Photovoltaic power prediction and fault diagnosis system based on digital twinning

Abstract

The invention relates to a photovoltaic technology, in particular to a digital twinning-based photovoltaic power prediction and fault diagnosis system, which comprises a server, and is characterized in that the server carries out data acquisition through a data acquisition module and creates a virtual model corresponding to a photovoltaic system by using a virtual model creation module; the server maps the collected data into a virtual model through a data mapping module, and generates a digital twin model corresponding to the photovoltaic system by using a digital twin model generation module based on the virtual model of the mapped data; according to the technical scheme provided by the invention, the defects that in the prior art, the photovoltaic power generation power is difficult to accurately predict, and accurate fault positioning and fault diagnosis are difficult to carry out on the photovoltaic module can be effectively overcome.

Description

Technical Field

[0001] The present invention relates to photovoltaic technology, and in particular to a photovoltaic power prediction and fault diagnosis system based on digital twins. Background Art

[0002] With the continuous development of new energy technologies, new energy power generation, mainly photovoltaics, has gradually become an important part of my country's power supply. As the proportion of photovoltaic and other new energy installed capacity continues to increase, the uncertainty of its output power has brought about a series of scheduling and operation problems. Accurately predicting photovoltaic power generation is an effective means to reduce the impact of uncertainty. Photovoltaics are unstable energy sources with random fluctuations. Large-scale photovoltaic grid connection will inevitably affect the stability of the power grid system. The power dispatching system needs to effectively predict the photovoltaic power generation power in the future to reasonably allocate the output of power generation equipment and ensure the stable and reliable operation of the power grid system.

[0003] At present, the photovoltaic power prediction method has not been perfected. Due to the neglect of weather factors and the influence of cloud movement, the overall prediction accuracy is low, and there is a large difference between the photovoltaic power prediction results and the actual photovoltaic power generation. If the power dispatch is directly based on the photovoltaic power prediction results, it will inevitably cause large fluctuations in the operation of the power grid system. In addition, how to locate and diagnose faults in photovoltaic modules is also an important direction worthy of research in the field of photovoltaic power generation. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention provides a photovoltaic power prediction and fault diagnosis system based on digital twins, which can effectively overcome the defects of the prior art that it is difficult to accurately predict the photovoltaic power generation power, and it is difficult to accurately locate and diagnose the faults of photovoltaic components.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0008] A photovoltaic power prediction and fault diagnosis system based on digital twins, comprising a server, wherein the server collects data through a data acquisition module, and uses a virtual model creation module to create a virtual model corresponding to the photovoltaic system, wherein the server maps the collected data to the virtual model through a data mapping module, and uses a digital twin model generation module to generate a digital twin model corresponding to the photovoltaic system based on the virtual model of the mapped data;

[0009] The server calculates the solar shading sequence of the next time period by using the ground-based cloud map collected by the data acquisition module through the data analysis and processing module, and comprehensively judges the impact of solar shading according to the solar shading sequence of the next time period, and simultaneously obtains the actual photovoltaic power generation of the previous time period collected by the data acquisition module. The server constructs a first power generation prediction model through the first power generation prediction model construction module, and inputs the weather data collected by the data acquisition module into the trained first power generation prediction model using the first power generation prediction result acquisition module to obtain the first power generation prediction result; the server constructs a second power generation prediction model through the second power generation prediction model construction module, and inputs the actual photovoltaic power generation of the previous time period and the solar shading sequence of the next time period into the trained second power generation prediction model using the second power generation prediction result acquisition module to obtain the second power generation prediction result; the server generates a photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result through the power generation prediction result generation module;

[0010] The server compares the photovoltaic power generation prediction results of each photovoltaic component with the actual photovoltaic power generation power through the fault detection module, and determines whether the photovoltaic component has a fault based on the comparison result. The server constructs a fault diagnosis model through the fault diagnosis model, and uses the fault diagnosis result generation module to diagnose the fault type of the faulty photovoltaic component based on the trained fault diagnosis model to generate a fault diagnosis result. The server sends the fault diagnosis result to the client through the fault alarm module, receives the fault work order fed back by the client, and uses the fault work order issuance emergency repair module to issue the fault work order to the operation and maintenance workers for emergency repair.

