Photovoltaic power generation prediction method, device, equipment, storage medium and product

Abstract

The present application discloses a photovoltaic power generation prediction method, apparatus, device, storage medium, and product, relating to the field of photovoltaic power generation prediction technology. The method comprises: obtaining meteorological data and photovoltaic power generation data of a photovoltaic site within a first preset time period before the current moment; the meteorological data comprising first meteorological data within the first preset time period and second meteorological data within a second preset time period after the current moment; inputting all of the first meteorological data, the second meteorological data, and the photovoltaic power generation data into a photovoltaic power generation prediction model to obtain a photovoltaic power generation prediction result; the model is trained using multiple sample data, wherein different sample data correspond to different sample moments, and each sample data comprises photovoltaic power generation data and meteorological data within a first preset sample time period before the corresponding sample moment, and meteorological data within a second preset sample time period after the sample moment. The present application can improve the accuracy of photovoltaic power generation prediction.

Description

Technical Field

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

[0002] Photovoltaic power generation prediction plays an important role in guiding decisions such as the operation planning of photovoltaic power plants. In related technologies, historical operating data of photovoltaic power plants is usually directly input into a trained prediction model to predict power generation.

[0003] However, photovoltaic power generation has great uncertainty and is easily affected by changes in the meteorological environment (such as light intensity). It is difficult to guarantee the accuracy of the prediction results by relying solely on the historical operating data of photovoltaic power fields to predict power generation. Summary of the Invention

[0004] The main purpose of this application is to provide a photovoltaic power generation power prediction method, device, equipment, storage medium and product, aiming to solve the technical problem of low accuracy of photovoltaic power generation power prediction in related technologies.

[0005] To achieve the above objectives, this application proposes a photovoltaic power generation power prediction method, which includes:

[0006] Acquire meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes a plurality of first meteorological data within the first preset time period and a plurality of second meteorological data within a second preset time period after the current moment;

[0007] All first meteorological data, all second meteorological data and photovoltaic power generation data are input into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training multiple photovoltaic site sample data, different photovoltaic site sample data correspond to different sample moments, and each photovoltaic site sample data includes photovoltaic power generation data and meteorological data in a first preset sample time period before the corresponding sample moment, and meteorological data in a second preset sample time period after the sample moment.

[0008] In one embodiment, the first meteorological data and the second meteorological data both correspond to sampling times, and the first meteorological data and the second meteorological data both include total irradiance;

[0009] Before the step of inputting all the first meteorological data, all the second meteorological data, and the photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result, the method further includes:

[0010] For each sampling moment, statistical processing is performed on all total irradiances within a preset time window at the sampling moment to obtain first meteorological characteristic data;

[0011] For the first meteorological data or the second meteorological data at each sampling moment, the sampling moment corresponding to the maximum total irradiance within the preset statistical time period of the sampling moment is determined as the reference moment;

[0012] For each sampling moment, obtaining second meteorological characteristic data based on a time difference between the sampling moment and a reference moment;

[0013] The steps of inputting all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result include:

[0014] All first meteorological characteristic data, all second meteorological characteristic data, and photovoltaic power generation data are input into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result.

[0015] In one embodiment, the first meteorological data and the second meteorological data further include humidity parameters, wind parameters, rainfall, and ambient temperature;

[0016] Before the step of inputting all the first meteorological data, all the second meteorological data, and the photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result, the method further includes:

[0017] For each sampling moment of the first meteorological data or the second meteorological data, determining third meteorological characteristic data based on the product of the total irradiance and the humidity parameter, the wind parameter, and the rainfall; and / or

[0018] For the first meteorological data at each sampling moment, determining fourth meteorological characteristic data based on the temperature difference between the ambient temperature and the internal temperature of the photovoltaic device;

[0019] The steps of inputting all first meteorological characteristic data, all second meteorological characteristic data, and photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result include:

[0020] All first meteorological characteristic data, all second meteorological characteristic data, photovoltaic power generation data, and all third meteorological characteristic data and / or all fourth meteorological characteristic data are input into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result.

[0021] In one embodiment, the photovoltaic power generation prediction model includes a first photovoltaic power generation prediction model and a second photovoltaic power generation prediction model;

[0022] The steps of inputting all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result include:

[0023] Inputting all first meteorological data, all second meteorological data, and photovoltaic power generation data into a first photovoltaic power generation power prediction model to obtain a first photovoltaic power generation power prediction result;

[0024] inputting all the first meteorological data, all the second meteorological data, and the photovoltaic power generation data into a second photovoltaic power generation power prediction model to obtain a second photovoltaic power generation power prediction result;

[0025] The photovoltaic power generation power prediction result is determined based on an average of the first photovoltaic power generation power prediction result and the second photovoltaic power generation power prediction result.

[0026] In one embodiment, the first photovoltaic power generation prediction model is obtained by training an optimized Transformer model, and the optimized Transformer model is obtained by replacing the decoder of the original Transformer model with a deep neural network DNN; the second photovoltaic power generation prediction model is obtained by training a CNN-GRU model.

[0027] In one embodiment, the photovoltaic power generation data includes photovoltaic array voltage, photovoltaic array current, grid-side three-phase voltage, grid-side three-phase current, power factor, power generation efficiency, and photovoltaic power generation power.

[0028] In addition, to achieve the above-mentioned purpose, the present application also provides a photovoltaic power generation power prediction device, which includes:

[0029] A data acquisition module is configured to acquire meteorological data and photovoltaic power generation data of a photovoltaic station within a first preset time period before a current moment; wherein the meteorological data includes a plurality of first meteorological data within the first preset time period and a plurality of second meteorological data within a second preset time period after the current moment;

[0030] The power prediction module is used to input all the first meteorological data, all the second meteorological data and photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training multiple photovoltaic site sample data, different photovoltaic site sample data correspond to different sample moments, and each photovoltaic site sample data includes photovoltaic power generation data and meteorological data in a first preset sample time period before the corresponding sample moment and meteorological data in a second preset sample time period after the sample moment.

[0031] In addition, to achieve the above-mentioned purpose, the present application also proposes a photovoltaic power generation prediction device, which includes: a memory, a processor, and a computer program stored in the memory and runnable on the processor, and the computer program is configured to implement the steps of the photovoltaic power generation prediction method as described above.

