A method and system for short-time prediction of regional photovoltaic power generation

Abstract

This invention discloses a method and system for short-term prediction of regional photovoltaic power generation. The method includes the following steps: obtaining the regional type of a preset area where photovoltaic power generation devices are installed; selecting, based on the regional type, several second meteorological elements that play a dominant role in the short-term prediction of regional photovoltaic power generation from several first preset meteorological elements; obtaining the meteorological data corresponding to the several second meteorological elements in the preset area at the current time; and performing short-term prediction of regional photovoltaic power generation based on a short-term photovoltaic power generation prediction model and in combination with the meteorological data to obtain the output power value of the photovoltaic power generation device. By fully considering the impact of special meteorological conditions in different regions on photovoltaic power generation, compared with general prediction methods, this targeted processing can more accurately reflect the actual situation, making the prediction results closer to the actual output power value of the photovoltaic power generation device and reducing prediction errors.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation prediction technology, and in particular to a method and system for short-term prediction of regional photovoltaic power generation. Background Technology

[0002] Renewable energy is an essential alternative to depleted fossil fuels and a crucial pathway to reducing greenhouse gas emissions such as carbon dioxide and nitrogen oxides. However, solar energy is underdeveloped globally. The sun can be considered a giant natural fusion reactor, providing far more energy to Earth than humanity needs. Solar energy possesses significant advantages such as cleanliness, safety, efficiency, and sustainability, playing a vital role in my country's energy transition and serving as the primary renewable energy source. Currently, photovoltaic (PV) power generation is experiencing unprecedented development opportunities, offering a clean, quiet, and low-cost form of electricity. However, because PV power generation is correlated with weather conditions (such as solar irradiance, temperature, and cloud cover), the operation of PV power plants is challenging. This often leads to unpredictability and randomness in PV power generation, damaging grid stability. Correspondingly, the goal of PV power generation forecasting is to predict future PV power generation with the highest accuracy. PV power generation forecasting methods are mainly divided into two categories: physical model-based methods and data-based methods. The latter does not require predicting light amplitude, has simple modeling, and lower prediction costs, thus gaining widespread popularity. Generally, data-based methods include statistical time series analysis methods, traditional machine learning methods, and deep learning methods. This invention uses long short-term memory neural networks in deep learning methods for prediction.

[0003] Accurate photovoltaic (PV) power generation forecasting plays a crucial role in improving grid connection efficiency, reducing carbon emissions, and promoting the coordinated development of the economy and PV power generation. The land-atmosphere system has a significant impact on PV power generation, and different geographical and climatic parameters have varying quantitative effects on PV power generation forecasting. However, existing research lacks comprehensive coverage of a wide range of geographical and meteorological parameters. For example, the importance of meteorological elements differs significantly in coastal and plain areas characterized by high electricity demand, mountainous areas with relatively dispersed demand, or desert areas where supply exceeds demand.

[0004] Current research typically selects highly correlated meteorological elements from common meteorological factors such as radiation, temperature, wind speed, and cloud cover as model input parameters. A deep learning neural network model is then built, with training and testing sets divided. After multiple iterations of training and parameter adjustments, accuracy is tested. However, this parameter selection method does not consider the specific topographic and climatic conditions of a particular region, does not comprehensively cover and rank the correlation of meteorological elements across multiple levels, and lacks input of heterogeneous meteorological elements that are strongly correlated with local conditions such as altitude, topography, and climate. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for short-term prediction of regional photovoltaic power generation. By introducing a photovoltaic power generation prediction model based on meteorological element ranking, the model quantitatively ranks the correlation between a large number of meteorological elements at multiple levels in various fields of the region and photovoltaic power generation. It then selects important meteorological elements that are highly correlated with photovoltaic power generation in different regions as input parameters for the prediction model. This achieves a high correlation between photovoltaic power generation and the characteristics of the region, reduces prediction errors, and improves the accuracy of short-term forecasts.

[0006] To address the aforementioned technical problems, a first aspect of this invention provides a method for short-time prediction of regional photovoltaic power generation, comprising the following steps:

[0007] Obtain the region type of the preset area where photovoltaic power generation devices are installed;

[0008] Based on the aforementioned regional type, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation are selected from several first preset meteorological elements.

