Photovoltaic power generation power prediction method and system

Abstract

The invention discloses a photovoltaic power generation power prediction method and system. The method comprises the following steps: acquiring a photovoltaic power data set; constructing a photovoltaic power generation dynamic graph based on the photovoltaic power data set in combination with the environment variables; performing a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power, and performing a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power; and combining the space representation and the time representation to obtain the space-time representation of the photovoltaic power, and obtaining a photovoltaic power prediction result according to the space-time representation of the photovoltaic power. According to the method, a dynamic association equation between photovoltaic power and various environment variables in a photovoltaic power generation system is established; the method achieves the precise prediction of the photovoltaic power in the photovoltaic power generation system through the construction of the time-space joint prediction, solves the problem of the impact on the power prediction from the complex dynamic changes of all environmental factors in the photovoltaic power generation process, and improves the accuracy and efficiency of the online monitoring and energy management of the photovoltaic system.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent distribution networks, and particularly to a photovoltaic power prediction method and system. Background Art

[0002] The widespread application of solar energy can reduce carbon emissions, improve energy security, and promote the green transformation of the economy. However, solar power generation is affected by climate and weather, with volatility and intermittency, challenging grid stability and increasing operation difficulty. Therefore, it is very necessary to carry out photovoltaic power prediction. Accurate prediction can address instability issues, ensure grid stability, optimize scheduling and resource allocation, improve energy efficiency, reduce operating costs, and lay the foundation for the development of intelligent power grids.

[0003] In the field of photovoltaic power prediction, there are certain problems in the existing technologies. Traditional machine learning methods such as linear regression, support vector machines, decision trees, and random forests are widely used. Although linear regression is simple and efficient, it has poor ability to handle non-linear relationships; support vector machines are suitable for high-dimensional data, but the training time is long; decision trees are easy to understand, but prone to overfitting; random forests integrate decision trees to improve accuracy, but are complex. Neural network methods such as long short-term memory networks (LSTMs), convolutional neural networks (CNNs), and deep neural networks (DNNs) can better handle complex non-linear relationships and time series features. LSTMs can capture long-term dependencies, but the training time is long; CNNs improve performance through feature extraction and have high requirements for data preprocessing; DNNs have strong modeling capabilities, but the training is complex and prone to overfitting. Generally speaking, traditional methods have high computational efficiency and interpretability, and neural network methods have great potential in processing complex data and improving accuracy. However, both types of methods have limitations in handling non-linear relationships, long training time, overfitting, etc. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a photovoltaic power prediction method and system to solve the problems existing in traditional machine learning and neural network methods in photovoltaic power prediction, such as limited ability to handle non-linear relationships, long training time, overfitting, etc.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides a photovoltaic power prediction method, including:

[0008] Obtain a photovoltaic power data set;

[0009] Based on the photovoltaic power data set, construct a photovoltaic power generation dynamic graph in combination with environmental variables;

[0010] Perform a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power, and perform a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power;

[0011] Combine the spatial representation and the temporal representation to obtain a spatio-temporal representation of the photovoltaic power, and obtain a photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power.

[0012] As a preferred embodiment of the photovoltaic power generation prediction method of the present invention, wherein: performing a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power includes:

[0013] The first operation establishes a first equation and solves the first equation to obtain the state of each node of the photovoltaic power dynamic graph at different time steps;

[0014] Aggregate the node states to obtain a spatial representation of the photovoltaic power.

[0015] As a preferred embodiment of the photovoltaic power generation prediction method of the present invention, wherein: performing a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power includes:

[0016] The second operation establishes a second equation, and based on the second equation, uses a temporal convolutional network to extract the time series information of different time steps of the photovoltaic power dynamic graph;

[0017] Aggregate the time series information to obtain a temporal representation of the photovoltaic power.

[0018] As a preferred embodiment of the photovoltaic power generation prediction method of the present invention, wherein: combining the spatial representation and the temporal representation to obtain a spatio-temporal representation of the photovoltaic power includes:

[0019] Use the time series information obtained by the second operation as the initial state of each node in the first operation to obtain a spatio-temporal representation of the photovoltaic power.

[0020] As a preferred embodiment of the photovoltaic power generation prediction method of the present invention, wherein: obtaining a photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power includes:

[0021] Decode the generated spatio-temporal representation through a fully connected layer to obtain a photovoltaic power prediction result.

