A photovoltaic power generation prediction method and system based on deep learning

Abstract

The present invention discloses a photovoltaic power prediction method and system based on deep learning, which relates to the field of photovoltaic power generation technology. The steps of the method include: collecting power generation environment data sets and satellite spectral images, and collecting power output data corresponding to the photovoltaic power generation system; dividing the photovoltaic power plant into regions to obtain vegetation-affected areas and vegetation-unaffected areas; extracting satellite spectral images of the vegetation-affected areas, obtaining vegetation growth trend influence coefficients and environmental influence coefficients in the vegetation-affected areas through the power generation environment data sets and the satellite spectral images of the vegetation-affected areas, and recording the corresponding weather conditions; setting a weather dynamic power generation prediction model to predict photovoltaic power generation; real-time monitoring of the vegetation growth trend influence coefficient, and determining whether to issue an early warning to the administrator. The present invention solves the problem of large errors in photovoltaic power generation prediction due to ecological reasons, and ensures the reliability and efficiency of the photovoltaic power generation prediction process.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and specifically to a photovoltaic power generation power prediction method and system based on deep learning. Background Art

[0002] With the continuous growth of global energy demand and the increasing awareness of environmental protection, the development and utilization of renewable energy has become the focus of global attention. As a clean and renewable form of energy, solar photovoltaic power generation has been widely studied and applied worldwide due to its advantages such as abundant resources, environmental friendliness and wide distribution. However, the output power of photovoltaic power generation systems is affected by many factors, such as weather conditions, light intensity, temperature, etc., and shows strong uncertainty and volatility. For example, around photovoltaic power stations, if new vegetation appears due to human or natural reasons, and these vegetation are not considered in the initial planning, it will cause the problem of power generation panels being blocked. This uncertainty has brought severe challenges to the dispatching and operation of the power grid, affecting the safe and stable operation and economic benefits of the power grid.

[0003] Traditional photovoltaic power prediction methods are mostly based on statistical models or physical models. These methods have problems such as low prediction accuracy and poor adaptability when dealing with complex and nonlinear photovoltaic power prediction problems. With the rapid development of artificial intelligence technology, deep learning, as a powerful machine learning technology, has achieved remarkable results in many fields with its powerful feature extraction and model learning capabilities. Applying deep learning technology to photovoltaic power prediction is expected to break through the limitations of traditional methods and improve prediction accuracy and practicality.

[0004] For example, the existing Chinese patent with publication number CN117154724B discloses a photovoltaic power generation power prediction method, including the following steps: step one: collecting power data of the photovoltaic power generation system; step two: generating an initial optical power sampling series based on the power data of the photovoltaic power generation system; step three: generating a cumulative sequence based on the initial optical power sampling series; step four: generating a mean sequence based on the cumulative sequence; step five: generating a prediction first-order model based on the mean sequence; step six: generating a new optical power prediction model based on step five; step seven: predicting photovoltaic power generation based on the new optical power prediction model generated in step six. However, this design has a large amount of calculation and a complex process, and does not take into account the problem of vegetation shading at the edge of the power plant. For this reason, the present invention provides a photovoltaic power generation power prediction method and system based on deep learning. Summary of the invention

[0005] The purpose of the present invention is to provide a photovoltaic power generation power prediction method and system based on deep learning to solve the existing problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a photovoltaic power generation prediction method based on deep learning, comprising the following steps:

[0007] S1. Obtain power generation environment data sets through sensors installed in photovoltaic power generation systems, collect satellite spectral images of photovoltaic power plants through satellite images, and collect power output data corresponding to photovoltaic power generation systems;

[0008] S2. Divide the photovoltaic power plant into regions to obtain vegetation-affected areas and vegetation-unaffected areas;

[0009] S3, extracting the satellite spectral image of the vegetation impact area, obtaining the vegetation growth trend influence coefficient and the environmental influence coefficient of the vegetation impact area through the power generation environment data set and the satellite spectral image of the vegetation impact area, and recording the corresponding weather conditions;

[0010] S4, setting a weather dynamic power generation prediction model, taking the output data in step S3 as input, and predicting the photovoltaic power generation;

[0011] S5. Real-time monitoring of the vegetation growth trend influence coefficient. When the vegetation growth trend influence coefficient is greater than a set vegetation growth trend threshold, an early warning is issued to the administrator, and corresponding measures are taken.

