Photovoltaic power prediction method

Abstract

The invention relates to a photovoltaic power prediction method. The method comprises the following steps: acquiring training data, wherein the training data comprises a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers; the photovoltaic training parameters comprise a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter is a photovoltaic parameter which is continuous with the first photovoltaic parameter in time and is later than the first photovoltaic parameter in time; the training photovoltaic power is the actual photovoltaic power within the time corresponding to the second photovoltaic parameter; inputting the first photovoltaic training parameter into the initial prediction model to obtain first predicted photovoltaic power; the first predicted photovoltaic power is the predicted photovoltaic power within the time corresponding to the second photovoltaic parameter; determining a loss value based on the first predicted photovoltaic power and the first photovoltaic power; and optimizing parameters of the initial prediction model based on the loss value to obtain a target prediction model. By adopting the method, the target prediction model with relatively high accuracy in the aspect of photovoltaic power prediction can be obtained through training.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic, in particular to a photovoltaic power prediction method. BACKGROUND

[0002] With the popularization of photovoltaic power generation technology, photovoltaic power prediction technology has emerged. Photovoltaic power prediction refers to predicting the output power of a photovoltaic power generation system through various technical means to improve the accuracy of resource scheduling of the power system and ensure the stable operation of the power grid.

[0003] With the acceleration of global energy transformation, the installed capacity of photovoltaic power generation continues to expand, and the randomness and intermittency of photovoltaic power generation technology have brought severe challenges to power grid dispatching. If the prediction accuracy of photovoltaic power prediction technology cannot be improved, it will be difficult for the power grid dispatching department to ensure the stability of the power grid. SUMMARY

[0004] Therefore, it is necessary to provide a photovoltaic power prediction method, device, computer equipment, computer readable storage medium and computer program product capable of training a target prediction model with high accuracy in photovoltaic power prediction to solve the above technical problems.

[0005] In a first aspect, the present application provides a photovoltaic power prediction method, comprising:

[0006] obtaining training data, the training data comprising a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers; the photovoltaic training parameters comprising a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter being a photovoltaic parameter that is continuous in time with the first photovoltaic parameter and later in time than the first photovoltaic parameter; the training photovoltaic power being the actual photovoltaic power in the time corresponding to the second photovoltaic parameter;

[0007] inputting the first photovoltaic training parameter into an initial prediction model to obtain a first predicted photovoltaic power; the first photovoltaic training parameter being one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power being the predicted photovoltaic power in the time corresponding to the second photovoltaic parameter;

[0008] determining a loss value based on the first predicted photovoltaic power and the first photovoltaic power; the first photovoltaic power being one of the plurality of training photovoltaic powers corresponding to the first photovoltaic training parameter;

[0009] optimizing the parameters of the initial prediction model based on the loss value to obtain a target prediction model.

[0010] In a second aspect, the present application also provides a photovoltaic power prediction device, comprising:

[0011] The acquisition module is configured to acquire training data, the training data comprising a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers; the photovoltaic training parameters comprising a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter being a photovoltaic parameter that is continuous in time with the first photovoltaic parameter and later in time than the first photovoltaic parameter; the training photovoltaic power being an actual photovoltaic power in a time corresponding to the second photovoltaic parameter;

[0012] The first determination module is configured to input the first photovoltaic training parameter into an initial prediction model to obtain a first predicted photovoltaic power; the first photovoltaic training parameter being one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power being a predicted photovoltaic power in a time corresponding to the second photovoltaic parameter;

[0013] The second determination module is configured to determine a loss value based on the first predicted photovoltaic power and a first photovoltaic power; the first photovoltaic power being one of the plurality of training photovoltaic powers corresponding to the first photovoltaic training parameter;

[0014] The third determination module is configured to optimize parameters of the initial prediction model based on the loss value to obtain a target prediction model.

[0015] In a third aspect, a computer device is provided. The computer device includes a memory and a processor. The memory stores a computer program. The processor implements some or all of the steps described in any method of the first aspect of the embodiments of the present application when executing the computer program.

[0016] In a fourth aspect, a computer readable storage medium is provided. The computer readable storage medium stores a computer program. The computer program, when executed by a processor, implements some or all of the steps described in any method of the first aspect of the embodiments of the present application.

[0017] In a fifth aspect, a computer program product is provided. The computer program product includes a computer program. The computer program, when executed by a processor, implements some or all of the steps described in any method of the first aspect of the embodiments of the present application.

[0018] The photovoltaic power prediction method, device, computer device, computer readable storage medium and computer program product obtain training data, the training data includes a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers; the photovoltaic training parameters include first photovoltaic parameters and second photovoltaic parameters; the second photovoltaic parameters are photovoltaic parameters that are continuous in time with the first photovoltaic parameters and later in time than the first photovoltaic parameters; the training photovoltaic power is the actual photovoltaic power in the corresponding time of the second photovoltaic parameters; the first photovoltaic training parameters are input into an initial prediction model to obtain first predicted photovoltaic power; the first photovoltaic training parameters are one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power is the predicted photovoltaic power in the corresponding time of the second photovoltaic parameters; a loss value is determined based on the first predicted photovoltaic power and the first photovoltaic power; the first photovoltaic power is one of the plurality of training photovoltaic powers corresponding to the first photovoltaic training parameters; and parameters of the initial prediction model are optimized based on the loss value to obtain a target prediction model. The photovoltaic power prediction method provided in the present application can train a target prediction model with high accuracy in photovoltaic power prediction, because the photovoltaic training parameters include the first photovoltaic parameters and the second photovoltaic parameters that are continuous in time, and the parameters of the initial prediction model are optimized based on the loss value determined based on the first predicted photovoltaic power and the first photovoltaic power. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without creative effort.

[0020] Figure 1 An application environment diagram of the photovoltaic power prediction method in an embodiment;

[0021] Figure 2 A flowchart of the photovoltaic power prediction method in an embodiment;

[0022] Figure 3 A structure diagram of the initial prediction model in an embodiment;

[0023] Figure 4 A structure block diagram of the photovoltaic power prediction device in an embodiment;

[0024] Figure 5 An internal structure diagram of the computer device in an embodiment;

[0025] Figure 6 An internal structure diagram of the computer device in another embodiment. DETAILED DESCRIPTION

[0026] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0027] The photovoltaic power prediction method provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 . The terminal 102 communicates with the server 104 through a network. The data storage system can store data required to be processed by the server 104. The data storage system can be integrated on the server 104, or placed on a cloud or other network server. The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, a projection device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be a stand-alone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. It is easy to understand that, in addition to obtaining the target prediction model by using the photovoltaic power prediction method provided by the embodiments of the present application, the application of the target prediction model can also be applied to an application environment as shown in Figure 1 .

