Photovoltaic power prediction method, model training method, device, equipment and medium

Abstract

The present disclosure relates to a photovoltaic power prediction method, model training method, device, equipment and medium. Real-time weather forecast data of a target photovoltaic power station is obtained; the photovoltaic power prediction model of the target photovoltaic power station is used to process the real-time weather forecast data to obtain the target predicted photovoltaic power of the target photovoltaic power station; wherein the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, and the data volume of the first training data is much larger than the data volume of the second training data. In this way, the model suitable for the target photovoltaic power station is used to predict the photovoltaic power generation power, thereby improving the prediction accuracy.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a photovoltaic power prediction method, a model training method, a device, a equipment and a medium. Background Art

[0002] With the rapid development of photovoltaic power generation technology, many regions have built or renovated many photovoltaic power stations and used them for photovoltaic power generation. However, due to the short operation time of the newly built or renovated photovoltaic power stations, it is impossible to accumulate sufficient historical operation data for users to conduct photovoltaic power prediction and analysis.

[0003] Therefore, it is necessary to provide a photovoltaic power prediction method to accurately predict the power generation of a newly built or rebuilt photovoltaic power station. Summary of the invention

[0004] In order to solve the above technical problems, the present disclosure provides a photovoltaic power prediction method, a model training method, an apparatus, a device and a medium.

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

[0006] Obtain real-time weather forecast data for the target photovoltaic power station;

[0007] Using the photovoltaic power prediction model of the target photovoltaic power station, the real-time weather forecast data is processed to obtain the target predicted photovoltaic power of the target photovoltaic power station;

[0008] Among them, the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0009] In a second aspect, the present disclosure provides a method for training a photovoltaic power prediction model, the method comprising:

[0010] Acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station;

[0011] Pre-training the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model;

[0012] Based on the first training data, update the parameters of the adaptive learner in the first model to obtain a second model after the parameters are updated;

[0013] Fine-tune the parameters of the adaptive learner in the second model based on the second training data to determine a photovoltaic power prediction model for the target photovoltaic power station;

[0014] The data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0015] In a third aspect, the present disclosure provides a photovoltaic power prediction device, the device comprising:

[0016] The first acquisition module is used to obtain real-time weather forecast data of the target photovoltaic power station;

[0017] A photovoltaic power prediction module, used to process the real-time weather forecast data using the photovoltaic power prediction model of the target photovoltaic power station to obtain a target predicted photovoltaic power of the target photovoltaic power station;

[0018] Among them, the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0019] In a fourth aspect, the present disclosure provides a training device for a photovoltaic power prediction model, the device comprising:

[0020] A second acquisition module is used to acquire first training data of the source photovoltaic power station and second training data of the target photovoltaic power station;

[0021] A pre-training module, used to pre-train the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model;

[0022] A parameter updating module, used for updating the parameters of the adaptive learner in the first model based on the first training data, and obtaining a second model after the parameters are updated;

[0023] A parameter fine-tuning module, configured to fine-tune the parameters of the adaptive learner in the second model based on the second training data, and determine a photovoltaic power prediction model for the target photovoltaic power station;

[0024] The data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0025] In a fifth aspect, an embodiment of the present disclosure further provides an electronic device, the device comprising:

[0026] one or more processors;

[0027] a storage device for storing one or more programs,

[0028] When one or more programs are executed by one or more processors, the one or more processors implement the method provided by the first aspect or the second aspect.

[0029] In a sixth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided in the first aspect or the second aspect when executed by a processor.

[0030] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0031] A photovoltaic power prediction method, model training method, device, equipment and medium of the embodiments of the present disclosure obtain real-time weather forecast data of a target photovoltaic power station; use the photovoltaic power prediction model of the target photovoltaic power station to process the real-time weather forecast data to obtain the target predicted photovoltaic power of the target photovoltaic power station; wherein the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold. Therefore, when it is necessary to predict the photovoltaic power of a target photovoltaic power station with a short operating time, such as a newly built or expanded one, the parameters of the adaptive learner and the transfer learner in the initial model are roughly adjusted using the first training data with a larger data volume, and then the adaptive learner is fine-tuned using the second training data with a smaller data volume, so that the photovoltaic power prediction model can be adapted to predict the photovoltaic power of a newly built or renovated target photovoltaic power station, thereby ensuring accurate prediction of the generated power of the target photovoltaic power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0034] Figure 1 A schematic diagram of a photovoltaic power prediction method provided by an embodiment of the present disclosure;

[0035] Figure 2 A schematic diagram of a flow chart of a photovoltaic power prediction model training method provided in an embodiment of the present disclosure;

[0036] Figure 3 A schematic diagram of the structure of a photovoltaic power prediction device provided by an embodiment of the present disclosure;

[0037] Figure 4 A schematic diagram of the structure of a photovoltaic power prediction model training device provided in an embodiment of the present disclosure;

[0038] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0041] At present, when it is necessary to predict the photovoltaic power of a target photovoltaic power station that has been in operation for a short time, such as a newly built or rebuilt one, the photovoltaic power prediction model is first trained using the historical data of the source photovoltaic power station that has been in operation for a long enough time, and then the photovoltaic power prediction model is migrated to the scene of the target photovoltaic power station to achieve the photovoltaic power prediction of the target photovoltaic power station that has been in operation for a short time using the photovoltaic power prediction model. However, due to the differences between different photovoltaic power stations, it is difficult to accurately predict the power generation of the target photovoltaic power station through model migration.

