTBF), and other data.

[0071] Fault feedback information refers to the information reported by users to the server regarding current faults at charging stations. Since users typically provide feedback to the server via text, audio, or other means, this feedback information is usually unstructured data. Unstructured data lacks a predefined data model or fixed structure and cannot be directly stored and managed using traditional relational databases (two-dimensional table format). Its formats are diverse, including text, images, audio, video, social media content, emails, and web documents.

[0072] For example, the fault feedback message is: after charging for 10 minutes, the current suddenly drops to 0, and charging stops.

[0073] For another example, the fault feedback information is: the power of the charging pile is sometimes high and sometimes low, and the charging speed is unstable.

[0074] Charging pile logs refer to all information recorded between the charging pile and the charging network management system, including various operations and status changes during the charging process. Charging pile logs can be used for fault diagnosis, system maintenance, and performance monitoring. Fault diagnosis involves quickly locating the cause of charging pile failures by analyzing error information in the logs, improving maintenance efficiency. System maintenance involves performing system maintenance and optimization based on the charging pile's operating status and operation history recorded in the logs. Performance monitoring involves evaluating the charging pile's working status and efficiency based on the performance reflected in the logs.

[0075] In some embodiments, the charging pile log may include charging process records, error information records, user operation records, device status records, etc. The charging process records may include charging start and end times or changes in charging power, the error information may include various error codes and reasons that occur during the charging process, the user operation records may include user behavior, such as starting charging and stopping charging, and the device status records may include information such as the hardware status and software version of the charging pile.

[0076] S22. Based on the target data collected, predict the charging pile faults and obtain the fault prediction results.

[0077] In this step, the fault prediction result is obtained by using the accurate judgment capability of supervised learning to predict faults in the charging pile. Users can decompose the charging pile system structure from top to bottom, for example, first breaking it down into system-level modules such as power supply module, communication module, charging pile control module, charging pile interface module, user interaction module, environmental safety module, and software module. Then, sub-modules of each module are extracted, such as AC / DC converters, capacitors, inductors, relays, contactors, and control boards. Finally, common fault types of each sub-module can be extracted by combining historical fault data and standard documents, such as relay sticking, communication failure, and capacitor short circuit. Through this bottom-up inductive manual annotation method, all possible fault types of the charging pile can be identified, achieving complete coverage of all fault types. The server can use all the identified possible fault types to predict charging pile faults, thus avoiding the server's inability to identify unknown fault types that have not been identified, thereby ensuring the reliability and accuracy of the fault prediction results obtained by the server.

[0078] In some embodiments, please refer to Figure 3 S22 includes:

[0079] S221. Extract the features of the target data to obtain target feature information.

[0080] In this step, the target feature information is the feature information used to predict charging pile faults. In some embodiments, the target feature information includes time-series feature information, which includes, but is not limited to, features such as voltage, current, and power. It is understood that, to improve the data quality of the target acquisition data, in some embodiments, the server can first perform data preprocessing on the target acquisition data, and then extract features from the preprocessed target acquisition data. The data preprocessing methods may include, but are not limited to, data cleaning, data deduplication, missing value imputation, and outlier detection. For different types of data, the server can use different feature extraction methods to extract data features. For example, the server can use signal analysis methods such as wavelet transform and Fourier transform to extract current waveform features, and use dimensionality reduction methods such as LDA (Linear Discriminant Analysis) and PCA (Principal Component Analysis) to extract key feature variables.

[0081] S222. Based on the target feature information and the preset fault prediction model, predict the charging pile fault and obtain the fault prediction result.

[0082] In this step, the server can use the target feature information as input to the preset fault prediction model, and predict the charging pile fault based on the output of the preset prediction model to obtain the fault prediction result.

[0083] In some embodiments, the preset fault prediction model includes an anomaly detection sub-model and a fault type identification sub-model. The anomaly detection sub-model is used to detect whether the target feature information is abnormal. Its main purpose is to detect whether the timing features such as current and voltage are normal. Whether the current and voltage are normal is a sub-factor of fault type identification. The fault type identification sub-model is used to identify the fault type of the charging pile.