[0028] In an exemplary embodiment, as shown in Figure 2 , a photovoltaic power prediction method is provided. Taking the terminal in Figure 1 as an example, the method includes the following steps 202 to 208. Wherein:

[0029] Step 202, obtaining training data, the training data including a plurality of photovoltaic training parameters and a plurality of training photovoltaic power; the photovoltaic training parameters including a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter being a photovoltaic parameter that is continuous in time with the first photovoltaic parameter and later in time than the first photovoltaic parameter; the training photovoltaic power being an actual photovoltaic power in a time corresponding to the second photovoltaic parameter.

[0030] The training data can be obtained from historical data of a photovoltaic power generation system. Optionally, the photovoltaic power generation system can be a photovoltaic power plant.

[0031] Optionally, the first photovoltaic parameter can comprise a first weather variable, a first theoretical irradiance, a first photovoltaic component temperature, and a second photovoltaic power.

[0032] Optionally, the second photovoltaic parameter can comprise a second weather variable, a second theoretical irradiance, and a second photovoltaic component temperature.

[0033] Since in the model training process, the parameters of the initial prediction model are optimized to obtain the final target prediction model based on the loss value between the first predicted photovoltaic power output by the initial prediction model and one of the training photovoltaic powers, the photovoltaic power parameter is not included in the second photovoltaic parameter.

[0034] The second photovoltaic parameter is continuous in time with the first photovoltaic parameter, meaning that the starting time corresponding to the second photovoltaic parameter immediately follows the ending time corresponding to the first photovoltaic parameter.

[0035] Optionally, the time length corresponding to the second photovoltaic parameter is less than the time length corresponding to the first photovoltaic parameter.

[0036] For example, assuming that the time corresponding to the first photovoltaic parameter is from June 1, 2024 to August 31, 2024, the time corresponding to the second photovoltaic parameter can be from September 1, 2024 to September 3, 2024. By ensuring that the second photovoltaic parameter is continuous in time with the first photovoltaic parameter, it is beneficial to ensure that the first predicted photovoltaic power obtained by the initial prediction model based on the first photovoltaic training parameter can fully consider the photovoltaic power characteristics within the time length corresponding to the first photovoltaic parameter and within the time length corresponding to the second photovoltaic parameter, that is, it can more fully consider the continuity of photovoltaic power characteristics over time, and has higher accuracy.

[0037] Step 204: inputting the first photovoltaic training parameter into the initial prediction model to obtain the first predicted photovoltaic power; the first photovoltaic training parameter is one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power is the predicted photovoltaic power within the time corresponding to the second photovoltaic parameter.

[0038] Optionally, the initial prediction model refers to a prediction model in a non-converged state, that is, the parameters have not been optimized at this time.

[0039] For easy understanding, the first photovoltaic training parameter comprises the corresponding first photovoltaic parameter and the second photovoltaic parameter.

[0040] The first predicted photovoltaic power refers to the result predicted by the initial prediction model for the photovoltaic power within the time corresponding to the second photovoltaic parameter.

[0041] In step 206, a loss value is determined based on the first predicted photovoltaic power and the first photovoltaic power. The first photovoltaic power is one of the training photovoltaic powers corresponding to the first photovoltaic training parameter.

[0042] The loss value is a quantitative index for measuring the difference between the first predicted photovoltaic power and the first photovoltaic power.

[0043] The first photovoltaic power is the photovoltaic power in the time corresponding to the first photovoltaic training parameter.

[0044] In step 208, the parameters of the initial prediction model are optimized based on the loss value to obtain a target prediction model.

[0045] In an exemplary embodiment, the optimization of the parameters of the initial prediction model based on the loss value to obtain the target prediction model includes: optimization of the parameters of the initial prediction model based on a target loss function and the loss value to obtain the target prediction model.

[0046] Optionally, the target loss function can be a rectified linear unit (ReLU) function.

[0047] As can be easily understood, the parameters of the initial prediction model are iteratively optimized based on the loss value until the loss value converges, and then the target prediction model is obtained.

[0048] In the photovoltaic power prediction method, the training data includes a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers. The photovoltaic training parameters include a first photovoltaic parameter and a second photovoltaic parameter. The second photovoltaic parameter is a photovoltaic parameter that is time-continuous with the first photovoltaic parameter and later in time than the first photovoltaic parameter. The training photovoltaic power is the actual photovoltaic power in the time corresponding to the second photovoltaic parameter. The first photovoltaic training parameter is input into an initial prediction model to obtain a first predicted photovoltaic power. The first photovoltaic training parameter is one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power is the predicted photovoltaic power in the time corresponding to the second photovoltaic parameter. A loss value is determined based on the first predicted photovoltaic power and the first photovoltaic power. The first photovoltaic power is one of the training photovoltaic powers corresponding to the first photovoltaic training parameter. The parameters of the initial prediction model are optimized based on the loss value to obtain a target prediction model. The photovoltaic power prediction method provided by the present application can train a target prediction model with high accuracy in photovoltaic power prediction by optimizing the parameters of the initial prediction model based on the loss value determined based on the first predicted photovoltaic power and the first photovoltaic power.

[0049] In an exemplary embodiment, the method further includes:

[0050] obtaining to-be-predicted data;

[0051] based on the to-be-predicted data, using a target prediction model to predict photovoltaic power, to obtain target photovoltaic power.

[0052] In this embodiment, since the first photovoltaic training parameters input into the initial prediction model include the first photovoltaic parameters and the second photovoltaic parameters, correspondingly, the to-be-predicted data also includes the photovoltaic parameters in the preset historical time and the photovoltaic parameters in the preset future time; the photovoltaic parameters in the preset future time are continuous in time with the photovoltaic parameters in the preset historical time, and are later in time than the photovoltaic parameters in the preset historical time.

[0053] The target photovoltaic power refers to the photovoltaic power corresponding to the time of the predicted photovoltaic parameters in the preset future time.

