New energy vehicle charging pile fault diagnosis method based on deep learning

Abstract

The invention relates to the field of machine learning, and discloses a charging pile fault diagnosis method, which comprises the following steps: S1, preparing an environment, building a deep learning development environment, and providing a necessary technical environment for the training and deployment of a subsequent model; and S2, data acquisition and processing: installing a current sensor, a voltage sensor and a temperature sensor on a new energy automobile charging detection vehicle, and acquiring operation data of the charging pile in real time. And S3, model definition: a fault diagnosis model based on deep learning is constructed, a fully connected neural network (DNN) is adopted, and the model comprises an input layer, a hidden layer and an output layer. The key operation data of the charging pile are collected in real time through the sensor, the fault state is intelligently judged in combination with the deep learning model, rapid and accurate fault positioning and diagnosis are achieved, and the maintenance efficiency of the charging pile is remarkably improved. The system can automatically extract data features and analyze fault types, and manual intervention is reduced.

Description

Technical Field

[0001] The present invention relates to the field of machine learning, and in particular to a method for diagnosing faults of charging piles for new energy vehicles. Background Art

[0002] With the global emphasis on environmental protection, the popularization of new energy vehicles has become an inevitable trend, especially the rapid growth of electric vehicles, which has accelerated the construction of charging facilities. According to statistics, by 2020, my country is expected to add more than 12,000 centralized charging and swapping stations and more than 4.8 million decentralized charging piles. There are many types of charging piles, including AC charging piles and DC charging piles, and the technical level and functions of various types of charging piles are constantly improving. However, charging piles will have various faults during operation, affecting charging efficiency and safety. These faults may come from design defects of the equipment itself, aging of components, and the influence of environmental factors.

[0003] At present, fault diagnosis mainly relies on manual inspection and experience judgment, lacking systematic monitoring methods and intelligent diagnostic tools. This method is not only inefficient, but also easily affected by human factors, resulting in missed detection and misjudgment. Although some charging piles are equipped with basic self-diagnosis functions, they are often limited to simple status monitoring and cannot fully analyze the cause of the fault, especially in complex fault situations, and often cannot provide effective diagnostic support. Therefore, developing an intelligent fault diagnosis method based on deep learning to improve the accuracy and real-time performance of fault detection has become an urgent problem to be solved. A new energy vehicle charging pile fault diagnosis method based on deep learning is proposed. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the deficiencies in the prior art, the present invention provides a new energy vehicle charging pile fault diagnosis method based on deep learning to solve the above problems.

[0006] (II) Technical solution

[0007] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a new energy vehicle charging pile fault diagnosis method based on deep learning, comprising the following steps:

[0008] S1. Environmental Preparation

[0009] By installing Baidu's PaddlePaddle framework, build a deep learning development environment and ensure that Python and related dependency modules are installed on the system. Complete the installation with the following command:

[0010] pip install paddlepaddle

[0011] This step provides the necessary technical environment for subsequent model training and deployment.

[0012] S2. Data Collection and Processing

[0013] Current sensors, voltage sensors and temperature sensors are installed on new energy vehicle charging detection vehicles to collect real-time operation data of charging piles, including key parameters such as voltage, current and temperature, and record working status and fault history data to form a data set. The original data is cleaned, denoised, normalized, and missing values ​​are filled and abnormal values ​​are detected to ensure the integrity and consistency of the data set.

[0014] S3. Model definition

[0015] A fault diagnosis model based on deep learning is constructed using a fully connected neural network (DNN). The model includes an input layer, a hidden layer, and an output layer. The data is processed by activation functions ReLU and Sigmoid to achieve the status classification of the charging pile. Some model code examples are as follows:

[0016] import paddle

[0017] import paddle.nn as nn

[0018] class ChargingFaultNet(nn.Layer):

[0019] def __init__(self):

[0020] super(ChargingFaultNet,self).__init__()

[0021] self.fc1 = nn.Linear(3,64)

[0022] self.fc2 = nn.Linear(64,32)

[0023] self.fc3 = nn.Linear(32,1)

[0024] def forward(self,x):

[0025] x=nn.functional.relu(self.fc1(x))

[0026] x=nn.functional.relu(self.fc2(x))

[0027]