[0084] In some embodiments, the server can acquire historical data, including historical fault data, extract features from the historical data to obtain feature sampling information, and train an anomaly detection sub-model and a fault type identification sub-model based on the feature sampling information.

[0085] In this embodiment, historical data collection includes charging pile-side historical data and server-side historical data. Charging pile-side historical data refers to historical data collected by the charging pile through IoT devices, sensors, smart meters, and other devices or equipment. Server-side historical data refers to historical data stored on the server. Feature sampling information may include user behavior feature information, numerical feature information, and status feature information. User behavior feature information may include startup method, number of operations, abnormal interaction logs, etc. Numerical feature information may include voltage anomaly labels, mean, range, skewness, etc., predicted from charging voltage / current / power time-series data. Status feature information may include whether the module overheated, whether communication was interrupted, duration, and time of occurrence, etc.

[0086] In some embodiments, the server can associate feature sampling information with anomaly detection labels to obtain a first sample dataset, and train a first model based on the first sample dataset to obtain an anomaly detection sub-model.

[0087] In this embodiment, the anomaly detection label is a label used to identify the normal state of a feature. The anomaly detection label can include normal labels and anomaly labels. The normal label is a label indicating that the feature is normal, and the anomaly label is a label indicating that the feature is abnormal. The first model can be any suitable detection model. In some embodiments, the first model is an LSTM (Long Short-Term Memory) model.

[0088] In some embodiments, the server can associate feature sampling information with fault type labels to obtain a second sample dataset, and train a second model based on the second sample dataset to obtain a fault type identification sub-model.

[0089] In this embodiment, the fault type label is a label used to identify the fault type of the feature. As mentioned earlier, the fault type label can be created by sorting out all possible fault types. Thus, the fault type label can include labels covering various fault types, such as relay adhesion labels, communication fault labels, capacitor short circuit labels, insulation aging labels, over-temperature protection labels, meter abnormality labels, communication interruption labels, power module fault labels, etc. It can be understood that when there is feature sampling information superimposed by multiple faults, the feature sampling information can correspond to multiple fault type labels. The second model can be any suitable identification model. In some embodiments, the second model is an XGBoost (eXtremeGradient Boosting) model. XGBoost is a multi-class classification model based on Gradient Boosting Decision Trees (GBDT). XGBoost can be used not only for multi-class classification tasks but also for regression, ranking, and anomaly detection. In multi-class classification, XGBoost constructs multiple decision trees, each of which partitions the input data and assigns a probability score to each class. Finally, XGBoost performs a weighted average of the probability scores from all decision trees to determine the probability that the input data belongs to each class.

[0090] In some embodiments, please refer to Figure 4 S222 includes:

[0091] S2221. Input the target feature information into the anomaly detection sub-model to obtain the anomaly detection result and anomaly detection feature information, including the anomaly detection label.

[0092] S2222. Input the anomaly detection feature information into the fault type identification sub-model to obtain the fault type identification result.

[0093] S2223. Verify the fault type identification result based on the anomaly detection result, and obtain the fault prediction result.

[0094] In S2221, the anomaly detection result is used to indicate whether the target feature information is abnormal or related to a fault. For example, for the detection of current feature information, the anomaly detection result can be "current normal" or "current abnormal." The anomaly detection feature information is target feature information with an anomaly detection label. The anomaly detection label can be determined based on the anomaly detection result. For example, when the anomaly detection result is "current abnormal," the anomaly detection feature information can be current feature information with a current abnormality label. Since a single current / voltage value cannot reflect whether it is "abnormal" or "fault-related," electronic devices can use current sequence features or voltage sequence features as input to the anomaly detection sub-model. This can improve the accuracy of the anomaly detection result output by the anomaly detection sub-model and increase the confidence of the anomaly detection feature information.

