Multi-source meteorological driving regional photovoltaic power generation power prediction method and system

Abstract

The application discloses a multi-source meteorological driving regional photovoltaic power generation power prediction method and system, relates to the photovoltaic power generation power prediction technical field, and the method comprises the steps of obtaining multi-source meteorological data of a target region, constructing a meteorological space-time grid; constructing a photovoltaic space grid, and determining a capacity weight; establishing a power grid mapping relationship between the meteorological space-time grid and the photovoltaic space grid; performing each time scale power prediction to obtain the predicted power value of each grid unit; and performing weighted aggregation on the predicted power value of each grid unit to obtain the regional photovoltaic power generation power prediction result.The application solves the technical problems that the multi-source meteorological data is not fully utilized, the different time scale prediction results are discontinuous, and the prediction accuracy and stability are insufficient under complex weather in the prior art, achieves the scale prediction of the regional photovoltaic power generation power, and effectively improves the stability and accuracy of the regional photovoltaic power prediction under complex weather conditions.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation prediction technology, specifically to a method and system for predicting regional photovoltaic power generation driven by multiple meteorological sources. Background Technology

[0002] In recent years, with the large-scale and clustered development of regional photovoltaic power plants, photovoltaic power generation has been significantly affected by fluctuations in meteorological conditions. Traditional single-point forecasting methods are difficult to adapt to the need for unified modeling of photovoltaic power plants across a large area. Existing technologies generally suffer from problems such as insufficient fusion of multi-source meteorological data, weak spatiotemporal feature extraction capabilities, and low matching degree between photovoltaic spatial distribution and meteorological grids. Furthermore, forecast results at different time scales are prone to time-series jumps and discontinuous connections, and the accuracy of regional power aggregation is insufficient. As a result, the accuracy, stability, and practicality of regional photovoltaic power generation forecasting under complex meteorological conditions cannot meet the actual engineering requirements of grid dispatching and new energy consumption.

[0003] Existing technologies suffer from insufficient utilization of multi-source meteorological data and discontinuous prediction results at different time scales, leading to technical problems such as inadequate prediction accuracy and stability under complex weather conditions. Summary of the Invention

[0004] This application provides a method and system for predicting regional photovoltaic power generation driven by multiple meteorological sources, which is used to address the technical problems in the prior art, such as insufficient utilization of multiple meteorological data and discontinuous prediction results at different time scales, resulting in insufficient prediction accuracy and stability under complex weather conditions.

[0005] In view of the above problems, this application provides a method and system for predicting regional photovoltaic power generation driven by multiple meteorological sources.

[0006] A first aspect of this application provides a multi-source meteorological-driven method for predicting regional photovoltaic power generation, the method comprising: Multi-source meteorological data of the target area is acquired, and spatiotemporal scale features are extracted and aggregated from the multi-source meteorological data to construct a meteorological spatiotemporal grid. Based on the structural distribution and operation data of photovoltaic power plants, a photovoltaic spatial grid is constructed, and capacity weights are determined according to the photovoltaic installed capacity within each grid. A power grid mapping relationship is established between the meteorological spatiotemporal grid and the photovoltaic spatial grid. According to the power grid mapping relationship, power prediction at each time scale is performed using a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid unit. The predicted power values ​​of each grid unit are weighted and aggregated according to the capacity weights to obtain the predicted power generation result of the regional photovoltaic power generation.