[0011] Preferably, the data analysis and processing module calculates the solar occlusion sequence of the next period using the ground-based cloud map collected by the data collection module, and comprehensively determines the solar occlusion impact according to the solar occlusion sequence of the next period, including:

[0012] Based on the ground-based cloud image, the optical flow method is used to calculate the moving speed and direction of the cloud layer relative to the sun. According to the moving speed and direction, the sun blocking situation at each second-level time interval in the next period is calculated to obtain the sun blocking sequence for the next period.

[0013] The solar occlusion impact is comprehensively judged according to the solar occlusion sequence of the next time period. If the solar occlusion impact is greater than a first preset threshold, a solar occlusion prompt message is sent to the server.

[0014] Preferably, when the server receives the sun shading prompt information, it starts the second power generation prediction result acquisition module, and at the same time controls the power generation prediction result generation module to generate the photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result;

[0015] When the server does not receive the sun shading prompt information, the second power generation prediction result acquisition module is closed, and the power generation prediction result generation module is controlled to use the first power generation prediction result as the photovoltaic power generation prediction result.

[0016] Preferably, before the first power generation prediction result acquisition module inputs the weather data collected by the data collection module into the trained first power generation prediction model to obtain the first power generation prediction result, it includes:

[0017] S11, establishing a first historical data set, and dividing the first historical data set into a training set, a validation set, and a test set according to a preset ratio;

[0018] S12, setting a loss function and an optimizer of the first power generation prediction model, and inputting the training set into the first power generation prediction model for model training;

[0019] S13, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0020] S14. If the loss value is less than the preset threshold, the model training ends, and the current first power generation prediction model is the trained first power generation prediction model; otherwise, the process returns to S12 and continues to use the training set for model training;

[0021] S15, inputting the validation set into the trained first power generation prediction model, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model;

[0022] S16, inputting the test set into the optimized first power generation prediction model to evaluate the model performance;

[0023] The first historical data set includes historical weather data with historical photovoltaic actual power generation tags, and the historical weather data includes historical solar radiation, historical sunshine duration, historical temperature, historical humidity and historical wind speed.

[0024] Preferably, the first power generation prediction result acquisition module inputs the weather data collected by the data collection module into the trained first power generation prediction model to obtain the first power generation prediction result, including:

[0025] Inputting the real-time weather data into the trained first power generation prediction model to obtain a first power generation prediction result;

[0026] Among them, real-time weather data includes real-time solar radiation, real-time sunshine duration, real-time temperature, real-time humidity and real-time wind speed.

[0027] Preferably, the second power generation prediction result acquisition module inputs the actual photovoltaic power generation of the previous period and the solar shading sequence of the next period into the trained second power generation prediction model to obtain the second power generation prediction result, including:

[0028] S21, inputting the actual photovoltaic power generation in the previous period and the solar shading sequence in the next period into the trained second power generation prediction model;

[0029] S22, performing wavelet maximum layer decomposition on the actual photovoltaic power generation in the previous period to obtain an approximate phasor sequence of the maximum layer and detailed phasor sequences of different levels;

[0030] S23, constructing a phasor sequence matrix of each phasor sequence, and inputting the phasor sequence matrix of each phasor sequence into a convolutional neural network to obtain a predicted phasor sequence of the next time period corresponding to the phasor sequence matrix of each phasor sequence;

[0031] S24, reconstructing all predicted phasor sequences of the next period using a wavelet reconstruction algorithm to obtain the actual photovoltaic power generation sequence of the previous period;

[0032] S25. Linearly fuse the actual photovoltaic power generation sequence of the previous period and the solar shading sequence of the next period to obtain a second power generation prediction result.

[0033] Preferably, the fault detection module compares the photovoltaic power generation prediction result of each photovoltaic module with the actual photovoltaic power generation power, and determines whether the photovoltaic module has a fault according to the comparison result, including:

[0034] The difference between the photovoltaic power generation prediction result and the actual photovoltaic power generation of the photovoltaic component is calculated. If the difference is greater than a second preset threshold, it is determined that the photovoltaic component is faulty.