[0012] The present invention is further improved in that the vegetation impact area is obtained by extracting the NDVI data set of all pixels of the satellite spectral image of the photovoltaic power plant, setting a vegetation coverage threshold, and listing the pixels in the NDVI data set that are greater than the vegetation coverage threshold in the vegetation impact area foreground image, separating the independent connected areas in the foreground image of the vegetation impact area through connected component analysis, setting a minimum area threshold, extracting the areas greater than the minimum area threshold in the independent connected areas for regional merging to obtain the vegetation area; setting a square structure element with a pixel number of N×N, placing the square structure element at the position of each pixel of the satellite spectral image, if the square structure element has an intersection with the vegetation area, marking the pixel at the center position of the square structure element as the impact area, merging the impact area and the vegetation area to obtain the vegetation impact area, and listing the remaining areas in the satellite spectral image of the photovoltaic power plant as vegetation unaffected areas.

[0013] The present invention is further improved in that the vegetation growth trend influence coefficient is obtained by extracting the satellite spectral image of the vegetation influence area every day within T days to obtain the vegetation index data set of the photovoltaic power plant within T days. NDVI i Represents the vegetation index of the PV power plant on the i-th day in the vegetation index dataset, NIR irepresents the near-infrared reflectance of the vegetation-affected area on the i-th day, Red i Represents the red light band reflectance of the vegetation-affected area on the i-th day, i∈(1,2,…,T), and the vegetation height is calculated by the vegetation index Where a, b and c represent the parameters obtained through regression analysis; the calculation formula of the vegetation growth trend influence coefficient is: Among them, α1 represents the weight of vegetation index, and α2 represents the weight of vegetation height.

[0014] A further improvement of the present invention is that the power generation environment data set includes the light intensity G under standard conditions. stc , the area of ​​the power generation panel assembly coa, the reference temperature of the power generation panel assembly T ref , calculate the illumination correction factor G cell Indicates the actual light intensity; calculates the temperature correction coefficient T of the photovoltaic panel cor =1+TC×(T cell -T ref ), where TC represents the rated efficiency decrease index of the power generation panel for every degree increase, T cell Indicates the actual temperature of the power generation panel assembly; calculates the wind speed pollution impact value by combining the satellite spectral image of the vegetation-unaffected area with the Gaussian diffusion model Where ws represents the actual wind speed, Gra represents the mean gray value of the area not affected by vegetation, represents the gray value gradient of the image in the area not affected by vegetation, x represents the position of the pixel, and the calculation formula of the environmental impact coefficient is:

[0015]

[0016] The present invention is further improved in that the weather dynamic power generation prediction model is based on a convolutional neural network, takes the vegetation growth trend influence coefficient as input data, the environmental influence coefficient and its components as important power generation influencing factors, and extracts the power output number and time series data corresponding to the photovoltaic power generation system to form a four-dimensional tensor as input; including an input layer, a processing layer, a crop impact value prediction layer, an impact threshold check layer, a dynamic routing layer and an output layer;

[0017] The input layer includes extracting weather conditions corresponding to the vegetation growth trend influence coefficient and the environmental influence coefficient, and the weather conditions include sunny and non-sunny days;

[0018] The processing layer processes the input data using two fully connected layers and then uses a concatenation layer to fuse and concatenate the input data into a vector X=Concatenate(We,vgt',Eni'), where Concatenate() represents a concatenation function, We represents a weather condition vector output by the fully connected layer, vgt' represents a vegetation growth trend influence coefficient vector output by the fully connected layer, and Eni' represents an environmental influence coefficient vector output by the fully connected layer;

[0019] The crop impact prediction layer processes the output vector of the processing layer by using the convolution layer and the pooling layer, predicts the vegetation growth trend impact coefficient by the fully connected layer, checks whether the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value by the impact threshold check layer, and if the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value, truncation is performed; if the predicted vegetation growth trend impact coefficient is less than or equal to the set vegetation growth maximum value, it is input to the dynamic routing layer;

[0020] The dynamic routing layer dynamically selects a processing path according to weather conditions and the output layer uses it as output.