[0007] A second aspect of this application provides a multi-source weather-driven regional photovoltaic power generation prediction system, the system comprising: The system includes a meteorological spatiotemporal grid construction module, which acquires multi-source meteorological data for the target area, extracts and aggregates spatiotemporal scale features from the multi-source meteorological data, and constructs a meteorological spatiotemporal grid. A photovoltaic spatial grid construction module is used to construct a photovoltaic spatial grid based on the structural distribution and operational data of photovoltaic power plants, and determines capacity weights according to the photovoltaic installed capacity within each grid. A power grid mapping relationship establishment module is used to establish a power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid. A predicted power value acquisition module is used to predict power at various time scales using a pre-constructed multi-scale power prediction model based on the power grid mapping relationship, obtaining the predicted power value for each grid unit. A power generation prediction result acquisition module is used to weighted aggregate the predicted power values ​​of each grid unit according to the capacity weights to obtain the predicted power generation result for the regional photovoltaic system.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: This method acquires multi-source meteorological data for the target area and constructs a meteorological spatiotemporal grid; it also constructs a photovoltaic spatial grid and determines capacity weights based on the installed photovoltaic capacity within each grid; establishes a power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid; uses a pre-constructed multi-scale power prediction model to predict power at various time scales, obtaining the predicted power value for each grid unit; and weights and aggregates the predicted power values ​​of each grid unit according to the capacity weights to obtain the predicted power generation of the regional photovoltaic system. This achieves large-scale prediction of regional photovoltaic power generation and effectively improves the stability and accuracy of regional photovoltaic power prediction under complex meteorological conditions. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A schematic diagram of the multi-source meteorological-driven regional photovoltaic power generation prediction method provided in this application embodiment; Figure 2 A schematic diagram of the structure of a multi-source weather-driven regional photovoltaic power generation prediction system provided in this application embodiment.

[0011] Figure labeling: Meteorological spatiotemporal grid construction module 10, photovoltaic spatial grid construction module 20, power grid mapping relationship establishment module 30, predicted power value acquisition module 40, power generation prediction result acquisition module 50. Detailed Implementation

[0012] This application provides a method and system for predicting regional photovoltaic power generation driven by multiple meteorological sources, which addresses the technical problems in the prior art where insufficient utilization of multiple meteorological data and discontinuous prediction results at different time scales lead to insufficient prediction accuracy and stability under complex weather conditions.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] Example 1, as Figure 1 As shown, this application provides a method for predicting regional photovoltaic power generation driven by multiple meteorological sources, the method comprising: Step S100: Obtain multi-source meteorological data of the target area, extract and aggregate the spatiotemporal scale features of the multi-source meteorological data, and construct a meteorological spatiotemporal grid.

[0015] Specifically, the process involves acquiring multi-source meteorological data for the target area, including numerical weather prediction data, satellite cloud imagery data, ground observation data, and all-sky imager data. This type of data contains meteorological information with different spatial and temporal resolutions. First, the multi-source meteorological data undergoes unified temporal processing, including resampling, missing value imputation, and outlier removal according to a preset time step. Then, the target area is spatially gridded according to a preset spatial resolution to establish meteorological basic grid units. Meteorological data with different spatial resolutions are mapped to the corresponding meteorological basic grid units. Multi-source meteorological data mapped to the same meteorological basic grid unit are then fused and aggregated to generate unified meteorological basic grid data. Finally, spatial statistical features and temporal variation features are extracted from the meteorological basic grid data, and these two types of features are combined and aggregated to generate a meteorological spatiotemporal grid feature set, thus completing the construction of the meteorological spatiotemporal grid.

[0016] Step S200: Based on the structural distribution and operation data of photovoltaic power plants, construct a photovoltaic spatial grid and determine the capacity weight according to the photovoltaic installed capacity in each grid.

[0017] Specifically, the target area is divided into regular grids according to a preset spatial resolution to establish initial photovoltaic spatial grid units. Each photovoltaic power station is accurately mapped to its corresponding initial photovoltaic spatial grid unit based on its geographical coordinates. At the same time, correlation analysis is conducted by combining the historical power time series of photovoltaic power stations to complete the grouping of power stations. Based on the photovoltaic grouping relationship, the initial photovoltaic spatial grid is calibrated and reconstructed to ensure that photovoltaic power stations within the same photovoltaic spatial grid are in the same group, and that photovoltaic power stations within the same group have consistent power response characteristics under similar meteorological conditions. This completes the construction of the photovoltaic spatial grid. On this basis, the photovoltaic installed capacity data within each photovoltaic spatial grid is statistically analyzed, and the capacity weight corresponding to each grid is determined based on this data.

[0018] Step S300: Establish the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid.

[0019] Specifically, the mapping weight coefficients between the meteorological spatiotemporal grid and the photovoltaic spatial grid are determined based on their spatial positional relationship. First, the spatial intersection weight, the spatial distance attenuation weight based on the preset distance attenuation function, and the upwind dynamic influence weight determined by the real-time wind direction are calculated separately between the meteorological spatiotemporal grid units and the photovoltaic spatial grid units. Then, the final mapping weight coefficients are determined by selecting a single weight or by weighting / producting multiple weights and normalizing them. A mapping weight matrix from the meteorological spatiotemporal grid units to the photovoltaic spatial grid units, which is dynamically updated with the real-time wind direction, is constructed. Based on these mapping weight coefficients, the meteorological features in the meteorological spatiotemporal grid are accurately mapped to the corresponding photovoltaic spatial grids, generating a set of meteorological driving features for each photovoltaic spatial grid. This establishes the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid.

