Photovoltaic power generation prediction method and device, and storage medium

Abstract

The invention relates to the technical field of photovoltaic power generation prediction, in particular to a photovoltaic power generation prediction method and device and a storage medium. The prediction method comprises the following steps: acquiring historical weather characteristics, equipment identification information and historical real generating capacity; converting the equipment identification information into numerical serial numbers and inputting the numerical serial numbers into a feature coding model to obtain equipment coding features; the historical weather features and the equipment coding features are fused and then input into a deep learning model to obtain photovoltaic predicted generating capacity; based on a preset physical estimation model, extracting a part of weather features and a part of coding features to calculate photovoltaic estimated generating capacity; and finally, calculating fusion loss based on the photovoltaic predicted power generation amount, the photovoltaic estimated power generation amount and the historical real power generation amount to perform model training, and predicting photovoltaic power generation based on the trained feature coding model and the deep learning model. According to the method provided by the invention, the model can be effectively predicted under the condition of less data or larger noise, and the generalization ability and robustness of the model are improved.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic power generation prediction, and particularly to a photovoltaic power generation prediction method, device and storage medium. Background Art

[0002] Photovoltaic power generation prediction is the core technical support for the user-side energy storage system to maximize economic performance, optimize operation efficiency, and achieve the most stable grid coordination. Due to differences in orientation, tilt angle, photovoltaic panel area, occlusion, dust, etc. of different photovoltaic devices, under the same weather characteristic conditions, the generated power may vary.

[0003] In the prior art, photovoltaic power generation prediction is a machine learning method driven by historical weather characteristic data and historical photovoltaic power generation data as data sources. This method, on the one hand, largely depends on a large amount of high-quality data, and the prediction accuracy is relatively low in the case of less data or more noise; on the other hand, it cannot distinguish the individuality of devices, is prone to prediction deviation, and the model generalization ability is poor. Summary of the Invention

[0004] The embodiments of the present application aim to provide a photovoltaic power generation prediction method, device and storage medium to solve the problems that in the prior art, it is difficult for a deep learning model to obtain accurate and reliable prediction results only relying on historical weather characteristic data and historical photovoltaic power generation data, and it cannot distinguish the individuality of devices, and is prone to prediction deviation.

[0005] To solve the above technical problems, the embodiments of the present application provide the following technical solutions: According to the first aspect of the present application, a photovoltaic power generation prediction method is provided, and the method includes: Obtain historical weather characteristics, device identification information, and historical actual power generation; Convert the device identification information into a numerical serial number and input it into a feature encoding model to obtain device encoding features; Fuse the historical weather characteristics and the device encoding features and input them into a deep learning model to obtain predicted photovoltaic power generation; Based on a preset physical estimation model, extract some weather characteristics from the historical weather characteristics and some encoding features from the device encoding features to calculate the estimated photovoltaic power generation; Calculate a fusion loss based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation, and train the deep learning model and the feature encoding model based on the fusion loss to obtain a trained deep learning model and a trained feature encoding model; Predict photovoltaic power generation based on the trained feature encoding model and the trained deep learning model.

[0006] Optionally, the feature encoding model includes an embedding layer, a batch normalization layer, a fully connected layer, and a Sigmoid normalization layer.

[0007] Optionally, the device encoding features include conversion efficiency, photovoltaic panel area, temperature coefficient, photovoltaic panel tilt angle, photovoltaic panel pitch angle, and several device characteristic features.

[0008] Optionally, extracting partial weather features from the historical weather features and extracting partial encoding features from the device encoding features based on the preset physical estimation model includes: Extracting radiation and temperature from the historical weather features based on the preset physical estimation model, and extracting conversion efficiency, photovoltaic panel area, and temperature coefficient from the device encoding features.

[0009] Optionally, the calculation formula of the physical estimation model is:

[0010] where, is the estimated photovoltaic power generation, is the conversion efficiency, is the radiation, is the photovoltaic panel area, is the temperature coefficient, is the temperature, is the preset reference temperature.

[0011] Optionally, the deep learning model includes a convolutional neural network module, a position encoding module, a long short-term memory network module, a Transformer module, and a fully connected layer.