[0009] Obtain real-time meteorological data corresponding to the plurality of second meteorological elements in the preset area at the current time;

[0010] Based on the short-time photovoltaic power generation prediction model, and combined with the real-time meteorological data, short-time prediction of regional photovoltaic power generation is performed to obtain the output power value of the photovoltaic power generation device.

[0011] Furthermore, the selection of several second meteorological elements from several first preset meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation includes:

[0012] Based on the same sampling period, historical data of several first-preset meteorological elements of a preset regional type and the output power value of photovoltaic power generation device at the corresponding time are obtained respectively.

[0013] Based on historical data of the first preset meteorological element and the output power value of the photovoltaic power generation device, calculate the correlation coefficient between each of the first preset meteorological element and the output power value of the photovoltaic power generation device.

[0014] For each regional type, the correlation coefficients of the several first preset meteorological elements are sorted in descending order, and several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation for the regional type are selected.

[0015] Further, the calculation of the correlation coefficient between each of the first preset meteorological elements and the output power value of the photovoltaic power generation device includes:

[0016] The correlation coefficient between the first preset meteorological element and the output power value of the photovoltaic power generation device is calculated based on the Pearson correlation coefficient.

[0017] The correlation coefficient r j The calculation formula is:

[0018]

[0019] Where j is the sequence number of multiple first-preset meteorological elements, X i These are the serial numbers of several historical data points for the first preset meteorological element. Y is the average of several historical data of the first preset meteorological element. i The historical data of the output power value of the photovoltaic power generation device corresponding to the i-th first preset meteorological element, where n is the number of historical data of the first preset meteorological element.

[0020] Furthermore, in the short-time photovoltaic power generation prediction model, the weight coefficient value of the second meteorological element is associated with the regional type.

[0021] Furthermore, the regional types include: coastal regions, plain regions, mountainous regions, plateau regions, and desert regions;

[0022] The first preset meteorological elements include: evapotranspiration, relative humidity, heat flux, wind speed, air temperature, global horizontal irradiance, wind direction, daily rainfall, air pressure, hail accumulation, and solar radiation intensity.

[0023] Furthermore, the corresponding second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation in the coastal area include: evapotranspiration, relative humidity, daily rainfall, and solar radiation intensity;

[0024] The corresponding second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation in the plains region include: solar radiation intensity, temperature, wind direction, and daily rainfall;

[0025] The secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the mountainous region include: solar radiation intensity, wind direction, and temperature.

[0026] The corresponding secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the plateau region include: snowfall, hail accumulation, solar radiation intensity, and temperature.

[0027] The secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the desert region include: solar radiation intensity, dust storms, temperature, and heat flux.

[0028] Furthermore, the short-time photovoltaic power generation prediction model is a two-layer LSTM multi-step prediction model built under the Tensorflow architecture;

[0029] The two-layer LSTM multi-step prediction model includes: an input layer, a first LSTM unit, and a second LSTM unit;

[0030] The input layer receives real-time meteorological data corresponding to the plurality of second meteorological elements and sends it to the first LSTM unit;

[0031] The first LSTM unit extracts features from the real-time meteorological data and sends the extracted first data hidden state information to the second LSTM unit. The first data hidden state information includes: the trend of real-time meteorological data of the second meteorological element changing over time, the correlation between several second meteorological elements, and the correlation between the second meteorological element and the output power value of the photovoltaic power generation device.

[0032] The second LSTM unit receives the first data hiding state information and performs feature extraction again to obtain the second data hiding state information;

[0033] The dual-layer LSTM multi-step prediction model calculates the output power value of the photovoltaic power generation device based on the second data hidden state information.

[0034] The activation functions of the first LSTM unit and the second LSTM unit are both hyperbolic non-tangent functions.

[0035] Furthermore, the data training set for the short-time photovoltaic power generation prediction model includes: ECMWFReanalysis v5 meteorological reanalysis data.

[0036] Accordingly, a second aspect of the present invention provides a regional photovoltaic power generation short-time prediction system, comprising:

[0037] The region type acquisition module is used to acquire the region type of a preset region where photovoltaic power generation devices are installed;

[0038] The element type acquisition module is used to select, based on the region type, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation from several first preset meteorological elements.

