A multi-source data driven variable mode photovoltaic micro-grid system fault online diagnosis method and system

Abstract

The application belongs to the technical field of energy, and discloses a multi-source data driven variable mode photovoltaic micro-grid system fault online diagnosis method and system, obtains original data and carries out outlier rejection and normalization processing; the processed data is subjected to signal screening and is divided into multiple subsets; time domain features are extracted from each subset and a feature set is generated by labeling; the feature set is input into multiple classification models for classification prediction to determine a target classification model; the target classification model is used by a digital signal processor to perform online classification monitoring on the fault to be detected, the diagnosis time is recorded, and the fault diagnosis result is indicated by an LED lamp. The application realizes continuous monitoring and instant analysis of the real-time state of the photovoltaic system, improves the fault detection speed and accuracy, ensures the correctness of the fault type judgment, quickly takes maintenance or adjustment measures, improves the system stability, and reduces downtime and operation loss.

Description

Technical Field

[0001] This invention belongs to, but is not limited to, the field of energy technology, and particularly relates to a method and system for online fault diagnosis of a multi-source data-driven variable mode photovoltaic microgrid system. Background Technology

[0002] In the energy sector, solar power generation is an important renewable energy technology that converts solar energy into electricity, achieving the efficient utilization of clean energy. Large-scale grid-connected photovoltaic (PV) power systems are commonly used to ensure efficient power delivery to users. However, failures in these systems can lead to not only economic and operational losses but also threats to environmental protection and grid stability. Therefore, implementing online fault classification and prediction is crucial for ensuring system stability and efficient operation. This helps in the timely detection and maintenance of the system, preventing potential faults and quickly resolving problems.

[0003] Fault diagnosis in photovoltaic microgrid systems mainly involves two aspects: the development of traditional fault diagnosis methods and data-driven fault classification systems. Early fault prediction typically employed physical equipment-based methods, using external instruments (such as thermal imaging devices) to monitor photovoltaic modules for fault diagnosis. However, this method has limitations; for example, the accuracy of fault detection drops significantly when there is insufficient sunlight or minor module mismatch. In contrast, data-driven methods collect operational data from photovoltaic modules through SCADA (Supervisory Control and Data Acquisition) systems, preprocessing and analyzing this data. This method is less dependent on sunlight conditions and ambient temperature, works effectively even at night, and the diagnostic and decision-making processes can be optimized through algorithm improvements.

[0004] Large-scale photovoltaic (PV) power generation systems are typically installed in remote areas far from city centers, making supervision and maintenance extremely inconvenient, especially in regions like Africa and the Middle East, where numerous PV systems are deployed in deserts, exposed to harsh environmental conditions for extended periods. These environmental factors increase the types and probability of failures, making the entire PV grid-connected system more susceptible to anomalies. Furthermore, the increasing frequency of fires globally in recent years means that if a fire breaks out in a PV system, the situation could worsen if timely feedback and intervention are not implemented. For the power system, a fire could lead to a complete system failure, severely impacting people's production and daily lives.

[0005] Photovoltaic system faults can be categorized into two types based on severity: acute faults and chronic faults. Acute faults, such as short circuits and open circuits, typically cause system output interruptions and require immediate resolution. Chronic faults, while having a less severe impact on the system, can still reduce system efficiency and stability if ignored over a long period. Furthermore, traditional MPPT (Maximum Power Point Tracking) algorithms struggle to accurately locate the maximum power point under low irradiance conditions, potentially leading to power point tracking oscillations or system instability. Voltage fluctuations under low irradiance conditions have a more significant impact on the system; switching to IPPT (Fixed Power Point Tracking) mode can improve system stability.

[0006] In summary, developing a multi-source data-driven online fault diagnosis method for variable-mode photovoltaic microgrid systems can not only address the shortcomings of existing technologies but also significantly improve the fault detection efficiency and operational reliability of photovoltaic systems, which is of great significance for promoting the development and application of photovoltaic technology.