[0012] Optionally, the input feature of the long short-term memory network module is the feature obtained by adding and fusing the output feature of the convolutional neural network module and the output feature of the position encoding module.

[0013] Optionally, calculating the fusion loss based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation includes: Calculating the physical constraint loss based on the predicted photovoltaic power generation and the estimated photovoltaic power generation; Calculating the model prediction loss based on the predicted photovoltaic power generation and the historical actual power generation; Calculating the fusion loss based on the physical constraint loss and the model prediction loss.

[0014] According to a second aspect of the present application, there is provided an electronic device, including at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the photovoltaic power generation prediction method described above.

[0015] According to a third aspect of the present application, there is provided a computer storage medium storing instructions or programs, which when executed by at least one processor, cause the at least one processor to execute the photovoltaic power generation prediction method described above.

[0016] The beneficial effects of the embodiments of the present application are as follows: Different from the prior art, in the embodiments of the present application, a photovoltaic power generation prediction method is provided. First, historical weather features, device identification information, and historical actual power generation are obtained, and the device identification information is converted into a numerical serial number and then input into a feature encoding model to obtain device encoding features; then, the historical weather features and the device encoding features are fused and input into a deep learning model to obtain predicted photovoltaic power generation, and based on a preset physical estimation model, partial weather features are extracted from the historical weather features and partial encoding features are extracted from the device encoding features to calculate the estimated photovoltaic power generation; finally, a fusion loss is calculated based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation, and the deep learning model and the feature encoding model are trained based on the fusion loss to obtain a trained deep learning model and a trained feature encoding model, and photovoltaic power generation is predicted based on the trained feature encoding model and the trained deep learning model. The method of the present application is different from traditional data-driven neural networks. In the learning process, not only the personalized features of the device are added, but also physical knowledge rules are used to guide the model in the learning process, so that the model can effectively predict even in the case of less data or more noise, improving the generalization ability and robustness of the model. Description of the Drawings

[0017] One or more embodiments are illustrated by way of example in the accompanying drawings, which do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements, unless otherwise stated, and the drawings in the drawings do not constitute a scale limitation.

[0018] Figure 1 It is a schematic diagram of the algorithm framework of a photovoltaic power generation prediction model provided by an embodiment of the present application; Figure 2 It is a structural diagram of the feature encoding model provided by an embodiment of the present application; Figure 3 It is a structural diagram of the deep learning model provided by an embodiment of the present application; Figure 4 It is a schematic flowchart of a photovoltaic power generation prediction method provided by an embodiment of the present application; Figure 5 It is a structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0019] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0020] In addition, the technical features involved in the various embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0021] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0022] The key terms in the present application are explained as follows: 1. Data-driven: It is a method based on data analysis and machine learning, which trains a model through a large amount of data to make predictions or decisions.

[0023] 2. Physics-informed: It means integrating physical laws and prior knowledge into a machine learning model to improve the accuracy and interpretability of the model.

[0024] 3. Convolutional Neural Network (CNN): It is a deep learning model commonly used for image recognition and processing, which extracts local features through convolutional layers.

[0025] 4. Long Short-Term Memory (LSTM): It is a special recurrent neural network that can effectively handle long-term dependencies in time series data.

[0026] 5. Transformer: It is a deep learning model based on the self-attention mechanism, widely used in natural language processing and time series prediction, and can capture global dependencies.

[0027] 6. Positional Encoding: It is used in the Transformer model to represent the position information of each element in the input sequence, helping the model understand the sequence order.

[0028] 7. Loss Function: It is a function used to measure the difference between the model's predicted values and the true values. The model parameters are optimized by minimizing the loss function.

[0029] 8. Generalization Ability: It refers to the ability of the model to perform well on unseen data, that is, the model can adapt to new data and make accurate predictions.

[0030] 9. Embedding Layer: It is a neural network layer that converts discrete data (such as categorical variables) into continuous vectors for easier model processing and learning.

[0031] 10. Conversion Efficiency: It refers to the efficiency of a photovoltaic system in converting solar energy into electrical energy, usually expressed as a percentage.