[0054] In this embodiment, since the target prediction model has high accuracy in photovoltaic power prediction, when the to-be-predicted data is input into the target prediction model, the corresponding target photovoltaic power in the preset future time with high accuracy can be obtained.

[0055] In one exemplary embodiment, the above obtaining to-be-predicted data includes:

[0056] obtaining historical weather variables, historical photovoltaic module temperatures, historical photovoltaic power in a preset historical time, and predicted weather variables in a preset future time;

[0057] determining historical theoretical irradiance based on the historical weather variables;

[0058] determining predicted theoretical irradiance based on the predicted weather variables;

[0059] determining predicted photovoltaic module temperature based on the historical weather variables, the predicted weather variables, a static heat loss coefficient, a wind cooling coefficient, and an absorption efficiency of the photovoltaic module to solar radiation energy;

[0060] determining the historical weather variables, the historical theoretical irradiance, the historical photovoltaic module temperature, the historical photovoltaic power, the predicted weather variables, the predicted theoretical irradiance, and the predicted photovoltaic module temperature as the to-be-predicted data.

[0061] In this embodiment, the length of time corresponding to the preset future time is less than the length of time corresponding to the preset historical time.

[0062] Optionally, the preset historical time can be a historical 1-month time, a historical 3-month time, a historical 6-month time, or other historical time lengths.

[0063] Optionally, the preset future time can be a future 1-day time, a future 3-day time, a future 5-day time, or other future time lengths.

[0064] Optionally, the predicted weather variable can be obtained from a meteorological bureau platform, a weather station at the location of the photovoltaic power generation system, or other meteorological agencies.

[0065] The theoretical irradiance and the photovoltaic module temperature are variables that are manually calculated and added in order to reduce the model deviation caused by sensor errors or outliers, and to enhance the generalization ability of the target prediction model to different climate conditions. In other words, the historical theoretical irradiance and the predicted theoretical irradiance involved in the training process of the initial prediction model and the application process of the target prediction model are manually calculated, while the historical photovoltaic module temperature can be directly obtained from the historical data, and the predicted photovoltaic module temperature needs to be calculated.

[0066] The theoretical irradiance is used to provide an accurate irradiance reference, which is beneficial to remove the random fluctuation factors of the weather environment caused by the atmosphere and the weather, and to strengthen the physical consistency of the weather environment, thereby improving the adaptability of the target prediction model to different weather and different time periods, and reducing the fitting deviation of the target prediction model to climate change.

[0067] Optionally, the clear sky model (Ineichen model) can be used to calculate the historical theoretical irradiance and the predicted theoretical irradiance of the photovoltaic power generation system.

[0068] The static heat loss coefficient refers to a parameter used to describe the heat loss ability of the photovoltaic module in the photovoltaic power generation system under static conditions, where the static condition refers to the absence of wind cooling or other convection effects.

[0069] The wind cooling coefficient refers to a parameter used to describe the cooling effect of the wind on the photovoltaic module.

[0070] The photovoltaic module temperature can be understood as the temperature of the photovoltaic cell.

[0071] The absorption efficiency of the photovoltaic module to solar radiation energy refers to the ratio between the portion of solar radiation energy that can be converted into electrical energy by the photovoltaic module and the solar radiation energy absorbed by the photovoltaic module.

[0072] The photovoltaic module temperature is an important factor affecting the photovoltaic power. However, when predicting the photovoltaic power at a future time point at the current time point, the photovoltaic module temperature at the future time point cannot be directly obtained. Therefore, optionally, a wind speed coupled heat balance model can be used to calculate the predicted photovoltaic module temperature.

[0073] Optionally, the predicted photovoltaic module temperature can be determined based on the historical temperature average value in the historical weather variable, the predicted total irradiance in the predicted weather variable, the predicted wind speed in the predicted weather variable, the temperature rise adjustment term in the predicted weather variable, the static heat loss coefficient, the wind cooling coefficient, and the absorption efficiency of the photovoltaic module to solar radiation energy.

[0074] Exemplarily, the historical average temperature in the historical weather variable is represented as the predicted total irradiance in the predicted weather variable is represented as the predicted wind speed in the predicted weather variable is represented as the temperature rise adjustment term in the predicted weather variable is represented as the static heat loss coefficient is represented as the wind cooling coefficient is represented as the absorption efficiency of the photovoltaic module to the solar radiation energy is represented as the predicted photovoltaic module temperature can be represented by the following calculation formula: .

[0075] In the embodiment, the to-be-predicted data is determined by the historical weather variable, the historical theoretical irradiance, the historical photovoltaic module temperature, the historical photovoltaic power, the predicted weather variable, the predicted theoretical irradiance and the predicted photovoltaic module temperature, and the historical theoretical irradiance, the predicted theoretical irradiance and the predicted photovoltaic module temperature in the to-be-predicted data are all calculated variables, so that the data quality of the to-be-predicted data can be improved, and the accuracy and reliability of the target photovoltaic power in the future time predicted by the target prediction model can be improved.

[0076] In an exemplary embodiment, the first photovoltaic parameter includes the first weather variable, the first theoretical irradiance, the first photovoltaic module temperature and the second photovoltaic power, and the second photovoltaic parameter includes the second weather variable, the second theoretical irradiance and the second photovoltaic module temperature.

[0077] It is easily understood that, whether in the training process of the initial prediction model or in the application process of the target prediction model, the photovoltaic power in the future time relative to the historical time needs to be predicted, so in the training process of the initial prediction model, although the second photovoltaic parameter as the historical data actually includes the first photovoltaic power, the first photovoltaic power will not be input into the initial prediction model.

[0078] It should be noted that, since the training process of the initial prediction model corresponds to the application process of the target prediction model, the calculation method of the first theoretical irradiance in the first photovoltaic parameter and the second theoretical irradiance in the second photovoltaic parameter is the same as the calculation method of the historical theoretical irradiance in the to-be-predicted data and the predicted theoretical irradiance, and thus will not be described here.

[0079] In an exemplary embodiment, as shown in Figure 3 the initial prediction model 30 includes a time sequence module 302, a non-time sequence module 304 and a prediction module 306;

[0080] The inputting the first photovoltaic training parameter into the initial prediction model 30 to obtain the first predicted photovoltaic power comprises:

[0081] The time sequence module 302 is used to extract features of the first photovoltaic parameter in the first photovoltaic training parameter to obtain time sequence features.

[0082] The non-time sequence module 304 is used to extract features of the second photovoltaic parameter in the first photovoltaic training parameter to obtain non-time sequence features.