[0022] As a preferred embodiment of the photovoltaic power generation prediction method of the present invention, wherein: the first equation is expressed as:

[0023]

[0024] where, dH G (t) represents the change of the node state over time, and f(H G (t)) is a function of the node state.

[0025] As a preferred solution of the photovoltaic power prediction method described in the present invention, where: the second equation is expressed as:

[0026]

[0027] where, H T (t) represents the time series characteristics of each time step, TCN represents the temporal convolutional network, and Θ is a set of parameters.

[0028] In a second aspect, the present invention provides a photovoltaic power prediction system, including:

[0029] A data acquisition module for acquiring a photovoltaic power data set;

[0030] A processing module for constructing a photovoltaic power dynamic graph based on the photovoltaic power data set in combination with environmental variables; performing a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power, and performing a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power;

[0031] A prediction module for combining the spatial representation and the temporal representation to obtain a spatio-temporal representation of the photovoltaic power, and obtaining a photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power.

[0032] In a third aspect, the present invention provides a computing device, including:

[0033] A memory and a processor;

[0034] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, and when the computer-executable instructions are executed by the processor, the steps of the photovoltaic power prediction method are implemented.

[0035] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the photovoltaic power prediction method are implemented.

[0036] Compared with the prior art, the beneficial effects of the present invention: The present invention establishes a dynamic correlation equation between the photovoltaic power and various environmental variables in the photovoltaic power generation system; and through the construction of spatio-temporal joint prediction, realizes the accurate prediction of the photovoltaic power in the photovoltaic power generation system, solves the influence of the complex dynamic changes of various environmental factors on power prediction during the photovoltaic power generation process, and improves the accuracy and efficiency of on-line monitoring and energy management of the photovoltaic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0038] Figure 1 It is a schematic diagram of the overall process logic of the photovoltaic power prediction method according to an embodiment of the present invention;

[0039] Figure 2 It is a schematic diagram of the specific process logic of the photovoltaic power prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0041] Embodiment 1

[0042] Referring to Figure 1 - Figure 2 , an embodiment of the present invention provides a photovoltaic power prediction method, including:

[0043] S100: Obtain a photovoltaic power data set;

[0044] S102: Based on the photovoltaic power data set, construct a photovoltaic power dynamic graph in combination with environmental variables;

[0045] S104: Perform a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power, and perform a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power;

[0046] S106: Combine the spatial representation and the temporal representation to obtain a spatio-temporal representation of the photovoltaic power, and obtain a photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power.

[0047] It should be noted that the present invention establishes a dynamic correlation equation between the photovoltaic power and various environmental variables in the photovoltaic power generation system; and through the construction of spatio-temporal joint prediction, the accurate prediction of the photovoltaic power in the photovoltaic power generation system is realized, the influence of the complex dynamic changes of various environmental factors on the power prediction in the photovoltaic power generation process is solved, and the accuracy and efficiency of the online monitoring and energy management of the photovoltaic system are improved.

[0048] In the embodiment of the present application, the above step S102 includes the following sub-steps A1 - A2;

[0049] In A1: Preprocess the photovoltaic power dataset and environmental variables.

[0050] In the embodiment of the present application, the photovoltaic power data may include historical photovoltaic power data and real-time photovoltaic power data.

[0051] In a possible embodiment, the environmental variables may include solar irradiance, temperature, wind speed, etc.

[0052] In a possible embodiment, the preprocessing may include outlier processing, missing value processing, feature extraction, data standardization, etc.; the outlier processing method may be the Z-score method, the method based on the interquartile range, the isolation forest method, etc.; the missing value processing method may be the mean / median / mode filling method, the multiple imputation method, the nearest value filling method, etc.; the feature extraction method may be the principal component analysis method, the linear discriminant analysis method, the wavelet transform method, the autoencoder method, etc.; the data standardization method may be the min-max standardization method, the Z-score standardization method, the decimal scaling standardization method, etc.

