Offshore wind power low-frequency circuit breaker fault diagnosis method and related device

Abstract

The invention discloses an offshore wind power low-frequency circuit breaker fault diagnosis method and a related device, electrical parameter data, environmental data and mechanical data are fused and spliced into a multi-dimensional feature vector, various fault feature information of a low-frequency circuit breaker in a complex working environment is comprehensively captured, the fault sensitivity and recognition capability are improved, and the fault diagnosis accuracy is improved. The random forest algorithm is adopted to realize high-precision diagnosis under a small sample, limited low-frequency circuit breaker data is fully utilized, the overfitting problem is effectively avoided, the number of decision trees in the random forest is optimized, the classification performance of the model is improved, the ROC curve and the confusion matrix are adopted to evaluate the performance of the low-frequency circuit breaker fault diagnosis model, and the reliability of the low-frequency circuit breaker fault diagnosis model is improved. And the effectiveness and reliability of the fault diagnosis model are improved. The technical problems that an existing offshore wind power low-frequency circuit breaker fault diagnosis method based on a single data source is high in false detection rate and an offshore wind power low-frequency circuit breaker fault diagnosis method based on multi-modal data fusion has model overfitting are solved.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method and related apparatus for diagnosing faults in low-frequency circuit breakers for offshore wind power. Background Technology

[0002] Traditional fault diagnosis methods for low-frequency circuit breakers in offshore wind power typically rely on a single data source, such as extracting vibration spectrum features through Fast Fourier Transform (FFT) or using thresholding methods to monitor SCADA parameter exceedances. However, these methods struggle to capture the implicit correlations between multimodal data, resulting in false positive rates generally exceeding 20%. Existing multimodal data fusion methods also exhibit significant limitations. Early fusion methods directly spliced ​​together raw data from different modes, such as merging 30-dimensional SCADA parameters with 1024-dimensional vibration waveforms, ignoring the misalignment of data spatiotemporal scales. This led to the curse of dimensionality and overfitting during model training, increasing test set loss by approximately 50%. While later fusion strategies mitigated this problem by processing each modal data independently and then voting on the results, they lost fine-grained cross-modal feature interactions, such as the spatiotemporal correlation between transient impacts in vibration signals and localized temperature rises in infrared images. Therefore, improving the fault identification capability of low-frequency circuit breakers in offshore wind power and reducing the false positive rate is a pressing technical challenge for those skilled in the art. Summary of the Invention

[0003] This invention provides a fault diagnosis method and related apparatus for offshore wind power low-frequency circuit breakers, which solves the technical problems of high false detection rate in existing fault diagnosis methods for offshore wind power low-frequency circuit breakers based on a single data source and model overfitting in fault diagnosis methods for offshore wind power low-frequency circuit breakers based on multi-modal data fusion.

[0004] In view of this, the first aspect of the present invention provides a fault diagnosis method for low-frequency circuit breakers in offshore wind power, comprising:

[0005] Collect electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker. The electrical parameter data includes the opening and closing coil current, contact voltage, and arc energy. The environmental data includes salt spray concentration, temperature, and humidity.

[0006] The electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker are normalized and concatenated into a multi-dimensional feature vector, and then dimensionality reduction is performed to obtain the original dataset of the low-frequency circuit breaker.

[0007] The original dataset is divided into a training set and a test set, and the training set is further divided into multiple data subsets.

[0008] Decision trees for random forests are trained based on multiple subsets of data;

[0009] The number of decision trees in the random forest is optimized to obtain a trained random forest model;

[0010] The trained random forest model was tested using a test set to obtain a fault diagnosis model for low-frequency circuit breakers.

[0011] The performance of the fault diagnosis model for low-frequency circuit breakers was evaluated using ROC curves and confusion matrices.

[0012] Optionally, the decision trees of the random forest are trained based on multiple subsets of data, including:

[0013] The number of features in the decision tree nodes of a random forest is determined from multiple data subsets using k-fold cross-validation.

[0014] A single decision tree is trained using the number of features in a predetermined decision tree node.

[0015] Optionally, the training set can be divided into multiple data subsets, including:

[0016] The training set is divided into multiple data subsets using a sampling method with replacement.

[0017] A second aspect of the present invention provides a fault diagnosis device for low-frequency circuit breakers in offshore wind power, comprising:

[0018] The data acquisition module is used to collect electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker. The electrical parameter data includes the opening and closing coil current, contact voltage, and arc energy, while the environmental data includes salt spray concentration, temperature, and humidity.