[0021] A further improvement of the present invention is that the input layer also includes an embedding layer, and the weather conditions are converted into dense vectors through the embedding layer.

[0022] The present invention is further improved in that the dynamic routing layer also includes setting the weight w of the vegetation growth trend influence coefficient under sunny conditions. sunny =0.8 and weight w for non-sunny conditions cloudy = 0.2, create sunny day branch paths and non-sunny day branch paths, the sunny day branch path is represented by X sunny =sig(W sunny )×Concatenate(We,0.8vgt',0.2Eni'), the non-sunny branch path is represented by X cloudy =sig(W cloudy )×Concatenate(We,0.2vgt',0.8Eni'), where sig() represents the sigmoid activation function, W sunny represents the probability of sunny days in historical weather data, W cloudy Represents the probability of non-sunny days in historical weather data.

[0023] On the other hand, the present invention provides a photovoltaic power generation prediction system based on deep learning, comprising:

[0024] The data acquisition module is used to obtain the power generation environment data set through sensors installed in the photovoltaic power generation system, collect satellite spectral images of the photovoltaic power plant through satellite images, and collect power output data corresponding to the photovoltaic power generation system;

[0025] A feature screening module, used to extract photovoltaic power generation power influencing features according to the data of the data acquisition module;

[0026] The model building module is used to set up a weather dynamic power generation prediction model, taking the output data in the feature screening module as input to predict the photovoltaic power generation;

[0027] A further improvement of the present invention is that the feature screening module includes a region division unit and an influence coefficient calculation unit, the region division unit is used to divide the photovoltaic power plant into regions to obtain vegetation affected areas and vegetation unaffected areas; the influence coefficient calculation unit is used to calculate the vegetation growth trend influence coefficient and the environmental influence coefficient.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. The present invention firstly combines the sensor data, satellite spectral images and power output data of the photovoltaic power generation system to more comprehensively reflect the operating status and external environmental impact of the photovoltaic power generation system, thereby improving the accuracy and reliability of the prediction model;

[0030] 2. By finely dividing the photovoltaic power plant into different areas, distinguishing between vegetation-affected areas and unaffected areas, the impact of vegetation on power generation efficiency is analyzed in a targeted manner, thus improving the calculation accuracy;

[0031] 3. By calculating the vegetation growth trend influence coefficient, the prediction model can be dynamically adjusted to adapt to different vegetation coverage conditions. In addition, the introduction of the environmental impact coefficient further enhances the model's adaptability to different environmental conditions, ensuring that the prediction results are closer to the actual situation and achieving accurate prediction of photovoltaic power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of a photovoltaic power prediction method based on deep learning in the present invention;

[0033] Figure 2 This is a framework diagram of a photovoltaic power prediction system based on deep learning in the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0035] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0036] Example 1

[0037] Figure 1 A flowchart of a photovoltaic power generation prediction method based on deep learning disclosed in this embodiment is shown, and the steps are as follows:

[0038] S1. Obtain power generation environment data sets through sensors installed in photovoltaic power generation systems, collect satellite spectral images of photovoltaic power plants through satellite images, and collect power output data corresponding to photovoltaic power generation systems;

[0039] S2. Divide the photovoltaic power plant into regions to obtain vegetation-affected regions and vegetation-unaffected regions; the vegetation-affected regions are obtained by extracting the NDVI data set of all pixels of the satellite spectral image of the photovoltaic power plant, setting a vegetation coverage threshold, and listing pixels in the NDVI data set that are greater than the vegetation coverage threshold into the foreground image of the vegetation-affected region; separate independent connected regions in the foreground image of the vegetation-affected region through connected component analysis, set a minimum area threshold, extract regions greater than the minimum area threshold from the independent connected regions, and merge the regions to obtain vegetation regions; set a square structure element with a pixel number of N×N, and place the square structure element at the position of each pixel in the satellite spectral image; if the square structure element has an intersection with the vegetation region, mark the pixel at the center of the square structure element as the affected region, merge the affected region and the vegetation region to obtain the vegetation-affected region, and list the remaining regions in the satellite spectral image of the photovoltaic power plant as vegetation-unaffected regions.

