Construction method and application of photovoltaic theoretical generating capacity prediction model, and terminal

Abstract

The invention discloses a construction method and application of a photovoltaic theoretical generating capacity prediction model, and a terminal. The construction method comprises the following steps: taking the installed capacity of a station and historical weather information of each time period as feature parts of a training set; according to whether light is discarded in a time period, taking the historical actual generating capacity information as a training target part of training set data when light is not discarded, performing interpolation according to historical actual generating capacity of two adjacent time periods when light is discarded to obtain the training target part of the training set data, and if light is discarded in at least two adjacent time periods, outputting the training target part of the training set data; if so, abandoning the corresponding data; generating a training set; performing model training according to the training set to obtain a prediction model; the prediction accuracy and robustness are improved; the method is used for predicting the theoretical photovoltaic generating capacity, for the theoretical generating capacity of each time period, according to whether light is abandoned or not, the actual photovoltaic generating capacity or the larger value of the actual photovoltaic generating capacity and the predicted photovoltaic generating capacity is selected, it is ensured that the theoretical generating capacity is certainly larger than or equal to the actual value, and calculation of the theoretical generating capacity is more accurate.

Description

[0001] This case is a divisional application based on the invention patent with application date of May 30, 2024, application number 202410683471.0, and name “A method and terminal for calculating the theoretical power generation of photovoltaics” as the parent case. Technical Field

[0002] The present invention relates to the technical field of power generation calculation, and in particular to a method for constructing a photovoltaic theoretical power generation prediction model and its application and terminal. Background Art

[0003] In the daily operation of the photovoltaic storage charging and inspection station, due to insufficient remaining battery energy storage capacity or insufficient charging load, the photovoltaic power generation needs to stop working during operation, resulting in abandoned light. In order to accurately measure the amount of abandoned light, it is necessary to calculate the theoretical power generation of photovoltaic power generation every day to support the daily operation monitoring and optimization of the station.

[0004] The existing method for calculating the amount of photovoltaic abandoned light usually adopts the prediction method. However, when using the prediction method alone, there is a deviation between the theoretical prediction value and the actual value every hour, regardless of whether there is abandoned light in the time period. As a result, the calculation value of the theoretical abandoned light amount is not accurate enough, which brings certain deviations to daily operation monitoring. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for constructing a photovoltaic theoretical power generation prediction model and its application and terminal, so as to improve the accuracy of photovoltaic theoretical power generation calculation.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for constructing a photovoltaic theoretical power generation prediction model comprises the following steps:

[0008] Obtain the installed capacity of the station, historical weather information, and historical actual power generation information of the station;

[0009] The installed capacity of the station and the historical weather information of each period are used as the feature part of the training set;

[0010] According to whether the light is abandoned in the time period, if the light is not abandoned, the historical actual power generation information is used as the training target part of the training set data. If the light is abandoned, interpolation is performed according to the historical actual power generation of the two previous and next time periods to obtain the training target part of the training set data. If there are at least two adjacent time periods in which the light is abandoned, the corresponding data are discarded;

[0011] Generate training and test sets;

[0012] Model training is performed according to the training set to obtain a prediction model.

[0013] An application of a photovoltaic theoretical power generation prediction model includes the following steps:

[0014] S1. Obtain weather information by time period;

[0015] The weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena;

[0016] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0017] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period;

[0018] The predicted photovoltaic power generation is obtained by the prediction model obtained by the above-mentioned method for constructing a photovoltaic theoretical power generation prediction model, which is based on weather information and site installed capacity prediction.

[0019] A method for calculating theoretical photovoltaic power generation, comprising the steps of:

[0020] S1. Obtain weather information by time period;

[0021] The weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena;

[0022] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0023] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period;

[0024] The predicted photovoltaic power generation is obtained by a pre-trained prediction model based on weather information and station installed capacity;

[0025] The prediction model adopts a multi-layer perceptron network structure;

[0026] The multi-layer perceptron network structure includes an input layer, a connection layer, a fully connected layer and an output layer arranged in sequence;

[0027] An embedding layer is provided between the input layer and the connection layer to convert non-numerical features into numerical features;

[0028] The fully connected layer is set between the connection layer and the output layer, and the formula is expressed as:

[0029] Y = f(W*X+b);

[0030] Among them, X represents the input data, the shape is (n, m), n is the number of samples, and m is the input feature dimension;

[0031] W represents the connection weight matrix, with a shape of (m, k);

[0032] b is the bias vector, shape is (1, k);

[0033] k represents the number of output features;

[0034] f is the activation function applied on XW;

[0035] Y represents output;

[0036] S3. According to the theoretical power generation in each time period, the theoretical photovoltaic power generation of the day is assembled.