[0095] In S2222, the fault type identification result indicates whether the target feature information is associated with a specific fault type. For example, the fault type identification result could be relay sticking, communication failure, capacitor short circuit, insulation aging, over-temperature protection, meter malfunction, communication interruption, power module failure, etc. Since the target feature information is time-series feature information, it is not suitable to directly use it as input to the fault type identification sub-model. The ideal input parameter for the fault type identification sub-model is feature information with labels. The anomaly detection sub-model outputs anomaly detection feature information as target feature information with anomaly detection labels. Therefore, the server can directly use the anomaly detection feature information as input to the fault type identification sub-model. This not only ensures that the fault type identification sub-model outputs the fault type identification result correctly, but also, by using the anomaly detection feature information as a high-confidence input, improves the accuracy of the fault type identification result output by the fault type identification sub-model.

[0096] In S2223, for a specific fault type, the anomaly detection result and the fault type identification result should have a certain correlation. Therefore, the server can cross-verify the anomaly detection result and the fault type identification result to ensure that both the anomaly detection result and the fault type identification result are accurate. For example, if the anomaly detection result is voltage anomaly and the fault type identification result is power module fault, since voltage anomaly and power module fault are highly correlated, the server can assume that both voltage anomaly and power module fault have passed verification. That is, the anomaly detection result of voltage anomaly and the fault type identification result of power module fault are accurate. Therefore, this cross-verification method can improve the accuracy of fault prediction.

[0097] S23. Generate at least one fault inference link based on fault feedback information, charging pile logs and preset fault inference model.

[0098] In this step, the preset fault reasoning model is used to generate fault reasoning chains. A fault reasoning chain refers to the process of breaking down a complex fault problem into multiple sub-problems or a series of ordered steps, and then deriving the root cause of the fault through step-by-step reasoning. The preset fault reasoning model possesses powerful chain-like thinking, which significantly improves the model's reasoning ability and accuracy in practical applications. Especially when dealing with problems requiring precise reasoning, such as mathematical reasoning or logical judgment, chain-like thinking can reduce the error rate and improve the model's reliability.

[0099] In some embodiments, the server can acquire historical unstructured data, which includes a first part of data and a second part of data, including fault type labels. The server can acquire structured sample data obtained by manually annotating the first part of data, input the structured sample data and the second part of data into a third model to obtain a generalized corpus, and fine-tune a fourth model based on the generalized corpus to obtain a preset fault reasoning model. The fourth model is a sub-model of the third model.

[0100] In this embodiment, historical unstructured data refers to historical unstructured data with fault type labels, and structured sample data is a set of answer templates that have been manually compiled.

[0101] For example, a set of structured sample data is as follows:

[0102] Input: After charging for 10 minutes, the current suddenly drops to 0, and charging stops.

[0103] Thinking process:

[0104] 1) Obtain the current power curve and find that the current suddenly drops and does not recover;

[0105] 2) Check if there is an internal short circuit in the module or a failure in current output;

[0106] 3) Check the power module self-test log to determine if internal protection has been activated;

[0107] 4) Check the control command log to confirm whether the background control was forcibly terminated;

[0108] 5) Retrieve the user-side BMS logs to confirm whether the vehicle is restricted from continuing to charge.

[0109] Output: Charging module failure, system power failure, or BMS termination of charging.

[0110] For another example, here is another set of structured sample data:

[0111] Input: The power of the charging station fluctuates, and the charging speed is unstable.

[0112] Thinking process:

[0113] 1) Obtain the charging pile's current, voltage, and power data, and check the fluctuation range.

[0114] 2) Check if the mains input voltage is stable. If the voltage deviation exceeds ±10%.

[0115] 3) Check if the charging gun temperature is abnormal. Overheating may cause a decrease in power.

[0116] 4) Analyze the BMS data to check if the battery is actively limiting power.

[0117] Output: Abnormal grid voltage, excessively high charging pile temperature, or power limitation by the BMS.

[0118] In some embodiments, the third model is an open-source large model. An open-source large model refers to an artificial intelligence model whose source code, training data, and technical documentation are publicly available, allowing anyone to view, modify, and distribute it.

[0119] In some embodiments, the fourth model is the Llama8B model, which is a sub-model of the third model with parameter 8B. The server can replace the large model with the finely tuned sub-model, that is, generate the fault reasoning link through the sub-model instead of the large model, so that the fault reasoning link can better fit the actual fault diagnosis application scenario.