[0028] S4. Model training

[0029] The preprocessed data set is divided into a training set and a test set. The PaddlePaddle framework is used for model training. The binary cross entropy is used as the loss function, and the Adam optimizer optimizes the model parameters. The model weights are continuously adjusted during the training process to improve the accuracy. The training code example is as follows:

[0030]

[0031] S5. Model evaluation and deployment

[0032] Use the test set to evaluate the accuracy of the model, verify its classification ability, and deploy the trained model to the vehicle test platform to achieve real-time prediction. Use the following code to load the model and perform prediction:

[0033] model.set_state_dict(paddle.load('charging_fault_net.pdparams'))

[0034] model.eval()

[0035] S6. Fault diagnosis and feedback

[0036] The deep learning model is used to infer real-time data, determine whether the charging pile is faulty, and present the results through the VUE framework. The system has online learning capabilities and can adjust model parameters as new data is input to continuously optimize fault diagnosis results.

[0037] (III) Beneficial effects

[0038] Compared with the prior art, the present invention provides a new energy vehicle charging pile fault diagnosis method based on deep learning, which has the following beneficial effects:

[0039] 1. Real-time and accuracy: Through sensors, key operating data of charging piles are collected in real time, and deep learning models are used to make intelligent judgments on fault conditions, so as to achieve fast and accurate fault location and diagnosis, and significantly improve the maintenance efficiency of charging piles.

[0040] 2. System intelligence: The system can automatically extract data features and analyze fault types, reducing manual intervention; combined with the online learning mechanism, the model can be continuously optimized to adapt to the complex and changeable working environment of charging piles and improve the robustness of fault diagnosis.

[0041] 3. Visual result presentation: The fault diagnosis results are graphically displayed through the VUE framework, providing fault reports and maintenance suggestions to help operators quickly understand the problem and shorten decision-making time.

[0042] 4. Flexible deployment and scalability: The diagnostic model can be deployed on the vehicle test platform to facilitate on-site detection and use; through modular design, the system supports the data collection and diagnostic needs of various types of charging piles and has strong adaptability.

[0043] 5. Cost and maintenance optimization: By predicting and diagnosing potential faults in advance, the number of charging pile repairs and system downtime can be reduced, reducing operating costs, while extending equipment life and improving system reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0045] Figure 2 This is a schematic diagram of data collection for the present invention;

[0046] Figure 3 This is a schematic diagram of data cleaning of the present invention;

[0047] Figure 4 This is a schematic diagram for optimizing the model of the present invention;

[0048] Figure 5 This is a schematic diagram of the core service deployment of the present invention; DETAILED DESCRIPTION

[0049] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solution. The following description is only an example embodiment, based on which ordinary technicians in the field can make improvements without departing from the spirit of the present invention, and these improvements still fall within the scope of protection of the present invention.

[0050] S1. Data collection and monitoring

[0051] The built-in sensors in the vehicle detection platform collect the operating data of the new energy vehicle charging pile in real time. Including:

[0052] 1. Voltage collection: Use a high-precision voltage sensor to record the output voltage of the charging pile port in real time;

[0053] 2. Current collection: monitor the current changes during charging through the Hall current sensor;

[0054] 3. Resistance measurement: Combine current and voltage data, use Ohm's law to calculate the port resistance and analyze the insulation condition;

[0055] 4. Charge and discharge efficiency: record input and output power, and calculate efficiency to determine whether there is abnormal energy loss.

[0056] All sensor data are transmitted to the local storage device through the A / D conversion module, and the timestamp and geographic location information are attached.

[0057] Abnormal handling: If the signal is lost or the data deviates from the reasonable range during the sensor acquisition process, the system will trigger an alarm mechanism to prompt the operator to check the hardware connection or replace the sensor.

[0058] S2. Data cleaning and denoising

[0059] The collected raw data may contain noise or missing values. To improve the input quality of the diagnostic model, the following methods are used:

[0060] 1. Denoising: eliminate random noise through wavelet transform filtering algorithm;

[0061] 2. Missing value filling: Lagrange interpolation method is used to complete the discontinuous data;

[0062] 3. Normalization: All data are uniformly mapped to the range of [0,1] to reduce the dimensional differences between features.

[0063] After cleaning is completed, the system will perform a data consistency check. If any abnormality is found, it will return to S1 for re-collection.