[0032] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the photovoltaic power generation prediction method as described above are implemented.

[0033] In addition, to achieve the above-mentioned purpose, the present application also proposes a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the photovoltaic power generation power prediction method as described above are implemented.

[0034] One or more technical solutions proposed in this application have at least the following technical effects:

[0035] In the photovoltaic power generation prediction method proposed in this application, photovoltaic power generation data within a first preset time period before the current moment is combined with multiple first meteorological data within the first preset time period and multiple second meteorological data within a second preset time period after the current moment, and these are collectively used as inputs to a photovoltaic power generation prediction model to obtain a photovoltaic power generation prediction result. The addition of the first and second meteorological data can provide more comprehensive information on meteorological environmental changes, enabling the photovoltaic power generation prediction model to more comprehensively capture external factors that affect photovoltaic power generation, thereby better capturing the correlation between photovoltaic power generation and meteorological changes, and thus improving the accuracy of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0038] Figure 1 A schematic diagram of a process flow provided for the first embodiment of the photovoltaic power generation prediction method of the present application;

[0039] Figure 2 Schematic diagram of statistical processing of total irradiance within a preset time window for an example;

[0040] Figure 3 A schematic diagram for determining second meteorological characteristic data of an example;

[0041] Figure 4 This is a schematic diagram of the model prediction results of meteorological data and photovoltaic power generation data that have not been processed by feature engineering; Figure 4 (a) is a schematic diagram comparing the predicted results of the CNN-GRU model with the actual results. Figure 4 (b) is a schematic diagram comparing the predicted results and actual results of the ResNet-GRU model. Figure 4 (c) is a schematic diagram comparing the predicted results and actual results of the Encoder-DNN Transformer model. Figure 4 (d) Schematic diagram comparing the predicted results of the MLP model with the actual results;

[0042] Figure 5 This is a schematic diagram of the model prediction results based on meteorological characteristic data and photovoltaic power generation data; Figure 5 (a) is a schematic diagram comparing the predicted results of the CNN-GRU model with the actual results. Figure 5 (b) is a schematic diagram comparing the predicted results and actual results of the ResNet-GRU model. Figure 5 (c) is a schematic diagram comparing the predicted results and actual results of the Encoder-DNN Transformer model. Figure 5 (d) Schematic diagram comparing the predicted results of the MLP model with the actual results;

[0043] Figure 6 This is a schematic diagram of the module structure of the photovoltaic power prediction device according to an embodiment of the present application;

[0044] Figure 7 Schematic diagram of the device structure of the hardware operating environment involved in the photovoltaic power generation prediction method in the embodiment of the present application.

[0045] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0047] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0048] The main solution of the embodiment of the present application is: obtaining meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes multiple first meteorological data within the first preset time period and multiple second meteorological data within a second preset time period after the current moment; all first meteorological data, all second meteorological data and photovoltaic power generation data are input into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training multiple photovoltaic station sample data, different photovoltaic station sample data correspond to different sample moments, and each photovoltaic station sample data includes photovoltaic power generation data and meteorological data within the first preset sample time period before the corresponding sample moment and meteorological data within the second preset sample time period after the sample moment.

[0049] Photovoltaic power generation forecasting plays an important role in guiding decisions such as the operational planning of photovoltaic power plants. Related technologies typically use historical operating data from photovoltaic power plants to directly input trained prediction models to predict power generation. However, photovoltaic power generation exhibits significant uncertainty and is susceptible to changes in the meteorological environment (such as sunlight). Relying solely on historical operating data for power generation forecasting is difficult to guarantee accurate prediction results.

[0050] The present application provides a solution, which combines photovoltaic power generation data within a first preset time period before the current moment with multiple first meteorological data within the first preset time period and multiple second meteorological data within a second preset time period after the current moment, and uses them together as input to a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result; adding the first meteorological data and the second meteorological data can provide additional and more comprehensive information on meteorological environment changes, so that the photovoltaic power generation power prediction model can more comprehensively understand the external factors affecting photovoltaic power generation, thereby better capturing the correlation between photovoltaic power generation power and meteorological changes, and improving the accuracy of the prediction results.

[0051] It should be noted that the executor of this embodiment is a photovoltaic power generation prediction device, which can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, personal computer, etc., or an electronic device that can realize the above functions.

[0052] Based on this, the embodiment of the present application provides a photovoltaic power generation power prediction method, referring to Figure 1 , Figure 1 This is a flow chart of Example 1 of the photovoltaic power generation prediction method of this application.

[0053] In this embodiment, the photovoltaic power generation power prediction method includes steps S100 to S200:

[0054] Step S100, obtaining meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes multiple first meteorological data within the first preset time period and multiple second meteorological data within a second preset time period after the current moment.

[0055] Step S200, all the first meteorological data, all the second meteorological data and photovoltaic power generation data are input into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training multiple photovoltaic site sample data, different photovoltaic site sample data correspond to different sample moments, and each photovoltaic site sample data includes photovoltaic power generation data and meteorological data in a first preset sample time period before the corresponding sample moment and meteorological data in a second preset sample time period after the sample moment.

[0056] Specifically, in actual applications, photovoltaic stations usually have dedicated monitoring systems to record photovoltaic power generation data in real time; for example, SCADA (Supervisory Control And Data Acquisition). Through the corresponding API interface of the monitoring system, multiple photovoltaic power generation data within the first preset time period before the current moment can be directly extracted. Alternatively, the photovoltaic station will regularly record and store historical power generation data in a local database, or directly obtain photovoltaic power generation data within the first preset time period before the current moment through database query. Photovoltaic power generation data is generally related data that can reflect the operating status and power generation performance of the photovoltaic station, and may include but is not limited to photovoltaic array voltage, photovoltaic array current, grid-side three-phase voltage, grid-side three-phase current, power factor, power generation efficiency, photovoltaic power generation power and other data.