[0083] The prediction module 306 is used to predict the photovoltaic power based on the time sequence features and the non-time sequence features to obtain the first predicted photovoltaic power.

[0084] Since the environmental factors such as illumination and temperature in the historical time have time dynamic characteristics, the time sequence module 302 with time sequence feature extraction capability is particularly suitable for the first photovoltaic training parameter. Optionally, the time sequence module 302 including a temporal convolutional network (TCN) layer can be used to extract features of the first photovoltaic parameter in the first photovoltaic training parameter, that is, the TCN layer can be included in the time sequence module 302.

[0085] The TCN layer is composed of causal convolution, residual connection and dilated convolution. The causal convolution can ensure that the TCN layer only depends on the first photovoltaic parameter in the historical time when predicting the first predicted photovoltaic power in the future time. The residual connection can solve the gradient vanishing problem in the deep network through the skip connection to improve the training efficiency of the initial prediction model 30. The dilated convolution can expand the receptive field of the convolution kernel through interval sampling, which realizes the dependence relationship of the first photovoltaic parameter in a longer time range without increasing the computational complexity. It can be seen that, compared with the traditional recurrent neural network (RNN) or long short-term memory network (LSTM), the TCN layer can have better parallelism and training stability in the training process of the initial prediction model 30.

[0086] Since the factor that most directly affects the photovoltaic power in the future time is the weather condition corresponding to the time node, in addition to the time sequence module 302 for extracting the time sequence features of the first photovoltaic parameter in the historical time, a non-time sequence module 304 needs to be additionally arranged to extract the influence of other variables in the future time on the photovoltaic power. Optionally, the other variables in the future time include weather variables, manually calculated theoretical irradiance, and photovoltaic cell temperature, etc. Optionally, the non-time sequence module 304 including a Fully Convolutional Neural Network (FCNN) layer can be adopted to realize high-order feature extraction on the weather variable condition in the future time to obtain the non-time sequence features corresponding to the weather variable condition in the future time.

[0087] It should be noted that, in the process of training the initial prediction model 30, the second photovoltaic parameter is a photovoltaic parameter in the future time relative to the first photovoltaic parameter, but the first photovoltaic parameter and the second photovoltaic parameter are both photovoltaic parameters in the historical time relative to the current time node.

[0088] Exemplarily, in the case where the non-time sequence module 304 includes the FCNN layer, the FCNN layer includes multiple convolution layers, and the number of convolution layers is denoted as The weight of the lth convolution layer is denoted as The bias term of the lth convolution layer is denoted as The non-time sequence feature is denoted as The non-time sequence feature As a high-order abstract representation of the weather variable, the non-time sequence weather environment attribute information can be highlighted to generate high-level and abstract weather environment features for representing the weather environment state. Based on this, the non-time sequence feature can be represented by the following calculation formula: It can be seen that, by the feature extraction of the second photovoltaic parameter in the first photovoltaic training parameter through the multiple convolution layers of the non-time sequence module 304, the non-time sequence weather variable attribute information can be highlighted, and thus high-level and abstract non-time sequence features can be generated to represent the weather variable state in the future time through the non-time sequence features.

[0089] Optionally, the weights of the different layer convolution layers in the FCNN layer are continuously adjusted in the training process of the initial prediction model 30, and finally, the mapping relationship from the input to the output, i.e., the mapping relationship from the second photovoltaic parameter to the non-time sequence feature, is gradually clear and explicit through the multiple layer transformations of the FCNN layer.

[0090] In this embodiment, the first photovoltaic parameter is extracted using the time sequence module to obtain time sequence features, and the second photovoltaic parameter is extracted using the non-time sequence module to obtain non-time sequence features. Further, based on the time sequence features and the non-time sequence features, the photovoltaic power is predicted using the prediction module to obtain the first predicted photovoltaic power. Based on this, in the process of predicting the photovoltaic power in the future time, the effective complementation of information based on the time sequence features and the non-time sequence features is realized, and the second photovoltaic parameter obtained based on the non-time sequence features fully considers the weather variable state in the future time, thereby ensuring that the target prediction model obtained by training has high accuracy in photovoltaic power prediction.

[0091] In one exemplary embodiment, as shown in Figure 3 The time sequence module 302 includes a time sequence feature extraction layer 3022. The time sequence module 302 extracts the first photovoltaic parameter in the first photovoltaic training parameter to obtain time sequence features, including:

[0092] The time sequence feature extraction layer 3022 extracts the first photovoltaic parameter in the first photovoltaic training parameter to obtain time sequence features.

[0093] In this embodiment, when the time sequence feature extraction layer is a TCN layer, the TCN layer can fully capture the characteristics of the first photovoltaic parameter in time sequence to extract more abundant and more accurate time sequence features. The time sequence features can provide accurate time sequence information for the training process of the initial prediction model, and ensure that the target prediction model obtained based on the time sequence features has high accuracy in photovoltaic power prediction.

[0094] In one exemplary embodiment, as shown in Figure 3 The time sequence module 302 further includes a decomposition layer 3024.

[0095] The time sequence feature extraction layer 3022 extracts the first photovoltaic parameter in the first photovoltaic training parameter to obtain time sequence features, including:

[0096] The decomposition layer 3024 is used to perform variational modal decomposition on the first photovoltaic parameter to obtain a plurality of intrinsic modal functions. The values of the plurality of intrinsic modal functions satisfy a preset value condition.

[0097] The time sequence feature extraction layer 3022 extracts the plurality of intrinsic modal functions to obtain time sequence features.

[0098] In the formula, VMD refers to a self-adaptive signal processing method for decomposing a complex signal into a plurality of intrinsic mode functions (IMFs) with different center frequencies and bandwidths. The VMD decomposes the signal into a plurality of intrinsic mode functions by constructing a variational problem and using an optimization algorithm to find the optimal intrinsic mode functions, thereby realizing multi-scale decomposition of the signal.

[0099] In the plurality of intrinsic mode functions obtained by performing the VMD on the first photovoltaic parameter, each intrinsic mode function corresponds to a different frequency scale, and each intrinsic mode function can reflect the change of the first photovoltaic parameter in different frequency bands, which is conducive to revealing the hidden structural relationship in the first photovoltaic parameter.

[0100] The purpose of performing the VMD on the first photovoltaic parameter is to find a set of optimal intrinsic mode functions and center frequencies to minimize the objective function.