[0039] The meteorological data acquisition module is used to acquire real-time meteorological data corresponding to the plurality of second meteorological elements in the preset area at the current time.

[0040] The photovoltaic power prediction module is used to make short-term predictions of regional photovoltaic power generation based on the short-time photovoltaic power generation prediction model and combined with the real-time meteorological data, so as to obtain the output power value of the photovoltaic power generation device.

[0041] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the above-described method for short-term prediction of regional photovoltaic power generation.

[0042] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described method for short-time prediction of regional photovoltaic power generation.

[0043] The above-described technical solutions of the embodiments of the present invention have the following beneficial technical effects:

[0044] 1. By selecting the dominant second meteorological element based on the preset regional type and combining it with real-time meteorological data for prediction, the impact of special meteorological conditions in different regions on photovoltaic power generation is fully considered. Compared with general prediction methods, this targeted approach can more accurately reflect the actual situation, making the prediction results closer to the actual output power value of photovoltaic power generation devices and reducing prediction errors.

[0045] 2. Adjust the input meteorological elements for different regional types so that the short-term photovoltaic power generation prediction model can adapt to a variety of complex geographical environments, such as coastal areas, mountains, and deserts. Regardless of the region, the model can focus on key meteorological factors and avoid prediction deviations caused by irrelevant or secondary factors, thereby improving the applicability and stability of the model in different regions.

[0046] 3. Short-term forecasts based on real-time meteorological data ensure that the forecast results can reflect the current and upcoming power generation situation in a timely manner. This is crucial for the real-time scheduling of photovoltaic power plants and the immediate balance adjustment of the power system. It can help operators make reasonable decisions quickly, such as the charging and discharging control of energy storage devices and the optimization of power output, thereby improving the operating efficiency of the entire photovoltaic power generation system. Attached Figure Description

[0047] Figure 1 This is a flowchart of the regional photovoltaic power generation short-time prediction method provided in an embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of the dual-layer LSTM multi-step prediction model architecture provided in an embodiment of the present invention;

[0049] Figure 3 This is a block diagram of a regional photovoltaic power generation short-time prediction system module provided in an embodiment of the present invention.

[0050] Figure label:

[0051] 1. Regional type acquisition module; 2. Element type acquisition module; 3. Meteorological data acquisition module; 4. Photovoltaic power prediction module. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0053] Please refer to Figure 1 The first aspect of this invention provides a method for short-time prediction of regional photovoltaic power generation, comprising the following steps:

[0054] Step S100: Obtain the region type of the preset region where photovoltaic power generation devices are installed.

[0055] Specifically, the regional types include: coastal regions, plain regions, mountain regions, plateau regions, and desert regions; the first preset meteorological elements include: evapotranspiration, relative humidity, heat flux, wind speed, temperature, global horizontal radiation, wind direction, daily rainfall, air pressure, hail accumulation, and solar radiation intensity.

[0056] In coastal areas, photovoltaic (PV) power generation devices are located near the ocean and are affected by the marine climate. These devices must withstand salt spray corrosion, strong winds (including typhoons), and high humidity. They are typically installed on coastal mudflats, dikes, or land a certain distance from the coast, and their layout must consider safety distances and marine environmental factors. In plains areas, the terrain is flat and open, with relatively uniform distribution of meteorological elements. PV power generation devices can be deployed on a large scale and in a centralized manner, making installation and maintenance convenient and fully utilizing land resources. They can also be integrated with agriculture and other industries. In mountainous areas, the complex terrain and varying solar radiation reception necessitate the layout of PV power generation devices. If these are mostly located on sunny slopes, construction and maintenance become more difficult, requiring consideration of terrain bearing capacity, shading, equipment transportation, and installation. In plateau areas, strong solar radiation is beneficial for increasing power generation, but the variable weather, such as rapid cloud movement and accumulation, can lead to unstable solar radiation. Strong winds may affect the stability of PV supports, while blizzards may cover PV modules, reducing the area receiving solar radiation and thus lowering power generation. Additionally, low temperatures may affect the performance of PV cells, reducing power generation efficiency. Desert regions have long hours of sunshine and strong solar radiation, but they also experience frequent sandstorms, little rainfall, and large temperature differences between day and night. They can make use of vast land and abundant sunshine.

