Charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction

Abstract

The invention relates to a charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction, and the method comprises the following steps: 1, extracting multi-dimensional fault feature data information from a grid-side converter and a post-stage LLC circuit, carrying out the preprocessing of the multi-dimensional fault feature data information, and fusing the information into a high-dimensional feature vector; 2, training a machine learning model by using the fused high-dimensional feature vectors in the step 1 to obtain a trained charging pile grid-side converter fault diagnosis model; 3, evaluating the generalization ability of the fault diagnosis model of the charging pile grid-side converter in the step 2 through cross validation; and step 4, using the trained charging pile grid-side converter fault diagnosis model to diagnose the multi-dimensionally extracted fault features. According to the invention, the accuracy and speed of fault diagnosis can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis of electric energy conversion devices, and relates to a method for diagnosing faults of a charging pile grid-side converter, and in particular to a method for diagnosing faults of a charging pile grid-side converter based on multi-dimensional feature extraction. Background Art

[0002] As the main public charging facilities for electric vehicles, high-power DC charging piles can meet their needs for fast charging. At the same time, the rapid popularization of electric vehicles has also led to the large-scale construction of high-power DC charging piles. However, DC charging piles are in high voltage, high power, and high frequency operation during operation, and have a high potential failure rate, such as charging pile main circuit failure and charging module failure. Among them, the charging module is composed of a grid-side converter and an LLC DC-DC converter. Due to the harsh working conditions and the presence of power electronic devices that are prone to aging and failure, it has become the component with the highest probability of failure in the DC charging pile, accounting for more than 70%.

[0003] The fault diagnosis of the grid-side converter is generally performed based on the input three-phase current. This diagnostic method is too dependent on the input current and is prone to misdiagnosis.

[0004] In summary, the main challenge facing the fault diagnosis of the grid-side converter in the current charging pile is the scarcity of fault characteristics, which makes the fault diagnosis system have great limitations in accurately identifying faults.

[0005] Therefore, in order to solve the above problems, the present invention proposes a charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction.

[0006] After searching, no public documents of the prior art identical or similar to the present invention were found. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a charging pile grid-side converter fault diagnosis method based on multidimensional feature extraction. In order to improve the efficiency of charging pile fault diagnosis, the grid-side converter output voltage and battery-side charging current fault characteristics are introduced, and a multidimensional fusion fault diagnosis method is adopted to improve the accuracy and speed of fault diagnosis.

[0008] The present invention solves the practical problem by adopting the following technical solutions:

[0009] A charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction comprises the following steps:

[0010] Step 1: extract multi-dimensional fault feature data information from the grid-side converter and the subsequent LLC circuit, and pre-process the multi-dimensional fault feature data information to fuse it into a high-dimensional feature vector;

[0011] Step 2: Use the high-dimensional feature vector fused in step 1 to train the machine learning model to obtain a trained charging pile grid-side converter fault diagnosis model;

[0012] Step 3: Evaluate the generalization ability of the charging pile grid-side converter fault diagnosis model in step 2 through cross validation;

[0013] Step 4: Use the trained charging pile grid-side converter fault diagnosis model to diagnose the multi-dimensionally extracted fault features.

[0014] Moreover, the step 1 of extracting multi-dimensional fault feature data information from the fault diagnosis process of the grid-side converter includes:

[0015] Including the peak value and average value of the grid-side converter input current, the peak value, average value, zero value and instantaneous value of the grid-side converter output voltage, the distortion degree, abnormal frequency component and waveform kurtosis characteristics of the LLC circuit charging current on the battery side.

[0016] Moreover, the specific method of step 2 is:

[0017] Select support vector machine to train the model. During the model training process, adjust the algorithm's hyperparameters to optimize the model performance, and use training data to train the model to ensure that the model can effectively learn the characteristics of different fault modes and obtain a trained fault diagnosis model.

[0018] Moreover, the specific method of step 3 is:

[0019] K-fold cross validation is used to evaluate the generalization ability of the model to ensure that the model performs consistently on different data subsets. The confusion matrix is ​​used to analyze the classification results of the model and the accuracy, recall, and F1 score indicators are calculated to evaluate the performance of the model on different fault categories. The ROC curve is drawn to evaluate the performance of the classification model.

[0020] Moreover, the specific method of step 4 is:

[0021] The multi-dimensional extracted fault features are diagnosed through the trained charging pile grid-side converter fault diagnosis model; different fault categories are classified by adding labels, and the fault category labels and related fault descriptions are output to determine the fault type, and the fault accuracy is determined through the confusion matrix method.