[0040] S3. Extract the satellite spectral image of the vegetation impact area, obtain the vegetation growth trend influence coefficient and the environmental influence coefficient of the vegetation impact area through the power generation environment dataset and the satellite spectral image of the vegetation impact area, and record the corresponding weather conditions; the vegetation growth trend influence coefficient is obtained by extracting the satellite spectral image of the vegetation impact area every day within T days to obtain the vegetation index dataset of the photovoltaic power plant within T days NDVI i Represents the vegetation index of the PV power plant on the i-th day in the vegetation index dataset, NIR i represents the near-infrared reflectance of the vegetation-affected area on the i-th day, Red i Represents the red light band reflectance of the vegetation-affected area on the i-th day, i∈(1,2,…,T), and the vegetation height is calculated by the vegetation index Where a, b and c represent the parameters obtained through regression analysis; the calculation formula of the vegetation growth trend influence coefficient is: Among them, α1 represents the weight of vegetation index, and α2 represents the weight of vegetation height.

[0041] The power generation environment data set includes the light intensity G under standard conditions stc , the area of ​​the power generation panel assembly coa, the reference temperature of the power generation panel assembly T ref , calculate the illumination correction factor G cell Indicates the actual light intensity; calculates the temperature correction coefficient T of the photovoltaic panel cor =1+TC×(T cell -T ref ), where TC represents the rated efficiency decrease index of the power generation panel for every degree increase, T cell Indicates the actual temperature of the power generation panel assembly; calculates the wind speed pollution impact value by combining the satellite spectral image of the vegetation-unaffected area with the Gaussian diffusion model Where ws represents the actual wind speed, Gra represents the mean gray value of the area not affected by vegetation, represents the gray value gradient of the image in the area not affected by vegetation, x represents the position of the pixel, and the calculation formula of the environmental impact coefficient is:

[0042]

[0043] S4, setting a weather dynamic power generation prediction model, taking the output data in step S3 as input, and predicting the photovoltaic power generation power; the weather dynamic power generation prediction model is based on a convolutional neural network, takes the vegetation growth trend influence coefficient as input data, and the environmental impact coefficient and its components as important power generation influencing factors, and extracts the power output number and time series data corresponding to the photovoltaic power generation system to form a four-dimensional tensor as input; including an input layer, a processing layer, a crop impact value prediction layer, an impact threshold check layer, a dynamic routing layer and an output layer;

[0044] The input layer includes extracting weather conditions corresponding to the vegetation growth trend influence coefficient and the environmental influence coefficient, wherein the weather conditions include sunny days and cloudy days; and also includes an embedding layer, through which the weather conditions are converted into dense vectors;

[0045] The processing layer processes the input data using two fully connected layers and then uses a concatenation layer to fuse and concatenate the input data into a vector X=Concatenate(We,vgt',Eni'), where Concatenate() represents a concatenation function, We represents a weather condition vector output by the fully connected layer, vgt' represents a vegetation growth trend influence coefficient vector output by the fully connected layer, and Eni' represents an environmental influence coefficient vector output by the fully connected layer;

[0046] The crop impact prediction layer processes the output vector of the processing layer by using the convolution layer and the pooling layer, predicts the vegetation growth trend impact coefficient by the fully connected layer, checks whether the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value by the impact threshold check layer, and if the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value, truncation is performed; if the predicted vegetation growth trend impact coefficient is less than or equal to the set vegetation growth maximum value, it is input to the dynamic routing layer;

[0047] The dynamic routing layer dynamically selects the processing path according to the weather conditions and outputs it to the output layer as the output, and also includes setting the weight w of the vegetation growth trend influence coefficient under sunny conditions sunny =0.8 and weight w for non-sunny conditions cloudy = 0.2, create sunny day branch paths and non-sunny day branch paths, the sunny day branch path is represented by X sunny =sig(W sunny )×Concatenate(We,0.8vgt',0.2Eni'), the non-sunny branch path is represented by X cloudy =sig(W cloudy )×Concatenate(We,0.2vgt',0.8Eni'), where sig() represents the sigmoid activation function, W sunny represents the probability of sunny days in historical weather data, W cloudy Represents the probability of non-sunny days in historical weather data.