[0042] In order to improve the photovoltaic power prediction accuracy of a target photovoltaic power station that has been newly built or expanded and has a short operating time, the embodiments of the present disclosure provide a photovoltaic power prediction method, model training method, device, equipment and medium.

[0043] Combine the following Figure 1 The photovoltaic power prediction method provided in the embodiment of the present disclosure is described. In the embodiment of the present disclosure, the photovoltaic power prediction method can be executed by an electronic device. The electronic device can include a device with communication functions such as a tablet computer, a desktop computer, a laptop computer, etc., and can also include a device simulated by a virtual machine or a simulator.

[0044] Figure 1 A schematic flow chart of a photovoltaic power prediction method provided by an embodiment of the present disclosure is shown.

[0045] like Figure 1 As shown, the photovoltaic power prediction method may include the following steps.

[0046] S110. Obtain real-time weather forecast data for the target photovoltaic power station.

[0047] In this embodiment, when it is necessary to predict the power generation of a target photovoltaic power station that has been in operation for a short time, numerical weather forecast data for a period of time in the future after the current time is collected as real-time weather forecast data, and the power generation of the target photovoltaic power station is further predicted based on the real-time weather forecast data.

[0048] Among them, the target photovoltaic power station refers to a newly built or rebuilt photovoltaic power station that has been in operation for a short time, so that the target photovoltaic power station has only a small amount of training data for model training.

[0049] Among them, real-time weather forecast data refers to weather forecast data that can reflect the state of atmospheric movement in the future. Specifically, according to the actual atmospheric conditions, under certain initial and boundary conditions, numerical calculations are performed to solve the fluid mechanics and thermodynamics equations describing the weather evolution process, and weather forecast data that reflects the state of atmospheric movement in the future is obtained, and the weather forecast data is used as real-time weather forecast data.

[0050] Optionally, the real-time weather forecast data may include M types of meteorological characteristics at time 1 to time N, specifically including but not limited to solar irradiance, surface temperature, surface pressure, relative humidity, etc.

[0051] S120. Use the photovoltaic power prediction model of the target photovoltaic power station to process the real-time weather forecast data to obtain the target predicted photovoltaic power of the target photovoltaic power station; wherein the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0052] In this embodiment, before the real-time weather forecast data is input into the photovoltaic power prediction model, the data of N moments in the real-time weather forecast data are arranged in chronological order to obtain a one-dimensional sequence of numerical weather forecast, and then the data of M meteorological characteristics are aligned at the same time point to obtain two-dimensional numerical weather forecast data.

[0053] Furthermore, the two-dimensional numerical weather forecast data is input into a photovoltaic power prediction model, so that the two-dimensional numerical weather forecast data is processed based on the photovoltaic power prediction model to obtain a target predicted photovoltaic power of a target photovoltaic power station.

[0054] The photovoltaic power prediction model refers to a prediction model that is suitable for the target photovoltaic power station.

[0055] The first time threshold is a maximum value, which can be determined based on experience. The first quantity threshold is also a maximum value, which can be determined based on experience. It can be understood that since the first time that the source photovoltaic power station has been in operation is greater than the first time threshold, the source photovoltaic power station has sufficient first training data for model training, that is, the amount of the first training data is greater than the first quantity threshold.

[0056] The second time threshold is a minimum value, which can be determined based on experience. The second quantity threshold is also a minimum value, which can be determined based on experience. It can be understood that since the second time length of the target photovoltaic power station that has been in operation is less than the second time threshold, the model training can only be performed from a small amount of second training data of the target photovoltaic power station, that is, the amount of the second training data is less than the second quantity threshold.

[0057] The adaptive learner in the initial model can be understood as a numerical weather forecast embedding layer. It should be noted that the parameters of the adaptive learners corresponding to different photovoltaic power stations are quite different. Therefore, in the process of training the initial model, it is necessary to pre-train, update and fine-tune the parameters of the adaptive learner so that the trained photovoltaic power prediction model can be well adapted to the power generation prediction of the target photovoltaic power station.

[0058] The transfer learner in the initial model may include an encoder and an output layer. It should be noted that the parameters of the transfer learners corresponding to different photovoltaic power stations are slightly different. Therefore, in the process of training the initial model, after pre-training the transfer learner using the first training data of the source photovoltaic power station, the transfer learner can be used to process the real-time weather forecast data of the target photovoltaic power station to adapt to the photovoltaic power prediction of the target photovoltaic power station.

[0059] Specifically, when performing model training, since the amount of data of the first training data of the source photovoltaic power station is large enough, the initial model is first pre-trained using the first training data with a sufficiently large amount of data to obtain the pre-trained first model, and then the first training data is used to continue to update the parameters of the adaptive learner in the first model, so that the adaptive learner uses the first training data of the source photovoltaic power station to adjust the parameters twice to obtain an updated second model, and finally, the adaptive learner in the second model is continued to be fine-tuned using a smaller amount of second training data, so that the adaptive learner uses the second training data of the target photovoltaic power station to adjust again, thereby obtaining a photovoltaic power prediction model suitable for predicting the power generation power of the target wind farm.

[0060] In this embodiment, S120 includes but is not limited to the following methods: based on the adaptive learner in the photovoltaic power prediction model, the real-time weather forecast data is processed with meteorological features to obtain a target meteorological feature vector; based on the transfer learner in the photovoltaic power prediction model, the target meteorological feature vector is predicted with photovoltaic power to obtain a target predicted photovoltaic power.