[0007] Currently, the existing technologies have the following problems: physical detection methods require specific instruments, which are costly and cannot be used on a large scale; energy loss methods require continuous updates to the simulation model to reduce the output deviation from the actual photovoltaic array; IV curve-based methods cannot achieve real-time fault diagnosis due to the long measurement cycle of IV curves; there is limited research on online fault diagnosis methods for switching between MPPT and IPPT modes; and faults occurring at lower radiation levels are more likely to go undetected. Summary of the Invention

[0008] To address the problems of existing technologies, this invention provides a multi-source data-driven online fault diagnosis method and system for variable-mode photovoltaic microgrid systems. This effectively solves the problems of high cost, difficulty in large-scale application, and inability to achieve real-time fault diagnosis in existing technologies. This method enables continuous monitoring and immediate analysis of the real-time status of the photovoltaic system under different modes, significantly improving the speed and efficiency of fault detection, ensuring accurate fault type identification, and thus enabling rapid implementation of corresponding maintenance or adjustment measures. This improves system stability, reduces system downtime, lowers operational losses, and further enhances the reliability and operational efficiency of the photovoltaic microgrid system.

[0009] This invention is implemented as follows: a multi-source data-driven online fault diagnosis method for variable-mode photovoltaic microgrid systems, comprising:

[0010] S1: Obtain the raw data and perform outlier removal and normalization on the raw data;

[0011] S2: Filter the obtained processed data, divide the filtered signals into multiple subsets;

[0012] S3: Extract temporal features from each subset of multiple subsets, and label the temporal features to obtain the feature set;

[0013] S4: The feature set obtained in S3 is used as the input of multiple classification models. After classification prediction, the target classification model is determined.

[0014] S5: Based on the target classification model, the digital signal processor performs online classification monitoring of the fault to be detected and records the diagnosis time, and turns on a certain number and position of LEDs for the fault to be detected to obtain the fault diagnosis result.

[0015] Furthermore, in S1, obtaining the raw data includes:

[0016] The photovoltaic array output is generated by a solar photovoltaic array simulator. Faults are injected into the photovoltaic array. After multiple experiments, raw data is obtained from the sensor. The raw data includes two modes: MPPT and IPPT, and fault-free data and fault data.

[0017] Furthermore, in step S2, filtering the original signal includes:

[0018] Calculate the Euclidean distance between each signal in two modes, MPPT and IPPT, and use the Euclidean distance to filter the original signals, selecting the original signals with low similarity between the two modes as diagnostic signals.

[0019] Furthermore, the time-domain features include mean, standard deviation, skewness, kurtosis, maximum value, minimum value, peak-to-peak value, mean square value, amplitude factor, waveform factor, impact factor, margin factor, and energy.

[0020] Furthermore, in step S4, the feature set described in step S3 is used as input to multiple classification models for classification prediction, and then the target classification model is determined, including:

[0021] According to preset conditions, the fault-free data is labeled, and the fault data is then input into the multiple classification models for binary classification to obtain classification results, wherein the classification results include fault results and fault-free results.

[0022] Based on the classification results, the fault results are labeled and used as inputs to multiple classification models for multi-classification.

[0023] Furthermore, in step S5, a certain number and location of LEDs are activated for each type of fault to identify the fault at different locations.

[0024] Another objective of this invention is to provide a multi-source data-driven online fault diagnosis system for variable-mode photovoltaic microgrid systems, the system comprising:

[0025] The data preprocessing module is used to obtain raw data and perform outlier removal and normalization on the raw data;

[0026] The signal filtering module, connected to the data preprocessing module, is used to calculate the Euclidean distance between each signal of the two modes (MPPT and IPPT), and to filter the original signals using the Euclidean distance, selecting the original signals with low similarity between the two modes as diagnostic signals.

[0027] The feature extraction module, connected to the signal filtering module, is used to extract time-domain features from each subset of the multiple subsets, and to obtain a feature set by labeling the time-domain features.

[0028] The classification prediction module, connected to the feature extraction module, is used to take the feature set obtained by the feature extraction module as input to multiple classification models, perform classification prediction, and then determine the target classification model.