[0032] 11. Photovoltaic Panel Area: It refers to the effective area of a photovoltaic module. The larger the area, the higher the theoretical power generation.

[0033] 12. Temperature Coefficient: It refers to the ratio of the output power of a photovoltaic module to the change in temperature. It is usually negative, indicating that the power decreases when the temperature rises.

[0034] 13. Photovoltaic Panel Tilt Angle: It refers to the angle between the photovoltaic panel and the horizontal plane. A reasonable tilt angle can improve the power generation efficiency.

[0035] 14. Photovoltaic Panel Pitch Angle: It refers to the angle of the photovoltaic panel in the vertical direction, which affects its radiation reception efficiency.

[0036] 15. Serial Number (SN): It is the unique serial number of a device or component, used to identify and track the production, installation, and maintenance information of the product.

[0037] 16. Feature Encoding: It is the process of converting non-numerical features into numerical forms for easier processing by machine learning models.

[0038] 17. Dense Vector: A vector consisting of continuous values, usually used to represent features or embedding representations.

[0039] 18. Photovoltaic Panel Dust: refers to the dust accumulated on the photovoltaic panels, which will affect their efficiency in receiving radiation.

[0040] 19. Occlusion: This refers to the situation where photovoltaic panels are partially or completely covered by objects, resulting in a decrease in power generation efficiency.

[0041] 20. Generalization Ability: refers to the ability of a model to perform well on unseen data, that is, the model can adapt to new data and make accurate predictions.

[0042] 21. Robustness: refers to the ability of a model to maintain stable performance in the face of noise, outliers, or data changes.

[0043] 22. Adam (Adaptive Moment Estimation): A widely used deep learning optimization algorithm that dynamically adjusts the learning rate of each parameter by calculating the first-order moment (mean) and second-order moment (variance) of the gradient. AdamW is an improved version of Adam that modifies the weight decay implementation in the traditional Adam optimizer, providing better generalization performance in deep learning training.

[0044] An embodiment of the present application proposes a photovoltaic power generation prediction method that integrates physical information. This method uses physical laws to guide the training of a deep learning model, incorporates personalized features of the device during the training process, and introduces physical knowledge into the loss function, so that the model can not only learn from the data, but also comply with the physical constraints of the photovoltaic field. It can be effectively trained in the absence of data and noise, and produce predictions with high accuracy, thereby enhancing the robustness and generalization ability of the model.

[0045] Please refer to Figure 1 , Figure 1 This is a schematic diagram of the algorithm framework of a photovoltaic power generation prediction model provided in the embodiment of this application. Figure 1 As shown, the photovoltaic power generation prediction model 100 includes a first input module 10, a second input module 20, an identification digitization module 30, a feature encoding model 40, a deep learning model 50, a physical estimation module 60 and a loss calculation module 70.

[0046] The first input module 10 is used to obtain historical weather characteristics, including radiation, temperature, air pressure, wind speed, cloud cover, sun angle, etc.

[0047] The second input module 20 is used to obtain device identification information. In one embodiment, the device identification information is the device SN code, which is usually a string containing letters and numbers.

[0048] The identification numerical conversion module 30 is used to convert the device identification information into a numerical serial number. Since the device identification information contains letters and numbers and cannot be directly input into the neural network, it is necessary to convert it into a numerical serial number first, such as 0, 1, 2, ……, 9999.

[0049] The feature encoding model 40 is used to map the discrete numerical serial number into a continuous vector representation, then convert the high-dimensional vector representation into a low-dimensional dense vector, and obtain the device encoding feature through normalization processing. Mapping the discrete numerical serial number into a continuous vector representation not only facilitates the processing of the deep learning model but also captures the semantic information and similarity between categories. Converting the high-dimensional coefficient data into a low-dimensional dense vector effectively reduces the complexity and computational amount of the model. At the same time, as the input layer of the model, it provides a suitable data form for the subsequent deep learning model, enabling the model to learn richer feature representations and thus improving the overall performance. In one embodiment, the device encoding features include conversion efficiency, photovoltaic panel area, temperature coefficient, photovoltaic panel tilt angle, photovoltaic panel pitch angle, and several device characteristic features; the device characteristic features include electronic characteristics, photovoltaic panel dust, occlusion, etc.