[0035] Preferably, the fault diagnosis result generating module performs fault type diagnosis on the faulty photovoltaic assembly based on the trained fault diagnosis model, and before generating the fault diagnosis result, comprises:

[0036] S31, establishing a second historical data set, and dividing the second historical data set into a training set, a validation set, and a test set according to a preset ratio;

[0037] S32, setting the loss function and optimizer of the fault diagnosis model, and inputting the training set into the fault diagnosis model for model training;

[0038] S33, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0039] S34, if the loss value is less than the preset threshold, the model training is completed, and the current fault diagnosis model is the trained fault diagnosis model, otherwise return to S32, and continue to use the training set for model training;

[0040] S35, input the validation set into the trained fault diagnosis model, evaluate the generalization ability of the model by observing the performance on the validation set, and tune the hyperparameters and structure of the model;

[0041] S36, inputting the test set into the tuned fault diagnosis model to evaluate the model performance;

[0042] Among them, the second historical data set includes historical working status data with fault type labels, and the historical working status parameters include historical working voltage data, historical working current data, standard voltage data, standard current data, historical working voltage changes and historical working current changes of each photovoltaic component.

[0043] Preferably, the fault diagnosis result generating module performs fault type diagnosis on the faulty photovoltaic assembly based on the trained fault diagnosis model to generate a fault diagnosis result, including:

[0044] S41, inputting the real-time working status data into the trained fault diagnosis model, and determining whether the changes of the real-time working voltage data and the real-time working current data are periodic in the historical voltage changes and the historical current changes respectively;

[0045] S42, if the change of the real-time working voltage data is periodic in the historical voltage change situation, it is determined that there is a shadow fault, otherwise it is determined whether the real-time working current data is less than the standard current data;

[0046] S43, if the real-time working current data is less than the standard current data, it is determined that there is an open circuit fault, otherwise it is determined whether the real-time working voltage data is less than the standard voltage data;

[0047] S44, if the real-time working voltage data is less than the standard voltage data, it is determined that a short circuit fault exists;

[0048] The real-time working status data includes the real-time working voltage data and the real-time working current data of each photovoltaic module.

[0049] Preferably, the data collected by the data collection module include ground-based cloud maps, real-time weather data and real-time working status data of each photovoltaic component.

[0050] (III) Beneficial effects

[0051] Compared with the prior art, the photovoltaic power prediction and fault diagnosis system based on digital twins provided by the present invention has the following beneficial effects:

[0052] 1) When the server receives the sun shading prompt information, the second power generation prediction result acquisition module is started, and the power generation prediction result generation module is controlled to generate the photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result; when the server does not receive the sun shading prompt information, the second power generation prediction result acquisition module is closed, and the power generation prediction result generation module is controlled to use the first power generation prediction result as the photovoltaic power generation prediction result; a "selective prediction" mechanism is established according to the sun shading effect in the next period of time. When the sun shading effect is large, the photovoltaic power generation prediction result is generated based on the real-time weather data and the sun shading sequence; when the sun shading effect is small, the photovoltaic power generation prediction result is generated only based on the real-time weather data. This "selective prediction" mechanism can not only effectively improve the accuracy of the photovoltaic power generation prediction, but also can reduce the calculation amount of the photovoltaic power generation prediction based on the sun shading sequence as much as possible, that is, effectively improve the real-time performance of the photovoltaic power generation prediction;

[0053] 2) The fault detection module compares the photovoltaic power prediction results of each photovoltaic module with the actual photovoltaic power generation, and determines whether the photovoltaic module has a fault based on the comparison results. The fault diagnosis model is constructed. The fault diagnosis result generation module diagnoses the fault type of the faulty photovoltaic module based on the trained fault diagnosis model and generates a fault diagnosis result. On the one hand, by comparing the photovoltaic power prediction results of each photovoltaic module with the actual photovoltaic power generation, the fault location of the faulty photovoltaic module can be accurately performed. On the other hand, the real-time working status data of each photovoltaic module can be used based on the fault diagnosis model to perform accurate fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 It is a schematic diagram of the system of the present invention;

[0056] Figure 2 A schematic diagram of a process for generating photovoltaic power generation prediction results by a power generation prediction result generating module in the present invention;

[0057] Figure 3 A schematic diagram of a flow chart of the second power generation prediction result acquisition module in the present invention obtaining the second power generation prediction result;

[0058] Figure 4 The present invention is a schematic diagram of a process for the fault diagnosis result generating module to diagnose the fault type of a faulty photovoltaic component and generate a fault diagnosis result. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Photovoltaic power prediction and fault diagnosis system based on digital twin, such as Figure 1 As shown, it includes a server, which collects data through a data acquisition module and creates a virtual model corresponding to the photovoltaic system using a virtual model creation module. The server maps the collected data to the virtual model through a data mapping module, and uses a digital twin model generation module to generate a digital twin model corresponding to the photovoltaic system based on the virtual model of the mapped data.