[0057] The first meteorological data is essentially historical meteorological data for a period of time before the current moment, while the second meteorological data is predicted meteorological data for a period of time after the current moment. Both the first meteorological data and the second meteorological data can be obtained from public meteorological data sources (such as local meteorological stations, meteorological service providers, etc.). For example, the first meteorological data can be extracted from the historical monitoring data of the local meteorological station, and the second meteorological data can be extracted from the NWP (Numerical Weather Prediction) data of the local meteorological station. It should be noted that the lengths of the first preset time period and the second preset time period can be set according to actual forecast requirements. For example, in one example, in order to meet the short-term forecast requirements for photovoltaic power generation in the next few days, the first preset time period can be the 14-day period before the current moment, and the second preset time period can be the 7-day period after the current moment.

[0058] Then, all the acquired first meteorological data, all the acquired second meteorological data and photovoltaic power generation data are input into the photovoltaic power generation power prediction model. The photovoltaic power generation power prediction model is obtained by training with sample data from multiple photovoltaic sites. Different photovoltaic site sample data correspond to different sample moments, and each photovoltaic site sample data includes photovoltaic power generation data and meteorological data within a first preset sample time period before the corresponding sample moment and meteorological data within a second preset sample time period after the sample moment; the label value of the photovoltaic power generation power prediction model is the photovoltaic power generation power within the second preset sample time period after the sample moment; the first preset sample time period and the second preset sample time period here are respectively the same as the aforementioned first preset time period and second preset time period.

[0059] It can be understood that the photovoltaic power generation prediction model is trained using historical photovoltaic power generation data before the sample moment, as well as meteorological data that includes a complete time series of historical meteorological data before the sample moment and future meteorological data after the sample moment. It can comprehensively learn the nonlinear relationship between meteorological factors and photovoltaic power generation, avoid prediction inaccuracies caused by meteorological environment fluctuations, and effectively improve the accuracy of photovoltaic power generation prediction results. Therefore, the photovoltaic power generation data within the first preset time period before the current moment, the first meteorological data within the first preset time period, and the second meteorological data within the second preset time period after the current moment are input into the photovoltaic power generation prediction model obtained by training with the above-mentioned photovoltaic station sample data. It is possible to predict photovoltaic power generation in future time periods and obtain a relatively accurate photovoltaic power generation in the second preset time period after the current moment.

[0060] Directly utilizing photovoltaic power generation data within a first preset time period before the current moment, multiple first meteorological data within the first preset time period, and multiple second meteorological data within a second preset time period after the current moment for power prediction takes into account the impact of meteorological factors on photovoltaic power generation, and can improve the accuracy of photovoltaic power generation prediction to a certain extent. However, since there is a relatively complex nonlinear relationship between photovoltaic power generation and meteorological factors, and the changes in meteorological factors have certain temporal regularities, in order to further improve the accuracy of photovoltaic power generation prediction, steps A100 to A300 may be included before step S200 to perform feature engineering on the meteorological data to enhance the strong correlation between the meteorological data and photovoltaic power generation, so that the photovoltaic power generation prediction model can better learn the potential relationship between meteorological factors and photovoltaic power generation, so as to fully utilize meteorological data for prediction and improve the prediction accuracy of photovoltaic power generation.

[0061] Step A100 : For each sampling moment, statistical processing is performed on all total irradiances within a preset time window at the sampling moment to obtain first meteorological characteristic data.

[0062] Step A200: for each sampling moment of the first meteorological data or the second meteorological data, the sampling moment corresponding to the maximum total irradiance within the preset statistical time period of the sampling moment is determined as the reference moment.

[0063] Step A300: For each sampling moment, obtain second meteorological characteristic data based on the time difference between the sampling moment and the reference moment.

[0064] To ensure the effective construction of feature engineering data, the Spearman correlation coefficient method can be used to pre-analyze the correlation between various types of meteorological data and photovoltaic power generation data to obtain a data correlation heat map. The correlation heat map can intuitively show the correlation coefficient between each data and photovoltaic power generation. Among them, meteorological data includes but is not limited to total irradiance, humidity, wind speed, wind volume, atmospheric pressure, temperature, etc.; photovoltaic power generation data includes but is not limited to photovoltaic array voltage, photovoltaic array current, grid-side three-phase voltage, grid-side three-phase current, power factor, active power, reactive power, cumulative power generation, and photovoltaic power generation.

[0065] In this example, the inventors used the Spearman correlation coefficient method to analyze the various types of meteorological data and photovoltaic power generation data, finding that photovoltaic power generation has the highest correlation with total irradiance. Therefore, feature engineering data can be constructed around total irradiance. Data such as atmospheric pressure and cumulative power generation have very low correlations with photovoltaic power generation. Therefore, this data with low correlations with photovoltaic power generation can be deleted to reduce data processing.

[0066] The first meteorological data and the second meteorological data both have corresponding sampling times, and both the first meteorological data and the second meteorological data include total irradiance. Since photovoltaic output power has a certain time series regularity, time series-related meteorological characteristic data can be constructed based on total irradiance.

[0067] Specifically, for each sampling moment, all total irradiances within the preset time window of the sampling moment can be statistically processed by summing, averaging, median and maximum values ​​to obtain the first meteorological characteristic data; wherein, the preset time window can be a 0~24h time window of the day of the sampling moment, or a 1h time window of the hour of the sampling moment. It can be set according to data processing requirements and is not specifically limited here. Figure 2 This is a schematic diagram of the statistical processing of the total irradiance in a preset time window. Figure 2The chart shows the multiple sampling times from 9:15:00 on January 1, 2019 to 13:00:00 on January 1, 2019, and the actual power (i.e., photovoltaic output power) corresponding to each time. The sampling interval between two adjacent sampling times is 15 minutes. Taking the sampling time of 11:00:00 on January 1, 2019 as an example, Figure 2 The red dotted box in the middle indicates the preset time window (11:00:00~12:00:00) with a time length of one hour corresponding to the sampling time of 11:00:00 on January 1, 2019. The total irradiance corresponding to the four sampling times of 11:00:00, 11:15:00, 11:30:00 and 11:45:00 in the preset time window can be summed, averaged, median and maximum to determine the first meteorological characteristic data corresponding to the sampling time; similarly, the blue dotted box indicates a preset time window with a time length of three hours.