[0022] Advantages and beneficial effects of the present invention:

[0023] 1. The present invention proposes a charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction, which is dedicated to solving the problem of low diagnostic accuracy caused by insufficient fault features in the grid-side converter fault diagnosis process of the charging pile. This method comprehensively considers data from multiple dimensions in the multi-dimensional fusion fault feature extraction process to improve the accuracy of fault diagnosis. Specifically, this method combines multiple features of the output voltage of the grid-side converter in the charging pile and the charging current of the LLC circuit on the battery side to form a comprehensive fault feature representation, thereby more accurately perceiving and evaluating the overall operating status of the charging pile. This multi-dimensional feature fusion method can not only overcome the problem of low diagnostic accuracy caused by insufficient fault features in traditional methods, but also significantly improve the accuracy of fault diagnosis, ensuring that the charging pile can be identified promptly and accurately when a fault occurs.

[0024] 2. The present invention can fully mine and utilize various operating data of the grid-side converter in the charging pile by introducing multi-dimensional feature extraction technology, so that the fault diagnosis system can perform a more detailed analysis of the fault. This technology can not only extract more abundant fault features, but also effectively integrate data from different sources, thereby building a global understanding of the status of the charging pile. This global understanding helps to capture abnormal situations that may be overlooked by a single feature, and improves the sensitivity and accuracy of fault detection.

[0025] 3. In practical applications, the multidimensional feature extraction method of the present invention can also extract the actual fault characteristics of the grid-side converter in the charging pile, such as overload, abnormal temperature and voltage fluctuation, to perform more comprehensive fault fusion and provide more reliable diagnostic support. This not only optimizes the fault handling process of the charging pile, but also improves the overall safety and service stability of the charging pile, providing users with a more efficient and safe charging experience. And the present invention has good feasibility in actual charging pile fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a schematic diagram of the fault diagnosis principle of the charging pile grid-side converter based on multi-dimensional feature extraction of the present invention;

[0027] Figure 2 It is a curve diagram of input current, output voltage and battery-side LLC circuit charging current when the grid-side converter of the present invention operates normally;

[0028] Figure 3 It is a curve diagram of input current, output voltage and charging current of LLC circuit on battery side when a single tube failure occurs in the grid-side converter of the present invention at 0.4S;

[0029] Figure 4The graph is a graph of the input current, output voltage and charging current of the LLC circuit on the battery side when a double-tube fault occurs in the grid-side converter of the present invention at 0.4S. DETAILED DESCRIPTION

[0030] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings:

[0031] A charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction comprises the following steps:

[0032] Step 1: extract multi-dimensional fault feature data information from the grid-side converter and the subsequent LLC circuit, and pre-process the multi-dimensional fault feature data information to fuse it into a high-dimensional feature vector;

[0033] The step 1 of extracting multi-dimensional fault feature data information from the fault diagnosis process of the grid-side converter includes:

[0034] Including the peak value and average value of the grid-side converter input current, the peak value, average value, zero value and instantaneous value of the grid-side converter output voltage, the distortion degree, abnormal frequency component and waveform kurtosis characteristics of the LLC circuit charging current on the battery side.

[0035] In this embodiment, these data under each fault type represent different operating states of the system. These feature data are preprocessed, such as denoising and normalization, to ensure the quality and consistency of the data. These feature data are merged into a high-dimensional feature vector. This high-dimensional feature vector contains multi-dimensional information of input current, output voltage and charging current under different fault conditions, so that the model can comprehensively consider information in multiple dimensions for diagnosis.

[0036] Step 2: Use the high-dimensional feature vector fused in step 1 to train the machine learning model to obtain a trained charging pile grid-side converter fault diagnosis model;

[0037] The specific method of step 2 is:

[0038] A support vector machine (SVM) is selected to train the model. During the model training process, the hyperparameters of the algorithm are adjusted to optimize the model performance. The model is trained using training data to ensure that the model can effectively learn the characteristics of different fault modes and obtain a trained fault diagnosis model.

[0039] Step 3: Evaluate the generalization ability of the charging pile grid-side converter fault diagnosis model in step 2 through cross validation;

[0040] The specific method of step 3 is:

[0041] K-fold cross validation is used to evaluate the generalization ability of the model to ensure that the model performs consistently on different data subsets. The confusion matrix is ​​used to analyze the classification results of the model and the accuracy, recall, and F1 score indicators are calculated to evaluate the performance of the model on different fault categories. The ROC curve is drawn to evaluate the performance of the classification model.

