Fault fusion diagnosis method of distributed photovoltaic system and related device

Abstract

The invention discloses a fault fusion diagnosis method of a distributed photovoltaic system and a related device, and belongs to the technical field of fault diagnosis, and the method comprises the steps: collecting sensor time sequence data and an infrared image of the photovoltaic system, and carrying out the preprocessing; extracting an original time sequence feature and a frequency domain feature of the sensor time sequence data, inputting the original time sequence feature and the frequency domain feature into a pre-trained SVM model to obtain a second probability vector, and inputting the infrared image into a pre-trained CNN model to obtain a first probability vector; performing weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; and based on a preset threshold, performing confidence calculation on the weighted score vector, and after a confidence condition is satisfied, obtaining a fault category, and completing fault fusion diagnosis. According to the invention, the problem of misjudgment or missed judgment easily caused by direct output of a single diagnosis result during fault diagnosis in the prior art can be solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and particularly relates to a fault fusion diagnosis method and related device for a distributed photovoltaic system. Background Technique

[0002] As a key infrastructure in the field of renewable energy, in practical applications, the photovoltaic system faces complex and variable environmental conditions, such as light intensity fluctuations, drastic temperature changes, humidity differences, etc. These factors cause various faults to easily occur in core components such as photovoltaic modules, inverters, and brackets. Common fault types include component voltage imbalance, inverter grid connection failure, PV array overvoltage, insulation performance deterioration, leakage current exceeding the standard, grid parameter anomalies, and hardware aging. These faults not only directly affect the power generation efficiency of the system but may also pose safety hazards and cause economic losses.

[0003] For the fault diagnosis of photovoltaic systems, existing technologies mostly rely on single sensor data sources or simple multi-source data fusion methods. Traditional fusion diagnosis methods usually adopt a fixed weight allocation mechanism, that is, the same weights are assigned to various sensor data under different working conditions. This static weight strategy ignores the impact of environmental dynamic changes on sensor performance. Therefore, the fixed weight allocation mechanism is difficult to meet the diagnostic requirements under complex working conditions, resulting in fluctuating diagnostic accuracy and insufficient robustness. In addition, existing methods generally lack an effective credibility evaluation and feedback mechanism, often directly outputting diagnostic conclusions without quantitatively evaluating the confidence of the conclusions. When there is noise interference, data conflict, or model uncertainty in sensor data, the lack of credibility evaluation easily leads to misjudgment or missed judgment, reducing the overall performance of the diagnostic system. Summary of the Invention

[0004] The purpose of the present invention is to provide a fault fusion diagnosis method and related device for a distributed photovoltaic system to solve the problem that existing technologies directly output a single diagnostic result during fault diagnosis, which is prone to misjudgment or missed judgment.

[0005] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a fault fusion diagnosis method for a distributed photovoltaic system includes the following steps: Collect the sensor time-series data and infrared images of the photovoltaic system and perform preprocessing; Extract the original time-series features and frequency-domain features of the sensor time-series data, and input them into a pre-trained SVM model to obtain a second probability vector. At the same time, input the infrared image into a pre-trained CNN model to obtain a first probability vector; Perform weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; After calculating the confidence of the weighted score vector based on a preset threshold and meeting the confidence condition, a fault category is obtained, and the fault fusion diagnosis is completed.

[0006] In some embodiments, the step of collecting the sensor time-series data and infrared images of the photovoltaic system and performing preprocessing specifically includes: Removing the outliers in the sensor time-series data, performing normalization processing, and then adding Gaussian noise to obtain a time-series data set; Scaling the infrared image, cropping it to a unified size, and then performing random rotation and brightness adjustment to obtain an infrared image data set.

[0007] In some embodiments, the backbone network of the pre-trained CNN model adopts the ResNet-18 architecture, the number of neurons in the last fully connected layer is equal to the number of fault types, and a Dropout layer is added after each fully connected layer.

[0008] In some embodiments, during the pre-training of the SVM model, after every N fault diagnoses, the accuracy of the SVM model is recalculated using the diagnosis results of N / 2 times, and the class weights of the SVM model are adjusted. When the accuracy is lower than the preset accuracy threshold for several consecutive times, the SVM model is retrained until the maximum number of training times is reached.

[0009] In some embodiments, the step of obtaining a weighted score vector by weighted fusion of the first probability vector and the second probability vector specifically includes: Recording the confidences of the CNN model and the SVM model respectively through a sliding window; The confidences of the CNN model and the SVM model are used to adjust the weights through the Bayesian update formula to obtain updated dynamic weights; Based on the updated dynamic weights, the first probability vector and the second probability vector are weighted and fused to obtain a weighted score vector.