[0064] S3. Fault detection and determination

[0065] The cleaned data is input into the fault detection module, which contains the following functions:

[0066] 1. Real-time diagnosis: Use pre-trained deep learning models (such as DNN) to determine the status of the charging pile (normal / faulty);

[0067] 2. Fault classification: Features are extracted through convolutional neural network (CNN) to classify faults into voltage anomalies, current overloads, connection failures, etc.

[0068] The decision rule is based on the output probability of the model. If it exceeds the preset threshold (such as 0.8), it is considered a fault. For high-risk fault systems, an alarm is directly triggered and it is recommended to shut down the equipment.

[0069] S4. Model optimization and online learning

[0070] Through federated learning technology, the model can learn from multiple charging pile data without centralized data storage, protecting data privacy while enhancing the generalization ability of the model.

[0071] The optimization process includes:

[0072] 1. Collect on-site feedback data;

[0073] 2. Automatically mark new data;

[0074] 3. Update model weights and deploy to the vehicle server.

[0075] S5. System deployment and operation and maintenance

[0076] The trained model is deployed in the vehicle server in the form of a Docker image. The system supports automatic updates to ensure that the algorithm always maintains optimal performance. Deployment solutions include:

[0077] 1. Use Kubernetes to manage resource allocation of vehicle servers;

[0078] 2. Implement a real-time inference interface to receive sensor input and return diagnostic results via HTTP or WebSocket.

[0079] Working principle:

[0080] Data collection and real-time monitoring (S1)

[0081] The system relies on the vehicle detection platform to collect real-time operation data of new energy vehicle charging piles through high-precision current, voltage and temperature sensors. The sampling rate can reach 1kHz per second to ensure the capture of transient changes in the charging process.

[0082] The sensor data is converted by A / D and stored in the local device, with timestamp and geographic information attached for subsequent analysis. If the sensor signal is interrupted or the data exceeds the reasonable range, the system will trigger an alarm mechanism to prompt the operator to take inspection and maintenance measures.

[0083] Data preprocessing and denoising (S2)

[0084] To improve the accuracy of the fault detection model, the system performs preprocessing immediately after data collection is completed:

[0085] Use wavelet transform filtering algorithm to eliminate random noise in the signal;

[0086] Interpolation algorithms are used to fill missing values ​​to ensure data integrity;

[0087] The original data is normalized to unify the feature distribution and eliminate the influence of numerical differences on model judgment.

[0088] After preprocessing, the data is stored in the cache and verified by the verification mechanism. If the data verification fails, the system will return to the data acquisition module to obtain it again.

[0089] Fault detection and determination (S3)

[0090] The cleaned data is input into the deep learning fault diagnosis model. The model structure is as follows:

[0091] The first layer is the input layer, which accepts feature vectors such as voltage, current, and temperature;

[0092] The middle layer is a multi-layer fully connected neural network, which uses the activation function ReLU to extract nonlinear features;

[0093] The output layer uses the Sigmoid function to generate the fault judgment probability value.

[0094] A large amount of labeled data is used in the model training process, covering normal operation and various fault conditions to ensure that the model has good generalization ability. During reasoning, if the probability of failure exceeds a threshold (such as 80%), the system will determine it as a failure and trigger an early warning.

[0095] Model Optimization and Intelligent Learning (S4)

[0096] The system supports online learning capabilities, integrates operating data in different scenarios through federated learning technology, and dynamically adjusts model parameters. In actual operation, the system regularly obtains new data from sensors and uses online learning technology to enhance fault identification capabilities. Through the feedback mechanism, the system gradually improves the diagnostic accuracy under complex working conditions.

[0097] System deployment and operation maintenance (S5)

[0098] The diagnostic model is deployed in the vehicle server in the form of a lightweight Docker image and combined with Kubernetes for distributed management. After deployment, the system monitors sensor data in real time. If a failure occurs, a visual report is immediately generated through the VUE interface and a notification is sent to the management staff.

[0099] In abnormal situations, such as insufficient server resources or model inference failure, the system will switch to the backup server to ensure continuous service.

[0100] Troubleshooting and decision support

[0101] For identified faults, the system provides detailed classification and diagnostic reports, including fault type, possible causes and solution suggestions. Operators can quickly view the report content through the VUE interface and take repair measures according to the prompts. In addition, the system records all diagnostic history to provide data support for long-term equipment optimization and strategy formulation.