[0053] In the embodiment of the present application, the preprocessing includes outlier processing and data standardization operations;

[0054] Specifically, the interquartile range method is used to identify and process these outliers. After the outlier processing, the dimension of the data remains unchanged and is still represented as X ∈ R M×n , where M represents the number of historical power data and meteorological data, and n represents the number of samples;

[0055] In the embodiment of the present application, for the upper quartile (Q3) and lower quartile (Q1) of the data, calculate the interquartile range IQR = Q3 - Q1, and define the outliers as the data points less than Q1 - 1.5×IQR or greater than Q1 + 1.5×IQR. Then the interquartile range method is expressed by the formula:

[0056] IQR = Q3 - Q1

[0057] Lower bound = Q1 - 1.5×IQR

[0058] Upper bound = Q3 + 1.5×IQR

[0059] Specifically, using the Z-normalization method, subtract the mean value of each feature from its value and then divide by its standard deviation, so that the mean value of the normalized data is 0 and the standard deviation is 1. The dimension of the normalized data remains

[0060] The Z-normalization method is expressed by the formula:

[0061]

[0062] where, X ij is the j-th feature of the i-th sample in the original dataset, μ j is the mean value of feature j, and σ j is the standard deviation of feature j.

[0063] It should be noted that after the normalization process, the photovoltaic power data will have better adaptability, which helps to improve the convergence speed and prediction accuracy of the method of the present invention.

[0064] In A2: Construct a dynamic graph of photovoltaic power generation by using the preprocessed photovoltaic power dataset and environmental variables.

[0065] Specifically, the historical time series data of photovoltaic power generation, that is, the historical photovoltaic power data, can be regarded as a dynamic graph structure; each node represents different influencing factors, such as irradiance, temperature, wind speed, etc., and the edges represent the dependence relationships between these factors; by constructing the graph, an adjacency matrix A is generated to capture the relationships between variables.

[0066] In the embodiment of the present application, the generation process of the adjacency matrix A is as follows:

[0067] 1. Initialization: First, use the randomly initialized embedding matrix E 1 , E 2 ∈R N×d and the trainable parameter to generate two feature matrices M 1 , M 2 ∈R N×d ; It is expressed by the formula:

[0068]

[0069] where, β is a hyperparameter for adjusting the activation saturation.

[0070] 2. Calculation of the adjacency matrix: Using the above-generated feature matrices M 1 and M 2 , calculate the adjacency value A ij between node i and node j, which is achieved through the inner product difference of the non-linear activation function ReLU and the double feature matrix; the formula is as follows:

[0071]

[0072] In the embodiments of the present application, the relationships between nodes are extracted through two different network structures, and the adjacency matrix A is dynamically adjusted to reflect the stable dependence relationships between nodes.

[0073] It should be noted that through this dynamic optimization, the graph structure can be dynamically adjusted during the training process, capturing the complex dependence relationships in time series data and improving the accuracy of prediction.

[0074] In the embodiments of the present application, in step S104, the first operation is performed on the photovoltaic power dynamic graph to obtain the spatial representation of the photovoltaic power, including the following steps B1:

[0075] In B1: The first operation obtains the states of each node of the photovoltaic power dynamic graph at different time steps by establishing and solving the first equation; the node states are aggregated to obtain the spatial representation of the photovoltaic power.

[0076] In the embodiments of the present application, the continuous graph propagation (CGP) method is used, and the state change of the photovoltaic power under the action of various environmental variables such as solar irradiance, light intensity, temperature, etc. is represented by means of an ordinary differential equation (ODE), that is, the first equation is expressed as:

[0077]

[0078] After simplification, we get:

[0079]

[0080] Among them, I N is the identity matrix; A = D -1 / 2 AD -1 / 2 is the normalized adjacency matrix, where D is the degree matrix, dH G (t) represents the change of the node state over time, and f(H G (t)) is a function of the node state.

[0081] In a possible embodiment, the values of the adjacency relationships are determined by the interactions of variables such as photovoltaic power, solar irradiance, light intensity, and temperature; specifically, the element A of the adjacency matrix A ij represents the connection weight between nodes i and j, and these weights reflect the physical distance, environmental similarity, and the correlation of the above variables between them.