[0019] The data processing module is used to normalize the electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker and then concatenate them into a multi-dimensional feature vector, and perform dimensionality reduction processing to obtain the original dataset of the low-frequency circuit breaker.

[0020] The dataset partitioning module is used to divide the original dataset into training and test sets, and to further divide the training set into multiple data subsets.

[0021] The decision tree training module is used to train decision trees for random forests based on multiple subsets of data.

[0022] The optimization module is used to optimize the number of decision trees in the random forest to obtain a trained random forest model.

[0023] The model testing module is used to test the trained random forest model using a test set to obtain a low-frequency circuit breaker fault diagnosis model.

[0024] The evaluation module is used to evaluate the performance of the low-frequency circuit breaker fault diagnosis model using ROC curves and confusion matrices.

[0025] The fault diagnosis module is used to diagnose faults in offshore wind power low-frequency circuit breakers based on a low-frequency circuit breaker fault diagnosis model.

[0026] Optionally, the decision tree training module is specifically used for:

[0027] The number of features in the decision tree nodes of a random forest is determined from multiple data subsets using k-fold cross-validation.

[0028] A single decision tree is trained using the number of features in a predetermined decision tree node.

[0029] Optionally, the optimization module is specifically used for:

[0030] The number of decision trees in the random forest is optimized to obtain a well-trained random forest model, including:

[0031] The k-fold cross-validation method is used to optimize the number of decision trees in the random forest, and the optimal number of decision trees is determined to obtain the trained random forest model.

[0032] Optionally, the dataset partitioning module is specifically used for:

[0033] The original dataset is divided into a training set and a test set. The training set is further divided into multiple data subsets using a sampling method with replacement.

[0034] A third aspect of the present invention provides a fault diagnosis device for low-frequency circuit breakers in offshore wind power, the device comprising a processor and a memory:

[0035] The memory is used to store program code and transmit the program code to the processor;

[0036] The processor is used to execute the fault diagnosis method for offshore wind power low-frequency circuit breakers according to any one of the first aspects of the present invention, based on the instructions in the program code.

[0037] A fourth aspect of the present invention provides a computer-readable storage medium for storing program code for executing the fault diagnosis method for offshore wind power low-frequency circuit breakers according to any one of the first aspects of the present invention.

[0038] As can be seen from the above technical solutions, the present invention has the following advantages:

[0039] This invention provides a fault diagnosis method for low-frequency circuit breakers in offshore wind power systems. It integrates electrical parameter data, environmental data, and mechanical data, concatenating them into a multi-dimensional feature vector to comprehensively capture various fault characteristics of low-frequency circuit breakers under complex operating environments. This improves the sensitivity and identification capability of faults, reduces the false detection rate, and employs a random forest algorithm to achieve high-precision diagnosis with small samples. It fully utilizes limited low-frequency circuit breaker data, effectively avoiding overfitting, and optimizes the number of decision trees in the random forest to improve the model's classification performance. ROC curves and confusion matrices are used to evaluate the performance of the low-frequency circuit breaker fault diagnosis model, improving its effectiveness and reliability. This invention solves the technical problems of high false detection rates in existing fault diagnosis methods for offshore wind power low-frequency circuit breakers based on single data sources and model overfitting in fault diagnosis methods based on multi-modal data fusion. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating a method for diagnosing faults in low-frequency circuit breakers for offshore wind power, as provided in an embodiment of the present invention.

[0042] Figure 2 This is a logic block diagram of the offshore wind power low-frequency circuit breaker fault diagnosis method provided in the embodiments of the present invention;

[0043] Figure 3 This is a schematic diagram of the decision-making process of the random forest algorithm provided in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of a fault diagnosis device for a low-frequency circuit breaker in offshore wind power provided in an embodiment of the present invention. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] For easier understanding, please refer to Figure 1 and Figure 2 This invention provides an embodiment of a fault diagnosis method for low-frequency circuit breakers in offshore wind power, comprising:

[0047] Step 101: Collect electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker. The electrical parameter data includes the opening and closing coil current, contact voltage, and arc energy. The environmental data includes salt spray concentration, temperature, and humidity.