[0029] The online fault classification and monitoring module, connected to the classification and prediction module, is used to perform online fault classification and monitoring of the fault to be detected through a digital signal processor based on the target classification model, and to display the number and time of correct diagnosis of the fault data on a digital tube to obtain the fault diagnosis result.

[0030] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the online fault diagnosis method for multi-source data-driven variable mode photovoltaic microgrid system.

[0031] Another objective of this invention is to provide a computer-readable storage medium having a program stored thereon, characterized in that, when the program is executed by a processor, it is used to implement the steps of the online fault diagnosis method for the multi-source data-driven variable-mode photovoltaic microgrid system described above.

[0032] Another objective of this invention is to provide an information data processing terminal, characterized in that the information data processing terminal is used to implement the online fault diagnosis system for the multi-source data-driven variable mode photovoltaic microgrid system.

[0033] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0034] First, this invention solves the technical problems of high cost, inability to be used on a large scale, and difficulty in real-time fault diagnosis in the existing technology for online fault diagnosis of photovoltaic microgrid systems. It achieves the technical effects of enabling continuous monitoring and instant analysis of the real-time status of photovoltaic systems, improving fault detection speed, ensuring correct judgment of fault types, taking corresponding maintenance or adjustment measures more quickly and accurately, improving system stability, and reducing system downtime and operational losses.

[0035] Secondly, as supplementary evidence of the inventive step of the claims of this invention, it is also reflected in the following important aspects:

[0036] (1) The expected benefits and commercial value of the technical solution of this invention after transformation are as follows:

[0037] This invention significantly reduces downtime in photovoltaic microgrid systems through real-time fault diagnosis, thereby minimizing operational losses, improving system efficiency, and maximizing economic benefits. It enhances the stability and reliability of photovoltaic systems, promotes the widespread adoption and sustainable development of clean energy, helps reduce carbon emissions, and drives the application of green energy. This method is low-cost and scalable, and can be widely applied to large-scale photovoltaic power generation systems, especially suitable for deployment in remote areas or harsh environments, demonstrating broad market prospects.

[0038] (2) The technical solution of this invention fills a technical gap in the industry both domestically and internationally:

[0039] This invention is the first to combine multi-source data-driven approaches with variable-mode fault diagnosis, achieving precise switching and real-time diagnosis of MPPT and IPPT modes, significantly improving fault identification capabilities under low-irradiance conditions. A signal filtering method based on Euclidean distance is proposed to filter diagnostic signals with low similarity between modes, providing a new technical path for fault diagnosis of photovoltaic systems in complex environments. In the application of the classification model, a multi-layer classification architecture is introduced, enabling efficient classification of various fault types and improving diagnostic accuracy and efficiency. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the online fault diagnosis method for a multi-source data-driven photovoltaic microgrid system according to an embodiment of the present invention.

[0041] Figure 2 This is a diagram showing the signal filtering results using Euclidean distance according to an embodiment of the present invention.

[0042] Figure 3 This is a diagram showing the binary classification result using a CNN model in an embodiment of the present invention.

[0043] Figure 4 This is a diagram showing the binary classification results using the RF model in an embodiment of the present invention.

[0044] Figure 5 This is a diagram showing the result of binary classification using the KNN model in an embodiment of the present invention.

[0045] Figure 6 This is a diagram showing the result of binary classification using the DT model in an embodiment of the present invention;

[0046] Figure 7 This is a diagram showing the results of multi-class classification using a CNN model in an embodiment of the present invention.

[0047] Figure 8 This is a diagram showing the results of multi-class classification using the RF model in an embodiment of the present invention.

[0048] Figure 9 This is a diagram showing the results of multi-class classification using the KNN model in an embodiment of the present invention.

[0049] Figure 10 This is a diagram showing the results of multi-class classification using the DT model in an embodiment of the present invention.

[0050] Figure 11 This is a diagram showing the results of online binary classification of faults using the DT model in an embodiment of the present invention.