[0068] Establishing a preset time window and collecting meteorological data within it can capture complete meteorological information for the time period before and after the current moment. In this embodiment, total irradiance statistics at different time granularities (e.g., within a day or an hour) can indirectly reflect meteorological changes at different time granularities (e.g., within a day and an hour). This allows the photovoltaic power prediction model to more accurately learn the dynamic correlation between meteorological factors and power generation, and adapt to prediction needs at different time scales.

[0069] It is also possible to determine the maximum total irradiance within the preset statistical time period of the sampling moment for the first meteorological data or the second meteorological data at each sampling moment, and use the sampling moment corresponding to the maximum total irradiance as the reference moment; then calculate the time difference between the sampling moment and the reference moment, and use this time difference as the second meteorological characteristic data corresponding to the sampling moment. Figure 3 The second meteorological characteristic data is determined as a schematic diagram of an example; Figure 3 As shown, the maximum total irradiance on the day of sampling time A corresponds to sampling time B, and sampling time B is used as the reference time. The time difference between sampling time A and reference time B is the second meteorological characteristic data corresponding to sampling time A. The above preset statistical time period is generally a 24-hour time period on the day of the sampling time.

[0070] It is not difficult to understand that the length of time between each sampling moment and the sampling moment corresponding to the total irradiance peak (i.e., maximum total irradiance) of the day can reflect the movement trajectory of the sun. For similar times of the day, the solar irradiation, temperature and other conditions should be similar, and the photovoltaic power generation power should also be similar; therefore, the second meteorological characteristic data obtained through the time difference between the sampling moment and the reference moment can quantify the dynamic relationship between the solar movement law and the meteorological state, which helps to improve the accuracy of the prediction results.

[0071] By inputting all of the above-mentioned first meteorological characteristic data, all of the second meteorological characteristic data, and photovoltaic power generation data into the photovoltaic power generation prediction model, a more accurate photovoltaic power generation prediction result can be obtained. It should be noted that when using all of the first meteorological characteristic data, all of the second meteorological characteristic data, and photovoltaic power generation data obtained through feature engineering to predict photovoltaic power generation, the corresponding photovoltaic power generation prediction model should also be trained using photovoltaic site sample data processed through feature engineering. The feature engineering processing of photovoltaic site sample data can refer to the processing flow of the first meteorological characteristic data and the second meteorological characteristic data described above, and will not be repeated here.

[0072] In addition, in order to capture the interactive effects between different types of meteorological data, in a feasible implementation, steps A400 to A500 may be further included before step S200 to establish coupling relationships between different meteorological data:

[0073] Step A400: For each sampling moment of the first meteorological data or the second meteorological data, determine the third meteorological characteristic data based on the product of the total irradiance and the humidity parameter, the wind parameter and the rainfall. And / or

[0074] Step A500: For the first meteorological data at each sampling moment, fourth meteorological characteristic data is determined based on the temperature difference between the ambient temperature and the internal temperature of the photovoltaic device.

[0075] Specifically, the first meteorological data and the second meteorological data also include but are not limited to humidity parameters, wind parameters, rainfall, and ambient temperature data. Among them, humidity parameters include but are not limited to relative humidity; wind parameters include but are not limited to wind speed, wind direction, average wind speed, average wind direction, gust speed, and gust direction. Therefore, for each sampling moment, the third meteorological characteristic data may include but are not limited to the product of total irradiance and relative humidity, the product of total irradiance and wind speed, the product of total irradiance and wind direction, the product of total irradiance and average wind speed, the product of total irradiance and average wind direction, the product of total irradiance and gust speed, the product of total irradiance and gust direction, and the product of total irradiance and rainfall at that sampling moment. Through the interactive product between total irradiance and humidity parameters, wind parameters, rainfall and other data, the coupling effect between multiple meteorological factors can be captured to solve nonlinear problems.

[0076] For each sampling moment of the first meteorological data, the temperature difference between the ambient temperature and the internal temperature of the photovoltaic module can be calculated and used as the fourth meteorological characteristic data. It is easy to understand that the difference between the ambient temperature and the internal temperature of the photovoltaic module can reflect the degree of deviation between the actual operating temperature of the photovoltaic module and the environment. Generally speaking, the larger the temperature difference, the worse the heat dissipation of the photovoltaic module, and the photovoltaic power generation efficiency will decrease; the smaller the temperature difference, the better the heat dissipation of the photovoltaic module, and the higher the photovoltaic power generation efficiency. In other words, the temperature difference between the ambient temperature and the internal temperature of the photovoltaic module will also affect the photovoltaic power generation output. Therefore, adding the temperature difference as a fourth meteorological characteristic can help improve the accuracy of photovoltaic power generation output.

[0077] By inputting all the first meteorological characteristic data, all the second meteorological characteristic data, photovoltaic power generation data, and all the third meteorological characteristic data and / or all the fourth meteorological characteristic data obtained after the above-mentioned feature engineering processing into the photovoltaic power generation power prediction model, the corresponding photovoltaic power generation power prediction result can be obtained. Similarly, in this embodiment, when the photovoltaic power generation power is predicted using the first meteorological characteristic data, the second meteorological characteristic data, the third meteorological characteristic data, the fourth meteorological characteristic data and the photovoltaic power generation data, the corresponding photovoltaic power generation power prediction model should also be trained by the photovoltaic station sample data that has been processed by feature engineering. The feature engineering processing of the photovoltaic station sample data can refer to the processing flow of the first meteorological characteristic data, the second meteorological characteristic data, the third meteorological characteristic data and the fourth meteorological characteristic data, which will not be repeated here.

[0078] Furthermore, before constructing the first, second, third, and fourth meteorological characteristic data, the acquired photovoltaic power generation data, first and second meteorological data, etc., can be subjected to outlier processing and normalization. During outlier processing, considering that outliers generally have a short duration, outliers can be replaced with null values, and linear interpolation can then be used to interpolate the replaced null values. Outlier processing can improve data reliability and prevent abnormal noise data from affecting model prediction results.

[0079] Normalize the data after outlier processing. Specifically, use interval scaling to convert different types of data to the same dimension, avoiding data incomparability caused by different dimensions. For example, if data normalization is not performed, data with larger dimensions (such as total irradiance) may dominate model training, affecting the fair participation of other features. However, after normalization, the dimensional differences between different data are eliminated, ensuring that different features such as meteorological data and photovoltaic power generation data can contribute equally, thereby optimizing the model's prediction accuracy.