[0048] It should be noted that, in this embodiment of the invention, electrical parameter data of the low-frequency circuit breaker are collected, including the opening and closing coil current, contact voltage, and arc energy, and the time-frequency domain characteristics of the opening and closing coil current, contact voltage, and arc energy are extracted. Environmental data such as salt spray concentration, temperature, and humidity are collected to quantify the impact of environmental corrosion. Mechanical data include: contact opening distance, contact overtravel, and contact pressure.

[0049] Step 102: Normalize the electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker and concatenate them into a multi-dimensional feature vector. Then, perform dimensionality reduction processing to obtain the original dataset of the low-frequency circuit breaker.

[0050] It should be noted that the electrical, environmental, and mechanical data of the low-frequency circuit breaker are normalized and concatenated into a multi-dimensional feature vector. Principal component analysis or deep learning methods such as autoencoders are then used to reduce the dimensionality of this multi-dimensional feature vector, yielding the original dataset of the low-frequency circuit breaker. Since PCs or servers are difficult to deploy directly on the circuit breaker and cannot be implemented in embedded systems, the collected data can be wirelessly transmitted to the cloud for storage.

[0051] Step 103: Divide the original dataset into a training set and a test set, and divide the training set into multiple data subsets.

[0052] It should be noted that the original dataset is divided into two parts, corresponding to the test set and the training set, and then the training set is further divided into multiple data subsets. A sampling method with replacement is used to divide the training set into multiple data subsets to facilitate the determination of different parameters for subsequent model training stages.

[0053] Step 104: Train the decision tree of the random forest based on multiple data subsets.

[0054] It should be noted that k-fold cross-validation is used to determine the number of features in each node of the decision tree from multiple data subsets, thus preparing for the construction of the decision tree. A single decision tree is then trained using the determined number of features. The calculation principle is as follows: when optimizing the parameters of the decision tree using k-fold cross-validation, the original dataset M is first uniformly divided into k disjoint data subsets, satisfying M = ... ∪ ∪...∪ Furthermore, the intersection of any two subsets is empty. During the validation process, each subset is used as the test set in turn, and the remaining k-1 subsets are used as the training set: initially using... Build a decision tree model using the test set and the rest as the training set, and obtain performance metrics. Then rotated The test set is used for the test set, and the rest are used for the training set. This process continues until each subset is used as a test set once. Finally, the performance metrics obtained from k validations are averaged. As a basis for evaluating the overall performance of the model, by comparing the values ​​corresponding to different k values The optimal parameter configuration can be determined by the value, and this method effectively avoids the overfitting problem by making full use of the limited low-frequency circuit breaker data.

[0055] Step 105: Optimize the number of decision trees in the random forest to obtain a trained random forest model.

[0056] It should be noted that k-fold cross-validation is then used again to optimize the number of decision trees in the random forest, thereby determining the optimal number of decision trees and improving the model's performance and generalization ability. For example... Figure 3 As shown, based on the determined number of features and the optimal number of decision trees, the random forest model is trained to integrate the advantages of multiple decision trees and improve the accuracy of fault diagnosis.

[0057] Step 106: Test the trained random forest model using the test set to obtain the low-frequency circuit breaker fault diagnosis model.

[0058] It should be noted that the trained random forest model is tested using the test set partitioned in step 103 to obtain test results. If the test results show that the model meets the requirements, the trained random forest model is determined as the low-frequency circuit breaker fault diagnosis model; if the test results show that the model does not meet the requirements, the random forest model needs to be trained again.

[0059] Step 107: Use ROC curves and confusion matrix to evaluate the performance of the low-frequency circuit breaker fault diagnosis model.

[0060] It should be noted that the performance of the low-frequency circuit breaker fault diagnosis model is evaluated. The ROC (Receiver Operating Characteristic) curve and confusion matrix are used to measure the model's performance and assess its effectiveness and reliability in fault diagnosis. In evaluating model performance, the false positive rate (FPR) and true positive rate (TPR) are also used, and their calculation methods are as follows:

[0061]

[0062] Wherein, TP is to identify positive examples as positive examples, FP is to identify negative examples as positive examples, FN is to identify positive examples as negative examples, and TN is to identify negative examples as negative examples.

[0063] The curve plotted with the false positive rate on the x-axis and the true positive rate on the y-axis is called the Receiver Operating Characteristic (ROC) curve, which provides a more intuitive measure of a model's classification performance. A larger area under the curve indicates better classification performance, and vice versa. The area under the curve is represented by the AUC value; a perfect classification performance results in an AUC of 1, while poorer performance corresponds to a smaller AUC.