[0050] The deep learning model 50 takes as input the features obtained by fusing the historical weather features acquired by the first input module 10 and the device encoding features output by the feature encoding model 40, and outputs the predicted photovoltaic power generation. The specific network structure of the deep learning model 50 is described in detail below regarding Figure 3 the description.

[0051] [[ID=1--6]]The physical estimation module 60 is used to calculate the estimated photovoltaic power generation based on a preset physical estimation model by extracting some weather features from the historical weather features and some encoding features from the device encoding features. In one embodiment, the calculation formula of this physical estimation model is:

[0052] where is the estimated photovoltaic power generation, is the conversion efficiency, is the radiation, is the photovoltaic panel area, is the temperature coefficient, is the temperature, is the preset reference temperature.

[0053] According to the parameter information required by the physical estimation model, extracting some weather features from historical weather features and some coding features from device coding features includes: extracting radiation and temperature from historical weather features, and extracting conversion efficiency, photovoltaic panel area, and temperature coefficient from device coding features. It can be understood that those skilled in the art can also construct other physical estimation models and extract the required features from historical weather features and device coding features for calculation.

[0054] The loss calculation module 70 is used to calculate the fusion loss based on the predicted photovoltaic power generation, estimated photovoltaic power generation, and historical actual power generation. Specifically, calculate the physical constraint loss based on the predicted photovoltaic power generation and the estimated photovoltaic power generation, calculate the model prediction loss based on the predicted photovoltaic power generation and the historical actual power generation, and then calculate the fusion loss based on the physical constraint loss and the model prediction loss. This module is mainly applied in the training process of the photovoltaic power generation prediction model, aiming to adjust the relevant parameters of the deep learning model and the feature coding model according to the fusion loss, and provide guidance for the learning direction of the model.

[0055] In one embodiment, the calculation formula for the fusion loss is:

[0056] Where is the fusion loss, is the physical constraint loss, is the model prediction loss, , are the weight coefficients, is the number of samples, is the th true value of the sample, is the estimated value of the physical estimation model for the th sample, is the predicted value of the deep learning model for the th sample.

[0057] Please refer to Figure 2 , Figure 2 which is the structural diagram of the feature coding model 40 provided by the embodiment of the present application. As Figure 2As shown in the figure, the feature encoding model 40 includes an embedding layer, a batch normalization layer, a fully connected layer, and a sigmoid normalization layer. The input of the feature encoding model 40 is the numerical serial number converted from the device SN code. The embedding layer maps the discrete numerical serial number into a continuous vector representation through semantic vectorization. The batch normalization layer performs batch normalization on the continuous vector representation. The fully connected layer is used to extract core features and perform feature compression, converting the high-dimensional vector representation into a low-dimensional dense vector. The sigmoid normalization normalizes the low-dimensional dense vector to 0-1, corresponding to N device encoding features. The embodiment of the present application introduces an embedded feature encoding model to perform interpretable modeling on unknown device features, enhancing the interpretability and robustness of the model.

[0058] Please refer to Figure 3 , Figure 3 which is the structural diagram of the deep learning model 50 provided by the embodiment of the present application. As Figure 3 shown, the deep learning model 50 includes a CNN module, a position encoding module, an LSTM module, a Transformer module, and a fully connected layer.

[0059] Among them, the CNN module is used for local feature extraction, and its input feature is the feature after fusing the historical weather feature and the device encoding feature. In one embodiment, the structure of the historical weather feature is (B, 60, 50), where B is the sample size in days; 60 is the number of time intervals from 5 am to 9 pm at 15-minute intervals; 50 is 50 weather features. Assuming the number of device encoding features is 8, the feature structure after fusing the historical weather feature and the device encoding feature is (B, 60, 58). The processing process of the CNN module is as follows: 1) Adjust the dimensions of the input data, and exchange the data content of the second dimension and the third dimension. For example, adjust the input feature (B, 60, 58) to (B, 58, 60). 2) Perform convolution processing. For example, after performing convolution processing on the input feature (B, 58, 60) with a convolution kernel of 3 and an output channel of 128, the output feature structure is (B, 128, 60). 3) Perform batch normalization processing. 4) Restore the data dimensions, that is, exchange the data content of the second dimension and the third dimension again. For example, the output feature structure after restoring the input feature (B, 128, 60) is (B, 60, 128).