[0101] The preset value condition refers to a condition that minimizes the objective function of the VMD. The values of the plurality of intrinsic mode functions satisfy the preset value condition, which means that the values of the plurality of intrinsic mode functions satisfy the condition that minimizes the objective function of the VMD.

[0102] Exemplarily, the number of intrinsic mode functions is denoted as The intrinsic mode function is denoted as The center frequency is denoted as The partial derivative with respect to time t is denoted as The Dirac function representing a unit impulse is denoted as The kernel of the Hilbert transform used to generate the analytic signal is denoted as The complex analytic signal can be denoted as Based on this, the objective function can be denoted as Based on mathematical principles, it is easy to understand that the objective function is an energy function that sums all intrinsic mode functions to obtain the minimum value.

[0103] Optionally, the first photovoltaic parameter is subjected to variational mode decomposition using the decomposition layer 3024 to obtain a plurality of intrinsic mode functions. Specifically, the historical photovoltaic power in the first photovoltaic parameter is subjected to variational mode decomposition using the decomposition layer 3024 to obtain a plurality of intrinsic mode functions corresponding to the historical photovoltaic power with different center frequencies. Then, the plurality of intrinsic mode functions are combined with other parameter variables in the first photovoltaic parameter except the historical photovoltaic power (i.e., the first weather variable, the first theoretical irradiance, and the first photovoltaic component temperature) to obtain a plurality of variable functions. The plurality of variable functions are input into the time sequence feature extraction layer 3022 for feature extraction to obtain time sequence features.

[0104] In the case where the time sequence module 302 includes a TCN layer, the time sequence feature extraction layer 3022 can be the TCN layer. In this case, the time sequence feature extraction layer 3022 can be stacked by a plurality of residual units, each of which includes a causal dilated convolution component, a weight normalization component, an activation function component, a regularization component, and a 1*1 convolution component.

[0105] The causal dilated convolution component is used to perform a convolution operation combining the characteristics of causal convolution and dilated convolution. The dilated convolution expands the receptive field of the convolution kernel through interval sampling, which can capture the dependency of the first photovoltaic parameter in a longer time range without increasing the computational complexity. Based on this, the causal dilated convolution component can maintain causality and expand the receptive field, and is suitable for time sequence feature extraction.

[0106] The weight normalization component is used to normalize the weights of the convolution kernel. By reducing the problems of gradient vanishing and gradient explosion, the stability and efficiency of the training process are improved.

[0107] Optionally, the activation function is a ReLU function. The activation function is used to enable the neural network to learn complex patterns and relationships, and can be used to reduce the problem of gradient vanishing.

[0108] The regularization component is used to randomly discard the output of a portion of neurons during the training process of the initial prediction model 30 to prevent the final target prediction model from having the undesirable phenomenon of overfitting and improve the generalization ability of the target prediction model.

[0109] The 1*1 convolution component is used to adjust the channel number of the first photovoltaic parameter without changing the spatial dimension of the first photovoltaic parameter to ensure that the first photovoltaic parameter can be connected in residual.

[0110] For example, the intrinsic mode function of the first photovoltaic parameter is represented as The time sequence feature extraction layer 3022 is used to extract features from the plurality of intrinsic mode functions, which realizes feature coding through multiple layers of dilated convolution. The number of layers is represented as The expanded convolution operation is represented as , Finally, the feature encoding obtained through the multi-layer expanded convolution is The time series feature obtained by fusing the features of all the eigenmode functions can be represented as i is the number of eigenmode functions.

[0111] In this embodiment, the time series module includes a decomposition layer and a time series feature extraction layer. Thus, through the variational mode decomposition and the multi-layer convolution operation, the target prediction model obtained through training can effectively capture the time series features of the photovoltaic parameters in the historical time, improve the understanding and prediction ability of the target prediction model for time series data, and also improve the generalization ability of the target prediction model.

[0112] In an exemplary embodiment, as shown in Figure 3 the time series module 302 further includes a dynamic window layer 3026;

[0113] The above uses the decomposition layer 3024 to perform variational mode decomposition on the first photovoltaic parameter to obtain a plurality of eigenmode functions, including:

[0114] The dynamic window layer 3026 is used to filter data of the first photovoltaic parameter to obtain partial training data, and the first photovoltaic parameter includes the partial training data.

[0115] The decomposition layer 3024 is used to perform variational mode decomposition on the partial training data to obtain a plurality of eigenmode functions.

[0116] The first photovoltaic parameter includes the partial training data, that is, the data amount of the partial training data is less than the first photovoltaic parameter.

[0117] The dynamic window layer 3026 has different window lengths at different time nodes, that is, the dynamic window layer 3026 is a dynamic window that is adaptively adjusted with the time node.

[0118] Optionally, in the case that the first photovoltaic parameter includes the weather variable (i.e., the first weather variable) in the corresponding historical time, and the weather variable in the historical time includes the irradiance data, the corresponding dynamic window length of the dynamic window layer 3026 used for filtering data of the first photovoltaic parameter is determined by the minimum window length, the maximum window length, the irradiance variance at the corresponding time node, the response sensitivity, the response threshold, and the median of the historical rolling maximum variance.

[0119] Optionally, the corresponding time node can be a connection time node between the first photovoltaic parameter and the second photovoltaic parameter, that is, a time intersection between the time node of the first photovoltaic parameter and the second photovoltaic parameter. In the process of data screening of the first photovoltaic parameter by using the dynamic window layer 3026, the corresponding dynamic window length of the dynamic window layer 3026 is the dynamic window length under the corresponding time node.

[0120] Optionally, the minimum window length and the maximum window length are pre-set lengths.

[0121] Optionally, the response sensitivity and the response threshold are pre-set parameters and can be adjusted.

[0122] Exemplarily, the minimum window length is represented as The maximum window length is represented as The irradiance variance on the corresponding time node is represented as The response sensitivity is represented as The response threshold is represented as The median of the historical rolling maximum variance is represented as The rolling window size is represented as The median of the historical rolling maximum variance can be represented as The median of the historical rolling maximum variance can improve the anti-exception ability of the dynamic window length. The corresponding dynamic window length of the dynamic window layer 3026 is represented as The corresponding dynamic window length of the dynamic window layer 3026 can be represented by the following calculation formula: The calculation formula of the corresponding dynamic window length of the dynamic window layer 3026 combines the sigmoid nonlinear mapping, which ensures that the range of the corresponding dynamic window length of the dynamic window layer 3026 can be adaptively adjusted.