[0061] Among them, the adaptive learner is a structure composed of a single-layer fully connected neural network, and the adaptive learner is used to convert the real-time weather forecast data into a vector form to obtain the target meteorological feature vector.

[0062] Among them, based on the transfer learner in the photovoltaic power prediction model, the target meteorological feature vector is used to predict the photovoltaic power to obtain the target predicted photovoltaic power, which specifically includes: according to the encoder in the transfer learner, the target meteorological feature vector is meteorologically encoded to determine the meteorological encoding vector; according to the output layer in the transfer learner, the meteorological encoding vector is classified and processed to determine the target predicted photovoltaic power.

[0063] Optionally, the encoding layer in the transfer learner includes but is not limited to one or more of the following combinations: an encoder of a transformer model (for example, including a multi-head attention mechanism, a feedforward neural network residual connection and normalization, position encoding, etc.), a long short-term memory network (for example, a forget gate, an input gate, an update gate, and an output gate), and a gated recurrent neural network (for example, a reset gate and an update gate).

[0064] Optionally, the output layer in the transfer learner may include but is not limited to a single-layer fully connected neural network, so that the output layer outputs a photovoltaic power prediction sequence for the next N moments, and uses the photovoltaic power prediction sequence for the next N moments as the target predicted photovoltaic power.

[0065] A photovoltaic power prediction method according to an embodiment of the present disclosure obtains real-time weather forecast data of a target photovoltaic power station; uses a photovoltaic power prediction model of the target photovoltaic power station to process the real-time weather forecast data to obtain a target predicted photovoltaic power of the target photovoltaic power station; wherein the photovoltaic power prediction model uses first training data of a source photovoltaic power station to pre-train an adaptive learner and a transfer learner in an initial model respectively, and uses the first training data to update parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold. Therefore, when it is necessary to predict the photovoltaic power of a target photovoltaic power station with a short operating time, such as a newly built or expanded one, the parameters of the adaptive learner and the transfer learner in the initial model are roughly adjusted using the first training data with a larger data volume, and then the adaptive learner is fine-tuned using the second training data with a smaller data volume, so that the photovoltaic power prediction model can be adapted to predict the photovoltaic power of a newly built or renovated target photovoltaic power station, thereby ensuring accurate prediction of the generated power of the target photovoltaic power station.

[0066] Combine the following Figure 2 The training method of the photovoltaic power prediction model provided by the embodiment of the present disclosure is described. In the embodiment of the present disclosure, the training method of the photovoltaic power prediction model can be executed by an electronic device or a server.

[0067] Figure 2 A schematic flow chart of a photovoltaic power prediction model training method provided by an embodiment of the present disclosure is shown.

[0068] like Figure 2 As shown, the training method of the photovoltaic power prediction model may include the following steps.

[0069] S210: Acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station.

[0070] Among them, the first time that the source photovoltaic power station has been in operation is greater than the first time threshold, the second time that the target photovoltaic power station has been in operation is less than the second time threshold, the first time threshold is greater than the second time threshold, the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0071] In this embodiment, there may be multiple source photovoltaic power stations, and the first training data includes first measured weather forecast data and first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively.

[0072] Optionally, the first measured weather forecast data may be historical numerical weather forecast data of a source photovoltaic power station, and the first measured photovoltaic power may be used as label data for model training.

[0073] Optionally, the second measured weather forecast data may be historical numerical weather forecast data of the target photovoltaic power station, and the second measured photovoltaic power may also be used as label data for model training.

[0074] S220. Pre-train the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model.

[0075] In this embodiment, the specific implementation of S220 includes but is not limited to the following:

[0076] S2201, based on the adaptive learner in the initial model, performing meteorological feature processing on the first measured weather forecast data to obtain a first meteorological feature vector;

[0077] S2202: Based on the transfer learner in the initial model, predict the photovoltaic power of the first meteorological feature vector to obtain a first predicted photovoltaic power of each source photovoltaic power station;

[0078] S2203, determining a first average prediction error of the initial model according to a first predicted photovoltaic power of each source photovoltaic power station and a first measured photovoltaic power of each source photovoltaic power station;

[0079] S2204. Pre-train the adaptive learner and the transfer learner in the initial model according to the first average prediction error to obtain a first model.

[0080] Among them, the adaptive learner in S2201 is a structure composed of a single-layer fully connected neural network, and the adaptive learner is used to convert the first measured weather forecast data into a vector form to obtain a first meteorological feature vector.

[0081] Among them, S2202 specifically includes: according to the encoder in the transfer learner, meteorologically encoding the first meteorological feature vector to determine the meteorological encoding vector; according to the output layer in the transfer learner, classifying the meteorological encoding vector to determine the first predicted photovoltaic power.

[0082] Among them, the specific implementation process of S2203 may include: initializing the parameters of the initial model; calculating the photovoltaic prediction error of each source photovoltaic power station based on the first predicted photovoltaic power and the first measured photovoltaic power of each source photovoltaic power station; calculating the average value of the photovoltaic prediction errors of multiple source photovoltaic power stations to obtain the first average prediction error of the initial model.

[0083] Optionally, the photovoltaic prediction error of each source photovoltaic power station may be determined based on all or part of the first predicted photovoltaic power and all or part of the first measured photovoltaic power of each source photovoltaic power station.

[0084] Optionally, the photovoltaic prediction error of each source photovoltaic power station can be calculated by using a mean square error, a mean absolute error, or the like.