[0051] Figure 12 This is a time-based graph showing the results of online binary fault classification using the DT model in an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0053] like Figure 1 As shown, this embodiment of the invention provides a multi-source data-driven online fault diagnosis method for variable-mode photovoltaic microgrid systems, including:

[0054] S1: Obtain the raw data and perform outlier removal and normalization on the raw data;

[0055] S2: Filter the obtained processed data, divide the filtered signals into multiple subsets;

[0056] S3: Extract temporal features from each subset of multiple subsets, and label the temporal features to obtain the feature set;

[0057] S4: The feature set obtained in S3 is used as the input of multiple classification models. After classification prediction, the target classification model is determined.

[0058] S5: Based on the target classification model, the digital signal processor performs online classification monitoring of the fault to be detected and records the diagnosis time, and turns on a certain number and position of LEDs for the fault to be detected to obtain the fault diagnosis result.

[0059] Example 1

[0060] This embodiment provides a multi-source data-driven online fault diagnosis method for photovoltaic microgrid systems, the method comprising:

[0061] S1: Obtain the original data and perform outlier removal and normalization on the original data.

[0062] Furthermore, in S1, obtaining raw data includes: generating photovoltaic array output through a solar array simulator, injecting faults into the constructed photovoltaic microgrid system, conducting multiple experiments, and then obtaining raw data from sensors, wherein the raw data includes fault-free data and fault data.

[0063] Specifically, the process involves collecting and normalizing raw data. The steps are as follows: First, a photovoltaic array output is generated using a solar array simulator, and then faults are injected into the constructed photovoltaic microgrid system. Next, a fault is generated for each experiment, and fault data from multiple experiments is collected from sensors. For example, fault data (including no faults and seven different faults) from eight experiments under two modes (MPPT and IPPT) are collected using sensors. Preprocessing includes outlier removal and data normalization. After normalizing the raw data, the processed first set of data is obtained.

[0064] S2: Filter the processed data, divide the filtered signals, and obtain multiple subsets.

[0065] Furthermore, the original signals are screened, and the Euclidean distance between each signal of the two modes (MPPT and IPPT) is calculated. The original signals with low similarity between the two modes are selected as the first signal. This first signal is then divided into multiple subsets, meaning the normalized fault data is divided into multiple subsets for subsequent classification and prediction. In this embodiment, eight subsets are preferred, but the specific number can be set according to actual needs; no specific limitation is imposed in this embodiment. Figure 2 As shown, the Euclidean distance between each signal under the two modes (MPPT and IPPT) is calculated and a heat map is generated. It can be seen from the heat map that the Euclidean distance between the three signals Ic, Va, and Vb is the smallest under the two modes, that is, the dissimilarity of these three signals is the largest under the two modes. Therefore, these three signals are selected as the input signals for fault diagnosis of photovoltaic microgrid system under the different modes.

[0066] S3: Extract time-domain features from each subset of multiple subsets, and label the time-domain features to obtain the feature set.

[0067] Furthermore, the time-domain features include mean, standard deviation, skewness, kurtosis, maximum value, minimum value, peak-to-peak value, mean square value, amplitude factor, waveform factor, impact factor, margin factor, and energy.

[0068] Specifically, time-domain features are extracted from each subset of faults in S2. These time-domain features include mean, standard deviation, skewness, kurtosis, maximum value, minimum value, peak-to-peak value, mean square value, amplitude factor, waveform factor, impact factor, margin factor, and energy. These extracted fault features are then labeled to obtain a feature set.

[0069] S4: The feature set obtained in S3 is used as the input of multiple classification models. After classification prediction, the target classification model is determined.

[0070] Furthermore, in S4, the step of using the feature set obtained in S3 as input to multiple classification models for classification prediction and then determining the target classification model includes: labeling the fault-free data according to preset conditions; labeling the fault data and then inputting it into the multiple classification models for binary classification to obtain classification results, wherein the classification results include fault results and fault-free results; and labeling the fault results according to the classification results and using them as input to multiple classification models for multi-classification.

[0071] Specifically, the feature sets obtained in S3 are used as inputs to multiple classification models for classification prediction to determine the target classification model. The specific process is as follows: First, based on preset conditions, fault-free and faulty data are separated; that is, fault-free data is labeled, and faulty data is labeled. Then, the data is input into multiple classification models for binary classification to obtain classification results. These results include faulty and fault-free results. In this embodiment, four models are preferred, specifically CNN (Convolutional Neural Networks), KNN (K-Nearest Neighbors), RF (Random Forest), and DT (Decision Tree). The specific models can be set according to actual needs; no specific limitation is imposed in this embodiment. Finally, based on the classification results, faulty results are labeled and used as inputs to multiple classification models for multi-classification.