[0120] On the one hand, this embodiment generalizes a portion of the manually annotated structured sample data and the remaining unstructured data to an open-source large model, which avoids generating fault reasoning links from scratch during the fine-tuning of sub-models, thereby improving the accuracy of fault reasoning link generation. On the other hand, by generalizing the corpus and fine-tuning the adjustable sub-models in the large model, and using the fine-tuned sub-models as preset fault reasoning models for generating fault reasoning links, the fault reasoning links generated by the preset fault reasoning models can better fit the actual fault diagnosis application scenarios.

[0121] In some embodiments, each fault reasoning link includes a first type of fault information and / or a second type of fault information, wherein the first type of fault information and the second type of fault information are fault information at different levels, but are closely related when a certain specific fault occurs. For example, for a fault of transformer damage, the fault reasoning link generated by the preset fault reasoning model may be charging pause -> voltage abnormality -> insulation aging. In this fault reasoning link, the first type of fault information may be voltage abnormality, and the second type of fault information may be insulation aging.

[0122] In some embodiments, each fault inference link also includes a fault inference result, which includes the root cause of the fault. The root cause of the fault is the fundamental reason for the charging pile fault reported by the user. As mentioned above, if the charging pile fault reported by the user is a charging pause, the root cause of the charging pause may be a transformer failure.

[0123] It is understandable that a fault reasoning link may include only one of the first type of fault information and the second type of fault information, or it may include both the first type of fault information and the second type of fault information.

[0124] S24. Perform fault diagnosis on the charging pile based on the fault prediction results and at least one fault reasoning link.

[0125] In this step, the server can verify each fault inference link based on the fault prediction results, and can select the most reliable fault inference link from at least one fault inference link, thereby ensuring the reliability of the fault inference link, ensuring the accuracy of the fault inference results, and thus improving the accuracy of fault diagnosis.

[0126] Overall, on the one hand, this embodiment can respond to user feedback on fault information in a timely manner and use the fault feedback information for fault reasoning, which is conducive to enhancing the generalization ability of fault diagnosis and thus achieving more intelligent fault diagnosis. On the other hand, by predicting charging pile faults through supervised learning methods to obtain fault prediction results and verifying the fault reasoning link using the fault prediction results, it effectively combines the accurate judgment ability of supervised learning with the powerful chain reasoning ability of the reasoning model, ensuring the reliability of the fault reasoning link and thus ensuring the accuracy of the fault reasoning results, thereby improving the accuracy of fault diagnosis.

[0127] In some embodiments, please refer to Figure 5 S24 includes:

[0128] S241. Extract the fault type information of each fault inference link to obtain the candidate fault information corresponding to each fault inference link.

[0129] For example, here are three fault reasoning chains:

[0130] Charging paused -> voltage abnormality -> insulation aging

[0131] Unstable charging speed -> Abnormal current -> Charging module failure

[0132] Low charging efficiency -> Abnormal power -> Overheating during charging

[0133] In this step, as mentioned above, each candidate fault information group includes a first type of fault information and / or a second type of fault information. For the fault inference link "charging pause -> voltage abnormality -> insulation aging", the server can extract the two types of fault information, voltage abnormality and insulation aging, and use these two types of fault information as candidate fault information corresponding to the fault inference link.

[0134] S242. Perform fault diagnosis on the charging pile based on the fault prediction results and candidate fault information.

[0135] In some embodiments, S242 includes: filtering target fault information from candidate fault information based on fault prediction results, determining target fault inference links based on target fault information, obtaining fault inference results of target fault inference links, and performing fault diagnosis on charging piles based on fault inference results.

[0136] For example, as mentioned earlier, candidate fault information includes (abnormal voltage, insulation aging), (abnormal current, charging module failure), and (abnormal power, charging temperature too high). The inference result of candidate fault information (abnormal voltage, insulation aging) is transformer damage; the inference result of candidate fault information (abnormal current, charging module failure) is relay sticking; and the inference result of candidate fault information (abnormal power, charging temperature too high) is capacitor short circuit. The server can match the fault prediction result with the first type of fault information and / or the second type of fault information in the candidate fault information, and take the successfully matched candidate fault information as the target fault information. For example, if the target fault information is (abnormal voltage, insulation aging), the server can determine that the target fault inference link is "charging pause -> abnormal voltage -> insulation aging". Since the fault inference result of this target fault inference link is transformer damage, the server can determine that the fault in the charging pile may be transformer damage, thereby realizing the fault diagnosis of the charging pile.