[0064] Step 108: Based on the low-frequency circuit breaker fault diagnosis model, perform fault diagnosis on the low-frequency circuit breaker of offshore wind power.

[0065] It should be noted that after obtaining the fault diagnosis model of the low-frequency circuit breaker, the fault diagnosis model of the low-frequency circuit breaker of offshore wind power is used to diagnose the fault and obtain the fault diagnosis results.

[0066] This invention provides a fault diagnosis method for low-frequency circuit breakers in offshore wind power systems. This method integrates electrical parameter data, environmental data, and mechanical data, concatenating them into a multi-dimensional feature vector. This comprehensively captures various fault characteristics of low-frequency circuit breakers under complex operating environments, improving sensitivity and identification capabilities while reducing false detection rates. A random forest algorithm is employed to achieve high-precision diagnosis with small sample sizes, fully utilizing limited low-frequency circuit breaker data to effectively avoid overfitting. The number of decision trees in the random forest is optimized to improve the model's classification performance. ROC curves and confusion matrices are used to evaluate the performance of the low-frequency circuit breaker fault diagnosis model, enhancing its effectiveness and reliability. This invention addresses the technical problems of high false detection rates in existing fault diagnosis methods for offshore wind power low-frequency circuit breakers based on single data sources and model overfitting in fault diagnosis methods based on multi-modal data fusion.

[0067] For easier understanding, please refer to Figure 4 This invention provides an embodiment of a fault diagnosis device for low-frequency circuit breakers in offshore wind power, comprising:

[0068] The data acquisition module is used to collect electrical parameter data, environmental data, and mechanical data of the low-frequency circuit breaker. The electrical parameter data includes the opening and closing coil current, contact voltage, and arc energy, while the environmental data includes salt spray concentration, temperature, and humidity.

[0069] The data processing module is used to normalize the electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker and then concatenate them into a multi-dimensional feature vector, and perform dimensionality reduction processing to obtain the original dataset of the low-frequency circuit breaker.

[0070] The dataset partitioning module is used to divide the original dataset into training and test sets, and to further divide the training set into multiple data subsets.

[0071] The decision tree training module is used to train decision trees for random forests based on multiple subsets of data.

[0072] The optimization module is used to optimize the number of decision trees in the random forest to obtain a trained random forest model.

[0073] The model testing module is used to test the trained random forest model using a test set to obtain a low-frequency circuit breaker fault diagnosis model.

[0074] The evaluation module is used to evaluate the performance of the low-frequency circuit breaker fault diagnosis model using ROC curves and confusion matrices.

[0075] The fault diagnosis module is used to diagnose faults in offshore wind power low-frequency circuit breakers based on a low-frequency circuit breaker fault diagnosis model.

[0076] In one embodiment, the decision tree training module is specifically used for:

[0077] The number of features in the decision tree nodes of a random forest is determined from multiple data subsets using k-fold cross-validation.

[0078] A single decision tree is trained using the number of features in a predetermined decision tree node.

[0079] In one embodiment, the optimization module is specifically used for:

[0080] The number of decision trees in the random forest is optimized to obtain a well-trained random forest model, including:

[0081] The k-fold cross-validation method is used to optimize the number of decision trees in the random forest, and the optimal number of decision trees is determined to obtain the trained random forest model.

[0082] In one embodiment, the dataset partitioning module is specifically used for:

[0083] The original dataset is divided into a training set and a test set. The training set is further divided into multiple data subsets using a sampling method with replacement.

[0084] This invention provides an embodiment of a fault diagnosis device for low-frequency circuit breakers in offshore wind power, the device comprising a processor and a memory:

[0085] The memory is used to store program code and transmit the program code to the processor;

[0086] The processor is used to execute any one of the offshore wind power low-frequency circuit breaker fault diagnosis methods in the embodiments of the present invention according to the instructions in the program code.

[0087] This invention provides a computer-readable storage medium for storing program code for executing any of the offshore wind power low-frequency circuit breaker fault diagnosis methods described in the embodiments of the offshore wind power low-frequency circuit breaker fault diagnosis methods provided in this invention.

[0088] The offshore wind power low-frequency circuit breaker fault diagnosis device, equipment, and computer-readable storage medium provided in this invention are all used to execute the offshore wind power low-frequency circuit breaker fault diagnosis method provided in this invention. Their principles and the technical effects achieved are the same as those of the offshore wind power low-frequency circuit breaker fault diagnosis method provided in this invention, and will not be repeated here.