[0057] Step S200: Based on the regional type, select several second meteorological elements from several first preset meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation.

[0058] The impact of meteorological factors on photovoltaic power generation varies across different regions. Therefore, in the short-time photovoltaic power generation prediction model, the weight coefficient of the second meteorological factor is related to the regional type. Different regional types result in different types and weight values ​​of the second meteorological factor for predicting power generation through the short-time photovoltaic power generation prediction model.

[0059] Step S300: Obtain real-time meteorological data corresponding to several second meteorological elements in the preset area at the current time.

[0060] Step S400: Based on the short-time photovoltaic power generation prediction model and combined with real-time meteorological data, short-time prediction of regional photovoltaic power generation is carried out to obtain the output power value of the photovoltaic power generation device.

[0061] By acquiring real-time meteorological data of key secondary meteorological elements in a preset region at the current moment, and conducting short-time photovoltaic power generation forecasting based on a short-time photovoltaic power generation prediction model to obtain output power values, real-time information can be fully utilized, improving the timeliness and accuracy of forecasts. By utilizing multiple secondary meteorological elements that have a dominant influence on power generation, unnecessary data processing and model complexity are reduced, thus improving forecasting efficiency.

[0062] Specifically, in step S200, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation are selected from several first preset meteorological elements, including:

[0063] Step S210: Based on the same sampling period, acquire historical data of several first preset meteorological elements of a preset regional type and the output power value of the photovoltaic power generation device at the corresponding time.

[0064] Step S220: Based on the historical data of the first preset meteorological element and the output power value of the photovoltaic power generation device, calculate the correlation coefficient between each first preset meteorological element and the output power value of the photovoltaic power generation device.

[0065] By calculating the correlation coefficient between the first preset meteorological element and the output power value of the photovoltaic power generation device, the degree of influence of each meteorological element on the power generation can be accurately determined. This data-driven method eliminates the limitations of subjective judgment, ensuring that the selected second meteorological elements are closely related to the power generation, thus allowing the subsequent short-term forecasting model to focus on key factors and greatly improving the pertinence of the forecast.

[0066] Step S230: For each regional type, sort the correlation coefficients of several first preset meteorological elements in descending order, and select several second meteorological elements that play a dominant role in the short-term forecast of regional photovoltaic power generation for each regional type.

[0067] Selecting dominant factors based on correlation coefficients avoids incorporating large amounts of irrelevant or weakly correlated meteorological data into the prediction model, simplifying the model structure, reducing data processing volume and computational complexity, and lowering the risk of overfitting. The optimized model can utilize computational resources more efficiently, accelerate prediction speed, and more accurately reflect the intrinsic relationship between meteorological factors and power generation, improving prediction accuracy and reliability.

[0068] Correlation analysis and element selection were conducted separately for each regional type, fully considering the unique meteorological characteristics and environmental conditions of different regions (such as coastal areas, plains, mountains, deserts, and plateaus). This allows the prediction model to be customized according to the actual conditions of different regions, better adapting to regional differences. Regardless of the complex geographical environment, it can effectively capture key meteorological factors affecting photovoltaic power generation, further improving the universality and accuracy of regional photovoltaic power generation short-term prediction in different regions.

[0069] Further, in step S220, the correlation coefficient between each first preset meteorological element and the output power value of the photovoltaic power generation device is calculated, including:

[0070] The correlation coefficient between the first preset meteorological element and the output power value of the photovoltaic power generation device was calculated based on the Pearson correlation coefficient.

[0071] Correspondingly, the correlation coefficient r j The calculation formula is:

[0072]

[0073] Where j is the sequence number of multiple first-preset meteorological elements, X i These are the serial numbers of several historical data points for the first preset meteorological element. Y is the average of several historical data of the first preset meteorological element. i The historical data of the output power value of the photovoltaic power generation device corresponding to the i-th first preset meteorological element, where n is the number of historical data of the first preset meteorological element.