[0085] Among them, the specific implementation process of S2204 may include: using the first average prediction error to update all layers of the initial model E times, specifically in the process of updating the initial model, using verification data to verify the accuracy, so that when the accuracy decreases p times continuously, the iterative process is terminated, and the optimal parameters are used as the final parameters of the initial model, thereby obtaining a pre-trained first model. Optionally, 1≤E≤1000, p is an iteration number threshold set according to experience. The parameter update algorithm can adopt but is not limited to Stochastic Gradient Descent (SGD), Adaptive Gradient (AdaGrad), Adaptive Motion Estimation (Adam) and other algorithms.

[0086] In the above manner, the adaptive learner and the transfer learner of the initial model are pre-trained using the first training data of the source photovoltaic power station with sufficient data volume, so as to adjust the model parameters of each layer of the initial model to the greatest extent.

[0087] S230: Update the parameters of the adaptive learner in the first model based on the first training data to obtain a second model after the parameters are updated.

[0088] In this embodiment, the specific implementation of S230 includes but is not limited to the following:

[0089] S2301, based on the adaptive learner in the first model, performing meteorological feature processing on the first measured weather forecast data to obtain a second meteorological feature vector;

[0090] S2302: Based on the transfer learner in the first model, predict the photovoltaic power of the second meteorological feature vector to obtain a second predicted photovoltaic power of each source photovoltaic power station;

[0091] S2303: determining a second average prediction error of the first model according to the second predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station;

[0092] S2304, determining an average parameter gradient of the first model according to the second average prediction error;

[0093] S2305. Based on the average parameter gradient, update the parameters of the adaptive learner in the first model to obtain the second model.

[0094] In S2301, the adaptive learner in the first model continues to convert the first measured weather forecast data into a vector form to obtain a second meteorological feature vector.

[0095] Among them, S2302 specifically includes: meteorologically encoding the second meteorological feature vector according to the encoder of the transfer learner in the first model to determine the meteorological coding vector; classifying the meteorological coding vector according to the output layer of the transfer learner in the first model to determine the second predicted photovoltaic power.

[0096] Among them, the specific implementation process of S2303 may include: initializing the parameters of the first model by using zero initialization, random initialization, etc.; calculating the photovoltaic prediction error of each source photovoltaic power station based on the first predicted photovoltaic power and the first measured photovoltaic power of each source photovoltaic power station; calculating the average value of the photovoltaic prediction errors of multiple source photovoltaic power stations to obtain the second average prediction error of the first model.

[0097] Optionally, the photovoltaic prediction error of each source photovoltaic power station may be determined based on all or part of the first predicted photovoltaic power and all or part of the first measured photovoltaic power of each source photovoltaic power station.

[0098] Optionally, the photovoltaic prediction error of each source photovoltaic power station can be calculated by using a mean square error, a mean absolute error, or the like.

[0099] Among them, the specific implementation process of S2304 may include: updating the parameters of the adaptive learner in the first model according to the average parameter gradient, and determining the transfer learner after the parameter update in the first model; calculating the parameter update gradient of each source photovoltaic power station according to the adaptive learner after the parameter update and the adaptive learner before the update; calculating the average value of the parameter update gradients of multiple source photovoltaic power stations to determine the average parameter gradient of the first model.

[0100] Among them, the specific implementation process of S2305 may include: using the average parameter gradient and based on the first-order meta-learning algorithm (Reptile algorithm) to update the adaptive learner in the first model F times, specifically in the process of updating the adaptive learner in the first model, using the verification data to verify the accuracy, so that when the accuracy drops p times continuously, the iterative process is terminated, and the optimal parameters are used as the final parameters of the adaptive learner in the first model, thereby obtaining the second model after the parameter update. Optionally, 1≤F≤1000, p is the threshold of the number of iterations set according to experience.

[0101] In the above manner, the first training data of the source photovoltaic power station with sufficient data volume is used to update the parameters of the adaptive learner in the pre-trained first model, thereby further optimizing the model parameters of the adaptive learner in the first model, so that the model can migrate more valuable features from the source photovoltaic power station.

[0102] S240: fine-tune parameters of the adaptive learner in the second model based on the second training data to determine a photovoltaic power prediction model for the target photovoltaic power station.

[0103] In this embodiment, the specific implementation of S240 includes but is not limited to the following:

[0104] S2401, based on the adaptive learner in the second model, performing meteorological feature processing on the second measured weather forecast data to obtain a measured meteorological feature vector;

[0105] S2402, based on the transfer learner in the second model, predicting the photovoltaic power of the measured meteorological feature vector to obtain a third predicted photovoltaic power of the target photovoltaic power station;

[0106] S2403, determining a third average prediction error of the second model according to the third predicted photovoltaic power and the second measured photovoltaic power;

[0107] S2404: fine-tune the parameters of the adaptive learner in the second model according to the third average prediction error to obtain a photovoltaic power prediction model.

[0108] Among them, the adaptive learner in S2401 is a structure composed of a single-layer fully connected neural network, and the adaptive learner is used to convert the second measured weather forecast data into a vector form to obtain a measured meteorological feature vector.

[0109] Among them, the specific implementation process of S2402 may include: meteorologically encoding the measured meteorological feature vector according to the encoder in the transfer learner to determine the meteorological coding vector; classifying the meteorological coding vector according to the output layer in the transfer learner to determine the third predicted photovoltaic power.