[0010] In some embodiments, the step of obtaining a fault category after calculating the confidence of the weighted score vector and meeting the confidence condition specifically includes: When the confidence is greater than or equal to 0, output the fault category corresponding to the highest diagnostic score in the weighted score vector; When the confidence is less than 0, re-collect the sensor time-series data and infrared images of the photovoltaic system and perform the subsequent steps of preprocessing and the like.

[0011] In a second aspect, a fault fusion diagnosis system for a distributed photovoltaic system includes: A data acquisition module, configured to acquire the sensor time-series data and infrared images of the photovoltaic system and perform preprocessing; A dual-model fault diagnosis module, configured to extract the original time-series features and frequency-domain features of the sensor time-series data, and input them into a pre-trained SVM model to obtain a second probability vector, and at the same time input the infrared image into a pre-trained CNN model to obtain a first probability vector; A result fusion module, configured to perform weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; A confidence judgment module, configured to calculate the confidence of the weighted score vector based on a preset threshold, and obtain the fault category after meeting the confidence condition, thereby completing the fault fusion diagnosis.

[0012] In a third aspect, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable in the processor. When the processor executes the computer program, the steps of the fault fusion diagnosis method for a distributed photovoltaic system are implemented.

[0013] In a fourth aspect, a computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the fault fusion diagnosis method for a distributed photovoltaic system are implemented.

[0014] In a fifth aspect, a computer program product includes a computer program, characterized in that when the computer program is executed by a processor, the steps of the fault fusion diagnosis method for a distributed photovoltaic system are implemented.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention utilizes two heterogeneous data sources, namely sensor time-series data and infrared images. The time-series data reflects the time-series characteristics of the system operation state, while the infrared image captures the spatial characteristics such as the temperature distribution of the device. By processing the image and time-series data through two models, namely CNN and SVM, and finally performing weighted fusion, multi-granularity and multi-modal fault characteristics can be extracted, effectively making up for the information limitations of a single data source and improving the comprehensiveness and accuracy of diagnosis.

[0016] Furthermore, the present invention respectively records the confidence of the CNN model and the SVM model through a sliding window, and adjusts the weights through the Bayesian update formula to obtain updated dynamic weights; based on the updated dynamic weights, the first probability vector and the second probability vector are weighted and fused to obtain a weighted score vector. It can dynamically adjust the fusion weights of the CNN and SVM models according to the real-time working conditions and environmental conditions, ensuring that multi-source information can be optimally fused under different environments, and significantly improving the flexibility and environmental adaptability of diagnosis.

[0017] Furthermore, the present invention eliminates outliers from the time series data, normalizes it, and adds Gaussian noise to simulate the real data distribution and enhance the model's robustness to noise. The infrared images are scaled, cropped, randomly rotated, and their brightness is adjusted to increase data diversity and improve the model's generalization ability.

[0018] Furthermore, the CNN model of the present invention uses a ResNet-18 backbone network, whose residual connection structure effectively alleviates the problem of gradient vanishing and supports deeper feature extraction. A Dropout layer is added after each fully connected layer to randomly discard some neurons, prevent the model from overfitting, and improve the model's generalization performance and stability.

[0019] Furthermore, during the pre-training process of the SVM model, the model performance is evaluated regularly using recent diagnostic results, the class weights are adjusted dynamically, and re-training is triggered when the accuracy continuously drops below the threshold. This self-update mechanism ensures that the model can adapt to changes in the data distribution and maintain a high diagnostic accuracy in the long term.

[0020] Furthermore, the present invention calculates the confidence of the weighted score vector. The diagnostic result is output only when the confidence meets the preset threshold, otherwise, the re-acquisition and diagnostic processes are triggered. This mechanism effectively quantifies the reliability of the diagnostic result, avoids misjudgments and missed detections caused by low-quality data or model uncertainty, and significantly improves the diagnostic accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of a fault fusion diagnosis method for a distributed photovoltaic system provided in this embodiment; Figure 2 is a structural diagram of a fault fusion diagnosis system for a distributed photovoltaic system provided in this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To enable those skilled in the art to better understand the solution of the present invention, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. The content is an explanation of the present invention rather than a limitation.

[0023] It should be noted that the terms "including" and "having" and any variations thereof in the description and claims of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, systems, products, or devices.

[0024] This embodiment provides a fault fusion diagnosis method for a distributed photovoltaic system, including the following steps: S1, Data Acquisition and Preprocessing Collect the sensor time-series data and infrared images of the photovoltaic system in real-time monitoring. After removing the outliers in the sensor time-series data caused by sensor noise or communication interruption, perform Z-score normalization on the sensor time-series data, scale the pixel values of the infrared images using Min-Max to [0, 1], and crop them to a unified size.