[0074] Correspondingly, the second meteorological element playing a dominant role in short-term forecasting of regional photovoltaic (PV) power generation in coastal areas includes evapotranspiration, relative humidity, daily rainfall, and solar radiation intensity. Evapotranspiration intensity can serve as an indirect indicator of weather conditions and solar radiation levels. Strong evapotranspiration may indicate sunny weather and ample solar radiation, leading to an upward trend in PV power generation. Simultaneously, higher evapotranspiration may be accompanied by wind speeds, which are beneficial for heat dissipation from PV panels, improving their power generation efficiency. Higher relative humidity increases the water vapor content in the air, and water vapor absorbs and scatters solar radiation. When relative humidity rises, the intensity of solar radiation reaching the PV panels may weaken, resulting in a decrease in power generation. However, a sharp drop in relative humidity may indicate sunny weather, which is conducive to increased power generation. During rainfall, thick cloud cover and reduced solar radiation lead to a decrease in PV power generation. For PV power generation forecasting in coastal areas, the amount and temporal distribution of daily rainfall are important factors. If a large amount of rainfall is predicted for a given day, it can be determined that PV power generation will significantly decrease during the rainfall period. Solar intensity is one of the most direct factors determining the power output of photovoltaic (PV) power generation in coastal areas. Stronger solar intensity means that PV cells can receive more light energy for power generation, and the power output will increase accordingly. When making short-term forecasts, it is necessary to consider the impact of marine environmental factors (such as sea surface reflection, water vapor, sea fog, etc.) on solar intensity.

[0075] Similarly, the second meteorological element that plays a dominant role in short-term forecasting of regional photovoltaic (PV) power generation in plain areas includes: solar radiation intensity, temperature, wind direction, and daily precipitation. The second meteorological element that plays a dominant role in short-term forecasting of regional PV power generation in mountainous areas includes: solar radiation intensity, wind direction, and temperature. The second meteorological element that plays a dominant role in short-term forecasting of regional PV power generation in plateau areas includes: snowfall, hail accumulation, solar radiation intensity, and temperature. The second meteorological element that plays a dominant role in short-term forecasting of regional PV power generation in desert areas includes: solar radiation intensity, dust storms, temperature, and heat flux.

[0076] Please refer to Figure 2 In one specific embodiment of this invention, the short-time photovoltaic power generation prediction model is a two-layer LSTM multi-step prediction model built on the Tensorflow architecture. The Adam optimizer is used to reasonably set and tune parameters such as learning rate, batch size, and epochs, and the accuracy is evaluated using a test set. R... 2 The RMSE index is used to quantitatively evaluate the accuracy and magnitude of prediction errors.

[0077] Long Short-Term Memory (LSTM) is a recurrent neural network (RNN) algorithm used in deep learning, which performs better for processing time series data. LSTM inherits the recursive characteristics of RNNs while making full use of time series data, compensating for the shortcomings of RNNs such as vanishing and exploding gradients, as well as their insufficient long-term memory capacity. Its basic structure mainly includes an input gate, a forget gate, an output gate, and a memory unit. Where x... t Indicates the input, y t Indicates output, h t and c t It can be considered as a short-term state and a long-term state, g t It is a candidate value. The input gate accepts input information and updates the state of the memory cell according to different conditions. The forget gate determines the information to be discarded based on specific conditions, and the output gate determines the output content based on the input information and the memory cell. The working process of the LSTM cell is as follows: At each time step, the forget gate receives the current state x. t and the hidden layer state h of the previous time step t-1 The output of the forget gate is mapped to the interval [0,1] through the activation function σ. When the output of the forget gate is 0, it means that the information from the previous state is completely discarded; when the output is 1, the information from the previous state is fully retained. The input of the input gate is transformed by a nonlinear function and then superimposed with the output of the forget gate to obtain the updated memory unit c. t Finally, the output gate operates based on the nonlinear function, according to c.t Dynamically control the output h of the LSTM t (y t The calculation formulas between the variables are shown below:

[0078] i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i );

[0079] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0080] o t =σ(W xo x t +W ho h t-1 +W co c t +b o );

[0081] g t =tanh(W xc x t +W hc h t-1 +b c );

[0082] c(t) = f t c t-1 +i t g t ;

[0083] h t =y t =o t tanh(c t ).

