Photovoltaic intelligent operation and maintenance management method and system based on cloud computing

Abstract

The invention discloses a photovoltaic intelligent operation and maintenance management method and system based on cloud computing, and relates to the technical field of photovoltaic and artificial intelligence, and the method comprises the steps: collecting the operation data and environment data of photovoltaic equipment through a sensor installed in a photovoltaic power station, the environment data comprises temperature, humidity, wind speed and illumination intensity; the operation data and the environment data are subjected to preprocessing and feature extraction through a time sequence analysis method, a comprehensive feature vector is obtained, and the time sequence analysis method fuses an ARI MA model and an exponential smoothing model; and the comprehensive feature vector is inputted to a photovoltaic fault detection model, abnormal data and a fault mode are automatically identified, the photovoltaic fault detection model is an SHAP model, and feature interaction items are fused in the SHAP model to enhance the interpretability of the SHAP model. According to the method, the accuracy and interpretability of the model are enhanced, feature selection and dimension reduction are promoted, and the calculation efficiency of the model is improved.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic and artificial intelligence technology, and in particular to a photovoltaic intelligent operation and maintenance management method and system based on cloud computing. Background Art

[0002] As an efficient and clean energy, solar energy has become an important part of the future power source. Distributed photovoltaic power stations are developing rapidly under the situation of energy conservation, emission reduction and green development. The operation and maintenance of distributed photovoltaic power stations indirectly affects the power quality of the distribution network, resulting in the loss of power generation of distributed photovoltaic power stations.

[0003] The operation and maintenance of photovoltaic power stations requires a lot of manpower, material and financial resources, but the existing operation and maintenance management methods have certain deficiencies in cost control, and the operation and maintenance costs need to be further reduced. In addition, since the operation and maintenance effect of distributed photovoltaic power stations is largely limited by the professional skills and operation level of the operation and maintenance personnel, the accuracy of the evaluation results is affected, which indirectly affects the power generation quality and economic benefits of distributed photovoltaic power stations.

[0004] Although existing photovoltaic operation and maintenance management methods have introduced technologies such as artificial intelligence and big data, the degree of intelligence in fault diagnosis and other aspects is still insufficient, lacking explainability and low computational efficiency. In addition, the failure modes of photovoltaic power stations are diverse and complex, and it is still difficult for AI to accurately identify and predict them. Due to the complexity and uncertainty of AI technology, the stability of its system is difficult to guarantee. For example, it may be affected by data quality, noise and outliers, resulting in unstable performance.

[0005] Therefore, it is necessary to establish a more accurate and effective photovoltaic intelligent operation and maintenance management method and system based on cloud computing. Summary of the invention

[0006] In view of the above problems, the present invention is proposed to provide a photovoltaic intelligent operation and maintenance management method and system based on cloud computing that overcomes all or at least part of the above problems.

[0007] According to one aspect of the present invention, a photovoltaic intelligent operation and maintenance management method based on cloud computing is provided, comprising:

[0008] Collecting operation data and environmental data of photovoltaic equipment through sensors installed in photovoltaic power stations, wherein the operation data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity;

[0009] Preprocessing and feature extraction of the operation data and the environmental data by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model;

[0010] The comprehensive feature vector is input into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated in the SHAP model to enhance the interpretability of the SHAP model.

[0011] Furthermore, the preprocessing and feature extraction of the operation data and the environment data by the time series analysis method to obtain the comprehensive feature vector further includes:

[0012] Extracting feature vectors reflecting changes in operating status and environmental factors from the pre-processed operating data and environmental data;

[0013] The prediction result of the ARIMA model is used as the input of the exponential smoothing model, or the parameters of the ARIMA model and the exponential smoothing model are jointly estimated to adjust the feature vector to obtain the comprehensive feature vector.

[0014] Furthermore, the calculation formula of the comprehensive feature vector is:

[0015]

[0016] Among them, Y t is the comprehensive feature vector at time t, X t is the eigenvector at time t, X t-1 is the eigenvector at time t-1, α, β and μ t are the parameters and random error terms of the ARIMA model, respectively, and h t is the parameter of the exponential smoothing model at time t, and γ is the preset weight threshold of the exponential smoothing model.

[0017] Furthermore, the fusion feature interaction term in the SHAP model further includes:

[0018] Defining feature interaction terms that have a common influence on the prediction results of the SHAP model;

[0019] The Shapley value of the feature interaction term is calculated, and the prediction result of the SHAP model is explained using the Shapley value.