[0110] Among them, the specific implementation process of S2403 may include: initializing the parameters of the second model; calculating multiple photovoltaic prediction errors of the target photovoltaic power station based on multiple third predicted photovoltaic powers and multiple second measured photovoltaic powers of the target photovoltaic power station; calculating the average value of multiple photovoltaic prediction errors to obtain the third average prediction error of the second model.

[0111] Optionally, a plurality of photovoltaic prediction errors of the target photovoltaic power station may be determined based on all or part of the third predicted photovoltaic power of the target photovoltaic power station and all or part of the second measured photovoltaic power.

[0112] Optionally, multiple photovoltaic prediction errors of the target photovoltaic power station are calculated using mean square error, mean absolute error, etc.

[0113] Among them, the specific implementation process of S2404 may include: using the third average prediction error and adopting the Reptile algorithm and other methods to update the adaptive learner in the second model G times, specifically, in the process of updating the adaptive learner in the second model, using the verification data to verify the accuracy, so that when the accuracy drops p times continuously, the iterative process is terminated, and the optimal parameters are used as the final parameters of the adaptive learner in the second model, thereby obtaining the second model with fine-tuning parameters. Optionally, 1≤G≤1000, p is the threshold of the number of iterations set according to experience.

[0114] Through the above method, after the first training data of the source photovoltaic power station with sufficient data volume is used to update the parameters of the adaptive learner in the first model after the parameters are updated, the second training data of the target photovoltaic power station itself is continued to be used to continue to fine-tune the parameters of the adaptive learner in the second model, thereby improving the adaptability of the photovoltaic power prediction model on the target photovoltaic power station.

[0115] A training method for a photovoltaic power prediction model of an embodiment of the present disclosure obtains first training data of a source photovoltaic power station and second training data of a target photovoltaic power station; uses the first training data to pre-train an adaptive learner and a transfer learner in an initial model respectively, and obtains a pre-trained first model; updates the parameters of the adaptive learner in the first model based on the first training data, and obtains a second model after parameter update; fine-tunes the parameters of the adaptive learner in the second model after parameter update based on the second training data, and determines the photovoltaic power prediction model of the target photovoltaic power station; wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold. In the above manner, the parameters of the initial model are fully adjusted using the first training data of the source photovoltaic power station with sufficient data volume, so that the initial model can migrate more valuable features from the source photovoltaic power station, and then the second training data of the target photovoltaic power station itself is used to continue to fine-tune the parameters of the adaptive learner in the second model, so as to improve the adaptability of the photovoltaic power prediction model on the target photovoltaic power station, and finally improve the power prediction accuracy of the target photovoltaic power station using the photovoltaic power prediction model.

[0116] The present disclosure also provides a photovoltaic power prediction device for implementing the photovoltaic power prediction method. Figure 3In the embodiment of the present disclosure, the photovoltaic power prediction device may be an electronic device. The electronic device may include a tablet computer, a desktop computer, a laptop computer, or other device with a communication function, or may include a device simulated by a virtual machine or a simulator.

[0117] Figure 3 A schematic structural diagram of a photovoltaic power prediction device provided by an embodiment of the present disclosure is shown.

[0118] like Figure 3 As shown, the photovoltaic power prediction device 300 may include:

[0119] A first acquisition module 310 is used to acquire real-time weather forecast data of a target photovoltaic power station;

[0120] The photovoltaic power prediction module 320 is used to process the real-time weather forecast data using the photovoltaic power prediction model of the target photovoltaic power station to obtain the target predicted photovoltaic power of the target photovoltaic power station;

[0121] Among them, the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0122] A photovoltaic power prediction device according to an embodiment of the present disclosure obtains real-time weather forecast data of a target photovoltaic power station; uses a photovoltaic power prediction model of the target photovoltaic power station to process the real-time weather forecast data to obtain a target predicted photovoltaic power of the target photovoltaic power station; wherein the photovoltaic power prediction model uses first training data of a source photovoltaic power station to pre-train an adaptive learner and a transfer learner in an initial model respectively, and uses the first training data to update parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune parameters of the adaptive learner in the second model obtained by parameter update, wherein the data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold. Therefore, when it is necessary to predict the photovoltaic power of a target photovoltaic power station with a short operating time, such as a newly built or expanded one, the parameters of the adaptive learner and the transfer learner in the initial model are roughly adjusted using the first training data with a larger data volume, and then the adaptive learner is fine-tuned using the second training data with a smaller data volume, so that the photovoltaic power prediction model can be adapted to predict the photovoltaic power of a newly built or renovated target photovoltaic power station, thereby ensuring accurate prediction of the generated power of the target photovoltaic power station.

[0123] In some embodiments of the present disclosure, the photovoltaic power prediction module 320 includes:

[0124] A meteorological feature processing unit, configured to perform meteorological feature processing on the real-time weather forecast data based on an adaptive learner in the photovoltaic power prediction model to obtain a target meteorological feature vector;

[0125] A photovoltaic power prediction unit is used to perform photovoltaic power prediction on the target meteorological feature vector based on a transfer learner in the photovoltaic power prediction model to obtain the target predicted photovoltaic power.

[0126] In some embodiments of the present disclosure, the photovoltaic power prediction unit is specifically used to:

[0127] According to the encoder in the transfer learner, meteorological encoding is performed on the target meteorological feature vector to determine a meteorological encoding vector;

[0128] According to the output layer in the transfer learner, the meteorological coding vector is classified and processed to determine the target predicted photovoltaic power.