[0084] In weather forecasting, LSTM (Laser-Based Stroke Module) models can be built using past weather data and trained to predict future weather conditions. Short-term weather forecasting (STM) refers to predicting weather changes over the next few hours. Compared to traditional weather forecasting methods, LSTM-based STM forecasts can more accurately capture weather trends, especially when weather changes are drastic within a short period, thus improving forecast accuracy. By iteratively training the LSTM model, it is ensured that it can accurately capture dynamic changes and potential patterns in the data, enabling short-term predictions of photovoltaic power generation (within 72 hours) based on meteorological factors such as evaporation in different regions.

[0085] Specifically, first define the dimensions of the model's input layer to enable it to receive preprocessed meteorological data and historical power generation data. For example, if there are five types of meteorological data (wind speed, temperature, global horizontal irradiance, daily rainfall, and wind direction) and historical power generation data, and the time series length is t, then the dimension of the input layer might be set to (t, 6), where 6 represents the five types of meteorological data plus the power generation data. First LSTM layer: Set the number of units in the first LSTM layer (e.g., 128 units). It will receive the data from the input layer and process it. During processing, the LSTM units will update the cell state and hidden state according to their internal structure (including input gates, forget gates, output gates, etc.), thereby extracting preliminary features of the data. Second LSTM layer: Set the number of units in the second LSTM layer (e.g., 64 units). It will receive the results processed by the first LSTM layer. Similarly, it will process the data through its internal structure to further mine deeper features of the data. The dimension of the model output layer is set according to the prediction step size. If multi-step prediction is performed, such as predicting the photovoltaic power generation in the next 3 hours, then the dimension of the output layer is set to 3 to output the predicted values ​​at the next 3 time points.

[0086] In addition, to eliminate the influence of different seasons, data from the first 21 days of each month were used for training, and data from the remaining days were used for testing, to ensure the reliability and accuracy of the model's predictions of changes in surface shortwave radiation parameters.

[0087] Specifically, the two-layer LSTM multi-step prediction model includes: an input layer, a first LSTM unit, and a second LSTM unit.

[0088] The input layer receives real-time meteorological data corresponding to several second meteorological elements and sends it to the first LSTM unit. The first LSTM unit extracts features from the real-time meteorological data and sends the extracted first data hidden state information to the second LSTM unit. The first data hidden state information includes: the time-varying trend of the real-time meteorological data of the second meteorological elements, the correlation between several second meteorological elements, and the correlation between the second meteorological elements and the output power value of the photovoltaic power generation device. The second LSTM unit receives the first data hidden state information and performs feature extraction again to obtain the second data hidden state information. The two-layer LSTM multi-step prediction model calculates the output power value of the photovoltaic power generation device based on the second data hidden state information.

[0089] The activation functions of both the first LSTM unit and the second LSTM unit are hyperbolic non-tangent functions.

[0090] Furthermore, the training set for the short-term photovoltaic power generation prediction model includes ECMWF Reanalysis v5 meteorological reanalysis data. Historical reanalysis data is used as input parameters. After integrating and designing systematic meteorological parameters from multiple fields, meteorological parameters are initially screened for different regions, ranked by correlation coefficient, and several meteorological parameters are selected according to their correlation from highest to lowest (or proportionally). Meteorological elements with high correlation to photovoltaic power generation are selected as input parameters for their respective prediction models. The output data consists of 15-minute resolution power generation data for photovoltaic power plants under different meteorological conditions in different regions. Through dataset adjustment, data preprocessing, training and test set sample division, model construction, and iterative training, power generation predictions for 0-72 hours are performed, thereby achieving short-term, high-precision forecasts of photovoltaic power generation in different regions.

[0091] Supported by a large volume of historical meteorological data from various fields and at multiple levels, the aforementioned method for short-term photovoltaic power generation prediction utilizes an LSTM deep learning model to more effectively combine the altitude, topography, and climate conditions of different regions. It also addresses regional heterogeneity by avoiding the loss of some highly correlated meteorological elements that serve as important input parameters for the model, thus more accurately predicting the short-term power generation of photovoltaic power plants in real-world application scenarios. This method can achieve high-precision predictions for short periods of 0-72 hours, providing technical support and reference for photovoltaic power generation grid connection technology and grid decision-making.