[0020] Step S400: Based on the power grid mapping relationship, perform power prediction at each time scale using a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid cell.

[0021] Specifically, based on the established power grid mapping relationship, the meteorological driving features corresponding to the photovoltaic spatial grid units are extracted. At the same time, the installed capacity features of each grid unit and the historical power time series features of the photovoltaic power station are combined to form a comprehensive input feature set, which is then input into a pre-constructed multi-scale power prediction model. This model includes ultra-short-term, short-term, and medium-to-long-term prediction models, which can realize synchronous power prediction at multiple time scales. After the model outputs the initial predicted power values ​​of each photovoltaic spatial grid unit at different time scales, the prediction results at each time scale are continuously corrected or smoothed to eliminate prediction deviations between adjacent time scales, so as to keep the results consistent. Finally, the accurate predicted power values ​​of each grid unit at different time scales are determined.

[0022] Step S500: The predicted power values ​​of each grid cell are weighted and aggregated according to the capacity weight to obtain the predicted power generation results of the regional photovoltaic system.

[0023] Specifically, based on the capacity weight determined by the installed capacity within each photovoltaic spatial grid, the predicted power values ​​of each photovoltaic spatial grid unit at different time scales are weighted and aggregated. The predicted power values ​​of each grid unit are weighted and integrated through the capacity weights, and finally the photovoltaic power generation prediction results of the target area as a whole at the corresponding time scale are obtained.

[0024] In one possible implementation, step S300 further includes: Step S310: Determine the mapping weight coefficient between the grids based on the spatial positional relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid. The mapping weight coefficient is obtained based on one or more of the following: spatial intersection ratio, spatial distance attenuation relationship, or upwind sampling area determined by real-time wind direction.

[0025] Step S320: Based on the mapping weight coefficient, the meteorological features in the meteorological spatiotemporal grid are mapped to the corresponding photovoltaic spatial grid, generating a set of meteorological driving features corresponding to the photovoltaic spatial grid, and establishing the power grid mapping relationship.

[0026] Specifically, based on the spatial correspondence between meteorological spatiotemporal grid units in the meteorological spatiotemporal grid and photovoltaic spatial grid units in the photovoltaic spatial grid, the mapping weight coefficients between grids are determined. First, the spatial intersection weight is calculated by the proportion of the spatial overlap area between the meteorological spatiotemporal grid units and the photovoltaic spatial grid units. Then, based on the spatial distance between the center points of the two grid units, the spatial distance attenuation weight is obtained by combining a preset distance attenuation function. At the same time, the upstream direction of the photovoltaic spatial grid unit is determined according to the real-time wind direction, and a sampling corridor of a preset size is constructed. The meteorological spatiotemporal grid units in the corridor are assigned upwind dynamic influence weights. Finally, one of the weights corresponding to the upwind sampling area determined by the real-time wind direction is selected as the mapping weight coefficient, or multiple weights are combined, weighted / producted, and normalized before being used as the mapping weight coefficients between grids.

[0027] Using the established inter-grid mapping weight coefficients as the core quantification basis, the meteorological features extracted from each unit of the meteorological spatiotemporal grid are precisely weighted and mapped to the corresponding photovoltaic spatial grid units according to the weight allocation rules. The multi-source meteorological features mapped to the same photovoltaic spatial grid unit are integrated and normalized. The meteorological feature dimensions and feature values ​​corresponding to each photovoltaic spatial grid unit are systematically sorted out, and a meteorological driving feature set with uniform dimensions for each photovoltaic spatial grid is generated. Through this feature set, the feature layers of the meteorological spatiotemporal grid and the photovoltaic spatial grid are accurately correlated, and finally a quantifiable and traceable power grid mapping relationship between the two is established.

[0028] In one possible implementation, step S310 further includes: Step S311: Obtain the spatial intersection weight based on the proportion of spatial overlap area between the meteorological spatiotemporal grid units in the meteorological spatiotemporal grid and the photovoltaic spatial grid units in the photovoltaic spatial grid.