[0123] As can be easily understood, based on the calculation formula of the corresponding dynamic window length of the dynamic window layer 3026, in the weather conditions such as rain, heavy fog and the like with large irradiance change degree, the dynamic window length is small; in the weather conditions such as clear sky and the like with small irradiance change degree, the dynamic window length is large.

[0124] It should be noted that after obtaining the dynamic window lengths by self-adaption, the lengths of the dynamic window lengths are not the same, and cannot be applied to the subsequent training process of the initial prediction model 30. Therefore, a mask is constructed here to unify the input length to The corresponding dynamic window length of the dynamic window layer 3026 is The first time step is set to 0.

[0125] ​In the embodiment, based on the dynamic window layer, a dynamic time window with a smaller length is used to capture the rapid change of the first photovoltaic parameter in short-term fluctuation weather such as rain and fog, and a dynamic time window with a larger length is used to extract the smooth and long-term trend of the first photovoltaic parameter in sunny weather. Based on this, since the embodiment avoids the rough method of dividing the weather type in units of days in the traditional way, the dynamic window layer can adapt to different weather changes in the same day, that is, the dynamic window layer can dynamically adapt to short-term weather changes, adjust the granularity of feature extraction of the first photovoltaic parameter by using the adaptive time window method, to cope with different scales of short-term weather changes, improve the time domain sensitivity of the target prediction model trained, and enhance the pertinence of the target prediction model to the time sequence features.

[0126] In an exemplary embodiment, the above obtaining training data comprises:

[0127] obtaining initial training data;

[0128] dividing the initial training data according to a preset time window length to obtain a plurality of initial training data segments;

[0129] determining the mean and standard deviation of each initial training data segment within the respective preset time window length;

[0130] determining abnormal data in the initial training data based on each initial training data segment, the mean corresponding to each initial training data segment, and the standard deviation corresponding to each initial training data segment;

[0131] eliminating the abnormal data from the initial training data to obtain the training data.

[0132] wherein the initial training data refers to training data that has not been subjected to abnormal data elimination.

[0133] Optionally, the preset time window length is set by the user in advance. Optionally, the preset time window length can be a time window length of 1 week, a time window length of 10 days, a time window length of 1 month, or a time window length of other length.

[0134] As can be easily understood, the preset time window length is smaller than the time length corresponding to the first photovoltaic parameter.

[0135] The time window length of each initial training data segment in the plurality of initial training data segments is the same.

[0136] Optionally, the mean and the standard deviation of each initial training data segment in the respective preset time window length are determined, and the mean and the standard deviation of different types of parameters in the preset time window length are determined, so that in the process of removing abnormal data from the initial training data, abnormal data corresponding to different types of parameters can be removed to obtain training data without abnormal data of different types of parameters.

[0137] In an exemplary embodiment, the determination of the abnormal data in the initial training data based on the initial training data segments, the mean corresponding to each initial training data segment, and the standard deviation corresponding to each initial training data segment comprises: determining the product of the adjustment parameter and the standard deviation corresponding to each initial training data segment; determining the minimum observation data based on the difference between the mean of each initial training data segment and the product; determining the maximum observation data based on the sum of the mean corresponding to each initial training data segment and the product; and determining the abnormal data in the initial training data by determining the observation data outside the minimum observation data and the maximum observation data in each initial training data segment.

[0138] The adjustment parameter can be set according to the experience of model training, that is, it can be adjusted and optimized according to the loss value.

[0139] For example, each initial training data segment is represented as The mean corresponding to each initial training data segment is represented as The standard deviation corresponding to each initial training data segment is represented as The adjustment parameter is represented as The minimum initial training data segment is represented as The maximum initial training data segment is represented as The calculation formula of the minimum initial training data segment is represented as: The calculation formula of the maximum initial training data segment is represented as: .

[0140] In this embodiment, the abnormal data is determined from the initial training data and removed to obtain the training data, so that the abnormal data caused by noise in the initial training data can be avoided to interfere with the model training process, and the target prediction model with high accuracy in photovoltaic power prediction can be trained based on the training data after removing the abnormal data.

[0141] In an exemplary embodiment, the removal of the abnormal data from the initial training data to obtain the training data comprises:

[0142] The abnormal data is removed from the initial training data to obtain initial training data after removing abnormal data. The missing values in the initial training data after removing abnormal data are filled by a K-nearest neighbor algorithm (KNN) to obtain training data.

[0143] In this embodiment, there may be continuous abnormal data in the initial training data. In this case, if a simple interpolation method is used to fill the missing values, the filled values may have large errors. Therefore, the KNN algorithm is used to predict the missing values based on the initial training data after removing abnormal data, and the predicted filled values are used to replace the missing values, which can significantly improve the data quality of the training data, and thus the target prediction model trained based on the improved training data has high accuracy in photovoltaic power prediction.

[0144] In one example embodiment, as shown in FIG. 3, the prediction module 306 includes a feature fusion layer 3062 and a fully connected layer 3064. The prediction module 306 is used to predict the photovoltaic power based on the time series features and the non-time series features to obtain the first predicted photovoltaic power, including: Figure 3

[0145] The feature fusion layer 3062 is used to fuse the time series features and the non-time series features to obtain the fused features.

[0146] The fully connected layer 3064 is used to predict the photovoltaic power based on the fused features to obtain the first predicted photovoltaic power.

[0147] For the fully connected layer 3064, the input data is the fused features, and the output data is the first predicted photovoltaic power.

[0148] In one example embodiment, the time series features and the non-time series features are fused to obtain the fused features, including:

[0149] The first weight is determined based on the non-time series features.

[0150] The second weight is determined based on the first weight.

[0151] The fused features are obtained based on the first weight, the time series features, the second weight, and the non-time series features. The first weight corresponds to the time series features, and the second weight corresponds to the non-time series features.

[0152] The first weight is an attention weight learned based on the non-time series features.

[0153] Optionally, the sum of the first weight and the second weight is 1, i.e., the second weight = 1 - the first weight.

[0154] ​Optionally, the fusion feature can be a weighted sum result between the second weight, the time sequence feature, the first weight and the non-time sequence feature, that is, can be expressed as: fusion feature = first weight * time sequence feature + second weight * non-time sequence feature.