[0137] In some embodiments, the fault prediction results include anomaly detection results and fault type identification results.

[0138] As mentioned earlier, the anomaly detection result can be obtained from the output of the anomaly detection sub-model, and the fault type identification result can be obtained from the output of the fault type identification sub-model.

[0139] In some embodiments, the server can match the anomaly detection result with the first type of fault information and / or the second type of fault information to obtain a first matching degree, match the fault type identification result with the first type of fault information and / or the second type of fault information to obtain a second matching degree, and filter out the target fault information from the candidate fault information based on the first matching degree and the second matching degree.

[0140] For example, as mentioned above, if the anomaly detection result is voltage anomaly and the fault type identification result is insulation aging, for the candidate fault information (voltage anomaly, insulation aging), the server can match voltage anomaly with voltage anomaly and insulation aging respectively to obtain the first matching degree, and then match insulation aging with voltage anomaly and insulation aging respectively to obtain the second matching degree.

[0141] When filtering target fault information from candidate fault information based on the first matching degree and the second matching degree, the server can determine whether there is a completely matching and unique candidate fault information based on the first matching degree and the second matching degree. If so, the candidate fault information is determined as the target fault information. If not, the server calculates the comprehensive matching degree based on the first matching degree and the second matching degree, and filters the target fault information from at least one candidate fault information based on the comprehensive matching degree.

[0142] For example, as mentioned earlier, when voltage anomaly is matched with voltage anomaly and insulation aging respectively, voltage anomaly is successfully matched, so the server can determine that the first matching degree is 100%. When insulation aging is matched with voltage anomaly and insulation aging respectively, insulation aging is successfully matched, so the server can determine that the second matching degree is 100%. Since both the first and second matching degrees are 100%, the server can determine that the candidate fault information (voltage anomaly, insulation aging) is a completely matching candidate fault information, and that the candidate fault information is unique. Therefore, the server can determine that the candidate fault information is the target fault information.

[0143] In some embodiments, the server can directly add the first matching score and the second matching score to obtain the comprehensive matching score, or it can weight the first matching score and the second matching score separately and then sum them to obtain the comprehensive matching score. For example, the server can calculate the comprehensive matching score according to the following formula:

[0144] M = m1*α + m2*β

[0145] Where M is the weighted sum, i.e. the overall matching degree, m1 is the first matching degree, α is the weight coefficient of the first matching degree, m2 is the second matching degree, and β is the weight coefficient of the second matching degree.

[0146] Understandably, using a weighted summation method can highlight the more important matching degree in practical applications, thereby obtaining a more ideal overall matching degree. For example, if the first matching degree is more important, the weight coefficient α of the first matching degree can be set to be greater than the weight coefficient β of the second matching degree; if the second matching degree is more important, the weight coefficient α of the first matching degree can be set to be less than the weight coefficient β of the second matching degree.

[0147] It should be noted that in the above embodiments, there is no necessarily a certain order between the steps. Those skilled in the art can understand from the description of the embodiments of the present invention that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.

[0148] As another aspect of this invention, this embodiment provides a fault diagnosis device. The fault diagnosis device can be a software module, which includes several instructions stored in a memory. A processor can access the memory, invoke the instructions, and execute them to complete the fault diagnosis methods described in the various embodiments above.

[0149] In some embodiments, the fault diagnosis device can be constructed from hardware components. For example, the fault diagnosis device can be constructed from one or more chips, which can work in coordination to complete the fault diagnosis methods described in the various embodiments above. As another example, the fault diagnosis device can also be constructed from components such as general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, ARM (Acorn RISC Machine), programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of these components.

[0150] In some embodiments, please refer to Figure 6 The fault diagnosis device 600 provided in this embodiment of the invention includes a first acquisition module 601, a prediction module 602, a generation module 603 and a diagnosis module 604.