[0042] In this embodiment, step 3 helps to determine whether the model is overfitting or underfitting. The confusion matrix is ​​used to analyze the classification results of the model, and the accuracy, recall rate, and F1 score indicators are calculated to evaluate the performance of the model on different fault categories.

[0043] The ROC curve (receiver operating characteristic curve) is drawn to evaluate the performance of the classification model. The classification ability of the model is judged by the area under the curve (AUC) value. The closer the AUC value is to 1, the better the classification performance of the model.

[0044] When evaluating the model, focus on whether the model after feature fusion is more advantageous than using only a single feature. The effectiveness of feature fusion can be verified by comparing the performance indicators of the fused feature model with those of the single feature model.

[0045] Step 4: Use the trained charging pile grid-side converter fault diagnosis model to diagnose the multi-dimensionally extracted fault features;

[0046] The specific method of step 4 is:

[0047] The multi-dimensional extracted fault features are diagnosed through the trained charging pile grid-side converter fault diagnosis model; different fault categories are classified by adding labels, and the fault category labels and related fault descriptions are output to determine the fault type, and the fault accuracy is determined through the confusion matrix method.

[0048] Fault accuracy is an important performance indicator to determine whether the multidimensional fault diagnosis strategy is effective.

[0049] The working principle of step 4 is:

[0050] Collect relevant input current, output voltage operation data, and sensor data from the grid-side converter and the subsequent LLC converter. Use feature fusion to combine multiple features into a feature vector. Standardize or normalize the extracted feature values ​​to ensure that they are within the same scale range to avoid certain features from having too much impact on the model results. Pass the preprocessed feature data as input to the model;

[0051] Based on the prediction results of the model, determine the possible fault types of the equipment or system. Use machine learning methods to diagnose the fault features extracted from multiple dimensions. Add labels to classify different fault categories, output fault category labels and related fault descriptions to determine the fault type, and use confusion matrix to determine the fault accuracy.

[0052] The working principle of the present invention is:

[0053] like Figure 1 As shown in the figure, multiple key information is extracted from the fault diagnosis process of the grid-side converter, including the peak value and mean value of the input current of the grid-side converter, the peak value, mean value, zero value and instantaneous value of the output voltage of the grid-side converter, the waveform distortion, abnormal frequency component and waveform kurtosis of the LLC circuit charging current on the battery side. These data under each fault type represent different operating states of the system. These feature data are preprocessed, such as denoising and normalization, to ensure the quality and consistency of the data. These feature data are merged into a high-dimensional feature vector. This high-dimensional feature vector contains multi-dimensional information of the input current, output voltage and charging current under different fault conditions, so that the model can comprehensively consider information of multiple dimensions for diagnosis. Subsequently, the fused high-dimensional feature vector is used to train the machine learning model. Support vector machine (SVM) is selected to train the model. During the model training process, the hyperparameters of the algorithm are adjusted to optimize the model performance, and the model is trained using training data to ensure that the model can effectively learn the characteristics of different fault modes.

[0054] Evaluate the generalization ability of the model through cross-validation (such as K-fold cross-validation) to ensure that the model performs consistently on different data subsets. This helps determine whether the model is overfitting or underfitting. Use the confusion matrix to analyze the classification results of the model and calculate the accuracy, recall, and F1 score indicators to evaluate the performance of the model on different fault categories. Draw the ROC curve (receiver operating characteristic curve) to evaluate the performance of the classification model. The classification ability of the model is judged by the area under the curve (AUC) value. The closer the AUC value is to 1, the better the classification performance of the model.

[0055] Figure 2 The input current, output voltage and battery side LLC circuit charging current of the grid-side converter when it is operating normally. It can be seen that the input current of the grid-side converter should be in phase with the input voltage. This means that the waveforms of the current and voltage are synchronized in time, with a high power factor, and the DC charging pile is working in a healthy state. Figure 3 The input current, output voltage and battery-side LLC circuit charging current of the grid-side converter when a single-tube fault occurs at 0.4S. It can be seen that when a single-tube fault occurs, the input current is severely distorted and the harmonics are large. The output voltage and charging current contain harmonics of specific frequencies after the fault occurs. Figure 4 The input current, output voltage and battery-side LLC circuit charging current of the grid-side converter when a double-tube fault occurs at 0.4S. It can be seen that when a double-tube fault occurs, the input current zero-value platform (the time when the current is 0) is large, but the high-frequency harmonic content is low. The output voltage and charging current also produce harmonics of specific frequencies, but their frequencies are different from the harmonic levels of single-tube faults. Therefore, the input current zero-value platform size and the output voltage harmonic frequency and charging current harmonic frequency can be used as feature components for multi-dimensional fault feature extraction to provide a basis for subsequent fault analysis.