[0048] S5. Real-time monitoring of the vegetation growth trend influence coefficient. When the vegetation growth trend influence coefficient is greater than the set vegetation growth trend threshold, an early warning is issued to the administrator, and corresponding measures are taken, including weeding and consultation with relevant personnel or contacting professionals for management.

[0049] The threshold and weight may be set by default according to the present invention, or may be set by an operator.

[0050] Example 2

[0051] Figure 2The framework diagram of a photovoltaic power generation prediction system based on deep learning of the present invention is shown. Based on the same inventive concept as that of Example 1, the present invention provides a photovoltaic power generation prediction system based on deep learning, including:

[0052] The data acquisition module is used to obtain the power generation environment data set through sensors installed in the photovoltaic power generation system, collect satellite spectral images of the photovoltaic power plant through satellite images, and collect power output data corresponding to the photovoltaic power generation system;

[0053] A feature screening module is used to extract the photovoltaic power generation power impact characteristics according to the data of the data acquisition module; it includes a region division unit and an impact coefficient calculation unit, the region division unit is used to divide the photovoltaic power plant into regions to obtain vegetation impact areas and vegetation unaffected areas; the impact coefficient calculation unit is used to calculate the vegetation growth trend impact coefficient and the environmental impact coefficient;

[0054] The model building module is used to set up a weather dynamic power generation prediction model, taking the output data in the feature screening module as input to predict the photovoltaic power generation;

[0055] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0057] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0058] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0059] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A photovoltaic power prediction method based on deep learning, characterized by: The following steps are involved: S1. Obtain power generation environment data sets through sensors installed in photovoltaic power generation systems, collect satellite spectral images of photovoltaic power plants through satellite images, and collect power output data corresponding to photovoltaic power generation systems; S2. Divide the photovoltaic power plant into regions to obtain vegetation-affected areas and vegetation-unaffected areas; S3, extracting the satellite spectral image of the vegetation impact area, obtaining the vegetation growth trend influence coefficient and the environmental influence coefficient of the vegetation impact area through the power generation environment data set and the satellite spectral image of the vegetation impact area, and recording the corresponding weather conditions; S4, setting a weather dynamic power generation prediction model, taking the output data in step S3 as input, and predicting the photovoltaic power generation; S5. Real-time monitoring of the vegetation growth trend influence coefficient. When the vegetation growth trend influence coefficient is greater than a set vegetation growth trend threshold, an early warning is issued to the administrator, and corresponding measures are taken; The vegetation growth trend influence coefficient is obtained by extracting the satellite spectral image of the vegetation impact area every day within T days to obtain the vegetation index data set of the photovoltaic power plant within T days. , represents the vegetation index of the ith day in the vegetation index dataset of the photovoltaic power plant, represents the near-infrared band reflectance of the vegetation-affected area on the i-th day, represents the red light band reflectance of the vegetation-affected area on the i-th day, , calculate vegetation height through vegetation index , where a, b and c represent the parameters obtained through regression analysis; the calculation formula for the vegetation growth trend influence coefficient is: ,in, represents the weight of the vegetation index, represents the weight of vegetation height; The power generation environment dataset includes the light intensity under standard conditions , Power generation panel assembly area , the reference temperature of the power generation panel assembly , calculate the illumination correction factor , Indicates the actual light intensity; Calculate the temperature correction factor of photovoltaic panels ,in It means that the rated efficiency of the power generation panel decreases with each degree increase. Indicates the actual temperature of the power generation panel assembly; calculates the wind speed pollution impact value by combining the satellite spectral image of the vegetation-unaffected area with the Gaussian diffusion model ,in Indicates the actual wind speed, Indicates that vegetation does not affect the mean gray value of the region. represents the gray value gradient of the image in the area not affected by vegetation, x represents the position of the pixel, and the calculation formula of the environmental impact coefficient is: .