[0061] The server calculates the solar shading sequence of the next time period using the ground-based cloud map collected by the data acquisition module through the data analysis and processing module, and comprehensively judges the impact of solar shading based on the solar shading sequence of the next time period, and at the same time obtains the actual photovoltaic power generation of the previous time period collected by the data acquisition module. The server constructs a first power generation prediction model through the first power generation prediction model construction module, and uses the first power generation prediction result acquisition module to input the weather data collected by the data acquisition module into the trained first power generation prediction model to obtain the first power generation prediction result; the server constructs a second power generation prediction model through the second power generation prediction model construction module, and uses the second power generation prediction result acquisition module to input the actual photovoltaic power generation of the previous time period and the solar shading sequence of the next time period into the trained second power generation prediction model to obtain the second power generation prediction result; the server generates a photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result through the power generation prediction result generation module.

[0062] In the technical solution of the present application, the data collected by the data acquisition module includes ground cloud maps, real-time weather data and real-time working status data of each photovoltaic component.

[0063] ① The data analysis and processing module uses the ground-based cloud map collected by the data acquisition module to calculate the solar occlusion sequence for the next period, and comprehensively judges the solar occlusion impact based on the solar occlusion sequence for the next period, such as Figure 2 As shown, including:

[0064] Based on the ground-based cloud image, the optical flow method is used to calculate the moving speed and direction of the cloud layer relative to the sun. According to the moving speed and direction, the sun blocking situation at each second-level time interval in the next period is calculated to obtain the sun blocking sequence for the next period.

[0065] The solar occlusion impact is comprehensively judged according to the solar occlusion sequence of the next time period. If the solar occlusion impact is greater than a first preset threshold, a solar occlusion prompt message is sent to the server.

[0066] ② If Figure 2 As shown, when the server receives the sun shading prompt information, it starts the second power generation prediction result acquisition module, and at the same time controls the power generation prediction result generation module to generate the photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result;

[0067] When the server does not receive the sun shading prompt information, the second power generation prediction result acquisition module is closed, and the power generation prediction result generation module is controlled to use the first power generation prediction result as the photovoltaic power generation prediction result.

[0068] ③ The first power generation prediction result acquisition module inputs the weather data collected by the data acquisition module into the trained first power generation prediction model to obtain the first power generation prediction result, including:

[0069] S11, establishing a first historical data set, and dividing the first historical data set into a training set, a validation set, and a test set according to a preset ratio;

[0070] S12, setting a loss function and an optimizer of the first power generation prediction model, and inputting the training set into the first power generation prediction model for model training;

[0071] S13, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0072] S14. If the loss value is less than the preset threshold, the model training ends, and the current first power generation prediction model is the trained first power generation prediction model; otherwise, the process returns to S12 and continues to use the training set for model training;

[0073] S15, inputting the validation set into the trained first power generation prediction model, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model;

[0074] S16, inputting the test set into the optimized first power generation prediction model to evaluate the model performance;

[0075] The first historical data set includes historical weather data with historical photovoltaic actual power generation tags, and the historical weather data includes historical solar radiation, historical sunshine duration, historical temperature, historical humidity and historical wind speed.

[0076] The first power generation prediction result acquisition module inputs the weather data collected by the data collection module into the trained first power generation prediction model to obtain the first power generation prediction result, including:

[0077] Inputting the real-time weather data into the trained first power generation prediction model to obtain a first power generation prediction result;

[0078] Among them, real-time weather data includes real-time solar radiation, real-time sunshine duration, real-time temperature, real-time humidity and real-time wind speed.

[0079] ④ The second power generation prediction result acquisition module inputs the actual photovoltaic power generation of the previous period and the solar shading sequence of the next period into the trained second power generation prediction model to obtain the second power generation prediction result, such as Figure 3 As shown, including:

[0080] S21, inputting the actual photovoltaic power generation in the previous period and the solar shading sequence in the next period into the trained second power generation prediction model;

[0081] S22, performing wavelet maximum layer decomposition on the actual photovoltaic power generation in the previous period to obtain an approximate phasor sequence of the maximum layer and detailed phasor sequences of different levels;

[0082] S23, constructing a phasor sequence matrix of each phasor sequence, and inputting the phasor sequence matrix of each phasor sequence into a convolutional neural network to obtain a predicted phasor sequence of the next time period corresponding to the phasor sequence matrix of each phasor sequence;

[0083] S24, reconstructing all predicted phasor sequences of the next period using a wavelet reconstruction algorithm to obtain the actual photovoltaic power generation sequence of the previous period;

[0084] S25. Linearly fuse the actual photovoltaic power generation sequence of the previous period and the solar shading sequence of the next period to obtain a second power generation prediction result.