[0056] The innovation of the present invention lies in:

[0057] Since a single feature may not fully reflect the state of the system, especially in complex power electronic systems, changes in a single feature may not accurately indicate a specific fault type or location. Through feature fusion, the information of multiple features can be combined, allowing the diagnostic system to fully understand the fault status from multiple perspectives.

[0058] The present invention overcomes the problem that sensor measurement data is often affected by noise, resulting in uncertainty in feature data. By combining multiple features, the influence of noise on a single feature is reduced, thereby improving the robustness of diagnosis.

[0059] Some features may be redundant or highly correlated, which may lead to information duplication, increased computational complexity, and even affect the diagnosis results. Through dimensionality reduction methods or statistical fusion, redundant information can be removed, the feature set can be simplified, and the most diagnostically valuable information can be retained.

[0060] The present invention reduces the misdiagnosis rate and missed diagnosis rate of faults by extracting multi-dimensional data features;

[0061] The present invention can capture the interactive information between different features, which is beneficial for identifying complex fault modes;

[0062] The present invention can effectively resist the influence of noise and improve the overall robustness;

[0063] The present invention can capture early signs of failure in the system and, by analyzing subtle changes in multiple features, warn of potential failures in advance, thus reducing equipment downtime and maintenance costs.

[0064] The present invention can predict the development path of the fault, so as to take preventive maintenance measures and extend the service life of the equipment. It should be emphasized that the embodiments of the present invention are illustrative rather than restrictive, so the present invention includes but is not limited to the embodiments described in the specific implementation manner, and any other implementation manner derived by those skilled in the art based on the technical solution of the present invention also belongs to the scope of protection of the present invention.

Claims

1. A charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction, characterized in that: The following steps are involved: Step 1: extract multi-dimensional fault feature data information from the grid-side converter and the subsequent LLC circuit, and pre-process the multi-dimensional fault feature data information to fuse it into a high-dimensional feature vector; Step 2: Use the high-dimensional feature vector fused in step 1 to train the machine learning model to obtain a trained charging pile grid-side converter fault diagnosis model; Step 3: Evaluate the generalization ability of the charging pile grid-side converter fault diagnosis model in step 2 through cross validation; Step 4: Use the trained charging pile grid-side converter fault diagnosis model to diagnose the multi-dimensionally extracted fault features.

2. A charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction according to claim 1, characterized in that: The step 1 of extracting multi-dimensional fault feature data information from the fault diagnosis process of the grid-side converter includes: Including the peak value and average value of the grid-side converter input current, the peak value, average value, zero value and instantaneous value of the grid-side converter output voltage, the distortion degree, abnormal frequency component and waveform kurtosis characteristics of the LLC circuit charging current on the battery side.

3. A charging pile grid-side converter fault diagnosis method based on multi-dimensional feature extraction according to claim 1, characterized in that: The specific method of step 2 is: Select support vector machine to train the model. During the model training process, adjust the algorithm's hyperparameters to optimize the model performance, and use training data to train the model to ensure that the model can effectively learn the characteristics of different fault modes and obtain a trained fault diagnosis model.

4. The method for diagnosing faults of a charging pile grid-side converter based on multidimensional feature extraction according to claim 1 is characterized in that: The specific method of step 3 is: K-fold cross validation is used to evaluate the generalization ability of the model to ensure that the model performs consistently on different data subsets. The confusion matrix is ​​used to analyze the classification results of the model and calculate the accuracy, recall, and F1 score indicators to evaluate the performance of the model on different fault categories. Draw the ROC curve to evaluate the performance of the classification model.

5. The method for diagnosing faults of a charging pile grid-side converter based on multidimensional feature extraction according to claim 1, characterized in that: The specific method of step 4 is: The multi-dimensional extracted fault features are diagnosed through the trained charging pile grid-side converter fault diagnosis model; different fault categories are classified by adding labels, and the fault category labels and related fault descriptions are output to determine the fault type, and the fault accuracy is determined through the confusion matrix method.

Images