[0082] In an alternative embodiment, methods such as the Euler method, the fourth-order Runge-Kutta method, or the improved Euler method can be selected to solve the first equation to obtain the node state HG (t) Variation at each time step. Finally, by integrating the node states H at all time steps G (t i ), the final graphical module spatial representation is obtained

[0083] In the embodiments of the present application, in order to improve the calculation accuracy and handle various influencing factors such as solar irradiance, light intensity, and temperature, and predict under these complex and changeable conditions, the fourth-order Runge-Kutta method is used as the ODE solver to handle common non-stiff problems. At the same time, in order to cope with the drastic changes that may occur in the photovoltaic system, during the steeper periods, the fourth-order Runge-Kutta method can use a larger step size to obtain higher calculation efficiency while ensuring accuracy. Using the ODE solver to solve the differential equation at the above-mentioned moment, the calculation is expressed by the formula:

[0084]

[0085] Among them, k1 represents the change in the calculation state at the current time step t i The change in the calculation state, k1 = f(t i , H G (t i )); k2 represents the change in the calculation state at the midpoint time The change in the calculation state, k3 represents the change in the calculation state at the midpoint time again, k4: represents the change in the calculation state at the end time t i +h, k4 = f(t i +h, H G (t i ) + hk3); f(t i , H G ) = (A - I N )H G (t) is the right-hand side of the ordinary differential equation, and h is the step size, representing the time interval t i -t i-1 . In this way, the complex relationships between variables such as photovoltaic power, solar irradiance, light intensity, and temperature can be captured and integrated, thereby improving the prediction accuracy of photovoltaic power.

[0086] Obtain the final graphical module space Expressed as:

[0087]

[0088] Among them, Φ i is a function matrix for node state transformation, used to integrate the relevant information between nodes and generate the final spatial dependence relationship.

[0089] In another possible embodiment, methods such as physics-based parameter estimation methods, data-driven machine learning regression methods, and hybrid modeling methods combining the two can be used to establish the first equation;

[0090] In one possible embodiment, taking the data-driven machine learning regression method as an example, a large amount of photovoltaic power data and corresponding environmental variable data such as solar irradiance, light intensity, and temperature are collected. Machine learning algorithms such as support vector regression and random forest regression are used to learn these data, and the complex relationship between environmental variables and photovoltaic power is mined, so as to construct the specific form of the function f(H G (t)), and then a complete first equation is established.

[0091] In the embodiment of the present application, the second operation is performed on the photovoltaic power dynamic diagram in the above step S104 to obtain the time representation of the photovoltaic power, including the following steps C1:

[0092] In C1: The second operation establishes a second equation. Based on the second equation, a temporal convolutional network is used to extract the time series information of different time steps of the photovoltaic power dynamic diagram; the time series information is aggregated to obtain the time representation of the photovoltaic power.

[0093] In the embodiment of the present application, the continuous time aggregation (CTA) method is used to establish a second equation based on ordinary differential equations (ODEs) to represent the temporal dynamic changes of photovoltaic power and environmental variables such as solar irradiance, light intensity, and temperature. Specifically, the second equation is expressed as:

[0094]

[0095] where H T (t) represents the time series features of each time step, TCN represents the temporal convolutional network, and Θ is a set of parameters.

[0096] In the embodiment of the present application, the temporal convolutional network includes a filtering convolution f C and a gated convolution f G , which are used to extract meaningful time information from the input sequence and control the transmission of the information flow. The filtering convolution captures features of different time scales through convolution kernels W m of different sizes, while the gated convolution controls the transmission of the information flow through an activation function, so that various periodic features such as daily cycles, weekly cycles, and seasonal cycles can be effectively captured. Finally, this time series information is aggregated into a complete temporal dynamic representation for improving the prediction accuracy of photovoltaic power;

[0097] The calculation formula is as follows:

[0098] TCN(H T (t), Θ) = f C (H T (t), Θ C ) ⊙ f G (H T (t), Θ G )

[0099] where ⊙ represents the element-wise multiplication operation, and f C and f G represent the filtering convolution and the gated convolution respectively, and their parameters are Θ C and Θ G .

[0100] It should be noted that when designing these convolution modules, the method of the present invention takes into account the periodic characteristics of the photovoltaic system data, the daily cycle, such as the change of light intensity between day and night; the seasonal cycle, such as the irradiance difference between summer and winter. By introducing filtering convolutions f C with different convolution kernel sizes, features on different time scales can be extracted, while the gated convolution f G ensures the effective capture of the complex interaction between short-term fluctuations and long-term trends by controlling the flow of information. Continuous time aggregation summarizes this information extracted from the time series into a complete temporal dynamic representation, which serves as the temporal part of the photovoltaic power prediction. This method can not only reduce the number of parameters, but also avoid the problem of gradient disappearance that may occur in traditional methods, thereby improving the computational efficiency and prediction accuracy of the model.