[0060] The position encoding module is used to provide the model with chronological information, which is crucial for capturing the periodicity and trend of photovoltaic power generation. Photovoltaic power generation follows the laws of the sun's movement and has obvious daily periodicity. The position encoding module can provide the model with the absolute position information of each time point, helping the model understand the order and periodicity in the time series. At the same time, the model can more accurately learn the relationship between photovoltaic power generation and time position, thereby improving the prediction accuracy. Specifically, the formula of the position encoding module is as follows:

[0061] Among them, is the time step position index, is the index, 10000 is the frequency base, used to control the frequencies of the sine function and the cosine function, dim is the total dimension of features. For example, when the input feature structure of the position encoding module is (B, 60, 128), is 0, 1,..., 59; is 0, 1,..., 63; dim is 128.

[0062] The LSTM module is used to capture the long-term dependencies in the time series data and prevent gradient vanishing and gradient explosion. Its input feature is the feature after fusing the output features of the CNN module and the output features of the position encoding module. In one embodiment, the structure of the feature extracted by the CNN module is (B, 60, 128), and the structure of the feature extracted by the position encoding module is (B, 60, 128). The two are fused by addition, and the output feature structure is (B, 60, 128). The output feature structure of the LSTM module is the same as the input feature structure, which is also (B, 60, 128).

[0063] The Transformer module is used to capture the global dependencies in the time series data based on the attention mechanism. In one embodiment, the output feature structure and the input feature structure of the Transformer module are both (B, 60, 128).

[0064] The fully connected layer is used to perform linear transformation and dimensionality reduction on the high-dimensional output features to convert them into the final prediction results. In one embodiment, the input feature structure of the fully connected layer is (B, 60, 128), and the output feature structure is (B, 60, 1). Further, removing the last dimension of the output data gives the photovoltaic power generation prediction result (B, 60), where 60 represents the predicted photovoltaic power generation every 15-minute interval.

[0065] Please refer to Figure 4 , Figure 4 which is a schematic flowchart of a photovoltaic power generation prediction method provided by an embodiment of the present application. The method includes: Step S401, obtaining historical weather features, device identification information, and historical actual power generation.

[0066] Specifically, collect the historical predicted meteorological data, historical power generation power data, relevant time data, etc. of the photovoltaic power generation station to obtain historical weather features with dimensions of A B C and historical weather features with dimensions of A The historical actual power generation of B. Among them, A is the number of days, B is the number of sunshine time intervals per day, and C is the number of weather characteristics. Historical weather characteristics may include rainfall, temperature, humidity, radiation, air pressure, wind speed, etc. The historical actual power generation is the actual power generation corresponding to the historical predicted weather characteristics at the corresponding date and time interval.

[0067] In one embodiment, the device identification information is the device SN code, usually a string, containing letters and numbers.

[0068] Step S402: Convert the device identification information into a numerical serial number and then input it into the feature encoding model to obtain the device encoding feature.

[0069] Since the device identification information contains letters and numbers and cannot be directly input into the neural network, it is necessary to first convert it into a numerical serial number, such as 0, 1, 2,..., 9999. The feature encoding model maps the discrete numerical serial number into a continuous vector representation, then converts the high-dimensional vector representation into a low-dimensional dense vector, and obtains the device encoding feature through normalization processing. In one embodiment, the device encoding features include conversion efficiency, photovoltaic panel area, temperature coefficient, photovoltaic panel tilt angle, photovoltaic panel pitch angle, and several device characteristic features; the device characteristic features include electronic characteristics, photovoltaic panel dust, occlusion, etc.

[0070] Step S403: Fuse the historical weather characteristics and the device encoding features and then input them into the deep learning model to obtain the photovoltaic predicted power generation.