[0092] Accordingly, please refer to Figure 3 A second aspect of the present invention provides a regional photovoltaic power generation short-time prediction system, comprising:

[0093] Region type acquisition module 1 is used to acquire the region type of a preset region where photovoltaic power generation devices are installed;

[0094] The element type acquisition module 2 is used to select, based on the regional type, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation from several first preset meteorological elements.

[0095] Meteorological data acquisition module 3 is used to acquire meteorological data corresponding to several second meteorological elements in a preset area at the current time.

[0096] Photovoltaic power prediction module 4 is used to make short-term predictions of regional photovoltaic power generation based on a short-time photovoltaic power generation prediction model and combined with meteorological data, so as to obtain the output power value of the photovoltaic power generation device.

[0097] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the aforementioned regional photovoltaic power generation short-time prediction method.

[0098] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described method for short-time prediction of regional photovoltaic power generation.

[0099] This invention aims to protect a method and system for short-term prediction of regional photovoltaic power generation. The method includes the following steps: obtaining the regional type of a preset area where photovoltaic power generation devices are installed; selecting, based on the regional type, several second meteorological elements that play a dominant role in the short-term prediction of regional photovoltaic power generation from several first preset meteorological elements; obtaining the meteorological data corresponding to the several second meteorological elements in the preset area at the current time; and performing short-term prediction of regional photovoltaic power generation based on a short-term photovoltaic power generation prediction model and the meteorological data to obtain the output power value of the photovoltaic power generation device. The above technical solution has the following effects:

[0100] 1. By selecting the dominant second meteorological element based on the preset regional type and combining it with real-time meteorological data for prediction, the impact of special meteorological conditions in different regions on photovoltaic power generation is fully considered. Compared with general prediction methods, this targeted approach can more accurately reflect the actual situation, making the prediction results closer to the actual output power value of photovoltaic power generation devices and reducing prediction errors.

[0101] 2. Adjust the input meteorological elements for different regional types so that the short-term photovoltaic power generation prediction model can adapt to a variety of complex geographical environments, such as coastal areas, mountains, and deserts. Regardless of the region, the model can focus on key meteorological factors and avoid prediction deviations caused by irrelevant or secondary factors, thereby improving the applicability and stability of the model in different regions.

[0102] 3. Short-term forecasts based on real-time meteorological data ensure that the forecast results can reflect the current and upcoming power generation situation in a timely manner. This is crucial for the real-time scheduling of photovoltaic power plants and the immediate balance adjustment of the power system. It can help operators make reasonable decisions quickly, such as the charging and discharging control of energy storage devices and the optimization of power output, thereby improving the operating efficiency of the entire photovoltaic power generation system.

[0103] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0105] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for short-time prediction of regional photovoltaic power generation, characterized in that, Includes the following steps: Obtain the region type of the preset area where photovoltaic power generation devices are installed; Based on the aforementioned regional type, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation are selected from several first preset meteorological elements. Obtain real-time meteorological data corresponding to the plurality of second meteorological elements in the preset area at the current time; Based on the short-time photovoltaic power generation prediction model, combined with the real-time meteorological data, short-time prediction of regional photovoltaic power generation is carried out to obtain the output power value of the photovoltaic power generation device. The selection of several second meteorological elements from several first preset meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation includes: Based on the same sampling period, historical data of several first-preset meteorological elements of a preset regional type and the output power value of photovoltaic power generation device at the corresponding time are obtained respectively. Based on historical data of the first preset meteorological element and the output power value of the photovoltaic power generation device, calculate the correlation coefficient between each of the first preset meteorological element and the output power value of the photovoltaic power generation device. For each regional type, the correlation coefficients of several first preset meteorological elements are sorted in descending order, and several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation for the regional type are selected. The short-time photovoltaic power generation prediction model is a two-layer LSTM multi-step prediction model built under the Tensorflow architecture. The two-layer LSTM multi-step prediction model includes: an input layer, a first LSTM unit, and a second LSTM unit; The input layer receives real-time meteorological data corresponding to the plurality of second meteorological elements and sends it to the first LSTM unit; The first LSTM unit extracts features from the real-time meteorological data and sends the extracted first data hidden state information to the second LSTM unit. The first data hidden state information includes: the real-time meteorological data of the second meteorological element changing trend over time, the correlation between several second meteorological elements, and the correlation between the second meteorological element and the output power value of the photovoltaic power generation device. The second LSTM unit receives the first data hiding state information and performs feature extraction again to obtain the second data hiding state information; The dual-layer LSTM multi-step prediction model calculates the output power value of the photovoltaic power generation device based on the second data hidden state information. The activation functions of the first LSTM unit and the second LSTM unit are both hyperbolic non-tangent functions.