[0155] Optionally, the first weight can be determined based on the non-time sequence feature by a Sigmoid function.

[0156] Exemplarily, the non-time sequence feature at the t time node is expressed as , the time sequence feature at the t time node is expressed as , the fusion feature is expressed as , the first weight is expressed as , in the case that the first weight is determined based on the non-time sequence feature by a Sigmoid function, the first weight can be further expressed as , based on which, in order to realize that the first weight corresponds to the time sequence feature and the second weight corresponds to the non-time sequence feature, the fusion feature can be expressed by the following formula: .

[0157] In the embodiment, the fusion feature is obtained based on the first weight, the time sequence feature, the second weight and the non-time sequence feature, and the first weight corresponds to the time sequence feature and the second weight corresponds to the non-time sequence feature, so that the feature weight corresponding to the time sequence feature and the non-time sequence feature is dynamically adjusted by using the attention weight, so that the fusion feature obtained finally can deeply mine the feature characteristics of the time sequence feature and the non-time sequence feature, and also ensure that the time sequence feature can effectively complement the non-time sequence feature in information aspect, so as to ensure that the target prediction model obtained finally has high accuracy in photovoltaic power prediction.

[0158] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0159] Based on the same inventive concept, the embodiments of the present application also provide a photovoltaic power prediction device for implementing the photovoltaic power prediction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more photovoltaic power prediction device embodiments provided below can refer to the limitations of the photovoltaic power prediction method described above, which will not be repeated here.

[0160] In one exemplary embodiment, as shown in Figure 4 A photovoltaic power prediction device is provided, comprising: an acquisition module 402, a first determination module 404, a second determination module 406, and a third determination module 408, wherein:

[0161] The acquisition module 402 is configured to acquire training data, the training data comprising a plurality of photovoltaic training parameters and a plurality of training photovoltaic powers; the photovoltaic training parameters comprising a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter being a photovoltaic parameter that is continuous in time with the first photovoltaic parameter and later in time than the first photovoltaic parameter; the training photovoltaic power being an actual photovoltaic power in the time corresponding to the second photovoltaic parameter.

[0162] The first determination module 404 is configured to input the first photovoltaic training parameter into an initial prediction model to obtain a first predicted photovoltaic power; the first photovoltaic training parameter being one of the plurality of photovoltaic training parameters, and the first predicted photovoltaic power being a predicted photovoltaic power in the time corresponding to the second photovoltaic parameter.

[0163] The second determination module 406 is configured to determine a loss value based on the first predicted photovoltaic power and a first photovoltaic power; the first photovoltaic power being one of the plurality of training photovoltaic powers corresponding to the first photovoltaic training parameter.

[0164] The third determination module 408 is configured to optimize parameters of the initial prediction model based on the loss value to obtain a target prediction model.

[0165] In one exemplary embodiment, the initial prediction model comprises a time sequence module, a non-time sequence module, and a prediction module; the first determination module 404 is specifically configured to use the time sequence module to extract features of the first photovoltaic parameter in the first photovoltaic training parameter to obtain time sequence features; use the non-time sequence module to extract features of the second photovoltaic parameter in the first photovoltaic training parameter to obtain non-time sequence features; and use the prediction module to predict the photovoltaic power based on the time sequence features and the non-time sequence features to obtain the first predicted photovoltaic power.

[0166] In one exemplary embodiment, the time sequence module comprises a time sequence feature extraction layer; and the first determination module 404 is specifically configured to use the time sequence feature extraction layer to extract features of the first photovoltaic parameter in the first photovoltaic training parameter to obtain the time sequence features. In one exemplary embodiment, the non-time sequence module comprises a non-time sequence feature extraction layer; and the first determination module 404 is specifically configured to use the non-time sequence feature extraction layer to extract features of the second photovoltaic parameter in the first photovoltaic training parameter to obtain the non-time sequence features.

[0167] In an example embodiment, the time sequence module further comprises a decomposition layer; the first determining module 404 is specifically configured to perform variational modal decomposition on the first photovoltaic parameter using the decomposition layer to obtain a plurality of intrinsic modal functions; values of the plurality of intrinsic modal functions satisfy a preset value condition; and perform feature extraction on the plurality of intrinsic modal functions using a time sequence feature extraction layer to obtain the time sequence feature.

[0168] In an example embodiment, the time sequence module further comprises a dynamic window layer; the first determining module 404 is specifically configured to perform data screening on the first photovoltaic training parameter using the dynamic window layer to obtain partial training data, and the first photovoltaic parameter comprises the partial training data; and perform variational modal decomposition on the partial training data using the decomposition layer to obtain the plurality of intrinsic modal functions.

[0169] In an example embodiment, the obtaining module 402 is specifically configured to obtain initial training data; divide the initial training data according to a preset time window length to obtain a plurality of initial training data segments; determine a mean value and a standard deviation of each initial training data segment within a corresponding preset time window length; determine abnormal data in the initial training data based on each initial training data segment, the mean value corresponding to each initial training data segment, and the standard deviation corresponding to each initial training data segment; and remove the abnormal data from the initial training data to obtain the training data.

[0170] In an example embodiment, the prediction module comprises a feature fusion layer and a fully connected layer; the first determining module 404 is specifically configured to use the feature fusion layer to perform feature fusion on the time sequence feature and the non-time sequence feature to obtain a fusion feature; and use the fully connected layer to predict the photovoltaic power based on the fusion feature to obtain the first predicted photovoltaic power.

[0171] In an example embodiment, the first determining module 404 is specifically configured to determine a first weight based on the non-time sequence feature; determine a second weight based on the first weight; obtain the fusion feature based on the first weight, the time sequence feature, the second weight, and the non-time sequence feature; the first weight corresponds to the time sequence feature, and the second weight corresponds to the non-time sequence feature.

[0172] In an example embodiment, the third determining module 408 is further configured to obtain to-be-predicted data; and use the target prediction model to predict the photovoltaic power based on the to-be-predicted data to obtain a target photovoltaic power.

[0173] In an exemplary embodiment, the third determining module 408 is specifically used to acquire historical weather variables, historical photovoltaic module temperature, historical photovoltaic power, and predicted weather variables within a preset historical time period; determine historical theoretical irradiance based on historical weather variables; determine predicted theoretical irradiance based on historical weather variables and predicted weather variables; determine predicted photovoltaic module temperature based on historical weather variables, predicted weather variables, static heat loss coefficient, wind cooling coefficient, and the photovoltaic module's absorption efficiency of solar radiation energy; and determine historical weather variables, historical theoretical irradiance, historical photovoltaic module temperature, historical photovoltaic power, predicted weather variables, predicted theoretical irradiance, and predicted photovoltaic module temperature as data to be predicted.