[0025] Randomly rotate and adjust the brightness of the scaled and cropped infrared images for data augmentation; add Gaussian noise to the normalized sensor time-series data, and finally obtain the time-series dataset and infrared image dataset.

[0026] S2, Dual-Model Training and Validation S2.1, Construct a CNN (Convolutional Neural Network) Model The backbone network adopts the ResNet-18 architecture. Replace the last fully connected layer with a fully connected layer whose number of neurons is equal to the number of fault types, retain the original convolutional layer structure to extract image features, and add a Dropout layer after the fully connected layer to prevent overfitting.

[0027] Divide the enhanced infrared image dataset in the S1 stage into a training set, a validation set, and a test set, and convert the fault type labels into one-hot vectors, which are paired and stored with the infrared image data.

[0028] Input the training set into the CNN model for training; First, initialize the training parameters, set the initial learning rate, define the decay period, set the early stopping patience period, and record the current best validation set loss and best validation set accuracy; Start from the 1st epoch until the preset maximum epoch number is reached or the early stopping condition is triggered.

[0029] If the current validation accuracy is higher than the historical best accuracy, save the weights of the current model, and the previously saved best model file to ensure that the model with the best performance is always retained.

[0030] If the preset maximum epoch number is reached and the early stopping is not triggered, complete the training of all epochs. If the early stopping condition is triggered, terminate the training in advance. After the training is completed, obtain the best CNN model.

[0031] S2.2, Construct an SVM (Support Vector Machine) Model Extract the original time-series features and frequency-domain features of the enhanced time-series dataset in S1, and splice them into a high-dimensional feature vector, where the original time-series features are the calculated mean and variance, and the frequency-domain features are extracted through Fourier transform.

[0032] Hyperparameter optimization: The RBF kernel parameters and regularization parameters are selected through grid search. Five-fold cross-validation is used for verification, and the parameter combination with the highest average verification accuracy is selected as the optimal hyperparameters.

[0033] Initialize the SVM model with the optimized hyperparameters and adopt the one-against-one (OvO) strategy to handle multi-classification problems. Set class weights according to the proportion of various samples in the training set to alleviate the class imbalance problem.

[0034] During the training process, after the SVM model makes 100 diagnoses, recalculate the model accuracy with the latest 50 diagnosis results and adjust the class weights; if the accuracy of the trained SVM model is continuously lower than the preset accuracy threshold three times during weight update, repeat the above steps of feature extraction and hyperparameter optimization for model training.

[0035] S3, Real-time fusion diagnosis Input the real-time collected fault data. In the best CNN model, input the infrared image and output the first probability vector; in the SVM model, input the time series feature vector and output the second probability vector; Conduct initial weight allocation by analyzing the key indicators of accuracy and F1-score of the CNN model and the SVM model on the validation set; Set a fixed update period, specifically set the period by the number of diagnoses. Record the confidence levels of the most recent N fault diagnoses of the CNN model and the SVM model respectively through a sliding window. After each diagnosis, update the confidence level data in the sliding window. Based on the confidence level data, adjust the weights through the Bayesian update formula to obtain the updated dynamic weights; Finally, based on the updated dynamic weights, perform weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; S4, Confidence verification and optimization Calculate the confidence level according to the weighted score vector and the preset threshold. The confidence level is the difference between the highest diagnosis score in the weighted score vector and the preset threshold. If the confidence level ≥ 0, accept the current diagnosis result and output the fault category corresponding to the highest diagnosis score. If the confidence level < 0, do not accept the current diagnosis result and re-collect data for fault fusion diagnosis.

[0036] As Figure 2 shown, this embodiment also provides a fault fusion diagnosis system for a distributed photovoltaic system, including: A data acquisition module for collecting the sensor time series data and infrared images of the photovoltaic system and performing preprocessing; A dual-model fault diagnosis module, which is used to extract the original time-series features and frequency-domain features of the sensor time-series data, input them into a pre-trained SVM model to obtain a second probability vector, and at the same time input the infrared image into a pre-trained CNN model to obtain a first probability vector; A result fusion module, which is used to perform weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; A confidence judgment module, which is used to calculate the confidence of the weighted score vector based on a preset threshold, and after meeting the confidence condition, obtain the fault category to complete the fault fusion diagnosis.