[0151] The first acquisition module 601 is used to acquire target data, fault feedback information and charging pile logs. The prediction module 602 is used to predict charging pile faults based on the target data and obtain fault prediction results. The generation module 603 is used to generate at least one fault reasoning link based on fault feedback information, charging pile logs and preset fault reasoning model. The diagnosis module 604 is used to diagnose the charging pile faults based on the fault prediction results and at least one fault reasoning link.

[0152] In some embodiments, please refer to Figure 7 The prediction module 602 includes a first extraction unit 6021 and a prediction unit 6022.

[0153] The first extraction unit 6021 is used to extract the features of the target data to obtain target feature information, and the prediction unit 6022 is used to predict the charging pile fault based on the target feature information and the preset fault prediction model to obtain the fault prediction result.

[0154] In some embodiments, the preset fault prediction model includes an anomaly detection sub-model and a fault type identification sub-model. The prediction unit 6022 is specifically used to: input target feature information into the anomaly detection sub-model to obtain anomaly detection results and anomaly detection feature information, the anomaly detection feature information including anomaly detection labels; input the anomaly detection feature information into the fault type identification sub-model to obtain fault type identification results; verify the fault type identification results based on the anomaly detection results to obtain fault prediction results.

[0155] In some embodiments, please refer to Figure 8 The diagnostic module 604 includes a second extraction unit 6041 and a diagnostic unit 6042.

[0156] The second extraction unit 6041 is used to extract the fault information of each fault inference link and obtain the candidate fault information corresponding to each fault inference link. The diagnosis unit 6042 is used to perform fault diagnosis on the charging pile based on the fault prediction results and the candidate fault information.

[0157] In some embodiments, the diagnostic unit 6042 is specifically used to: filter out target fault information from candidate fault information based on fault prediction results, determine target fault reasoning links based on target fault information, obtain fault reasoning results of target fault reasoning links, and perform fault diagnosis on charging piles based on fault reasoning results.

[0158] In some embodiments, please refer to Figure 9 The fault diagnosis device 600 also includes a second acquisition module 605, a third acquisition module 606, a generalization module 607, and a fine-tuning module 608.

[0159] The second acquisition module 605 is used to acquire historical unstructured data, which includes a first part of data and a second part of data. The first part of data and the second part of data include fault type labels. The third acquisition module 606 is used to acquire structured sample data obtained by manually annotating the first part of data. The generalization module 607 is used to input the structured sample data and the second part of data into the third model to obtain generalized corpus. The fine-tuning module 608 is used to fine-tune the fourth model based on the generalized corpus to obtain a preset fault reasoning model. The fourth model is a sub-model of the third model.

[0160] It should be noted that the above-described fault diagnosis device can execute the fault diagnosis method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in the embodiments of the fault diagnosis device can be found in the fault diagnosis method provided in the embodiments of the present invention.

[0161] Please see Figure 10 , Figure 10This is a schematic diagram of the hardware structure of a server provided in an embodiment of the present invention. For example... Figure 10 As shown, server 200 includes one or more processors 201 and memory 202. Figure 10 Take a processor 201 as an example.

[0162] Processor 201 is configured to support the computer device in performing the corresponding functions in the methods described in the above method embodiments. Processor 201 may be a Central Processing Unit (CPU), a Network Processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an Application Specific Integrated Circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The aforementioned PLD may be a Complex Programmable Logic Device (CPLD), a Field-Programmable Gate Array (FPGA), a Generic Array Logic (GAL), or any combination thereof.

[0163] Memory 202 is used to store program code. Memory 202 may include volatile memory (VM), such as random access memory (RAM); memory may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 202 may also include combinations of the above types of memory.

[0164] The memory 202 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the fault diagnosis method in the embodiments of the present invention. The processor 201 executes various functional applications and data processing of the fault diagnosis method and fault diagnosis device by running the non-volatile software programs, instructions, and modules stored in the memory 202, that is, it realizes the functions of each module or unit of the fault diagnosis method and fault diagnosis device provided in the above method embodiments.