[0085] In the above technical solution, when the server receives the sun shading prompt information, the second power generation prediction result acquisition module is started, and the power generation prediction result generation module is controlled to generate the photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result; when the server does not receive the sun shading prompt information, the second power generation prediction result acquisition module is closed, and the power generation prediction result generation module is controlled to use the first power generation prediction result as the photovoltaic power generation prediction result; a "selective prediction" mechanism is established according to the sun shading effect in the next time period. When the sun shading effect is large, the photovoltaic power generation prediction result is generated based on the real-time weather data and the sun shading sequence; when the sun shading effect is small, the photovoltaic power generation prediction result is generated only based on the real-time weather data. This "selective prediction" mechanism can not only effectively improve the accuracy of the photovoltaic power generation prediction, but also can minimize the calculation amount of the photovoltaic power generation prediction based on the sun shading sequence, that is, effectively improve the real-time performance of the photovoltaic power generation prediction.

[0086] like Figure 1 As shown, the server compares the predicted photovoltaic power generation results of each photovoltaic module with the actual photovoltaic power generation through the fault detection module, and determines whether the photovoltaic module is faulty based on the comparison results. The server builds a fault diagnosis model through the fault diagnosis model, and uses the fault diagnosis result generation module to diagnose the fault type of the faulty photovoltaic module based on the trained fault diagnosis model to generate a fault diagnosis result. The server sends the fault diagnosis result to the client through the fault alarm module, receives the fault work order fed back by the client, and uses the fault work order issuance emergency repair module to issue the fault work order to the operation and maintenance workers for emergency repair.

[0087] ① The fault detection module compares the predicted photovoltaic power generation results of each photovoltaic module with the actual photovoltaic power generation, and determines whether there is a fault in the photovoltaic module based on the comparison results, such as Figure 4 As shown, including:

[0088] The difference between the photovoltaic power generation prediction result and the actual photovoltaic power generation of the photovoltaic component is calculated. If the difference is greater than a second preset threshold, it is determined that the photovoltaic component is faulty.

[0089] ② The fault diagnosis result generation module diagnoses the fault type of the faulty photovoltaic module based on the trained fault diagnosis model. Before generating the fault diagnosis result, it includes:

[0090] S31, establishing a second historical data set, and dividing the second historical data set into a training set, a validation set, and a test set according to a preset ratio;

[0091] S32, setting the loss function and optimizer of the fault diagnosis model, and inputting the training set into the fault diagnosis model for model training;

[0092] S33, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0093] S34, if the loss value is less than the preset threshold, the model training is completed, and the current fault diagnosis model is the trained fault diagnosis model, otherwise return to S32, and continue to use the training set for model training;

[0094] S35, input the validation set into the trained fault diagnosis model, evaluate the generalization ability of the model by observing the performance on the validation set, and tune the hyperparameters and structure of the model;

[0095] S36, inputting the test set into the tuned fault diagnosis model to evaluate the model performance;

[0096] Among them, the second historical data set includes historical working status data with fault type labels, and the historical working status parameters include historical working voltage data, historical working current data, standard voltage data, standard current data, historical working voltage changes and historical working current changes of each photovoltaic component.

[0097] The fault diagnosis result generation module diagnoses the fault type of the faulty PV module based on the trained fault diagnosis model and generates fault diagnosis results, such as Figure 4 As shown, including:

[0098] S41, inputting the real-time working status data into the trained fault diagnosis model, and determining whether the changes of the real-time working voltage data and the real-time working current data are periodic in the historical voltage changes and the historical current changes respectively;

[0099] S42, if the change of the real-time working voltage data is periodic in the historical voltage change situation, it is determined that there is a shadow fault, otherwise it is determined whether the real-time working current data is less than the standard current data;

[0100] S43, if the real-time working current data is less than the standard current data, it is determined that there is an open circuit fault, otherwise it is determined whether the real-time working voltage data is less than the standard voltage data;

[0101] S44, if the real-time working voltage data is less than the standard voltage data, it is determined that a short circuit fault exists;

[0102] The real-time working status data includes the real-time working voltage data and the real-time working current data of each photovoltaic module.