[0037] The division of modules in the embodiments of the present invention is illustrative. It is only a logical function division. In actual implementation, there may be other division methods. In addition, in each embodiment of the present invention, each functional module may be integrated in a processor, or may exist separately physically, or two or more modules may be integrated in one module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0038] In this embodiment, a computer device is also provided. The computer device includes a processor and a memory. The memory is used to store a computer program (in this embodiment, the computer program includes a calculation component and an iteration component, and can perform model calculation and model update). The computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiments of the present invention can be used for the operation of a fault fusion diagnosis method for a distributed photovoltaic system.

[0039] This embodiment also provides a storage medium, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and, of course, the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and this storage space stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in this storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the fault fusion diagnosis method for a distributed photovoltaic system in the above embodiment.

[0040] This embodiment also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by the processor, it implements the corresponding steps of the fault fusion diagnosis method for a distributed photovoltaic system in the above embodiment.

[0041] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0042] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0043] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one process Figure 1 one process or a plurality of processes and / or boxes Figure 1 the functions specified in one box or a plurality of boxes.

[0044] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 one process or a plurality of processes and / or boxes Figure 1 the functions specified in one box or a plurality of boxes.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific embodiments of the present invention. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A fault fusion diagnosis method for a distributed photovoltaic system, characterized in that, The following steps are involved: Collect sensor time series data and infrared images of photovoltaic systems and perform preprocessing; Extracting original time series features and frequency domain features of the sensor time series data, and inputting them into a pre-trained SVM model to obtain a second probability vector, and simultaneously inputting the infrared image into a pre-trained CNN model to obtain a first probability vector; Performing weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; Based on a preset threshold, after the confidence calculation is performed on the weighted score vector and the confidence condition is met, the fault category is obtained and the fault fusion diagnosis is completed.

2. The fault fusion diagnosis method of a distributed photovoltaic system according to claim 1, characterized in that The step of collecting the sensor time series data and infrared images of the photovoltaic system and performing preprocessing specifically includes: Eliminate outliers in the sensor time series data, perform normalization processing, and add Gaussian noise to obtain a time series data set; The infrared images are scaled and cropped to a uniform size, and then randomly rotated and the brightness is adjusted to obtain an infrared image dataset.

3. A fault fusion diagnosis method for a distributed photovoltaic system according to claim 1, characterized in that The backbone network of the pre-trained CNN model adopts the ResNet-18 architecture, the number of neurons in the last fully connected layer is equal to the number of fault types, and a Dropout layer is added after each fully connected layer.

4. A fault fusion diagnosis method for a distributed photovoltaic system according to claim 1, characterized in that, During the pre-training of the SVM model, after each N fault diagnoses are performed, the accuracy of the SVM model is recalculated using N / 2 diagnosis results, and the category weight of the SVM model is adjusted. When the accuracy is lower than the preset accuracy threshold for several consecutive times, the SVM model is retrained until the maximum number of training times is reached.

5. A fault fusion diagnosis method for a distributed photovoltaic system according to claim 1, characterized in that The step of performing weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector specifically includes: The confidence of the CNN model and the SVM model are recorded respectively through a sliding window; The confidence of the CNN model and the SVM model is adjusted by the Bayesian update formula to obtain the updated dynamic weight; Based on the updated dynamic weight, the first probability vector and the second probability vector are weightedly fused to obtain a weighted score vector.

6. The fault fusion diagnosis method of a distributed photovoltaic system according to claim 1, characterized in that, The step of obtaining the fault category after the confidence calculation is performed on the weighted score vector and the confidence condition is satisfied specifically includes: When the confidence level is greater than or equal to 0, outputting the fault category corresponding to the highest diagnosis score in the weighted score vector; When the confidence level is less than 0, the sensor time series data and infrared images of the photovoltaic system are collected again and preprocessed and the subsequent steps are performed.

7. A fault fusion diagnosis system for a distributed photovoltaic system, characterized in that, include: Data acquisition module, used to collect sensor time series data and infrared images of photovoltaic systems and perform preprocessing; A dual-model fault diagnosis module, used for extracting original time series features and frequency domain features of the sensor time series data, and inputting them into a pre-trained SVM model to obtain a second probability vector, and simultaneously inputting the infrared image into a pre-trained CNN model to obtain a first probability vector; A result fusion module, used for performing weighted fusion on the first probability vector and the second probability vector to obtain a weighted score vector; The confidence judgment module is used to calculate the confidence of the weighted score vector based on a preset threshold, and after meeting the confidence condition, obtain the fault category to complete the fault fusion diagnosis.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable in the processor. When the processor executes the computer program, the steps of the fault fusion diagnosis method for a distributed photovoltaic system according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the fault fusion diagnosis method for a distributed photovoltaic system according to any one of claims 1 to 6 are implemented.

10. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the fault fusion diagnosis method for a distributed photovoltaic system according to any one of claims 1 to 6 are implemented.

Images