[0029] Step S312: Calculate the spatial distance attenuation weight based on the spatial distance between the center points of the meteorological spatiotemporal grid unit and the photovoltaic spatial grid unit, using a preset distance attenuation function.

[0030] Step S313: Determine the upstream direction of the photovoltaic spatial grid unit based on the real-time wind direction, and construct a sampling corridor with a preset length and width in the upstream direction. Assign upwind dynamic influence weights to the meteorological spatiotemporal grid units located in the sampling corridor.

[0031] Step S314: Determine the mapping weight coefficient using one or more of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight.

[0032] Specifically, spatial overlay analysis is performed on each meteorological spatiotemporal grid cell in the meteorological spatiotemporal grid and each photovoltaic spatial grid cell in the photovoltaic spatial grid. The spatial overlap area between the two is accurately calculated. Then, the ratio of the overlap area to the total area of ​​the corresponding photovoltaic spatial grid cell is calculated. This ratio is used as the spatial intersection weight between the meteorological spatiotemporal grid cell and the photovoltaic spatial grid cell, thereby quantifying the influence weight of the spatial overlap degree of the two grid cells on the meteorological feature mapping.

[0033] First, the geographic center coordinates of each meteorological spatiotemporal grid unit in the meteorological spatiotemporal grid and each photovoltaic spatial grid unit in the photovoltaic spatial grid are extracted. The actual spatial distance between the center points of each pair of grid units is accurately calculated. Then, this spatial distance is substituted into a preset distance attenuation function for quantitative calculation. The spatial distance attenuation weight of each meteorological spatiotemporal grid unit to the corresponding photovoltaic spatial grid unit is obtained through function operation. This quantitatively characterizes the degree to which meteorological influence attenuates with the increase of spatial distance, and realizes the weight assignment of the distance dimension.

[0034] First, real-time wind direction monitoring data of the target area is acquired. Based on this wind direction data, the meteorological upstream direction of each photovoltaic spatial grid unit in the photovoltaic spatial grid is accurately determined. Then, according to the preset length and width parameters, a dedicated meteorological sampling corridor is constructed in the upstream direction corresponding to each photovoltaic spatial grid unit to delineate the effective influence range of meteorological features. Subsequently, for meteorological spatiotemporal grid units that fall within the sampling corridor range in the meteorological spatiotemporal grid, the corresponding upwind dynamic influence weight is assigned according to their position in the corridor. Meteorological spatiotemporal grid units that do not fall within the sampling corridor are not assigned this weight. This is to conform to the actual law of meteorological elements spreading with wind direction and realize the dynamic weight assignment of wind field dimension.

[0035] Based on the actual needs of meteorological characteristics, photovoltaic layout, and power prediction in the target area, one of the calculated spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight is selected and directly used as the mapping weight coefficient between meteorological spatiotemporal grid units and photovoltaic spatial grid units. If multiple weights are selected, the selected weights are weighted or multiplied, and the calculation results are normalized to construct a mapping weight matrix from meteorological spatiotemporal grid units to photovoltaic spatial grid units. The element values ​​in this matrix are used as the final mapping weight coefficients to ensure the rationality and adaptability of the weight assignment.

[0036] In one possible implementation, step S314 further includes: When one of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight is used, the weight used is used as the mapping weight coefficient.

[0037] When multiple weights among the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight are used, the multiple weights are weighted and combined or multiplied and normalized to construct a mapping weight matrix from meteorological spatiotemporal grid unit to photovoltaic spatial grid unit, wherein the element values ​​of the mapping weight matrix are used as the mapping weight coefficients.

[0038] Specifically, when determining the mapping weight coefficients between meteorological spatiotemporal grid units and photovoltaic spatial grid units, if any one of the following is selected—spatial intersection weight, spatial distance attenuation weight, or upwind dynamic influence weight—the calculation result of this type of weight can be directly assigned to the weight association dimension of the meteorological spatiotemporal grid unit and the photovoltaic spatial grid unit, and used as the mapping weight coefficient corresponding to the two. No additional weight combination calculation and normalization processing are required; the quantification result of this single weight can be directly used as the weight basis for the inter-grid meteorological feature mapping.