[0071] Specifically, the historical weather characteristics and the device encoding features are concatenated and fused. For example, if the number of historical weather characteristics is 50 and the number of device encoding features is 8, then the number of features after fusing the historical weather characteristics and the device encoding features is 58.

[0072] Step S404: Based on a preset physical estimation model, extract some weather characteristics from the historical weather characteristics and some encoding features from the device encoding features to calculate the photovoltaic estimated power generation.

[0073] In one embodiment, the calculation formula of the preset physical estimation model is:

[0074] Among them, is the photovoltaic estimated power generation, is the conversion efficiency, is the radiation, is the photovoltaic panel area, is the temperature coefficient, is the temperature, is the preset reference temperature.

[0075] According to the parameter information required by the physical estimation model, extracting some weather features from historical weather features and some coding features from device coding features includes: extracting radiation and temperature from historical weather features, and extracting conversion efficiency, photovoltaic panel area, and temperature coefficient from device coding features.

[0076] Step S405: Calculate the fusion loss based on the predicted photovoltaic power generation, estimated photovoltaic power generation, and historical actual power generation, and train the deep learning model and the feature coding model based on the fusion loss to obtain a trained deep learning model and a trained feature coding model.

[0077] Specifically, first calculate the physical constraint loss based on the predicted photovoltaic power generation and the estimated photovoltaic power generation, and calculate the model prediction loss based on the predicted photovoltaic power generation and the historical actual power generation. Then calculate the fusion loss based on the physical constraint loss and the model prediction loss. This module is mainly applied in the training process of the photovoltaic power generation prediction model, aiming to adjust the relevant parameters of the deep learning model and the feature coding model according to the fusion loss, and provide guidance for the learning direction of the model.

[0078] In one embodiment, the calculation formula for the fusion loss is:

[0079] Where, is the fusion loss, is the physical constraint loss, is the model prediction loss, 、 are weight coefficients, is the number of samples, is the th sample's true value, is the estimated value of the physical estimation model for the th sample, is the predicted value of the deep learning model for the th sample.

[0080] In model training, based on the fusion loss, use the AdamW optimization algorithm to adjust the model parameters of the deep learning model and the feature coding model, and optimize and to make the smallest.

[0081] Step S406: Predict the photovoltaic power generation based on the trained feature coding model and the trained deep learning model.

[0082] When predicting the power generation of a photovoltaic energy storage device, first obtain the predicted weather characteristics of the date to be predicted and the device identification information of the photovoltaic energy storage device. After converting the device identification information into a numerical serial number, encode it through a trained feature encoding model to obtain device encoding features. Then, fuse the predicted weather characteristics and the device encoding features and input them into a trained deep learning model to obtain the predicted photovoltaic power generation. Specifically, assume that the structure of the predicted weather characteristics for one day is (1, 60, 50), and the structure of the device encoding features is (1, 8). The fused features (1, 60, 58) of the predicted weather characteristics and the device encoding features are input into the trained deep learning model, and the prediction result (1, 60) is output. Among them, 60 represents the predicted photovoltaic power generation for each 15-minute interval.

[0083] The photovoltaic power generation prediction method provided by the embodiments of the present application first obtains historical weather characteristics, device identification information, and historical actual power generation. After converting the device identification information into a numerical serial number, input it into a feature encoding model to obtain device encoding features. Then, fuse the historical weather characteristics and the device encoding features and input them into a deep learning model to obtain the predicted photovoltaic power generation. Extract some weather characteristics from the historical weather characteristics and some encoding features from the device encoding features, and calculate the estimated photovoltaic power generation using a physical equation based on the partial weather characteristics and the partial encoding features. Finally, calculate the fusion loss based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation to train the deep learning model and the feature encoding model, obtain a trained deep learning model and a trained feature encoding model, and predict the photovoltaic power generation based on the trained feature encoding model and the trained deep learning model. The method of the present application, different from traditional data-driven neural networks, not only adds the personalized features of the device during the learning process, but also uses physical knowledge rules to guide the model during the learning process, improving the generalization ability of the model, improving the prediction accuracy in the case of less data or more noise, and enhancing the robustness of the model.