[0080] The photovoltaic power generation prediction model can be trained using common neural network models, such as MLP (Multilayer Perceptron), Light Gradient Boosting Machine (LighGBM), and Recurrent Neural Network (RNN). Alternatively, it can be trained using a hybrid architecture of common neural networks, such as the CNN-GRU model, ResNet-GRU model, and Encoder-DNNTransformer model.

[0081] The CNN-GRU model is a hybrid architecture combining a CNN (Convolutional Neural Network) and a GRU (Gated Recurrent Unit). The CNN-GRU model uses a CNN for feature extraction, employing a two-stage convolutional feature abstraction module. The first-stage module consists of three one-dimensional convolutional layers and a downsampling unit. The first convolutional layer uses a 5×1 convolution kernel with 192 channels for feature extraction. The next layers are compressed using 1×1 convolution kernels with 160 channels, and the final layer uses a 1×1 convolution kernel with 96 channels. This layer is followed by a one-dimensional max pooling layer with a sampling window length of 3 and a sliding stride of 2 for sequence length reduction. The second-stage module is a feature mapping unit. The first layer uses a 3×1 convolution kernel with 192 channels, followed by 1×1 convolution kernels with 192 channels in the middle layers, and the final layer uses a 1×1 convolution kernel with 10 channels to project high-dimensional features into photovoltaic power generation indicators. The output data matrix processed by this multi-stage convolutional architecture carries data features that more clearly reflect changes in photovoltaic output power. After the GRU completes feature extraction, the feature data matrix is ​​input into the prediction layer. The prediction layer of the CNN-GRU model uses a stacked two-layer GRU network structure. The GRU is a variant of the LSTM (Long Short-Term Memory) network, which removes the output gate, resulting in a simpler overall structure.

[0082] The ResNet-GRU model is a hybrid architecture of the ResNet (Residual Network) and the GRU. ResNet, as the feature extraction component, uses fully connected layers with batch normalization. The ResNet network structure consists of four RUs (Residual Units) with 128 feature vectors, six RUs with 256 feature vectors, and three RUs with 512 feature vectors. After ResNet extracts features from the model's input data, it outputs the extracted features to a two-layer stack of GRUs for prediction.

[0083] The Encoder-DNN Transformer model is an optimized Transformer model. The Encoder-DNN Transformer is a hybrid architecture model that replaces the decoder of the original Transformer model with a DNN (Deep Neural Network). In the original Transformer model, the encoder portion contains 66.4% of all trained parameters. The decoder, a symmetrical structure in the original Transformer model, typically accounts for half of the Transformer's total parameters. Removing the decoder significantly reduces model complexity, lowering the computational overhead of training and inference, and thus reducing the complexity of the gradient propagation path during training. Furthermore, the encoder is primarily used for positional encoding. Retaining the encoder ensures strong temporal correlation in photovoltaic power prediction. Positional encoding explicitly injects temporal information, compensating for the limited local receptive field of models such as CNNs and RNNs. The Transformer model's multi-head attention mechanism can avoid the vanishing gradient problem of RNNs and the locality limitations of CNNs.

[0084] The Encoder-DNN Transformer model does not use any recurrent or convolutional layers. It only involves the attention mechanism and common neural network layers such as the embedding layer, the DNN's Dense layer, and the normalization layer. Although the Transformer model has a large number of parameters, it is easy to parallelize, and models built on it are faster to train. When making predictions, the Encoder-DNN Transformer model first performs positional embedding to encode the position of each feature data in the feature vector. Fixed-position embedding can be used here. The encoder then performs encoding. The encoder takes a batch of vectors represented as feature ID sequences as input and encodes each feature data into a 512-dimensional representation. The encoder can be stacked three times. The data processed by the encoder is directly input into the multi-layer Dense layer of the DNN for prediction.

[0085] When predicting photovoltaic power generation, raw data such as the first meteorological data, the second meteorological data, and the photovoltaic power generation data, or meteorological data and photovoltaic power generation data that have undergone feature engineering processing, such as the first meteorological characteristic data, the second meteorological characteristic data, the third meteorological characteristic data, and the fourth meteorological characteristic data, can be input into a single photovoltaic power generation prediction model to directly obtain the photovoltaic power generation prediction result. Alternatively, in a feasible embodiment, the photovoltaic power generation prediction model can also include the first photovoltaic power generation prediction model and the second photovoltaic power generation prediction model; in this embodiment, step S200 can specifically include steps S210 to S230:

[0086] Step S210: input all first meteorological data, all second meteorological data and photovoltaic power generation data into a first photovoltaic power generation power prediction model to obtain a first photovoltaic power generation power prediction result.

[0087] Step S220: input all first meteorological data, all second meteorological data and photovoltaic power generation data into a second photovoltaic power generation power prediction model to obtain a second photovoltaic power generation power prediction result.

[0088] Step S230 : determining a photovoltaic power generation prediction result based on an average of the first photovoltaic power generation prediction result and the second photovoltaic power generation prediction result.

[0089] Specifically, the first photovoltaic power generation prediction model and the second photovoltaic power generation prediction model are two different models. The model structures may be different: the first photovoltaic power generation prediction model can be trained by an optimized Transformer model, which replaces the decoder of the original Transformer model with a deep neural network DNN; the second photovoltaic power generation prediction model is trained by a CNN-GRU model. Alternatively, the training data may be different: for example, the first photovoltaic power generation prediction model and the second photovoltaic power generation prediction model are both trained by an optimized Transformer model, but the first photovoltaic power generation prediction model is trained by photovoltaic power generation data and meteorological data in 2020, and the second photovoltaic power generation prediction model is trained by photovoltaic power generation data and meteorological data in 2023.