[0129] It should be noted that Figure 3 The photovoltaic power prediction device 400 shown can perform Figure 1 The various steps in the method embodiment shown in the figure are implemented Figure 1The various processes and effects in the method embodiment shown are not described in detail here.

[0130] The embodiment of the present disclosure also provides a photovoltaic power prediction model training device for implementing the photovoltaic power prediction model training method described above. Figure 4 In the embodiment of the present disclosure, the training device of the photovoltaic power prediction model may be an electronic device or a server.

[0131] Figure 4 A schematic structural diagram of a photovoltaic power prediction model training device provided by an embodiment of the present disclosure is shown.

[0132] like Figure 4 As shown, the photovoltaic power prediction model training device 400 may include:

[0133] A second acquisition module 410 is used to acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station;

[0134] A pre-training module 420 is used to pre-train the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model;

[0135] A parameter updating module 430, configured to update the parameters of the adaptive learner in the first model based on the first training data, and obtain a second model after the parameters are updated;

[0136] A parameter fine-tuning module 440 is used to fine-tune the parameters of the adaptive learner in the second model based on the second training data to determine the photovoltaic power prediction model of the target photovoltaic power station;

[0137] The data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0138] The training device of the photovoltaic power prediction model of the embodiment of the present disclosure obtains the first training data of the source photovoltaic power station and the second training data of the target photovoltaic power station; uses the first training data to pre-train the adaptive learner and the transfer learner in the initial model respectively, and obtains the first model after pre-training; based on the first training data, the adaptive learner in the first model is updated with parameters, and the second model after the parameters are updated is obtained; based on the second training data, the adaptive learner in the second model after the parameters are updated is fine-tuned to determine the photovoltaic power prediction model of the target photovoltaic power station; wherein, the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold. In the above manner, the parameters of the initial model are fully adjusted using the first training data of the source photovoltaic power station with sufficient data volume, so that the initial model can migrate more valuable features from the source photovoltaic power station, and then the second training data of the target photovoltaic power station itself is used to continue to fine-tune the parameters of the adaptive learner in the second model, so as to improve the adaptability of the photovoltaic power prediction model on the target photovoltaic power station, and finally improve the power prediction accuracy of the target photovoltaic power station using the photovoltaic power prediction model.

[0139] In some embodiments of the present disclosure, there are multiple source photovoltaic power stations, and the first training data includes first measured weather forecast data and first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively; the second training data includes second measured weather forecast data and second measured photovoltaic power of the target photovoltaic power station.

[0140] In some embodiments of the present disclosure, the pre-training module 420 is specifically used to:

[0141] Based on the adaptive learner in the initial model, the first measured weather forecast data is subjected to meteorological feature processing to obtain a first meteorological feature vector;

[0142] Based on the transfer learner in the initial model, predicting the photovoltaic power of the first meteorological feature vector to obtain a first predicted photovoltaic power of each source photovoltaic power station;

[0143] Determining a first average prediction error of the initial model according to the first predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station;

[0144] The adaptive learner and the transfer learner in the initial model are pre-trained according to the first average prediction error to obtain the first model.

[0145] In some embodiments of the present disclosure, the parameter updating module 430 is specifically used to:

[0146] Based on the adaptive learner in the first model, the first measured weather forecast data is subjected to meteorological feature processing to obtain a second meteorological feature vector;

[0147] Based on the transfer learner in the first model, predicting the photovoltaic power of the second meteorological feature vector to obtain a second predicted photovoltaic power of each source photovoltaic power station;

[0148] determining a second average prediction error of the first model according to the second predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station;

[0149] determining an average parameter gradient of the first model according to the second average prediction error;

[0150] Based on the average parameter gradient, update the parameters of the adaptive learner in the first model to obtain the second model.

[0151] In some embodiments of the present disclosure, the parameter fine-tuning module 440 is specifically used to:

[0152] Based on the adaptive learner in the second model, the second measured weather forecast data is subjected to meteorological feature processing to obtain a measured meteorological feature vector;

[0153] Based on the transfer learner in the second model, the measured meteorological feature vector is used to predict photovoltaic power to obtain a third predicted photovoltaic power of the target photovoltaic power station;

[0154] determining a third average prediction error of the second model according to the third predicted photovoltaic power and the second measured photovoltaic power;

[0155] Fine-tune the parameters of the adaptive learner in the second model according to the third average prediction error to obtain the photovoltaic power prediction model.

[0156] It should be noted that Figure 4 The photovoltaic power prediction model training device 400 shown can be executed Figure 2 The various steps in the method embodiment shown in the figure are implemented Figure 2 The various processes and effects in the method embodiment shown are not described in detail here.

[0157] Figure 5 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown.

[0158] like Figure 5 As shown, the electronic device may include a processor 501 and a memory 502 storing computer program instructions.

[0159] Specifically, the processor 501 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0160] The memory 502 may include a large capacity memory for information or instructions. By way of example and not limitation, the memory 502 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 502 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 502 may be inside or outside the integrated gateway device. In a particular embodiment, the memory 502 is a non-volatile solid-state memory. In a particular embodiment, the memory 502 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (Electrically Erasable Programmable ROM, EEPROM), an electrically rewritable ROM (EAROM) or a flash memory, or a combination of two or more of these.

[0161] The processor 501 reads and executes the computer program instructions stored in the memory 502 to perform the steps of the photovoltaic power prediction method provided in the embodiment of the present disclosure.

[0162] In one example, the electronic device may further include a transceiver 503 and a bus 504. Figure 5 As shown, the processor 501, the memory 502 and the transceiver 503 are connected via a bus 504 and communicate with each other.