[0020] Furthermore, the expression of the feature interaction term is:

[0021]

[0022] Among them, ω i is the weight parameter, representing the feature vector x i The contribution to the prediction results, x i is the value of the i-th feature, γij is the interaction parameter, representing the feature x i and feature x j The degree of influence of the interaction between the two features on the prediction results is used to measure the joint influence of two features on the prediction results when they appear at the same time, and n is the number of features.

[0023] Furthermore, the ARIMA model includes an outlier processing mechanism, and the outliers are adjusted according to the outlier processing mechanism to reduce the impact of the outliers on the prediction results of the ARIMA model.

[0024] Furthermore, the calculation formula of the outlier processing mechanism is:

[0025] y(t)=ω0+ω1x1(t)+ω2x2(t)+...+ω n x n (t)

[0026] Among them, y(t) is the predicted value at time point t, ω0 is the intercept, ω1x1(t)+ω2x2(t)+...+ω n x n (t) is a linear combination term, ω1 to ω n is the regression coefficient, x1(t) to x n (t) is the characteristic variable in the sequence data.

[0027] Furthermore, a comprehensive score is obtained by quantitatively scoring the operating status, fault type, fault frequency, fault impact range, number of maintenance records, environmental factors and equipment data of the photovoltaic power station;

[0028] The photovoltaic devices of the photovoltaic power station are divided into high priority, medium priority and low priority levels according to the comprehensive score.

[0029] Furthermore, the calculation formula of the comprehensive score is:

[0030] Score=w1×State+w2×Fault Type +w3×Fault Frequency +w4×Fault Impact +w5×Env+w6×Equipment+w7×Time+w8×Maintenance+w9×Usage+w 10 ×Operational Efficiency

[0031] Among them, Score is the comprehensive score, w1 to w 10 is the weight of each quantitative indicator, State is the operating state of the photovoltaic equipment, Fault TypeFor different types of faults, Fault Frequency is the fault frequency, Fault Impact is the fault impact range, Env is the environmental factor, Equipment is the physical state of the equipment, Time is the time, Maintenance is the number of maintenance records, Usage is the actual usage of the equipment, and OperationalEfficiency is the performance of the equipment.

[0032] According to another aspect of the present invention, a photovoltaic intelligent operation and maintenance management system based on cloud computing is provided, comprising:

[0033] A data acquisition module, used to collect operating data and environmental data of photovoltaic equipment through sensors installed in the photovoltaic power station, wherein the operating data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity;

[0034] A feature extraction module, used for preprocessing and feature extraction of the operation data and the environment data by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model;

[0035] A fault detection module is used to input the comprehensive feature vector into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated in the SHAP model to enhance the interpretability of the SHAP model.

[0036] According to the solution provided by the present invention, the operation data and environmental data of the photovoltaic equipment are collected by sensors installed in the photovoltaic power station, wherein the operation data include current, voltage and power, and the environmental data include temperature, humidity, wind speed and light intensity; the operation data and environmental data are preprocessed and feature extracted by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model; the comprehensive feature vector is input into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated in the SHAP model to enhance the interpretability of the SHAP model. The present invention enhances the accuracy and interpretability of the model, while promoting feature selection and dimensionality reduction, and improving the computational efficiency of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0038] Figure 1 A schematic diagram of a process of a photovoltaic intelligent operation and maintenance management method based on cloud computing according to an embodiment of the present invention is shown;

[0039] Figure 2 The schematic diagram shows the structure of the photovoltaic intelligent operation and maintenance management system based on cloud computing according to the embodiment of the present invention. DETAILED DESCRIPTION

[0040] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0041] Figure 1 The schematic diagram of the process of the photovoltaic intelligent operation and maintenance management method based on cloud computing in an embodiment of the present invention is shown. Specifically, the following steps are included:

[0042] Step S101, collecting operation data and environmental data of photovoltaic equipment through sensors installed in the photovoltaic power station, wherein the operation data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity.

[0043] In a photovoltaic power station, select appropriate locations to install sensors. These sensors can be existing commercial products or customized according to specific needs. Ensure that the sensors can accurately measure the required operating data and environmental data. The operating data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity. Install sensors that can monitor data such as current, voltage, power, temperature, humidity, wind speed and light intensity in real time in key areas and equipment of the photovoltaic power station. It helps to improve the operation and maintenance efficiency and management level of the photovoltaic power station and reduce the operation and maintenance costs. Select a suitable installation location to ensure that the sensors can accurately measure the required data. For example, for the measurement of current, voltage and power, install the sensor near the photovoltaic panel to directly measure the output current and voltage of the panel. For the measurement of temperature, humidity, wind speed and light intensity, install the sensor in the key areas of the photovoltaic power station, such as the surface of the panel, the bracket structure, etc. You can choose the appropriate sensor according to the type of data and accuracy requirements to be measured. For example, for the measurement of current and voltage, you can choose current transformers and voltage sensors; for the measurement of temperature and humidity, you can choose temperature sensors and humidity sensors; for the measurement of wind speed and light intensity, you can choose wind speed sensors and light sensors.