2. The photovoltaic power generation prediction method based on deep learning according to claim 1, characterized in that: The vegetation impact area is obtained by extracting the NDVI data set of all pixels of the satellite spectral image of the photovoltaic power plant, setting a vegetation coverage threshold, and listing the pixels in the NDVI data set that are greater than the vegetation coverage threshold into the foreground image of the vegetation impact area. The independent connected areas in the foreground image of the vegetation impact area are separated by connected component analysis, and a minimum area threshold is set. The areas greater than the minimum area threshold in the independent connected areas are extracted for regional merging to obtain the vegetation area. A square structure element with a pixel number of N×N is set, and the square structure element is placed at the position of each pixel of the satellite spectral image. If the square structure element has an intersection with the vegetation area, the pixel at the center position of the square structure element is marked as the impact area, and the impact area and the vegetation area are merged to obtain the vegetation impact area, and the remaining areas in the satellite spectral image of the photovoltaic power plant are listed as vegetation unaffected areas.

3. The photovoltaic power generation prediction method based on deep learning according to claim 2 is characterized in that: The weather dynamic power generation prediction model is based on a convolutional neural network, takes the vegetation growth trend influence coefficient as input data, the environmental impact coefficient and its components as important power generation influencing factors, and extracts the power output number and time series data corresponding to the photovoltaic power generation system to form a four-dimensional tensor as input; including an input layer, a processing layer, a crop impact value prediction layer, an impact threshold check layer, a dynamic routing layer and an output layer; The input layer includes extracting weather conditions corresponding to the vegetation growth trend influence coefficient and the environmental influence coefficient, and the weather conditions include sunny and non-sunny days; The processing layer processes the input data using two fully connected layers and then uses a concatenation layer to fuse and concatenate the input data into a vector , represents the concatenation function, Describes the weather condition vector output by the fully connected layer, Represents the vegetation growth trend influence coefficient vector output by the fully connected layer, Represents the environmental impact coefficient vector output by the fully connected layer; The crop impact prediction layer processes the output vector of the processing layer by using the convolution layer and the pooling layer, predicts the vegetation growth trend impact coefficient by the fully connected layer, checks whether the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value by the impact threshold check layer, and if the predicted vegetation growth trend impact coefficient is greater than the set vegetation growth maximum value, truncation is performed; if the predicted vegetation growth trend impact coefficient is less than or equal to the set vegetation growth maximum value, it is input to the dynamic routing layer; The dynamic routing layer dynamically selects a processing path according to weather conditions and the output layer uses it as output.

4. The photovoltaic power generation prediction method based on deep learning according to claim 3 is characterized by: The input layer also includes an embedding layer, through which the weather conditions are converted into dense vectors.

5. The photovoltaic power generation prediction method based on deep learning according to claim 4 is characterized in that: The dynamic routing layer also includes setting the weight of the vegetation growth trend influence coefficient under sunny conditions and weights for non-sunny conditions , create sunny branch paths and non-sunny branch paths, the sunny branch path is expressed as , the non-sunny branch path is expressed as ,in, represents the sigmoid activation function, represents the probability of sunny days in historical weather data, Represents the probability of non-sunny days in historical weather data.

6. A photovoltaic power generation prediction system based on deep learning, used to execute a photovoltaic power generation prediction method based on deep learning as claimed in any one of claims 1 to 5, characterized in that: include: The data acquisition module is used to obtain the power generation environment data set through sensors installed in the photovoltaic power generation system, and collect satellite spectral images of the photovoltaic power plant through satellite images, and collect power output data corresponding to the photovoltaic power generation system; A feature screening module, used to extract photovoltaic power generation power influencing features according to the data of the data acquisition module; The model building module is used to set up the weather dynamic power generation prediction model, taking the output data in the feature screening module as input to predict the photovoltaic power generation.

7. A photovoltaic power generation prediction system based on deep learning according to claim 6, characterized in that: The feature screening module includes a region division unit and an influence coefficient calculation unit. The region division unit is used to divide the photovoltaic power plant into regions to obtain vegetation-affected regions and vegetation-unaffected regions; the influence coefficient calculation unit is used to calculate the vegetation growth trend influence coefficient and the environmental influence coefficient.

Images