[0103] In the above technical scheme, the fault detection module compares the photovoltaic power prediction results of each photovoltaic module with the actual photovoltaic power generation, and determines whether the photovoltaic module has a fault based on the comparison results. The fault diagnosis model is constructed. The fault diagnosis result generation module diagnoses the fault type of the faulty photovoltaic module based on the trained fault diagnosis model and generates a fault diagnosis result. On the one hand, by comparing the photovoltaic power prediction results of each photovoltaic module with the actual photovoltaic power generation, the fault location of the faulty photovoltaic module can be accurately performed. On the other hand, the real-time working status data of each photovoltaic module can be used to perform accurate fault diagnosis based on the fault diagnosis model.

[0104] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. Photovoltaic power prediction and fault diagnosis system based on digital twin, characterized by: The system comprises a server, wherein the server collects data through a data collection module, and creates a virtual model corresponding to a photovoltaic system by using a virtual model creation module, wherein the server maps the collected data to the virtual model by using a data mapping module, and generates a digital twin model corresponding to the photovoltaic system based on the virtual model of the mapped data by using a digital twin model generation module; The server calculates the solar shading sequence of the next time period by using the ground-based cloud map collected by the data acquisition module through the data analysis and processing module, and comprehensively judges the impact of solar shading according to the solar shading sequence of the next time period, and simultaneously obtains the actual photovoltaic power generation of the previous time period collected by the data acquisition module. The server constructs a first power generation prediction model through the first power generation prediction model construction module, and inputs the weather data collected by the data acquisition module into the trained first power generation prediction model using the first power generation prediction result acquisition module to obtain the first power generation prediction result; the server constructs a second power generation prediction model through the second power generation prediction model construction module, and inputs the actual photovoltaic power generation of the previous time period and the solar shading sequence of the next time period into the trained second power generation prediction model using the second power generation prediction result acquisition module to obtain the second power generation prediction result; the server generates a photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result through the power generation prediction result generation module; The server compares the photovoltaic power generation prediction results of each photovoltaic component with the actual photovoltaic power generation power through the fault detection module, and determines whether the photovoltaic component has a fault based on the comparison result. The server constructs a fault diagnosis model through the fault diagnosis model, and uses the fault diagnosis result generation module to diagnose the fault type of the faulty photovoltaic component based on the trained fault diagnosis model to generate a fault diagnosis result. The server sends the fault diagnosis result to the client through the fault alarm module, receives the fault work order fed back by the client, and uses the fault work order issuance emergency repair module to issue the fault work order to the operation and maintenance workers for emergency repair.

2. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 1 is characterized in that: The data analysis and processing module calculates the solar occlusion sequence of the next period using the ground-based cloud map collected by the data collection module, and comprehensively determines the solar occlusion impact according to the solar occlusion sequence of the next period, including: Based on the ground-based cloud image, the optical flow method is used to calculate the moving speed and direction of the cloud layer relative to the sun. According to the moving speed and direction, the sun blocking situation at each second-level time interval in the next period is calculated to obtain the sun blocking sequence for the next period. The solar occlusion impact is comprehensively judged according to the solar occlusion sequence of the next time period. If the solar occlusion impact is greater than a first preset threshold, a solar occlusion prompt message is sent to the server.

3. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 2 is characterized in that: When the server receives the sun shading prompt information, it starts the second power generation prediction result acquisition module, and at the same time controls the power generation prediction result generation module to generate the photovoltaic power generation prediction result based on the first power generation prediction result and the second power generation prediction result; When the server does not receive the sun shading prompt information, the second power generation prediction result acquisition module is closed, and the power generation prediction result generation module is controlled to use the first power generation prediction result as the photovoltaic power generation prediction result.

4. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 3 is characterized in that: The first power generation prediction result acquisition module inputs the weather data collected by the data collection module into the trained first power generation prediction model to obtain the first power generation prediction result, including: S11, establishing a first historical data set, and dividing the first historical data set into a training set, a validation set, and a test set according to a preset ratio; S12, setting a loss function and an optimizer of the first power generation prediction model, and inputting the training set into the first power generation prediction model for model training; S13, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information; S14. If the loss value is less than the preset threshold, the model training ends, and the current first power generation prediction model is the trained first power generation prediction model; otherwise, the process returns to S12 and continues to use the training set for model training; S15, inputting the validation set into the trained first power generation prediction model, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model; S16, inputting the test set into the optimized first power generation prediction model to evaluate the model performance; The first historical data set includes historical weather data with historical photovoltaic actual power generation tags, and the historical weather data includes historical solar radiation, historical sunshine duration, historical temperature, historical humidity and historical wind speed.

5. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 4 is characterized in that: The first power generation prediction result acquisition module inputs the weather data collected by the data collection module into the trained first power generation prediction model to obtain the first power generation prediction result, including: Inputting the real-time weather data into the trained first power generation prediction model to obtain a first power generation prediction result; Among them, real-time weather data includes real-time solar radiation, real-time sunshine duration, real-time temperature, real-time humidity and real-time wind speed.

6. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 3 is characterized in that: The second power generation prediction result acquisition module inputs the actual photovoltaic power generation of the previous period and the solar shading sequence of the next period into the trained second power generation prediction model to obtain the second power generation prediction result, including: S21, inputting the actual photovoltaic power generation in the previous period and the solar shading sequence in the next period into the trained second power generation prediction model; S22, performing wavelet maximum layer decomposition on the actual photovoltaic power generation in the previous period to obtain an approximate phasor sequence of the maximum layer and detailed phasor sequences of different levels; S23, constructing a phasor sequence matrix of each phasor sequence, and inputting the phasor sequence matrix of each phasor sequence into a convolutional neural network to obtain a predicted phasor sequence of the next time period corresponding to the phasor sequence matrix of each phasor sequence; S24, reconstructing all predicted phasor sequences of the next period using a wavelet reconstruction algorithm to obtain the actual photovoltaic power generation sequence of the previous period; S25. Linearly fuse the actual photovoltaic power generation sequence of the previous period and the solar shading sequence of the next period to obtain a second power generation prediction result.

7. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 1, characterized in that: The fault detection module compares the photovoltaic power generation prediction result of each photovoltaic module with the actual photovoltaic power generation power, and determines whether the photovoltaic module has a fault according to the comparison result, including: The difference between the photovoltaic power generation prediction result and the actual photovoltaic power generation of the photovoltaic component is calculated. If the difference is greater than a second preset threshold, it is determined that the photovoltaic component is faulty.

8. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 7 is characterized in that: The fault diagnosis result generating module performs fault type diagnosis on the faulty photovoltaic component based on the trained fault diagnosis model, and before generating the fault diagnosis result, it includes: S31, establishing a second historical data set, and dividing the second historical data set into a training set, a validation set, and a test set according to a preset ratio; S32, setting the loss function and optimizer of the fault diagnosis model, and inputting the training set into the fault diagnosis model for model training; S33, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information; S34, if the loss value is less than the preset threshold, the model training is completed, and the current fault diagnosis model is the trained fault diagnosis model, otherwise return to S32, and continue to use the training set for model training; S35, input the validation set into the trained fault diagnosis model, evaluate the generalization ability of the model by observing the performance on the validation set, and tune the hyperparameters and structure of the model; S36, inputting the test set into the tuned fault diagnosis model to evaluate the model performance; Among them, the second historical data set includes historical working status data with fault type labels, and the historical working status parameters include historical working voltage data, historical working current data, standard voltage data, standard current data, historical working voltage changes and historical working current changes of each photovoltaic component.

9. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 8, characterized in that: The fault diagnosis result generation module performs fault type diagnosis on the faulty photovoltaic components based on the trained fault diagnosis model to generate a fault diagnosis result, including: S41, inputting the real-time working status data into the trained fault diagnosis model, and determining whether the changes of the real-time working voltage data and the real-time working current data are periodic in the historical voltage changes and the historical current changes respectively; S42, if the change of the real-time working voltage data is periodic in the historical voltage change situation, it is determined that there is a shadow fault, otherwise it is determined whether the real-time working current data is less than the standard current data; S43, if the real-time working current data is less than the standard current data, it is determined that there is an open circuit fault, otherwise it is determined whether the real-time working voltage data is less than the standard voltage data; S44, if the real-time working voltage data is less than the standard voltage data, it is determined that a short circuit fault exists; The real-time working status data includes the real-time working voltage data and the real-time working current data of each photovoltaic module.

10. The photovoltaic power prediction and fault diagnosis system based on digital twin according to claim 1, characterized in that: The data collected by the data collection module include ground cloud maps, real-time weather data and real-time working status data of each photovoltaic module.

Images