[0044] Step S102, preprocessing and feature extraction are performed on the operation data and the environment data by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model.

[0045] Specifically, the collected operation data and environmental data are cleaned and sorted, and outliers and missing values ​​are removed to ensure the quality and availability of the data. The time series analysis method is used to extract the features of the preprocessed data to obtain a comprehensive feature vector. The ARIMA model and the exponential smoothing model are combined to form a comprehensive time series analysis method, which can fully utilize the advantages of the ARIMA model in time series analysis and the advantages of the exponential smoothing model in processing non-stationary data.

[0046] Furthermore, the preprocessing and feature extraction of the operation data and the environment data by the time series analysis method to obtain the comprehensive feature vector further includes:

[0047] Extracting feature vectors reflecting changes in operating status and environmental factors from the pre-processed operating data and environmental data;

[0048] The prediction result of the ARIMA model is used as the input of the exponential smoothing model, or the parameters of the ARIMA model and the exponential smoothing model are jointly estimated to adjust the feature vector to obtain the comprehensive feature vector.

[0049] Specifically, according to the characteristics and business needs of the photovoltaic power station, select features related to the operating status and changes in environmental factors, such as statistical features such as the mean and variance of current, voltage, and power, and temporal features such as the trend, seasonality, and periodicity of temperature, humidity, wind speed, and light intensity.

[0050] The ARIMA model is a time series prediction model, and its full name is the Autoregressive Integrated Moving Average Model, which predicts future values ​​from the past and present values ​​of the time series. The advantage of the ARIMA model is that the model is simple and only requires endogenous variables without the need for other exogenous variables. In this embodiment, the prediction results of the ARIMA model are used as the input of the exponential smoothing model, which utilizes the high capture ability of the ARIMA model for short-term fluctuations and the predictive ability of the exponential smoothing model for long-term trends. Alternatively, the parameters of the ARIMA model and the exponential smoothing model are jointly estimated to adjust the eigenvector to obtain a better comprehensive eigenvector.

[0051] Furthermore, the calculation formula of the comprehensive feature vector is:

[0052]

[0053] Among them, Y t is the comprehensive feature vector at time t, X t is the eigenvector at time t, X t-1 is the eigenvector at time t-1, α, β and μ t are the parameters and random error terms of the ARIMA model, respectively. The random error term μ t It can help improve the fitting accuracy of the model and reduce the deviation of the model. t For the exponential smoothing model parameters at time t, select a suitable h t , so that the model can capture the long-term trend in the data, and γ is the preset weight threshold of the exponential smoothing model.

[0054] Furthermore, the ARIMA model includes an outlier processing mechanism, and the outliers are adjusted according to the outlier processing mechanism to reduce the impact of the outliers on the prediction results of the ARIMA model.

[0055] In this embodiment, it is necessary to identify outliers in the data. Outliers usually refer to data points that are obviously inconsistent with most data points. According to the nature and cause of the outliers, an appropriate processing strategy is selected. If the outliers deviate seriously from the normal range and cannot be reasonably explained, these data points are directly discarded; if the outliers are caused by data collection errors or recording errors, they are corrected.

[0056] Furthermore, the calculation formula of the outlier processing mechanism is:

[0057] y(t)=ω0+ω1x1(t)+ω2x2(t)+...+ω n x n (t)

[0058] Among them, y(t) is the predicted value at time point t, ω0 is the intercept, ω1x1(t)+ω2x2(t)+...+ω n x n (t) is a linear combination term, ω1 to ω n is the regression coefficient, x1(t) to x n (t) is the characteristic variable in the sequence data.

[0059] Step S103, inputting the comprehensive feature vector into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated into the SHAP model to enhance the interpretability of the SHAP model.

[0060] In this embodiment, the SHAP (SHapley Additive exPlanations) model is an explanatory model for explaining the prediction results of the machine learning model. The SHAP model is selected as a photovoltaic fault detection model for automatically identifying abnormal data and fault modes. In the SHAP model, the interpretability and accuracy of the model are enhanced by integrating feature interaction terms. The feature interaction terms take into account the interactions between different features and can more comprehensively explain the model prediction results. In practical applications, the SHAP model is regularly updated according to the operating status and data of the photovoltaic power station that is continuously monitored, which can further improve its accuracy and efficiency in identifying abnormal data and fault modes.