[0163] The bus 504 includes hardware, software, or both. For example, but not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 504 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0164] The following is an embodiment of a computer-readable storage medium provided in an embodiment of the present disclosure. The computer-readable storage medium and the photovoltaic power prediction method of the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the computer-readable storage medium, reference can be made to the embodiment of the above-mentioned photovoltaic power prediction method.

[0165] This embodiment provides a storage medium containing computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to perform a photovoltaic power prediction method. The method includes:

[0166] Obtain real-time weather forecast data for the target photovoltaic power station;

[0167] Using the photovoltaic power prediction model of the target photovoltaic power station, the real-time weather forecast data is processed to obtain the target predicted photovoltaic power of the target photovoltaic power station;

[0168] Among them, the photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, wherein the first time that the source photovoltaic power station has been in operation is greater than the first time threshold, the second time that the target photovoltaic power station has been in operation is less than the second time threshold, the first time threshold is greater than the second time threshold, the data amount of the first training data is greater than the first data amount threshold, the data amount of the second training data is less than the second data amount threshold, and the first data amount threshold is greater than the second data amount threshold.

[0169] The following is an embodiment of a computer-readable storage medium provided in an embodiment of the present disclosure. The computer-readable storage medium and the training method of the photovoltaic power prediction model in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the computer-readable storage medium, reference can be made to the embodiment of the training method of the photovoltaic power prediction model in the above-mentioned embodiments.

[0170] This embodiment provides a storage medium containing computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to perform a training method for a photovoltaic power prediction model. The method includes:

[0171] Acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station;

[0172] Pre-training the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model;

[0173] Based on the first training data, update the parameters of the adaptive learner in the first model to obtain a second model after the parameters are updated;

[0174] Fine-tune the parameters of the adaptive learner in the second model after the parameter update based on the second training data to determine the photovoltaic power prediction model of the target photovoltaic power station;

[0175] Among them, the first time that the source photovoltaic power station has been in operation is greater than the first time threshold, the second time that the target photovoltaic power station has been in operation is less than the second time threshold, the first time threshold is greater than the second time threshold, the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, and the first data volume threshold is greater than the second data volume threshold.

[0176] Of course, the storage medium containing computer executable instructions provided in the embodiment of the present disclosure is not limited to the above method operations, and can also execute related operations in the photovoltaic power prediction method or photovoltaic power prediction model training method provided in any embodiment of the present disclosure.

[0177] Through the above description of the implementation methods, the technical personnel in the relevant field can clearly understand that the present disclosure can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present disclosure can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer cloud platform (which can be a personal computer, server, or network cloud platform, etc.) to execute the photovoltaic power prediction method or photovoltaic power prediction model training method provided in each embodiment of the present disclosure.

[0178] Note that the above are only preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art will understand that the present disclosure is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present disclosure. Therefore, although the present disclosure is described in more detail through the above embodiments, the present disclosure is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present disclosure, and the scope of the present disclosure is determined by the scope of the appended claims.

Claims

1. A photovoltaic power prediction method, characterized in that: include: Obtain real-time weather forecast data for the target photovoltaic power station; Using the photovoltaic power prediction model of the target photovoltaic power station, the real-time weather forecast data is processed to obtain the target predicted photovoltaic power of the target photovoltaic power station; The photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, the first data volume threshold is greater than the second data volume threshold, the number of the source photovoltaic power stations is multiple, the first training data includes the first measured weather forecast data and the first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively; the second training data includes the second measured weather forecast data and the second measured photovoltaic power of the target photovoltaic power station; The photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, including: based on the adaptive learner in the initial model, the first measured weather forecast data is processed with meteorological features to obtain a first meteorological feature vector; based on the transfer learner in the initial model, the photovoltaic power is predicted for the first meteorological feature vector to obtain the first predicted photovoltaic power of each source photovoltaic power station; according to the first predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station, the first average prediction error of the initial model is determined; according to the first average prediction error, the adaptive learner and the transfer learner in the initial model are pre-trained to obtain the first model.

2. The method according to claim 1, wherein the photovoltaic power prediction model of the target photovoltaic power station is used to process the real-time weather forecast data to obtain the target predicted photovoltaic power of the target photovoltaic power station, comprising: Based on the adaptive learner in the photovoltaic power prediction model, the real-time weather forecast data is processed with meteorological features to obtain a target meteorological feature vector; Based on the transfer learner in the photovoltaic power prediction model, photovoltaic power prediction is performed on the target meteorological feature vector to obtain the target predicted photovoltaic power.

3. The method according to claim 2, wherein the photovoltaic power prediction is performed on the target meteorological feature vector based on the transfer learner in the photovoltaic power prediction model to obtain the target predicted photovoltaic power, comprising: According to the encoder in the transfer learner, meteorological encoding is performed on the target meteorological feature vector to determine a meteorological encoding vector; According to the output layer in the transfer learner, the meteorological coding vector is classified and processed to determine the target predicted photovoltaic power.