[0165] The memory 202 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the fault diagnosis device. In some embodiments, the memory 202 may optionally include memory remotely located relative to the processor, which can be connected to the fault diagnosis device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0166] The one or more modules are stored in the memory 202. When executed by the one or more processors 201, they perform the fault diagnosis method in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0167] This invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the method described in the foregoing embodiments.

[0168] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0169] Finally, it should be noted that the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. These embodiments are not intended to impose additional limitations on the content of the present invention; their purpose is to provide a more thorough and comprehensive understanding of the disclosure of the present invention. Furthermore, within the framework of the present invention, the above-mentioned technical features can be combined with each other, and many other variations of different aspects of the present invention as described above exist, all of which are considered to be within the scope of the present invention specification. Moreover, those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A fault diagnosis method, characterized in that, include: Acquire target data, fault feedback information, and charging pile logs; Based on the target data collected, charging pile faults are predicted, and fault prediction results are obtained. At least one fault reasoning link is generated based on the fault feedback information, the charging pile log, and the preset fault reasoning model. The fault reasoning link is a process of breaking down a complex fault problem into multiple sub-problems or a series of ordered steps, and deriving the root cause of the fault through step-by-step reasoning. The preset fault reasoning model is a model used to generate fault reasoning links. It is pre-built in the following way: acquiring historical unstructured data, wherein the historical unstructured data is historical unstructured data with fault type labels, including a first part of data and a second part of data, wherein the first part of data and the second part of data include fault type labels; Obtain structured sample data, which is obtained by manually annotating the first part of the data and serves as a manually compiled answer template; The structured sample data and the second part of the data are input into the third model to obtain the generalized corpus; Based on the generalized corpus, the fourth model is fine-tuned to obtain the preset fault reasoning model; The fourth model is a sub-model of the third model; The fault diagnosis of the charging pile is performed based on the fault prediction result and at least one fault inference link, including: extracting fault information of each fault inference link to obtain candidate fault information corresponding to each fault inference link; filtering target fault information from the candidate fault information based on the fault prediction result; determining the target fault inference link based on the target fault information; obtaining the fault inference result of the target fault inference link, and diagnosing the charging pile based on the fault inference result.

2. The fault diagnosis method according to claim 1, characterized in that, The step of predicting charging pile faults based on the target collected data and obtaining the fault prediction result includes: Extract the features from the target data to obtain target feature information; Based on the target feature information and the preset fault prediction model, the fault of the charging pile is predicted, and the fault prediction result is obtained.

3. The fault diagnosis method according to claim 2, characterized in that, The preset fault prediction model includes an anomaly detection sub-model and a fault type identification sub-model. The step of predicting charging pile faults based on the target feature information and the preset fault prediction model to obtain the fault prediction result includes: The target feature information is input into the anomaly detection sub-model to obtain anomaly detection results and anomaly detection feature information, wherein the anomaly detection feature information includes anomaly detection labels. The anomaly detection feature information is input into the fault type identification sub-model to obtain the fault type identification result; The fault type identification result is verified based on the anomaly detection result to obtain the fault prediction result.

4. The fault diagnosis method according to claim 1, characterized in that, The fault prediction result includes anomaly detection result and fault type identification result. Each candidate fault information includes first type fault information and / or second type fault information. The step of filtering target fault information from the candidate fault information based on the fault prediction result includes: The anomaly detection results are matched with the first type of fault information and / or the second type of fault information to obtain a first matching degree; The fault type identification result is matched with the first type of fault information and / or the second type of fault information to obtain the second matching degree; Target fault information is selected from the candidate fault information based on the first matching degree and the second matching degree.

5. The fault diagnosis method according to claim 4, characterized in that, The step of filtering target fault information from the candidate fault information based on the first matching degree and the second matching degree includes: Based on the first matching degree and the second matching degree, determine whether there is a completely matching and unique candidate fault information; If so, then the candidate fault information is determined to be the target fault information; If not, then calculate the comprehensive matching degree based on the first matching degree and the second matching degree, and filter the target fault information from the candidate fault information based on the comprehensive matching degree.

6. A server, characterized in that, The system includes a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, and the processor, when executing the one or more computer programs, causing the server to implement the fault diagnosis method as described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the fault diagnosis method as described in any one of claims 1 to 5.

Images