[0061] Furthermore, the fusion feature interaction term in the SHAP model further includes:

[0062] Defining feature interaction terms that have a common influence on the prediction results of the SHAP model;

[0063] The Shapley value of the feature interaction term is calculated, and the prediction result of the SHAP model is explained using the Shapley value.

[0064] In this embodiment, it is necessary to determine which features have interactions, which can be accomplished by, for example, statistical methods or other machine learning methods, which are not limited herein. Consider the correlation, synergy or dependency between features to identify feature interaction items that have a common impact on the prediction results. The specific form and mathematical expression of the feature interaction items can be determined by constructing specific functions and introducing new variables. Then, the Shapley value of the feature interaction item is calculated to measure the contribution of each feature interaction item to the model prediction results. The Shapley value is a concept in cooperative game theory, which is used to measure the contribution of each participant in the cooperative game to the value created by the cooperation. It distributes its share of income in the cooperation by measuring the marginal contribution of each participant in different alliances. The core idea is that the contribution of a participant should be measured by its marginal contribution in all possible alliances. The calculation method of the Shapley value includes traversing all possible subsets, calculating the expected marginal contribution of each member, and finally calculating its Shapley value based on the expected marginal contribution of each member. The Shapley value can fairly distribute the weight of each feature interaction item in the prediction result, taking into account all possible feature combinations. By analyzing the size and changing trend of the Shapley value, we can understand which feature interaction items have a significant impact on the prediction results, as well as the direction and degree of their impact.

[0065] Furthermore, the expression of the feature interaction term is:

[0066]

[0067] Among them, ω i is the weight parameter, representing the feature vector x i The contribution to the prediction results, x i is the value of the i-th feature, γ ij is the interaction parameter, representing the feature x i and feature x j The degree of influence of the interaction between the two features on the prediction results is used to measure the joint influence of two features on the prediction results when they appear at the same time, and n is the number of features.

[0068] Furthermore, a comprehensive score is obtained by quantitatively scoring the operating status, fault type, fault frequency, fault impact range, number of maintenance records, environmental factors and equipment data of the photovoltaic power station;

[0069] The photovoltaic devices of the photovoltaic power station are divided into high priority, medium priority and low priority levels according to the comprehensive score.

[0070] In this embodiment, a reasonable weight and scoring range can be set for each consideration factor according to the actual situation and expert opinions. For example, a good operating status can be given a higher score, and a high fault frequency can be given a lower score. Each factor is quantitatively scored according to the set standard. For example, a good operating status can be scored as 80, and 10 points are deducted for each fault frequency. The score of each factor is multiplied by its corresponding weight, and then the total is added to get the comprehensive score. For example, the operating status score is 80*0.3=24 points, and the fault frequency deduction is 10*0.2=-2 points, then the comprehensive score is 24+(-2)=22 points. According to the comprehensive score, set a threshold to divide the priority level of the equipment. For example, high priority: 90-100 points; medium priority: 70-89 points; low priority: 0-69 points. According to the divided priority levels, formulate corresponding maintenance and overhaul plans. For high-priority equipment, more frequent overhaul and maintenance can be arranged; for low-priority equipment, the frequency of maintenance can be appropriately reduced.

[0071] Furthermore, the calculation formula of the comprehensive score is:

[0072] Score=w1×State+w2×Fault Type +w3×Fault Frequency +w4×Fault Impact

[0073] +w5×Env+w6×Equipment+w7×Time+w8×Maintenance+w9×Usage+w 10 × Operational Efficiency, where Score is the comprehensive score, w1 to w 10 is the weight of each quantitative indicator, State is the operating state of the photovoltaic equipment, Fault Type For different types of faults, Fault Frequency is the fault frequency, Fault Impact is the fault impact range, Env is the environmental factor, Equipment is the physical state of the equipment, Time is the time, Maintenance is the number of maintenance records, Usage is the actual usage of the equipment, and OperationalEfficiency is the performance of the equipment.

[0074] According to the solution provided by the present invention, the operation data and environmental data of the photovoltaic equipment are collected by sensors installed in the photovoltaic power station, wherein the operation data include current, voltage and power, and the environmental data include temperature, humidity, wind speed and light intensity; the operation data and environmental data are preprocessed and feature extracted by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model; the comprehensive feature vector is input into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated in the SHAP model to enhance the interpretability of the SHAP model. The present invention enhances the accuracy and interpretability of the model, while promoting feature selection and dimensionality reduction, and improving the computational efficiency of the model.