[0090] All first meteorological data, all second meteorological data, and photovoltaic power generation data are input into the first photovoltaic power generation prediction model and the second photovoltaic power generation prediction model, respectively, to obtain corresponding first photovoltaic power generation power prediction results and second photovoltaic power generation power prediction results. Similarly, all first meteorological data, all second meteorological data, all third meteorological characteristic data, all fourth meteorological characteristic data, and photovoltaic power generation data that have undergone feature engineering can also be input into the first photovoltaic power generation prediction model and the second photovoltaic power generation prediction model, respectively, to obtain corresponding first photovoltaic power generation power prediction results and second photovoltaic power generation power prediction results. The average of the first photovoltaic power generation prediction results and the second photovoltaic power generation prediction results is then used as the final photovoltaic power generation power prediction result. Combining the prediction results of the two models to determine the photovoltaic power generation prediction result can reduce the prediction bias of a single model, making the prediction result closer to the true value.

[0091] It is not difficult to understand that the photovoltaic power generation prediction method provided in the embodiment of the present application combines the photovoltaic power generation data within the first preset time period before the current moment with multiple first meteorological data within the first preset time period and multiple second meteorological data within the second preset time period after the current moment, and uses them together as inputs of the photovoltaic power generation power prediction model to obtain photovoltaic power generation power prediction results; the added first meteorological data and second meteorological data can provide more comprehensive information on meteorological environment changes, so that the photovoltaic power generation power prediction model can more comprehensively obtain external factors affecting photovoltaic power generation power, thereby better capturing the correlation between photovoltaic power generation power and meteorological changes, so as to improve the accuracy of the prediction results.

[0092] In order to help understand the photovoltaic power generation power prediction method in this embodiment, the photovoltaic power generation data and meteorological data of a photovoltaic station in 2020 were used to construct a data set for model prediction verification.

[0093] It should be noted that this PV station includes four PV arrays: PV1, PV2, PV3, and PV4. PV power generation data includes: sampling time, PV1 voltage, PV1 current, PV2 voltage, PV2 current, PV3 voltage, PV3 current, PV4 voltage, PV4 current, input power, grid three-phase voltage (Ua, Ub, Uc), grid three-phase current (Ia, Ib, Ic), power factor, efficiency, PV power generation, PV internal temperature, etc.; meteorological data includes ambient temperature, humidity, atmospheric pressure, wind speed, wind direction, average wind speed, average wind direction, gust speed, gust direction, rainfall, total irradiance, and irradiance_POA. This example primarily uses four models for prediction: the CNN-GRU model, the ResNet-GRU model, the Encoder-DNN Transformer model, and the MLP model. The model inputs are meteorological data and PV power generation data from the past two weeks at different sampling times, as well as meteorological data for the next week. The model output is the PV power generation for the next week.

[0094] Figure 4 This is a schematic diagram of the model prediction results for meteorological data and photovoltaic power generation data that have not been processed by feature engineering. Figure 4 (a) Schematic diagram comparing the predicted results and actual results of the CNN-GRU model; Figure 4 (b) Schematic diagram comparing the predicted results and actual results of the ResNet-GRU model; Figure 4 (c) Schematic diagram comparing the predicted results and actual results of the Encoder-DNN Transformer model; Figure 4 (d) is a schematic diagram comparing the predicted results of the MLP model with the actual results. Figure 4 The blue curve in the middle is the actual photovoltaic output curve, which is used to represent the actual value of photovoltaic power generation. The green curve is the neural network prediction curve, which is used to represent the photovoltaic power generation prediction value output by the model. Figure 4 It can be seen that the predicted value of photovoltaic power generation is close to the actual value of photovoltaic power generation to a large extent. Therefore, the photovoltaic power generation prediction based on comprehensive meteorological data can achieve better prediction results.

[0095] Furthermore, feature engineering processing is performed on the original meteorological data of the photovoltaic station, and the first meteorological feature data obtained may include: total irradiance, the sum of daily total irradiance (here the sum of daily total irradiance means the sum of total irradiance at all sampling moments in the time window of the sampling moment, and the same applies to the daily average value of total irradiance, the median value of daily total irradiance, and the maximum value of daily total irradiance in the following text), the average value of daily total irradiance, the median value of daily total irradiance, the maximum value of daily total irradiance, and the sum of hourly total irradiance (here the sum of hourly total irradiance means the sum of total irradiance at all sampling moments in the time window of the hour of the sampling moment, and the average value of hourly total irradiance, the median value of hourly total irradiance in the following text The second meteorological characteristic data may include the time difference between the sampling time and the sampling time corresponding to the maximum total irradiance of the day; the third meteorological characteristic data include the product of total irradiance and relative humidity, the product of total irradiance and wind speed, the product of total irradiance and wind direction, the product of total irradiance and average wind speed, the product of total irradiance and average wind direction, the product of total irradiance and gust speed, the product of total irradiance and gust direction, the product of total irradiance and rainfall; the fourth meteorological characteristic data include the temperature difference between ambient temperature and photovoltaic internal temperature.

[0096] The feature-engineered data was also used to train four models: a CNN-GRU model, a ResNet-GRU model, an Encoder-DNN Transformer model, and an MLP model. The model inputs consisted of meteorological data (including the first, second, third, and fourth meteorological data) and photovoltaic power generation data from the past two weeks at different sampling times, as well as meteorological data for the week ahead of the sampling time. The model output was the photovoltaic power generation power for the next week.

[0097] Figure 5 This is a schematic diagram of the model prediction results based on meteorological characteristic data and photovoltaic power generation data. Figure 5 (a) Schematic diagram comparing the predicted results and actual results of the CNN-GRU model; Figure 5 (b) Schematic diagram comparing the predicted results and actual results of the ResNet-GRU model; Figure 5 (c) Schematic diagram comparing the predicted results and actual results of the Encoder-DNN Transformer model; Figure 5 (d) is a schematic diagram comparing the predicted results of the MLP model with the actual results. Figure 5 The blue curve in the middle is the actual photovoltaic output curve, which is used to represent the actual value of photovoltaic power generation. The green curve is the neural network prediction curve, which is used to represent the photovoltaic power generation prediction value output by the model. Figure 4 and Figure 5 It can be seen that compared with Figure 4 The predicted value of photovoltaic power generation and the actual value of photovoltaic power generation, Figure 5 The predicted photovoltaic power generation value shown is more consistent with the actual photovoltaic power generation value. This shows that meteorological data processed through feature engineering can further significantly improve the model's prediction results.