[0072] For example, the features obtained in S3 are used as input to the DT model for classification prediction. In this invention, faults are first binary classified to distinguish between fault-free and faulty systems, and then 14 types of faults are classified and predicted. Fault classification is performed using four prediction models, with the specific prediction steps as follows: First, when classifying faults, fault-free systems are labeled with the same label (label 1), and all fault data are labeled with the same label (label 2). These are then input into the four models for binary classification to initially indicate whether a fault has occurred. After binary classification, different labels are assigned to each fault type: labels 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, and 14. These labels are then input into the four models for multi-class classification. Figure 3 , 4 As shown in Figures 5 and 6, three models—CNN, RF, KNN, and DT—were used to perform binary classification on the dataset for verification and comparison. The results showed that the DT model had the most ideal diagnostic performance.

[0073] like Figure 7 , 8 As shown in Figures 9 and 10, the dataset was subjected to multi-class classification using three models: CNN, RF, KNN, and DT. The results were verified and compared, and the DT model was found to have the most ideal diagnostic performance.

[0074] The experiment was divided into four groups: one group obtained the binary and multi-class confusion matrices of the CNN model, another group obtained the binary and multi-class confusion matrices of the KNN model, another group obtained the binary and multi-class confusion matrices of the RF model, and the third group obtained the binary and multi-class confusion matrices of the DT model. The main objective was to compare the classification accuracy of the four models and select the model with the best classification performance for building an online diagnostic system. The DT model was also compared with classification results from other researchers to determine the model with the best classification performance for use in building the online diagnostic system.

[0075] like Figure 3 , 4 As shown in Figures 5 and 6, in the confusion matrix of the binary classification task, the DT model, after signal filtering and feature extraction, has a faster diagnosis time than the other three models. Similarly, in Figure 7 , 8 In the confusion matrices of the multi-classification task shown in Figures 9 and 10, the DT model outperforms the other three models in terms of both accuracy and time after applying the same signal filtering and feature extraction methods.

[0076] Consider the validation of various models on the same feature dataset. Based on the original data, useful signals were selected and multiple time-domain features were extracted. Therefore, the accuracy and time efficiency of the classification model depend crucially on the contribution of the signals and features to the model. During model training, adjustments to the model parameters ensured the model's effectiveness.

[0077] This invention can be implemented through hardware, software, or a combination of both. Hardware implementation may employ dedicated logic, while the software portion may be stored in memory and executed by a microprocessor or dedicated hardware. Those skilled in the art will understand that this invention can be implemented using computer instructions and / or processor control code, and can be stored on various media such as disks, CD / DVD-ROMs, programmable memories, or data signal carriers. The devices and modules can be implemented using hardware circuits such as integrated circuits, logic chips, semiconductors, and programmable hardware, or software executed by various types of processors, or a combination of both.

[0078] S5: Based on the target classification model, the digital signal processor performs online classification monitoring of the fault to be detected and records the diagnosis time, and turns on a certain number and position of LEDs for the fault to be detected to obtain the fault diagnosis result.

[0079] Furthermore, in S5, a certain number and location of LEDs are turned on for each type of fault to obtain fault diagnosis results.

[0080] Specifically, based on the target classification model, a digital signal processor is used to perform online classification monitoring of the fault to be detected and record the diagnosis time. The number of correctly diagnosed faults and the time are displayed on a digital tube, and a certain number and position of LEDs are turned on for the fault to be detected to obtain the fault diagnosis result.