[0089] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing a fault of a low-frequency circuit breaker of an offshore wind power, characterized in that, The method comprises the following steps: Collecting electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker, the electrical parameter data comprising closing and opening coil current, contact voltage and arc energy, the environmental data comprising salt mist concentration, temperature and humidity; Normalizing the electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker to splice them into a multi-dimensional feature vector, and performing dimension reduction processing to obtain an original data set of the low-frequency circuit breaker; Dividing the original data set into a training set and a test set, and dividing the training set into a plurality of data subsets; Training decision trees of a random forest based on the plurality of data subsets; Optimizing the number of decision trees of the random forest to obtain a trained random forest model; Testing the trained random forest model by using the test set to obtain a low-frequency circuit breaker fault diagnosis model; Performing performance evaluation on the low-frequency circuit breaker fault diagnosis model by using an ROC curve and a confusion matrix; Performing fault diagnosis on the offshore wind power low-frequency circuit breaker based on the low-frequency circuit breaker fault diagnosis model.

2. The offshore wind power low-frequency circuit breaker fault diagnosis method according to claim 1, characterized in that, The method of training decision trees of a random forest based on a plurality of data subsets comprises the following steps: Determining the number of features in a decision tree node of the random forest from the plurality of data subsets by using a k-fold cross-validation method; Training a single decision tree by using the determined number of features in the decision tree node.

3. The offshore wind power low-frequency circuit breaker fault diagnosis method according to claim 2, characterized in that, The method of optimizing the number of decision trees of the random forest to obtain a trained random forest model comprises the following steps: Optimizing the number of decision trees in the random forest by using a k-fold cross-validation method to determine an optimal number of decision trees, and obtaining a trained random forest model.

4. The offshore wind power low-frequency circuit breaker fault diagnosis method according to claim 1, characterized in that, The method of dividing the training set into a plurality of data subsets comprises the following step: Dividing the training set into a plurality of data subsets by using a sampling method with replacement.

5. A marine wind power low frequency circuit breaker fault diagnosis device, characterized in that, The method comprises the following steps: A data collection module is configured to collect electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker, the electrical parameter data comprising closing and opening coil current, contact voltage and arc energy, the environmental data comprising salt mist concentration, temperature and humidity; A data processing module is configured to normalize the electrical parameter data, environmental data and mechanical data of the low-frequency circuit breaker to splice them into a multi-dimensional feature vector, and perform dimension reduction processing to obtain an original data set of the low-frequency circuit breaker; A data set division module is configured to divide the original data set into a training set and a test set, and divide the training set into a plurality of data subsets; A decision tree training module is configured to train decision trees of a random forest based on the plurality of data subsets; An optimization module is configured to optimize the number of decision trees of the random forest to obtain a trained random forest model; A model testing module is configured to test the trained random forest model by using the test set to obtain a low-frequency circuit breaker fault diagnosis model; An evaluation module is configured to perform performance evaluation on the low-frequency circuit breaker fault diagnosis model by using an ROC curve and a confusion matrix; A fault diagnosis module is configured to perform fault diagnosis on the offshore wind power low-frequency circuit breaker based on the low-frequency circuit breaker fault diagnosis model.

6. The marine wind power low frequency circuit breaker fault diagnostic device according to claim 5, characterized in that, The decision tree training module is specifically configured to: Determine the number of features in a decision tree node of the random forest from the plurality of data subsets by using a k-fold cross-validation method; Train a single decision tree by using the determined number of features in the decision tree node.

7. The marine wind power low frequency circuit breaker fault diagnostic device according to claim 6, characterized in that, The optimization module is specifically configured to: The number of decision trees of the random forest is optimized to obtain a trained random forest model, including: The number of decision trees in the random forest is optimized by using k-fold cross validation method to determine the optimal number of decision trees, and a trained random forest model is obtained.

8. The marine wind power low frequency circuit breaker fault diagnostic device according to claim 5, characterized in that, The data set division module is specifically used for: Dividing the original data set into a training set and a test set, and dividing the training set into multiple data subsets by using a sampling method with replacement.

9. A marine wind power low frequency circuit breaker fault diagnostic device, characterized by The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the offshore wind power low-frequency circuit breaker fault diagnosis method according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium is used to store program code for executing the offshore wind power low-frequency circuit breaker fault diagnosis method.

Images