[0174] Each module in the aforementioned photovoltaic power prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0175] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores training data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a photovoltaic power prediction method.

[0176] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6The computer device shown in the figure includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. 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 operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, mobile cellular network, near field communication (NFC) or other technologies. The computer program is executed by the processor to realize a photovoltaic power prediction method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0177] Those skilled in the art can understand that, Figure 6 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0178] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the steps in the above method embodiments.

[0179] In one exemplary embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to realize the steps in the above method embodiments.

[0180] In one exemplary embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to realize the steps in the above method embodiments.

[0181] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0182] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. In the embodiments provided in the present application, any reference to memory, database or other medium can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0183] Any technical features in the above embodiments can be combined, and for the sake of brevity, not all possible combinations are described above, however, any combination of these technical features is deemed to be within the scope of the present application.

[0184] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A photovoltaic power prediction method, characterized in that, The method includes: Acquire training data, which includes multiple photovoltaic training parameters and multiple training photovoltaic powers; the photovoltaic training parameters include a first photovoltaic parameter and a second photovoltaic parameter; the second photovoltaic parameter is a photovoltaic parameter that is temporally continuous with the first photovoltaic parameter but later in time; the training photovoltaic power is the actual photovoltaic power within the time corresponding to the second photovoltaic parameter. The first photovoltaic training parameter is input into the initial prediction model to obtain the first predicted photovoltaic power; the first photovoltaic training parameter is one of a plurality of photovoltaic training parameters, and the first predicted photovoltaic power is the photovoltaic power predicted within the time period corresponding to the second photovoltaic parameter; The loss value is determined based on the first predicted photovoltaic power and the first photovoltaic power; the first photovoltaic power is one of the multiple trained photovoltaic powers that corresponds to the first photovoltaic training parameter. The parameters of the initial prediction model are optimized based on the loss value to obtain the target prediction model.

2. The method according to claim 1, characterized in that, The initial prediction model includes a time-series module, a non-time-series module, and a prediction module; The step of inputting the first photovoltaic training parameters into the initial prediction model to obtain the first predicted photovoltaic power includes: The time-series module is used to extract features from the first photovoltaic parameters in the first photovoltaic training parameters to obtain time-series features; The non-time-series module is used to extract features from the second photovoltaic parameters in the first photovoltaic training parameters to obtain non-time-series features. Based on the time-series features and the non-time-series features, the prediction module is used to predict the photovoltaic power to obtain the first predicted photovoltaic power.

3. The method according to claim 2, characterized in that, The time-series module includes a time-series feature extraction layer; the time-series module is used to extract features from the first photovoltaic parameters in the first photovoltaic training parameters to obtain time-series features, including: The time-series feature extraction layer is used to extract features from the first photovoltaic parameters in the first photovoltaic training parameters to obtain time-series features.

4. The method according to claim 3, characterized in that, The timing module also includes a decomposition layer; The time-series feature extraction layer is used to extract features from the first photovoltaic parameters in the first photovoltaic training parameters to obtain time-series features, including: The first photovoltaic parameter is subjected to variational mode decomposition using the decomposition layer to obtain multiple intrinsic mode functions; the values ​​of the multiple intrinsic mode functions satisfy preset value conditions; The temporal feature extraction layer is used to extract features from the multiple intrinsic mode functions to obtain temporal features.

5. The method according to claim 4, characterized in that, The timing module also includes a dynamic window layer; The first photovoltaic parameters are subjected to variational mode decomposition using the decomposition layer to obtain multiple intrinsic mode functions, including: The dynamic window layer is used to filter the data of the first photovoltaic training parameters to obtain partial training data. The first photovoltaic training parameters include partial training data. The decomposition layer is used to perform variational mode decomposition on the portion of the training data to obtain multiple intrinsic mode functions.

6. The method according to claim 1, characterized in that, The acquisition of training data includes: Obtain initial training data; The initial training data is divided into multiple initial training data segments according to the preset time window length. Determine the mean and standard deviation of each initial training data segment within its respective preset time window length; Based on each of the initial training data segments, the mean of each of the initial training data segments, and the standard deviation of each of the initial training data segments, abnormal data in the initial training data are determined. The abnormal data is removed from the initial training data to obtain the training data.

7. The method according to claim 2, characterized in that, The prediction module includes a feature fusion layer and a fully connected layer; based on the time-series features and the non-time-series features, the prediction module is used to predict photovoltaic power to obtain a first predicted photovoltaic power, including: The feature fusion layer is used to fuse the temporal features and the non-temporal features to obtain fused features. Based on the fusion characteristics, the photovoltaic power is predicted using the fully connected layer to obtain the first predicted photovoltaic power.

8. The method according to claim 7, characterized in that, The temporal features and the non-temporal features are fused to obtain fused features, including: Based on the aforementioned non-temporal features, a first weight is determined; Based on the first weight, determine the second weight; The fused feature is obtained based on the first weight, the temporal feature, the second weight, and the non-temporal feature; the first weight corresponds to the temporal feature, and the second weight corresponds to the non-temporal feature.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: Obtain the data to be predicted; Based on the data to be predicted, the photovoltaic power is predicted using the target prediction model to obtain the target photovoltaic power.

10. The method according to claim 9, characterized in that, The acquisition of the data to be predicted includes: Acquire historical weather variables, historical photovoltaic module temperature, historical photovoltaic power within a preset historical time period, as well as predicted weather variables within a preset future time period; Historical theoretical irradiance is determined based on the aforementioned historical weather variables; The theoretical irradiance is determined based on the predicted weather variables; Based on the historical weather variables, the predicted weather variables, the static heat loss coefficient, the wind cooling coefficient, and the absorption efficiency of the photovoltaic module to solar radiation energy, the predicted temperature of the photovoltaic module is determined. The historical weather variables, the historical theoretical irradiance, the historical photovoltaic module temperature, the historical photovoltaic power, the predicted weather variables, the predicted theoretical irradiance, and the predicted photovoltaic module temperature are determined as the data to be predicted.

Images