[0037] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0038] A terminal for calculating theoretical photovoltaic power generation includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:

[0039] S1. Obtain weather information by time period;

[0040] The weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena;

[0041] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0042] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period;

[0043] The predicted photovoltaic power generation is obtained by a pre-trained prediction model based on weather information and station installed capacity;

[0044] The prediction model adopts a multi-layer perceptron network structure;

[0045] The multi-layer perceptron network structure includes an input layer, a connection layer, a fully connected layer and an output layer arranged in sequence;

[0046] An embedding layer is provided between the input layer and the connection layer to convert non-numerical features into numerical features;

[0047] The fully connected layer is set between the connection layer and the output layer, and the formula is expressed as:

[0048] Y = f(W*X+b);

[0049] Among them, X represents the input data, the shape is (n, m), n is the number of samples, and m is the input feature dimension;

[0050] W represents the connection weight matrix, with a shape of (m, k);

[0051] b is the bias vector, shape is (1, k);

[0052] k represents the number of output features;

[0053] f is the activation function applied on XW;

[0054] Y represents output;

[0055] S3. According to the theoretical power generation in each time period, the theoretical photovoltaic power generation of the day is assembled.

[0056] A terminal comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for constructing a photovoltaic theoretical power generation prediction model or the application of a photovoltaic theoretical power generation prediction model are implemented.

[0057] The beneficial effects of the present invention are as follows: a method for constructing a photovoltaic theoretical power generation prediction model and its application and terminal of the present invention realize the construction of a training set of the prediction model and train the prediction model, wherein, during the period of abandoned light, since part of the photovoltaic power generation is "abandoned", directly using the actual power generation may affect the accuracy of the model, so new data is inserted by interpolation to replace the photovoltaic power generation during the period of abandoned light, so that even if there is no "complete" photovoltaic power generation data record in these periods, the model can learn relevant information, thereby improving the accuracy and robustness of the prediction; when it is applied to predicting the theoretical photovoltaic power generation, for the theoretical power generation of each period, in the period without abandoned light, the actual photovoltaic power generation is directly used to ensure accuracy, and in the period of abandoned light, the larger value of the actual photovoltaic power generation and the predicted photovoltaic power generation is taken to ensure that the theoretical power generation will be greater than or equal to the actual value, so that the theoretical power generation calculation is more accurate and more in line with the actual operational use needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A flow chart of a method for calculating theoretical photovoltaic power generation according to an embodiment of the present invention;

[0059] Figure 2 This is a structural diagram of a terminal for calculating theoretical photovoltaic power generation according to an embodiment of the present invention;

[0060] Figure 3 This is a flowchart of a method for calculating theoretical photovoltaic power generation according to an embodiment of the present invention;

[0061] Figure 4 This is an example diagram of a prediction model building process of a method for calculating theoretical photovoltaic power generation according to an embodiment of the present invention;

[0062] Figure 5 This is an example diagram of a prediction model structure of a method for calculating theoretical photovoltaic power generation according to an embodiment of the present invention;

[0063] Description of labels:

[0064] 1. A terminal for calculating theoretical photovoltaic power generation; 2. A processor; 3. A memory. DETAILED DESCRIPTION

[0065] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

[0066] Please refer to Figure 1 , Figure 3 and Figure 4 , a method for calculating the theoretical photovoltaic power generation, comprising the steps of:

[0067] S1. Obtain weather information by time period;

[0068] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0069] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period;

[0070] The predicted photovoltaic power generation is obtained by a pre-trained prediction model based on weather information and station installed capacity;

[0071] S3. According to the theoretical power generation in each time period, the theoretical photovoltaic power generation of the day is assembled.

[0072] From the above description, it can be seen that the beneficial effects of the present invention are: a method and terminal for calculating the theoretical photovoltaic power generation of the present invention, for the theoretical power generation of each time period, directly uses the actual photovoltaic power generation in the time period without abandonment to ensure accuracy, and in the time period of abandonment, takes the larger value of the actual photovoltaic power generation and the predicted photovoltaic power generation, ensuring that the theoretical power generation will definitely be greater than or equal to the actual value, making the theoretical power generation calculation more accurate and more in line with actual operational needs.

[0073] Furthermore, the training of the prediction model includes the steps of:

[0074] Obtain the installed capacity of the station, historical weather information, and historical actual power generation information of the station;

[0075] The installed capacity of the station and the historical weather information of each period are used as the feature part of the training set;

[0076] According to whether the light is abandoned in the time period, if the light is not abandoned, the historical actual power generation information is used as the training target part of the training set data. If the light is abandoned, interpolation is performed according to the historical actual power generation of the two previous and next time periods to obtain the training target part of the training set data. If there are at least two adjacent time periods in which the light is abandoned, the corresponding data are discarded;

[0077] Generate training set;

[0078] Model training is performed according to the training set to obtain a prediction model.