[0081] The established model is used for online fault classification via a Digital Signal Processing (DSP) board. To achieve this, 14 fault types (including IGBT faults, feedback sensor faults, intermittent voltage sag faults, photovoltaic array mismatch faults, photovoltaic array open-circuit faults, and two different PI controller faults) under two classified modes (MPPT) are displayed on the online system for easy differentiation. Specifically, the number and time of correct fault diagnosis for each of the 14 fault types under the two classified modes are displayed, and different alarms are given via LEDs. For example, the 14 faults are replaced with different codes, and different positions and numbers of LEDs are activated when different faults occur, displayed on the LEDs to facilitate differentiation of faults in different locations. In the online fault detection system, after new fault data is input, the classification model determines the fault type and displays the number and time of correct diagnoses, activating different positions and numbers of LEDs for the detection system. The monitoring system then takes corresponding measures (disconnection or alarm) to facilitate maintenance by grid personnel.

[0082] Furthermore, in S5, the classification model is deployed online using Python software and Code Composer Studio (CCS), specifically as follows:

[0083] 1. Train the DT classification model in Python and save the trained model and test set data;

[0084] 2. Convert the model and test data saved in step 1 into C language code;

[0085] 3. Embed the C code from step 2 into the CCS project and compile it;

[0086] 4. After compilation is complete and there are no errors, burn the created CCS project into the digital signal processor (DSP) for online fault diagnosis.

[0087] like Figure 11 , 12 As shown, the selected online fault diagnosis model DT model was deployed on a digital signal processor for online fault diagnosis. The test set contained 140 faults, and the number of correct diagnoses was also 140, resulting in a 100% accuracy rate and a time of 0.0003 milliseconds.

[0088] This invention better understands and adapts to complex operating environments and can dynamically switch modes based on real-time operating conditions. For example, it employs a hybrid tracking algorithm that adaptively switches between MPPT and IPPT, maximizing the energy output efficiency of photovoltaic systems, improving operational stability and adaptability, while extending equipment lifespan and reducing maintenance costs. Its application can significantly improve the overall performance and economic benefits of photovoltaic systems under complex environments and diverse load demands. Building upon traditional fault classification methods, this invention establishes an online fault classification and prediction method. Through sensor data such as voltage, current, temperature, and illuminance, and the DT machine learning algorithm, it enables continuous monitoring and immediate analysis of the photovoltaic system's real-time status. This not only significantly improves the speed of fault detection but also ensures accurate fault type identification through precise fault classification, enabling faster and more accurate implementation of corresponding maintenance or adjustment measures, reducing system downtime and operational losses.

[0089] This invention, by utilizing data from multiple sensors and environmental information, enables a better understanding and adaptation to complex operating environments. For example, in photovoltaic systems operating in deserts or other harsh environments, the new technology can distinguish between performance fluctuations caused by environmental factors (such as sandstorms, extreme temperature changes, etc.) and actual equipment failures. Furthermore, it can dynamically switch modes based on real-time operating conditions; for instance, a hybrid tracking algorithm that can adaptively switch between MPPT and IPPT can maximize the energy output efficiency of the photovoltaic system, improve operational stability and adaptability, while extending equipment lifespan and reducing maintenance costs.

[0090] Furthermore, regarding the technological advancement and methodological originality, compared to traditional methods, this invention can more comprehensively and accurately analyze and process complex datasets to identify and classify system faults in real time. The originality of this technology lies in its comprehensiveness and multi-dimensional data analysis capabilities, making fault diagnosis more accurate and efficient. Regarding practical benefits and economic value: implementing this invention can significantly reduce the maintenance costs and downtime of photovoltaic systems, improving system operating efficiency and power output. In the long run, this technology, which improves efficiency and reliability, not only saves system operators significant costs but also helps enhance the market competitiveness of renewable energy and promotes the wider application of green energy. Regarding environmental impact and social value: by improving the reliability and efficiency of photovoltaic systems, this invention indirectly reduces energy waste and environmental pollution caused by system faults. On the other hand, the MPPT algorithm may not effectively find the maximum power point under low irradiance conditions, potentially leading to power tracking oscillations or system instability. Under low irradiance conditions, voltage fluctuations have a greater impact on the system; switching to IPPT mode during actual operation can improve operational stability. More importantly, this technology enhances the system's adaptability to harsh environmental conditions, especially in regions prone to extreme weather events, which is of great significance for global sustainable development and environmental protection. Regarding its role in promoting technological progress: The implementation of this invention promotes technological advancements in related technical fields, particularly in the integrated application of data processing, machine learning, and photovoltaic technology. This interdisciplinary technological innovation not only optimizes existing engineering practices but also provides new ideas and methods for research and applications in other high-tech fields.