[0084] According to the embodiments of the present application, an electronic device is provided, such as Figure 5 , which is a schematic structural diagram of an electronic device provided by the embodiments of the present application. The electronic device 100 may include a processor 10, a communication interface 30, a memory 20, and a communication bus. Among them, the processor 10, the communication interface 30, and the memory 20 complete communication with each other through the communication bus. The processor 10 may call the logical instructions in the memory 20 to execute the above-mentioned photovoltaic power generation prediction method.

[0085] In addition, when the logical instructions in the above-mentioned memory 20 are implemented in the form of software functional units and sold or used as independent products, they can be stored in several computer-readable storage media. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the aforementioned photovoltaic power generation prediction method of this application. The aforementioned storage media include: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0086] According to an embodiment of this application, there is provided a computer-readable storage medium, the type of which is as described above. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the aforementioned photovoltaic power generation prediction method.

[0087] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the technical solution, in essence, or the part that contributes to the related art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disks, optical discs, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] As described above, the above are only the specific embodiments of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. A photovoltaic power generation prediction method, characterized in that, The method includes: Obtaining historical weather features, device identification information, and historical actual power generation; Converting the device identification information into a numerical serial number and inputting it into a feature encoding model to obtain device encoding features; Fusing the historical weather features and the device encoding features and inputting them into a deep learning model to obtain predicted photovoltaic power generation; Based on a preset physical estimation model, extracting some weather features from the historical weather features and some encoding features from the device encoding features to calculate the estimated photovoltaic power generation; Calculating a fusion loss based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation, and training the deep learning model and the feature encoding model based on the fusion loss to obtain a trained deep learning model and a trained feature encoding model; Predicting photovoltaic power generation based on the trained feature encoding model and the trained deep learning model.

2. The method according to claim 1, wherein The feature encoding model includes an embedding layer, a batch normalization layer, a fully connected layer, and a Sigmoid normalization layer.

3. The method according to claim 1, wherein The device encoding features include conversion efficiency, photovoltaic panel area, temperature coefficient, photovoltaic panel tilt angle, photovoltaic panel pitch angle, and several device characteristic features.

4. The method according to claim 1, characterized in that, The extracting some weather features from the historical weather features and some encoding features from the device encoding features based on a preset physical estimation model includes: Based on a preset physical estimation model, extracting radiation and temperature from the historical weather features, and extracting conversion efficiency, photovoltaic panel area, and temperature coefficient from the device encoding features.

5. The method according to claim 4, characterized in that, The calculation formula of the physical estimation model is: Among them, is the estimated photovoltaic power generation, is the conversion efficiency, is the radiation, is the area of the photovoltaic panel, is the temperature coefficient, is the temperature, is the preset reference temperature.

6. The method according to claim 1, characterized in that The deep learning model includes a convolutional neural network module, a position encoding module, a long short-term memory network module, a Transformer module, and a fully connected layer.

7. The method according to claim 6, characterized in that, The input feature of the long short-term memory network module is the feature obtained by adding and fusing the output feature of the convolutional neural network module and the output feature of the position encoding module.

8. The method according to any one of claims 1 to 7, characterized in that The calculating a fusion loss based on the predicted photovoltaic power generation, the estimated photovoltaic power generation, and the historical actual power generation includes: Calculating a physical constraint loss based on the predicted photovoltaic power generation and the estimated photovoltaic power generation; Calculating a model prediction loss based on the predicted photovoltaic power generation and the historical actual power generation; Calculating a fusion loss based on the physical constraint loss and the model prediction loss.

9. An electronic device, characterized in that, Including at least one processor and a memory communicatively connected to the at least one processor, the memory storing instructions executable by the at least one processor, and the instructions being executed by the at least one processor to enable the at least one processor to execute the photovoltaic power generation prediction method according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that, The computer storage medium stores instructions or programs, and when the instructions or programs are executed by at least one processor, the at least one processor is enabled to execute the photovoltaic power generation prediction method according to any one of claims 1 to 8.

Images