[0098] And from Figure 5 It can be seen that the prediction effects of the CNN-GRU model and the Encoder-DNN Transformer model are relatively better among the four models. Therefore, these two models can be combined as a combination model, and the photovoltaic power generation prediction results corresponding to the CNN-GRU model and the Encoder-DNN Transformer model are averaged as the photovoltaic power generation prediction result under the combination model. Table 1 shows the evaluation index results of different models under different inputs. The combination model in Table 1 is the average of the prediction results of the CNN-GRU model and the Encoder-DNN Transformer model; the original data refers to the meteorological data that has not been processed by feature engineering for prediction; the feature engineering data refers to the meteorological feature data obtained by feature engineering for prediction. The improvement rate represents the improvement of the evaluation index corresponding to the prediction based on the feature engineering data compared with the evaluation index corresponding to the prediction based on the original data.

[0099] Table 1

[0100]

[0101] The RMSE in Table 1 represents the root mean square error, R 2 is the coefficient of determination; these two indicators quantify the prediction performance of the model from the perspectives of error margin and model explanatory power. Among them, RMSE directly measures the average deviation between the predicted value and the true value. The smaller the value, the more accurate the prediction. 2 It represents the model’s ability to explain the fluctuation of the target variable (i.e., photovoltaic power generation), R 2 The larger the value, the better the model fits the data.

[0102] A vertical comparison in Table 1 clearly shows that the CNN-GRU model, the Encoder-DNN Transformer model, and their combined model all achieve relatively good prediction results. The CNN-GRU model's advantage in data processing lies in its ability to automatically extract high-dimensional features from the feature vector, reducing the complexity of subsequent model processing. Unlike conventional neural network models (such as the MLP model), which are prone to vanishing or exploding gradients when faced with long-term predictions, the Encoder-DNN Transformer model utilizes a multi-head attention mechanism, which avoids these problems. This allows the model to better adapt to long-term data sequences and learn complex patterns with different features at the same time point from the input data, improving the model's generalization ability. Therefore, a combined model can be formed by combining the CNN-GRU model and the Encoder-DNN Transformer model. The prediction results from the two optimal models can be averaged to form the final PV power prediction value, thereby improving the accuracy of the prediction results.

[0103] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the photovoltaic power generation power prediction method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0104] This application also provides a photovoltaic power generation prediction device, please refer to Figure 6 , the photovoltaic power generation power prediction device comprises:

[0105] The data acquisition module 10 is configured to acquire meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes a plurality of first meteorological data within the first preset time period and a plurality of second meteorological data within a second preset time period after the current moment;

[0106] The power prediction module 20 is used to input all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training the sample data of multiple photovoltaic sites, and the sample data of different photovoltaic sites correspond to different sample moments, and the sample data of each photovoltaic site includes the photovoltaic power generation data and meteorological data in the first preset sample time period before the corresponding sample moment and the meteorological data in the second preset sample time period after the sample moment.

[0107] The photovoltaic power prediction device provided in this application utilizes the photovoltaic power prediction method described in the aforementioned embodiments, thereby resolving the technical issue of low photovoltaic power prediction accuracy in related technologies. Compared to related technologies, the photovoltaic power prediction device provided in this application offers the same beneficial effects as the photovoltaic power prediction method described in the aforementioned embodiments. Other technical features of the photovoltaic power prediction device are the same as those disclosed in the aforementioned embodiments and are not further elaborated upon here.

[0108] The present application provides a photovoltaic power generation prediction device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the photovoltaic power generation prediction method in the above-mentioned embodiment one.

[0109] Reference below Figure 7 , which shows a schematic diagram of the structure of a photovoltaic power generation prediction device suitable for implementing the embodiments of the present application. The photovoltaic power generation prediction device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., as well as fixed terminals such as desktop computers. Figure 7 The photovoltaic power generation prediction device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0110] like Figure 7As shown, the photovoltaic power generation prediction device may include a processing device 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the photovoltaic power generation prediction device. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to I / O interface 1006: input device 1007, such as a touch screen, touchpad, keyboard, mouse, image sensor, microphone, etc.; output device 1008, such as a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003, such as a magnetic tape or hard disk; and communication device 1009. Communication device 1009 can allow the photovoltaic power generation prediction device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a photovoltaic power generation prediction device with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented or provided instead.

[0111] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0112] The photovoltaic power generation prediction device provided in this application utilizes the photovoltaic power generation prediction method described in the aforementioned embodiment, thereby resolving the technical issue of low photovoltaic power generation prediction accuracy in related technologies. Compared to related technologies, the photovoltaic power generation prediction device provided in this application achieves the same beneficial effects as the photovoltaic power generation prediction method described in the aforementioned embodiment. Other technical features of the photovoltaic power generation prediction device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0113] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0114] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0115] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, a computer program) stored thereon, and the computer-readable program instructions are used to execute the photovoltaic power generation power prediction method in the above-mentioned embodiment.

[0116] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0117] The computer-readable storage medium may be included in the photovoltaic power generation prediction device; or may exist independently without being assembled into the photovoltaic power generation prediction device.

[0118] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the photovoltaic power generation prediction device, the photovoltaic power generation prediction device: obtains meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes multiple first meteorological data within the first preset time period and multiple second meteorological data within a second preset time period after the current moment; all first meteorological data, all second meteorological data and photovoltaic power generation data are input into the photovoltaic power generation power prediction model to obtain the photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training multiple photovoltaic station sample data, different photovoltaic station sample data correspond to different sample moments, and each photovoltaic station sample data includes photovoltaic power generation data and meteorological data within the first preset sample time period before the corresponding sample moment and meteorological data within the second preset sample time period after the sample moment.

[0119] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0120] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0121] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0122] The computer-readable storage medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned photovoltaic power generation prediction method. This computer-readable storage medium can address the technical issue of low photovoltaic power generation prediction accuracy in related art. Compared to related art, the beneficial effects of the computer-readable storage medium provided in this application are similar to those of the photovoltaic power generation prediction method provided in the aforementioned embodiments, and are not further elaborated here.

[0123] The present application also provides a computer program product, comprising a computer program, which implements the steps of the photovoltaic power generation power prediction method as described above when the computer program is executed by a processor.