[0091] Third, as supporting evidence of the inventive step of the claims of this invention, the following important aspects are also reflected: The expected benefits and commercial value of the technical solution of this invention after transformation are as follows: 1. Reduced maintenance costs: Accurate fault prediction and diagnosis can help the operation and maintenance team make maintenance decisions in advance, avoiding large-scale damage and expensive emergency repairs. In the long run, this will significantly reduce the total cost of maintenance and repair, optimizing capital expenditure. 2. Increased energy output: Timely diagnosis and repair of faults can maximize the effectiveness and output of photovoltaic systems, avoid energy loss due to faults, and improve overall energy efficiency. 3. Expanded market potential and competitive advantage: Adopting advanced fault diagnosis technology can serve as a competitive advantage in the market, enhancing corporate brand image and market share. This technology can be provided to customers as a value-added service, especially in a market environment that increasingly focuses on renewable energy solutions. 4. Compliance with regulatory requirements and environmental protection standards: Reducing energy waste and improving system efficiency comply with increasingly stringent global environmental and energy efficiency standards. For companies operating in regions with increasingly stringent environmental regulations, this technology can help them comply with regulations and may enjoy government incentives.

[0092] The specific application areas or related products of this invention.

[0093] On the other hand, this embodiment also provides a multi-source data-driven online fault diagnosis system for variable-mode photovoltaic microgrid systems. In one specific implementation, the multi-source data-driven online fault diagnosis system for variable-mode photovoltaic microgrid systems includes:

[0094] The data preprocessing module is used for outlier removal and normalization of the data.

[0095] The signal filtering module is used to filter the processed data and divide the filtered signals using Euclidean distance as the selection criterion.

[0096] The feature extraction module is used to perform temporal feature extraction for each subset;

[0097] The classification prediction module is used to perform classification prediction by taking the extracted features as input to the classification model.

[0098] The online fault classification and monitoring module uses a digital signal processor to perform online fault classification and monitoring based on the obtained fault classification prediction model, records the diagnosis time, and activates a different number of LEDs for each type of fault to distinguish the fault location or type.

[0099] This invention establishes an online fault diagnosis system using Python and a DSP board by extracting multiple time-domain features and comparing four classification models. The system first trains the model by extracting multiple time-domain features, then tests the model on a dataset. The classification results and diagnosis times are compared with other systems. The model with the best final classification performance is then used for fault diagnosis via the DSP board. This online system makes a significant contribution to the fault monitoring and diagnosis of long-distance photovoltaic microgrid systems.

[0100] This invention also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a multi-source data-driven online fault diagnosis method.

[0101] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of a multi-source data-driven online fault diagnosis method.

[0102] This invention also provides an information data processing terminal, which is used to implement a multi-source data-driven online fault diagnosis system.

[0103] This invention first preprocesses the acquired raw data, including outlier removal and normalization. Then, the processed data is filtered. In this embodiment, Euclidean distance is used for signal filtering, resulting in filtered signals Ic, Va, and Vb. Subsequently, time-domain features are extracted from the filtered signals, including mean, standard deviation, skewness, kurtosis, maximum value, minimum value, peak-to-peak value, mean square value, amplitude factor, waveform factor, impulse factor, margin factor, and energy. The extracted features are then used as a feature set and input into a classification model for classification. This invention uses four models: CNN, KNN, RF, and DT. The classification results are compared, and the DT model outperforms the other three models in both accuracy and time. Therefore, the DT model is deployed on a digital signal processor for online fault classification, and the classification results and time are displayed on a digital tube. A certain number and location of LEDs are activated for each type of fault to distinguish them, enabling staff to take timely rescue measures.