[0079] From the above description, it can be seen that through the above steps, the training set of the prediction model is constructed and the prediction model is trained. In the period of abandoned light, since part of the photovoltaic power generation is "abandoned", directly using the actual power generation may affect the accuracy of the model. Therefore, new data is inserted by interpolation to replace the photovoltaic power generation in the period of abandoned light. In this way, even if there is no "complete" photovoltaic power generation data record in these periods, the model can learn relevant information, thereby improving the accuracy and robustness of the prediction.

[0080] Furthermore, the prediction model adopts a multi-layer perceptron network structure.

[0081] From the above description, it can be seen that the multi-layer perceptron network structure has the ability to handle complex nonlinear problems, can automatically learn data features, and has good privacy protection and high flexibility and scalability, making it more suitable for the prediction of photovoltaic power generation.

[0082] Furthermore, the weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0083] From the above description, it can be seen that weather information includes data information at the above multiple levels.

[0084] Furthermore, the input layers include irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0085] From the above description, it can be seen that according to the design of feature data, the input layer includes weather information, namely, irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0086] Furthermore, the fully connected layer consists of three layers.

[0087] From the above description, we can see that designing three fully connected layers can gradually extract and integrate useful information in the input data, thereby achieving advanced representation and prediction capabilities for the input data.

[0088] Furthermore, the dimensions of the three fully connected layers are 256 dimensions, 128 dimensions, and 64 dimensions, respectively.

[0089] From the above description, we can see that by gradually reducing the dimension of the fully connected layer (from 256 dimensions to 128 dimensions, and then to 64 dimensions), the model can extract more meaningful features in each layer, thereby improving the overall feature representation ability and helping to reduce the risk of overfitting, because smaller hidden layers can better control the complexity of the model.

[0090] Please refer to Figure 2 , a photovoltaic theoretical power generation calculation terminal, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0091] S1. Obtain weather information by time period;

[0092] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0093] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period;

[0094] The predicted photovoltaic power generation is obtained by a pre-trained prediction model based on weather information and station installed capacity;

[0095] S3. According to the theoretical power generation in each time period, the theoretical photovoltaic power generation of the day is assembled.

[0096] From the above description, it can be seen that the beneficial effects of the present invention are: a method and terminal for calculating the theoretical photovoltaic power generation of the present invention, for the theoretical power generation of each time period, directly uses the actual photovoltaic power generation in the time period without abandonment to ensure accuracy, and in the time period of abandonment, takes the larger value of the actual photovoltaic power generation and the predicted photovoltaic power generation, ensuring that the theoretical power generation will definitely be greater than or equal to the actual value, making the theoretical power generation calculation more accurate and more in line with actual operational needs.

[0097] Furthermore, the training of the prediction model includes the steps of:

[0098] Obtain the installed capacity of the station, historical weather information, and historical actual power generation information of the station;

[0099] The installed capacity of the station and the historical weather information of each period are used as the feature part of the training set;

[0100] According to whether the light is abandoned in the time period, if the light is not abandoned, the historical actual power generation information is used as the training target part of the training set data. If the light is abandoned, interpolation is performed according to the historical actual power generation of the two previous and next time periods to obtain the training target part of the training set data. If there are at least two adjacent time periods in which the light is abandoned, the corresponding data are discarded;

[0101] Generate training set;

[0102] Model training is performed according to the training set to obtain a prediction model.

[0103] From the above description, it can be seen that through the above steps, the training set of the prediction model is constructed and the prediction model is trained. In the period of abandoned light, since part of the photovoltaic power generation is "abandoned", directly using the actual power generation may affect the accuracy of the model. Therefore, new data is inserted by interpolation to replace the photovoltaic power generation in the period of abandoned light. In this way, even if there is no "complete" photovoltaic power generation data record in these periods, the model can learn relevant information, thereby improving the accuracy and robustness of the prediction.

[0104] Furthermore, the prediction model adopts a multi-layer perceptron network structure.

[0105] From the above description, it can be seen that the multi-layer perceptron network structure has the ability to handle complex nonlinear problems, can automatically learn data features, and has good privacy protection and high flexibility and scalability, making it more suitable for the prediction of photovoltaic power generation.

[0106] Furthermore, the weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0107] From the above description, it can be seen that weather information includes data information at the above multiple levels.