[0101] In the embodiment of the present application, step S106 described above includes the following steps D1 - D2:

[0102] In D1: Use the time series information obtained from the second operation as the initial state of each node in the first operation to obtain the spatio-temporal representation of the photovoltaic power;

[0103] Specifically, by organically combining the continuous graph propagation method and the continuous time aggregation method, modeling is carried out simultaneously in the time and space dimensions to obtain a more accurate photovoltaic power prediction result;

[0104] In the embodiment of the present application, each time state H T (t) in the continuous time aggregation process is used as the initial state of each node in the continuous graph propagation process, so as to achieve fully continuous joint modeling in time and space;

[0105] In the embodiment of the present application, the initial time state H(t) is calculated by the ordinary differential equation solver ODESolve 1 , and then another ordinary differential equation solver ODESolve 2Perform spatial propagation update in combination with the output of the temporal convolutional network, and finally obtain the spatial representation of the photovoltaic power;

[0106] In an alternative embodiment, the photovoltaic power prediction result can also be obtained by solving using the long short-term memory network method; specifically, organize historical photovoltaic power data, environmental variable data such as solar irradiance, temperature, and wind speed into a sequence in chronological order and input it into the long short-term memory network; the memory units and gating mechanisms inside the network can remember important information from past time steps and dynamically update this information according to the current input. Train the long short-term memory network with a large amount of training data, and adjust the weight parameters in the network so that the network can learn the complex non-linear relationship between photovoltaic power and environmental variables. During prediction, input new environmental variable data into the trained long short-term memory network, and the network will output the corresponding photovoltaic power prediction value.

[0107] In a possible embodiment, through two ordinary differential equation solvers ODESolve 1 and ODESolve 2 , the input time series can be converted into a spatial representation according to the following steps, and finally obtain H out :

[0108] 1. Initial state conversion:

[0109]

[0110] Among them, H(t) represents the dynamic state of the time series, and T cta is the time range of time aggregation;

[0111] The solution H(t) of the ordinary differential equation is obtained by the following formula.

[0112] 2. Time-space joint modeling:

[0113]

[0114] Among them, TCN(H(t), t, Θ) represents the output of the temporal convolutional network, which serves as the initial state of the CGP process, and P is a mapping function for transformation.

[0115] 3. Spatial propagation update:

[0116]

[0117] The above formula represents the dynamic change of spatial propagation, and numerical solution is performed through the ordinary differential equation solver ODESolve 2 In this way, the present invention can perform modeling synchronously in the time and space dimensions, and finally obtain a high-precision prediction result of the photovoltaic power.

[0118] It should be noted that the present invention combines the continuous graph propagation method and the continuous time aggregation method organically, enabling the present invention to capture the dynamic changes in the time series and the spatial dependence between photovoltaic power and weather characteristics, thereby improving the prediction accuracy.

[0119] In D2: The generated spatio-temporal representation is decoded through a fully connected layer to obtain the photovoltaic power prediction result.

[0120] It should be noted that the present invention innovatively models the continuous graph propagation method and the continuous time aggregation respectively from the spatial and time dimensions, and then combines them organically to achieve spatio-temporal joint modeling. There are multiple flexible methods for establishing and solving equations, which improves the adaptability and generalization ability; it can effectively capture the complex characteristics of photovoltaic power changes, including spatial dependence and various time period characteristics, and reduces the prediction error through the organic fusion of spatio-temporal information; the accurate prediction result can improve the online monitoring and energy management efficiency of the photovoltaic system, reduce the operation cost, and also enhance the acceptance and consumption capacity of the power system for photovoltaic electric energy, reduce the impact of its power fluctuation, and improve the stability of the power system.

[0121] The above is a schematic solution of a photovoltaic power generation prediction method of this embodiment. It should be noted that the technical solution of the photovoltaic power generation prediction system belongs to the same concept as the above-mentioned photovoltaic power generation prediction method. For the details not described in detail in the technical solution of the photovoltaic power generation prediction system in this embodiment, reference can be made to the description of the technical solution of the above-mentioned photovoltaic power generation prediction method.

[0122] Embodiment 2

[0123] This embodiment provides a photovoltaic power generation prediction system, including:

[0124] A data acquisition module, used to acquire a photovoltaic power data set;

[0125] A processing module, used to construct a photovoltaic power generation dynamic graph based on the photovoltaic power data set in combination with environmental variables; perform a first operation on the photovoltaic power dynamic graph to obtain the spatial representation of the photovoltaic power, and perform a second operation on the photovoltaic power dynamic graph to obtain the time representation of the photovoltaic power;

[0126] A prediction module, used to combine the spatial representation and the time representation to obtain the spatio-temporal representation of the photovoltaic power, and obtain the photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power.