[0104] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0105] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-source data-driven variable mode photovoltaic micro-grid system fault online diagnosis method, characterized in that, The method comprises: S1: obtaining original data, performing outlier rejection and normalization processing on the original data; S2: performing signal screening on the obtained processed data, dividing the screened signals and obtaining a plurality of subsets; S3: extracting time domain features from each of the plurality of subsets, respectively, and obtaining a feature set after labeling the time domain features; S4: inputting the feature set obtained in S3 into a plurality of classification models, respectively, performing classification prediction, and determining a target classification model; S5: according to the target classification model, performing online fault classification monitoring on the to-be-detected fault through a digital signal processor, recording the diagnosis time, and turning on a certain number and position of LED lights for the to-be-detected fault to obtain a fault diagnosis result; In S1, the original data is obtained by: generating photovoltaic array output through a solar photovoltaic array simulator, injecting faults in the photovoltaic array, performing multiple experiments, and obtaining original data from sensors, wherein the original data includes two modes, MPPT and IPPT, and fault-free data and fault data; In S2, the original signal is screened by: calculating the Euclidean distance between the original signals of the two modes MPPT and IPPT, and screening the original signals using the Euclidean distance to select the original signals with low similarity between the two modes as diagnostic signals.

2. The multi-source data-driven variable pattern photovoltaic microgrid system fault online diagnosis method of claim 1, wherein, The time domain features include mean, standard deviation, skewness, kurtosis, maximum value, minimum value, peak-to-peak value, mean square value, amplitude factor, waveform factor, impact factor, margin factor, and energy.

3. The multi-source data-driven variable pattern photovoltaic microgrid system fault online diagnosis method of claim 1, wherein, In S4, after inputting the feature set in S3 into a plurality of classification models, respectively, performing classification prediction, and determining a target classification model, the method comprises: According to a preset condition, the fault-free data is labeled, the fault data is labeled, and the labeled data is input into the plurality of classification models for binary classification to obtain a classification result, wherein the classification result includes fault results and fault-free results; According to the classification result, the fault results are labeled and input into a plurality of classification models for multi-classification.

4. The multi-source data-driven variable pattern photovoltaic microgrid system fault online diagnosis method of claim 1, wherein, In S5, a certain number and position of LED lights are turned on for each fault to determine the fault at different positions.

5. A multi-source data driven variable mode photovoltaic microgrid system fault online diagnosis system, characterized in that, The system comprises: a data preprocessing module for obtaining original data, performing outlier rejection and normalization processing on the original data, wherein the original data is obtained by: generating photovoltaic array output through a solar photovoltaic array simulator, injecting faults in the photovoltaic array, performing multiple experiments, and obtaining original data from sensors, wherein the original data includes two modes, MPPT and IPPT, and fault-free data and fault data; a signal screening module connected with the data preprocessing module, for calculating the Euclidean distance between the original signals of the two modes MPPT and IPPT, screening the original signals using the Euclidean distance, and selecting the original signals with low similarity between the two modes as diagnostic signals; The feature extraction module is connected with the signal screening module, and is configured to extract time domain features for each of the multiple subsets respectively, and obtain a feature set after labeling the time domain features; The classification prediction module is connected with the feature extraction module, and is configured to take the feature set obtained by the feature extraction module as an input of multiple classification models respectively, perform classification prediction, and determine a target classification model; The fault online classification monitoring module is connected with the classification prediction module, and is configured to perform fault online classification monitoring on a to-be-detected fault according to the target classification model through a digital signal processor, display a number of correct diagnoses and a time on a digital tube for the to-be-detected fault data, and obtain a fault diagnosis result. 6.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that, When the computer program is executed by the processor, the processor is caused to perform the steps of the multi-source data driven variable mode photovoltaic micro-grid system fault online diagnosis method according to any one of claims 1-4.

7. A computer-readable storage medium having stored thereon a program, characterized in that, When the program is executed by the processor, the steps of the multi-source data driven variable mode photovoltaic micro-grid system fault online diagnosis method according to any one of claims 1-4 are implemented.

8. An information data processing terminal, characterized by The information data processing terminal is configured to implement the multi-source data driven variable mode photovoltaic micro-grid system fault online diagnosis system according to claim 5.

Images