[0108] Furthermore, the input layers include irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0109] From the above description, it can be seen that according to the design of feature data, the input layer includes weather information, namely, irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

[0110] Furthermore, the fully connected layer consists of three layers.

[0111] From the above description, we can see that designing three fully connected layers can gradually extract and integrate useful information in the input data, thereby achieving advanced representation and prediction capabilities for the input data.

[0112] Furthermore, the dimensions of the three fully connected layers are 256 dimensions, 128 dimensions, and 64 dimensions, respectively.

[0113] From the above description, we can see that by gradually reducing the dimension of the fully connected layer (from 256 dimensions to 128 dimensions, and then to 64 dimensions), the model can extract more meaningful features in each layer, thereby improving the overall feature representation ability and helping to reduce the risk of overfitting, because smaller hidden layers can better control the complexity of the model.

[0114] The present invention provides a method and terminal for calculating the theoretical photovoltaic power generation, which are applicable to power stations with photovoltaic power generation and are used for calculating the theoretical photovoltaic power generation.

[0115] Please refer to Figure 1 and Figure 3 , Embodiment 1 of the present invention is:

[0116] A method for calculating theoretical photovoltaic power generation, comprising the steps of:

[0117] S1. Obtain weather information for different time periods.

[0118] In this embodiment, the weather data is obtained every hour, and the working status of the photovoltaic equipment is obtained every hour, forming a mark of whether the light is abandoned during the period.

[0119] S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value between the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period;

[0120] If there is no abandonment of light, the actual photovoltaic power generation is used as the theoretical power generation for this period.

[0121] In this embodiment, it is determined whether the photovoltaic power is abandoned every hour, and different theoretical power generation calculation methods are selected according to whether the photovoltaic power is abandoned. When the time period is marked as not abandoned, the actual photovoltaic power generation of the photovoltaic power period is directly calculated as the theoretical power generation; when the time period is marked as abandoned, the power generation is predicted based on weather data and the installed capacity of the site, and the predicted value is compared with the actual photovoltaic power generation of the time period, and the larger value of the two is taken as the theoretical power generation of the time period.

[0122] When the predicted power generation is greater than the actual power generation, the correction process can be expressed as:

[0123]

[0124] The predicted photovoltaic power generation is obtained by a pre-trained prediction model based on weather information and site installed capacity.

[0125] S3. According to the theoretical power generation in each time period, the theoretical photovoltaic power generation of the day is assembled.

[0126] In this embodiment, the theoretical photovoltaic power generation per hour obtained in step S2 is integrated to obtain the theoretical power generation per 24 hours.

[0127] Please refer to Figure 1 , Figure 4 and Figure 5 , Embodiment 2 of the present invention is:

[0128] A method for calculating the theoretical photovoltaic power generation is provided. In this embodiment, the construction of a prediction model is described.

[0129] In this embodiment, refer to Figure 5 , a prediction model is constructed using a multi-layer perceptron network structure (MLP), which includes an input layer, a connection layer, a fully connected layer and an output layer arranged in sequence.

[0130] In this embodiment, the input layer includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena. An embedding layer is also provided between the input layer and the connection layer to convert non-numerical features into numerical features. The fully connected layer is provided between the connection layer and the output layer, which are 256-dimensional, 128-dimensional and 64-dimensional fully connected layers respectively.

[0131] Dense: Dense layer, also called fully connected layer or affine layer. The specific formula is:

[0132] Y = f(W*X+b);

[0133] Among them, X represents the input data, the shape is (n, m), n is the number of samples, and m is the input feature dimension;

[0134] W represents the connection weight matrix, with a shape of (m, k), where the elements W in the W matrix ij Represents the connection weight between the i-th input feature and the j-th neuron;

[0135] b is the bias vector, shape is (1, k);

[0136] Where m represents the number of input features, k represents the number of output features. The shape of the bias vector b (1, k) means that there is a bias term on each output feature, and the number of bias terms is the same as the number of output features.

[0137] f is the activation function applied on XW, such as ReLU, Sigmoid, etc.;

[0138] Y represents output.

[0139] The training of the prediction model includes the steps of:

[0140] Obtain the installed capacity of the station, historical weather information, and historical actual power generation information of the station;

[0141] The installed capacity of the station and the historical weather information of each period are used as the feature parts of the training set.

[0142] In this embodiment, a photovoltaic prediction model training set is assembled based on the actual index values ​​of the weather each hour, including irradiance, temperature, humidity, air pressure, visibility, cloud cover, weather phenomena, site installed capacity, etc., as the feature part of the data set.