Claims

1. A photovoltaic intelligent operation and maintenance management method based on cloud computing, characterized in that: include: Collecting operation data and environmental data of photovoltaic equipment through sensors installed in photovoltaic power stations, wherein the operation data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity; Preprocessing and feature extraction of the operation data and the environmental data by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model; The comprehensive feature vector is input into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated into the SHAP model to enhance the interpretability of the SHAP model.

2. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 1 is characterized in that: The preprocessing and feature extraction of the operation data and the environment data by the time series analysis method to obtain a comprehensive feature vector further includes: Extracting feature vectors reflecting changes in operating status and environmental factors from the pre-processed operating data and environmental data; The prediction result of the ARIMA model is used as the input of the exponential smoothing model, or the parameters of the ARIMA model and the exponential smoothing model are jointly estimated to adjust the feature vector to obtain the comprehensive feature vector.

3. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 1 is characterized in that: The calculation formula of the comprehensive feature vector is: Among them, Y t is the comprehensive feature vector at time t, X t is the eigenvector at time t, X t-1 is the eigenvector at time t-1, α, β and μ t are the parameters and random error terms of the ARIMA model, respectively, and h t is the parameter of the exponential smoothing model at time t, and γ is the preset weight threshold of the exponential smoothing model.

4. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 1 is characterized in that: The fusion feature interaction term in the SHAP model further includes: Defining feature interaction terms that have a common influence on the prediction results of the SHAP model; The Shapley value of the feature interaction term is calculated, and the prediction result of the SHAP model is explained using the Shapley value.

5. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 4 is characterized in that: The expression of the feature interaction term is: Among them, ω i is the weight parameter, representing the feature vector x i The contribution to the prediction results, x i is the value of the i-th feature, γ ij is the interaction parameter, representing the feature x i and feature x j The degree of influence of the interaction between the two features on the prediction results is used to measure the joint influence of two features on the prediction results when they appear at the same time, and n is the number of features.

6. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 3 is characterized in that: The ARIMA model includes an outlier processing mechanism, and the outliers are adjusted according to the outlier processing mechanism to reduce the impact of the outliers on the prediction results of the ARIMA model.

7. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 6 is characterized in that: The calculation formula of the outlier processing mechanism is: y(t)=ω0+ω1x1(t)+ω2x2(t)+...+ω n x n (t) Among them, y(t) is the predicted value at time point t, ω0 is the intercept, ω1x1(t)+ω2x2(t)+...+ω n x n (t) is a linear combination term, ω1 to ω n is the regression coefficient, x1(t) to x n (t) is the characteristic variable in the sequence data.

8. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 1 is characterized in that: A comprehensive score is obtained by quantitatively scoring the operating status, fault type, fault frequency, fault impact range, maintenance record number, environmental factors and equipment data of the photovoltaic power station; The photovoltaic devices of the photovoltaic power station are divided into high priority, medium priority and low priority levels according to the comprehensive score.

9. The photovoltaic intelligent operation and maintenance management method based on cloud computing according to claim 8 is characterized in that: The calculation formula of the comprehensive score is: Score=w1×State+w2×Fault Type +w3×Fault Frequency +w4×Fault Impact +w5×Env+w6×Equipment+w7×Time+w8×Maintenance+w9×Usage+w 10 ×OperationalEfficiency Among them, Score is the comprehensive score, w1 to w 10 is the weight of each quantitative indicator, State is the operating state of the photovoltaic equipment, Fault Type For different types of faults, Fault Frequency is the fault frequency, Fault Impact is the fault impact range, Env is the environmental factor, Equipment is the physical state of the equipment, Time is the time, Maintenance is the number of maintenance records, Usage is the actual usage of the equipment, and OperationalEfficiency is the performance of the equipment.

10. A photovoltaic smart operation and maintenance management system based on cloud computing, the system is based on the photovoltaic smart operation and maintenance management method based on cloud computing as claimed in any one of claims 1 to 9, characterized in that: include: A data acquisition module, used to collect operating data and environmental data of photovoltaic equipment through sensors installed in the photovoltaic power station, wherein the operating data includes current, voltage and power, and the environmental data includes temperature, humidity, wind speed and light intensity; A feature extraction module, used for preprocessing and feature extraction of the operation data and the environment data by a time series analysis method to obtain a comprehensive feature vector, wherein the time series analysis method integrates an ARIMA model and an exponential smoothing model; A fault detection module is used to input the comprehensive feature vector into a photovoltaic fault detection model to automatically identify abnormal data and fault modes, wherein the photovoltaic fault detection model is a SHAP model, and feature interaction terms are integrated in the SHAP model to enhance the interpretability of the SHAP model.

Images