[0102] The above-described embodiments are used to illustrate the present invention, not to limit the present invention, so changes to the exemplified numerical values ​​or equivalent substitutions should still fall within the scope of the present invention.

[0103] From the above detailed description, it can be understood by those skilled in the art that the present invention can indeed achieve the aforementioned objectives and is in compliance with the provisions of the Patent Law.

[0104] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0105] It should be noted that the above description of the relevant process is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0106] The basic concepts have been described above. Obviously, for those of ordinary skill in the art who have read this application, the above invention disclosure is only for example and does not constitute a limitation of this application. Although not explicitly stated here, those of ordinary skill in the art may make various modifications, improvements and amendments to this application. Such modifications, improvements and amendments are suggested in this application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of this application.

[0107] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0108] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installation on an existing server or mobile device.

[0109] Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of the application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this method of the application should not be interpreted as reflecting the intention that the claimed object requires more features than those explicitly stated in each claim. On the contrary, the subject of the invention should have fewer features than the above single embodiment.

Claims

1. A new energy vehicle charging pile fault diagnosis method based on deep learning, characterized in that: S1. Environment preparation: Build a deep learning development environment by installing the PaddlePaddle framework and ensure that Python and related dependent modules are correctly configured; S2. Model definition: A charging pile status classification model is constructed using a fully connected neural network (DNN). The model includes an input layer, a hidden layer, and an output layer. Data processing and classification are achieved through activation functions ReLU and Sigmoid. S3. Data collection: The current, voltage and temperature sensors in the vehicle are used to collect the operating data during the charging process in real time to form a data set.

2. According to a new energy vehicle charging pile fault diagnosis method based on deep learning according to claim 1, it is characterized in that: The data acquisition module of step S3 also includes a data recording function, which can record the historical fault data of the charging pile and group and store the operating data under different working conditions for subsequent analysis and feature extraction.

3. According to a new energy vehicle charging pile fault diagnosis method based on deep learning according to claim 1, it is characterized in that: The data collection module in step S3 is combined with the GPS positioning module to add geographical location information to the collected data to form a multi-dimensional data set to provide support for regional fault mode analysis.

4. According to a new energy vehicle charging pile fault diagnosis method based on deep learning according to claim 1, it is characterized in that: S4. Data preprocessing: Clean, denoise and normalize the collected raw data, fill in missing values ​​and detect outliers; use Python combined with PaddlePaddle to segment the data to form standardized training sets and test sets.

5. The method for fault diagnosis of a new energy vehicle charging pile based on deep learning according to claim 1 is characterized in that: S5. Model training: Optimize model parameters through the Adam optimizer of the PaddlePaddle framework, use the binary cross entropy loss function to train the model, iteratively update the model weights, and improve the fault classification accuracy.

6. The method for fault diagnosis of a new energy vehicle charging pile based on deep learning according to claim 1 is characterized in that: The training model in step S5 has real-time adjustment capability and can be continuously trained and optimized according to newly added fault data, thereby improving the robustness and generalization ability of the diagnosis model.

7. The method for fault diagnosis of a new energy vehicle charging pile based on deep learning according to claim 1 is characterized in that: S6. Fault diagnosis: Use the trained deep learning model to infer the real-time input voltage, current, and temperature data, output the fault probability value through the Sigmoid function, and determine the fault type.

8. The method for fault diagnosis of a new energy vehicle charging pile based on deep learning according to claim 1, characterized in that: The diagnostic result of step S6 is visualized through VUE technology, including real-time data display, fault report generation and maintenance suggestion prompts, which is convenient for operators to understand and make decisions.

9. The method for diagnosing faults of new energy vehicle charging piles based on deep learning according to claim 1, characterized in that: S7. Model evaluation and deployment: Evaluate the model through the test set, calculate its classification accuracy, save the trained model to the vehicle test platform, and load the model in actual operation to achieve real-time prediction.

10. A new energy vehicle charging pile fault diagnosis method based on deep learning according to claim 1, characterized in that: S8. Feedback mechanism: The system has the ability to learn online. By continuously inputting new fault data, it can adaptively adjust the model parameters and gradually improve the accuracy and reliability of fault diagnosis.

Images