4. A training method for a photovoltaic power prediction model, characterized in that: include: Acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station, wherein the number of the source photovoltaic power stations is multiple, the first training data includes first measured weather forecast data and first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively; the second training data includes second measured weather forecast data and second measured photovoltaic power of the target photovoltaic power station; Pre-training the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model; Based on the first training data, update the parameters of the adaptive learner in the first model to obtain a second model after the parameters are updated; Fine-tune the parameters of the adaptive learner in the second model based on the second training data to determine a photovoltaic power prediction model for the target photovoltaic power station; The data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold; The using the first training data to pre-train the adaptive learner and the transfer learner in the initial model respectively to obtain the pre-trained first model includes: Based on the adaptive learner in the initial model, the first measured weather forecast data is subjected to meteorological feature processing to obtain a first meteorological feature vector; Based on the transfer learner in the initial model, predicting the photovoltaic power of the first meteorological feature vector to obtain a first predicted photovoltaic power of each source photovoltaic power station; Determining a first average prediction error of the initial model according to the first predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station; The adaptive learner and the transfer learner in the initial model are pre-trained according to the first average prediction error to obtain the first model.

5. The method according to claim 4, wherein updating the parameters of the adaptive learner in the first model based on the first training data to obtain the second model after the parameters are updated comprises: Based on the adaptive learner in the first model, the first measured weather forecast data is subjected to meteorological feature processing to obtain a second meteorological feature vector; Based on the transfer learner in the first model, performing photovoltaic power prediction on the second meteorological feature vector to obtain a second predicted photovoltaic power of each source photovoltaic power station; determining a second average prediction error of the first model according to the second predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station; determining an average parameter gradient of the first model according to the second average prediction error; Based on the average parameter gradient, update the parameters of the adaptive learner in the first model to obtain the second model.

6. The method according to claim 4, wherein fine-tuning parameters of the adaptive learner in the second model based on the second training data to determine the photovoltaic power prediction model of the target photovoltaic power station comprises: Based on the adaptive learner in the second model, the second measured weather forecast data is subjected to meteorological feature processing to obtain a measured meteorological feature vector; Based on the transfer learner in the second model, the measured meteorological feature vector is used to predict photovoltaic power to obtain a third predicted photovoltaic power of the target photovoltaic power station; determining a third average prediction error of the second model according to the third predicted photovoltaic power and the second measured photovoltaic power; Fine-tune the parameters of the adaptive learner in the second model according to the third average prediction error to obtain the photovoltaic power prediction model.

7. A photovoltaic power prediction device, characterized in that: include: The first acquisition module is used to obtain real-time weather forecast data of the target photovoltaic power station; A photovoltaic power prediction module, used to process the real-time weather forecast data using the photovoltaic power prediction model of the target photovoltaic power station to obtain a target predicted photovoltaic power of the target photovoltaic power station; The photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, and uses the first training data to update the parameters of the adaptive learner in the first model obtained by pre-training, and uses the second training data of the target photovoltaic power station to continue to fine-tune the parameters of the adaptive learner in the second model obtained by parameter update, the data volume of the first training data is greater than the first data volume threshold, the data volume of the second training data is less than the second data volume threshold, the first data volume threshold is greater than the second data volume threshold, the number of the source photovoltaic power stations is multiple, the first training data includes the first measured weather forecast data and the first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively; the second training data includes the second measured weather forecast data and the second measured photovoltaic power of the target photovoltaic power station; The photovoltaic power prediction model uses the first training data of the source photovoltaic power station to pre-train the adaptive learner and the transfer learner in the initial model respectively, including: based on the adaptive learner in the initial model, the first measured weather forecast data is processed with meteorological features to obtain a first meteorological feature vector; based on the transfer learner in the initial model, the photovoltaic power is predicted for the first meteorological feature vector to obtain the first predicted photovoltaic power of each source photovoltaic power station; according to the first predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station, the first average prediction error of the initial model is determined; according to the first average prediction error, the adaptive learner and the transfer learner in the initial model are pre-trained to obtain the first model.

8. A photovoltaic power prediction model training device, characterized in that: include: A second acquisition module is used to acquire first training data of a source photovoltaic power station and second training data of a target photovoltaic power station, wherein the number of the source photovoltaic power stations is multiple, the first training data includes first measured weather forecast data and first measured photovoltaic power corresponding to the multiple source photovoltaic power stations respectively; the second training data includes second measured weather forecast data and second measured photovoltaic power of the target photovoltaic power station; A pre-training module, used to pre-train the adaptive learner and the transfer learner in the initial model respectively using the first training data to obtain a pre-trained first model; A parameter updating module, used for updating the parameters of the adaptive learner in the first model based on the first training data, and obtaining a second model after the parameters are updated; A parameter fine-tuning module, configured to fine-tune the parameters of the adaptive learner in the second model based on the second training data, and determine a photovoltaic power prediction model for the target photovoltaic power station; The data volume of the first training data is greater than a first data volume threshold, the data volume of the second training data is less than a second data volume threshold, and the first data volume threshold is greater than the second data volume threshold; Pre-training modules are specifically used for: Based on the adaptive learner in the initial model, the first measured weather forecast data is subjected to meteorological feature processing to obtain a first meteorological feature vector; Based on the transfer learner in the initial model, predicting the photovoltaic power of the first meteorological feature vector to obtain a first predicted photovoltaic power of each source photovoltaic power station; Determining a first average prediction error of the initial model according to the first predicted photovoltaic power of each source photovoltaic power station and the first measured photovoltaic power of each source photovoltaic power station; The adaptive learner and the transfer learner in the initial model are pre-trained according to the first average prediction error to obtain the first model.

9. An electronic device, characterized in that: include: processor; A memory for storing executable instructions; The processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method described in any one of claims 1-3 or 4-6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method described in any one of claims 1 to 3 or 4 to 6.

Images