[0143] Depending on whether light is abandoned in a time period, when light is not abandoned, the historical actual power generation information is used as the training target part of the training set data. When light is abandoned, interpolation is performed based on the historical actual power generation of the two previous and next time periods to obtain the training target part of the training set data. If there are at least two adjacent time periods in which light is abandoned, the corresponding data are discarded.

[0144] In this embodiment, the power generation in each time period uses the actual photovoltaic power generation in the non-abandoned time period, and interpolates the average value of the actual photovoltaic power generation in the two previous and next time periods in the abandoned time period. Data with more than one consecutive abandoned time period are filtered out and not used.

[0145] Generate a training set.

[0146] In this embodiment, a training set is constructed based on the above steps, and a test set is constructed in the same manner.

[0147] Model training is performed according to the training set to obtain a prediction model.

[0148] Please refer to Figure 2 , Embodiment 3 of the present invention is:

[0149] A photovoltaic theoretical power generation calculation terminal 1 includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, the steps in a photovoltaic theoretical power generation calculation method of the above embodiment one or two are implemented.

[0150] To sum up, the present invention provides a method and terminal for calculating theoretical photovoltaic power generation. For the theoretical power generation in each time period, in the time period without abandoned light, the actual photovoltaic power generation is directly used, and the accuracy of the theoretical power generation is 100%. In order to ensure accuracy, in the time period of abandoned light, due to the manual shutdown or power reduction of the photovoltaic power generation, the photovoltaic power generation at this time cannot reach the limit power generation under the existing weather conditions. Therefore, the larger value of the actual photovoltaic power generation and the predicted photovoltaic power generation is taken to ensure that the theoretical power generation will be greater than or equal to the actual value, so that the theoretical power generation calculation is more accurate and more in line with the actual operation and use needs.

[0151] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for constructing a photovoltaic theoretical power generation prediction model, characterized in that: Including steps: Obtain the station's installed capacity, historical weather information, and historical actual power generation information; The installed capacity of the station and the historical weather information of each period are used as the feature parts of the training set; Based on whether solar power is abandoned in a time period, if solar power is not abandoned, the historical actual power generation information is used as the training target part of the training set data. If solar power is abandoned, the historical actual power generation information of the two previous and next time periods is interpolated to obtain the training target part of the training set data. If there are at least two adjacent time periods in which solar power is abandoned, the corresponding data is discarded; Generate training and test sets; Model training is performed based on the training set to obtain a prediction model.

2. The method for constructing a photovoltaic theoretical power generation prediction model according to claim 1, characterized in that: The prediction model adopts a multi-layer perceptron network structure; The multi-layer perceptron network structure includes an input layer, a connection layer, a fully connected layer and an output layer arranged in sequence; An embedding layer is provided between the input layer and the connection layer to convert non-numeric features into numerical features; The fully connected layer is set between the connection layer and the output layer, and the formula is expressed as: Y = f(W*X+b); Where X represents the input data, the shape is (n, m), n is the number of samples, and m is the input feature dimension; W represents the connection weight matrix, with a shape of (m, k); b is the bias vector, shape is (1, k); k represents the number of output features; f is the activation function applied on XW; Y represents output.

3. The method for constructing a photovoltaic theoretical power generation prediction model according to claim 2, characterized in that: The input layers include irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

4. A method for constructing a photovoltaic theoretical power generation prediction model according to claim 2, characterized in that: The fully connected layer consists of three layers.

5. The method for constructing a photovoltaic theoretical power generation prediction model according to claim 4, characterized in that: The dimensions of the three fully connected layers are 256, 128, and 64, respectively.

6. The method for constructing a photovoltaic theoretical power generation prediction model according to claim 1, characterized in that: The historical weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena.

7. An application of a photovoltaic theoretical power generation prediction model, characterized in that: Including steps: S1. Obtain weather information by time period; The weather information includes irradiance, temperature, humidity, air pressure, visibility, cloud cover and weather phenomena; S2. For each time period, determine whether to abandon the photovoltaic power generation. If the photovoltaic power generation is abandoned, select the larger value of the actual photovoltaic power generation and the predicted photovoltaic power generation as the theoretical power generation for the time period; If there is no abandonment of solar power, the actual photovoltaic power generation is used as the theoretical power generation for that period; The predicted photovoltaic power generation is obtained by a prediction model obtained by the method for constructing a photovoltaic theoretical power generation prediction model described in any one of claims 1 to 6 above, and is obtained based on weather information and site installed capacity prediction.

8. A terminal comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the processor implements the steps of the method for constructing a photovoltaic theoretical power generation prediction model or the application of a photovoltaic theoretical power generation prediction model as described in any one of claims 1 to 7.

Images