2. The method for short-time prediction of regional photovoltaic power generation according to claim 1, characterized in that, The calculation of the correlation coefficient between each of the first preset meteorological elements and the output power value of the photovoltaic power generation device includes: The correlation coefficient between the first preset meteorological element and the output power value of the photovoltaic power generation device is calculated based on the Pearson correlation coefficient. The correlation coefficient The calculation formula is: ； in, These are the sequence numbers of multiple pre-defined meteorological elements. These are the serial numbers of several historical data points for the first preset meteorological element. This is the average of several historical data points for the first preset meteorological element. In order to be with the first Historical data on the output power of photovoltaic power generation devices corresponding to the first preset meteorological element. This is the average of several historical data points for the output power of the photovoltaic power generation device corresponding to the first preset meteorological element. The amount of historical data for the first preset meteorological element.

3. The method for short-time prediction of regional photovoltaic power generation according to claim 1, characterized in that, In the short-time photovoltaic power generation prediction model, the weight coefficient value of the second meteorological element is associated with the regional type.

4. The method for short-time prediction of regional photovoltaic power generation according to claim 1, characterized in that, The geographical types include: coastal regions, plain regions, mountain regions, plateau regions, and desert regions; The first preset meteorological elements include: evapotranspiration, relative humidity, heat flux, wind speed, air temperature, global horizontal irradiance, wind direction, daily rainfall, air pressure, hail accumulation, snowfall, dust storms, and solar radiation intensity.

5. The method for short-time prediction of regional photovoltaic power generation according to claim 4, characterized in that, The corresponding second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation in the coastal areas include: evapotranspiration, relative humidity, daily rainfall, and solar radiation intensity. The corresponding second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation in the plains region include: solar radiation intensity, temperature, wind direction, and daily rainfall; The secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the mountainous region include: solar radiation intensity, wind direction, and temperature. The corresponding secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the plateau region include: snowfall, hail accumulation, solar radiation intensity, and temperature. The secondary meteorological elements that play a dominant role in short-term forecasting of regional photovoltaic power generation in the desert region include: solar radiation intensity, dust storms, temperature, and heat flux.

6. The method for short-time prediction of regional photovoltaic power generation according to claim 1, characterized in that, The training set for the short-time photovoltaic power generation prediction model includes: ECMWF Reanalysis v5 meteorological reanalysis data.

7. A regional photovoltaic power generation short-time prediction system, characterized in that, Predicting regional photovoltaic power generation based on the short-time prediction method for regional photovoltaic power generation as described in any one of claims 1-6 includes: The region type acquisition module is used to acquire the region type of a preset region where photovoltaic power generation devices are installed; The element type acquisition module is used to select, based on the region type, several second meteorological elements that play a dominant role in the short-term forecasting of regional photovoltaic power generation from several first preset meteorological elements. The meteorological data acquisition module is used to acquire real-time meteorological data corresponding to the plurality of second meteorological elements in the preset area at the current time. The photovoltaic power prediction module is used to make short-term predictions of regional photovoltaic power generation based on the short-time photovoltaic power generation prediction model and combined with the real-time meteorological data, so as to obtain the output power value of the photovoltaic power generation device.

8. An electronic device, characterized in that, include: At least one processor; And a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the regional photovoltaic power generation short-time prediction method as described in any one of claims 1-6.

Images