[0127] The above-mentioned each unit module can be embedded in the processor of the computer device in a hardware form or be independent of it, or can be stored in the memory of the computer device in a software form, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned each module.

[0128] Embodiment 3

[0129] This embodiment provides a computer device, which may be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a photovoltaic power prediction method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0130] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor, it realizes: obtaining a photovoltaic power data set; based on the photovoltaic power data set, combining environmental variables to construct a photovoltaic power dynamic graph; performing a first operation on the photovoltaic power dynamic graph to obtain a spatial representation of the photovoltaic power, performing a second operation on the photovoltaic power dynamic graph to obtain a temporal representation of the photovoltaic power; combining the spatial representation and the temporal representation to obtain a spatio-temporal representation of the photovoltaic power, and obtaining a photovoltaic power prediction result according to the spatio-temporal representation of the photovoltaic power.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A photovoltaic power prediction method, characterized in that, Including: Obtain a photovoltaic power dataset; Based on the photovoltaic power dataset, construct a dynamic graph of photovoltaic power generation in combination with environmental variables; Perform a first operation on the dynamic graph of photovoltaic power to obtain a spatial representation of photovoltaic power, and perform a second operation on the dynamic graph of photovoltaic power to obtain a temporal representation of photovoltaic power; Combine the spatial representation and the temporal representation to obtain a spatio-temporal representation of photovoltaic power, and obtain a photovoltaic power prediction result according to the spatio-temporal representation of photovoltaic power.

2. The photovoltaic power generation prediction method according to claim 1, wherein Performing a first operation on the dynamic graph of photovoltaic power to obtain a spatial representation of photovoltaic power includes: The first operation obtains the state of each node of the dynamic graph of photovoltaic power at different time steps by establishing a first equation and solving the first equation; Aggregate the node states to obtain a spatial representation of photovoltaic power.

3. The photovoltaic power generation prediction method according to claim 2, wherein Performing a second operation on the dynamic graph of photovoltaic power to obtain a temporal representation of photovoltaic power includes: The second operation extracts the time series information of different time steps of the dynamic graph of photovoltaic power by establishing a second equation and using a temporal convolutional network based on the second equation; Aggregate the time series information to obtain a temporal representation of photovoltaic power.

4. The photovoltaic power generation prediction method according to claim 3, wherein Combining the spatial representation and the temporal representation to obtain a spatio-temporal representation of photovoltaic power includes: Taking the time series information obtained by the second operation as the initial state of each node in the first operation to obtain a spatio-temporal representation of photovoltaic power.

5. The photovoltaic power generation prediction method according to claim 4, wherein Obtaining a photovoltaic power prediction result according to the spatio-temporal representation of photovoltaic power includes: Decode the generated spatio-temporal representation through a fully connected layer to obtain a photovoltaic power prediction result.

6. The photovoltaic power generation prediction method according to claim 5, characterized in that, The first equation is expressed as: where dH G (t) represents the change of the node state over time, and f(H G (t)) is a function of the node state.

7. The photovoltaic power generation prediction method according to claim 5 or 6, characterized in that The second equation is expressed as: Among them, H T (t) represents the time series features at each time step, TCN represents the temporal convolutional network, and Θ is the parameter set.

8. A system applying the photovoltaic power generation prediction method according to any one of claims 1-7, characterized in that, Including: A data acquisition module for obtaining a photovoltaic power dataset; A processing module for constructing a dynamic graph of photovoltaic power generation based on the photovoltaic power dataset in combination with environmental variables; performing a first operation on the dynamic graph of photovoltaic power to obtain a spatial representation of photovoltaic power, and performing a second operation on the dynamic graph of photovoltaic power to obtain a temporal representation of photovoltaic power; A prediction module for combining the spatial representation and the temporal representation to obtain a spatio-temporal representation of photovoltaic power, and obtaining a photovoltaic power prediction result according to the spatio-temporal representation of photovoltaic power.

9. An electronic device, including: A memory and a processor; The memory is used for storing computer-executable instructions, and the processor is used for executing the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the photovoltaic power generation prediction method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the photovoltaic power generation prediction method according to any one of claims 1 to 7 are implemented.

Images