[0124] The computer program product provided in this application can address the low accuracy of photovoltaic power generation prediction in related technologies. Compared to related technologies, the computer program product provided in this application has the same beneficial effects as the photovoltaic power generation prediction method provided in the above-mentioned embodiments, and will not be further elaborated here.

[0125] The above descriptions are only some embodiments of the present application and do not limit the scope of protection. All equivalent structural transformations made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the scope of protection.

Claims

1. A photovoltaic power generation power prediction method, characterized in that: The photovoltaic power generation power prediction method comprises: Acquire meteorological data and photovoltaic power generation data of the photovoltaic station within a first preset time period before the current moment; wherein the meteorological data includes a plurality of first meteorological data within the first preset time period and a plurality of second meteorological data within a second preset time period after the current moment; Inputting all of the first meteorological data, all of the second meteorological data, and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result; the photovoltaic power generation power prediction model is obtained by training a plurality of photovoltaic site sample data, different photovoltaic site sample data correspond to different sample moments, and each photovoltaic site sample data includes photovoltaic power generation data and meteorological data within a first preset sample time period before the corresponding sample moment, and meteorological data within a second preset sample time period after the sample moment; The first meteorological data and the second meteorological data both correspond to sampling times, and both the first meteorological data and the second meteorological data include total irradiance; Before the step of inputting all of the first meteorological data, all of the second meteorological data, and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result, the method further includes: For each sampling moment, performing statistical processing on all the total irradiances within a preset time window at the sampling moment to obtain first meteorological characteristic data; For each of the first meteorological data or the second meteorological data at the sampling moment, determining the sampling moment corresponding to the maximum total irradiance within the preset statistical time period within which the sampling moment is located as a reference moment; For each of the sampling moments, obtaining second meteorological characteristic data based on a time difference between the sampling moment and the reference moment; The step of inputting all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result comprises: All of the first meteorological characteristic data, all of the second meteorological characteristic data, and the photovoltaic power generation data are input into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result.

2. The photovoltaic power generation prediction method according to claim 1, wherein: The first meteorological data and the second meteorological data further include humidity parameters, wind parameters, rainfall and ambient temperature; Before the step of inputting all of the first meteorological data, all of the second meteorological data, and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result, the method further includes: For each of the first meteorological data or the second meteorological data at the sampling moment, determining third meteorological characteristic data based on the product of the total irradiance and the humidity parameter, the wind parameter, and the rainfall; and / or For each of the first meteorological data at the sampling moment, determining fourth meteorological characteristic data based on the temperature difference between the ambient temperature and the internal temperature of the photovoltaic device; The step of inputting all the first meteorological characteristic data, all the second meteorological characteristic data, and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result comprises: All the first meteorological characteristic data, all the second meteorological characteristic data, the photovoltaic power generation data, and all the third meteorological characteristic data and / or all the fourth meteorological characteristic data are input into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result.

3. The photovoltaic power generation prediction method according to any one of claims 1 to 2, characterized in that: The photovoltaic power generation prediction model includes a first photovoltaic power generation prediction model and a second photovoltaic power generation prediction model; The step of inputting all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into a photovoltaic power generation power prediction model to obtain a photovoltaic power generation power prediction result comprises: inputting all the first meteorological data, all the second meteorological data, and the photovoltaic power generation data into the first photovoltaic power generation power prediction model to obtain a first photovoltaic power generation power prediction result; inputting all the first meteorological data, all the second meteorological data and the photovoltaic power generation data into the second photovoltaic power generation power prediction model to obtain a second photovoltaic power generation power prediction result; The photovoltaic power generation power prediction result is determined based on an average of the first photovoltaic power generation power prediction result and the second photovoltaic power generation power prediction result.

4. The photovoltaic power generation prediction method according to claim 3, wherein: The first photovoltaic power generation prediction model is obtained by training an optimized Transformer model, and the optimized Transformer model is obtained by replacing the decoder of the original Transformer model with a deep neural network DNN; the second photovoltaic power generation prediction model is obtained by training a CNN-GRU model.

5. The photovoltaic power generation prediction method according to claim 1, wherein: The photovoltaic power generation data includes photovoltaic array voltage, photovoltaic array current, grid-side three-phase voltage, grid-side three-phase current, power factor, power generation efficiency and photovoltaic power generation power.

6. A photovoltaic power generation power prediction device, characterized in that: The photovoltaic power generation power prediction device comprises: a data acquisition module, configured to acquire meteorological data and photovoltaic power generation data of a photovoltaic station within a first preset time period before a current moment; wherein the meteorological data includes a plurality of first meteorological data within the first preset time period and a plurality of second meteorological data within a second preset time period after the current moment; a power prediction module, configured to input all of the first meteorological data, all of the second meteorological data, and the photovoltaic power generation data into a photovoltaic power generation prediction model to obtain a photovoltaic power generation prediction result; the photovoltaic power generation prediction model is trained using sample data from a plurality of photovoltaic sites, wherein the sample data from different photovoltaic sites correspond to different sample moments, and each of the photovoltaic site sample data includes photovoltaic power generation data and meteorological data within a first preset sample time period before the corresponding sample moment, and meteorological data within a second preset sample time period after the sample moment; The first meteorological data and the second meteorological data both correspond to sampling times, and both the first meteorological data and the second meteorological data include total irradiance; The data acquisition module is further configured to, for each sampling moment, perform statistical processing on all the total irradiances within a preset time window in which the sampling moment is located to obtain first meteorological characteristic data; for each of the first meteorological data or the second meteorological data at the sampling moment, determine the sampling moment corresponding to the maximum total irradiance within the preset statistical time period in which the sampling moment is located as a reference moment; and for each of the sampling moments, obtain second meteorological characteristic data based on the time difference between the sampling moment and the reference moment; The power prediction module is further configured to input all of the first meteorological characteristic data, all of the second meteorological characteristic data, and the photovoltaic power generation data into a photovoltaic power generation prediction model to obtain a photovoltaic power generation power prediction result.

7. A photovoltaic power generation power prediction device, characterized in that: The photovoltaic power generation prediction device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the photovoltaic power generation prediction method according to any one of claims 1 to 5.

8. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the photovoltaic power generation power prediction method according to any one of claims 1 to 5 are implemented.

9. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the photovoltaic power prediction method according to any one of claims 1 to 5 are implemented.

Images