[0039] When determining the mapping weight coefficients between meteorological spatiotemporal grid units and photovoltaic spatial grid units, if multiple weights among spatial intersection weights, spatial distance attenuation weights, and upwind dynamic influence weights are used simultaneously, a weighted combination or product combination operation is first performed on the selected weights according to the meteorological characteristic propagation law and the actual needs of photovoltaic power prediction. Then, the results after the combination operation are normalized to eliminate the dimensional differences between different weights and make the results within a reasonable value range. Based on the weight data after normalization, a mapping weight matrix is ​​constructed with meteorological spatiotemporal grid units as rows and photovoltaic spatial grid units as columns. The element value at each position in the matrix is ​​the final mapping weight coefficient between the corresponding meteorological spatiotemporal grid unit and photovoltaic spatial grid unit.

[0040] In one possible implementation, step S314 further includes: The mapping weight matrix is ​​dynamically updated over time based on the impact of real-time wind direction changes on meteorological propagation paths and cloud movement directions.

[0041] Specifically, in practical applications, dynamic meteorological data such as wind direction and cloud movement trajectory are continuously acquired by the real-time meteorological monitoring system of the target area. The impact of real-time wind direction changes on meteorological propagation paths and cloud movement directions is quantified into weight adjustment factors. Real-time iterative calculations are performed on the element values ​​related to the dynamic influence weight of the upwind direction in the mapping weight matrix. At the same time, the weight values ​​corresponding to the changes in the upstream sampling corridor range of the photovoltaic spatial grid unit caused by wind direction changes are recalculated. Combined with the basic values ​​of spatial intersection weight and spatial distance attenuation weight, the mapping weight matrix is ​​updated and corrected element by element. The entire mapping weight matrix is ​​dynamically refreshed according to the preset time step to ensure that the mapping weight coefficients corresponding to each meteorological spatiotemporal grid unit and photovoltaic spatial grid unit in the matrix always conform to the actual laws of real-time meteorological propagation and adapt to the changes in the meteorological characteristic influence range brought about by the cloud movement direction.

[0042] In one possible implementation, step S100 further includes: Step S110: Obtain multi-source meteorological data for the target area, wherein the multi-source meteorological data includes meteorological data with different spatial and temporal resolutions.

[0043] Step S120: Perform time-scale unification processing on the multi-source meteorological data, including resampling according to a preset time step, filling in missing values, and removing outliers.

[0044] Step S130: Divide the target area into spatial grids according to the preset spatial resolution and establish meteorological basic grid units.

[0045] Step S140: Map meteorological data with different spatial resolutions to the meteorological basic grid unit, and perform fusion and aggregation processing on multi-source meteorological data mapped to the same meteorological basic grid unit to generate unified meteorological basic grid data.

[0046] Step S150: Extract spatial statistical features and temporal variation features based on the meteorological basic grid data, combine and aggregate the spatial statistical features and temporal variation features to generate a meteorological spatiotemporal grid feature set, and construct the meteorological spatiotemporal grid.

[0047] Specifically, for the target area for photovoltaic power generation prediction, comprehensive collection of multi-source meteorological data, including numerical weather prediction data, satellite cloud image data, ground observation data, and all-sky imager data, is conducted. This type of meteorological data is acquired by different monitoring equipment and different forecasting systems, and naturally has different spatial and temporal resolutions, which can comprehensively cover various meteorological elements affecting photovoltaic power generation within the target area.

[0048] The acquired multi-source meteorological data undergoes unified time-scale preprocessing. Based on the preset time step required for photovoltaic power generation prediction, resampling is performed on meteorological data with different time resolutions to ensure consistency across data sources in the time dimension. Simultaneously, missing values ​​are filled in for data gaps in the time series data, and abnormal meteorological data are identified and removed through threshold judgment, statistical verification, and other methods, thereby improving the continuity, integrity, and reliability of multi-source meteorological data in the time series.

[0049] Based on the preset spatial resolution set for photovoltaic power generation prediction, the target area is divided and partitioned into regular spatial grids. The target area is discretized into several grid units of the same size and with regular arrangement according to a unified spatial scale. In this way, a meteorological basic grid unit for carrying multi-source meteorological data is established, forming a standardized spatial data carrier.

[0050] By employing spatial matching and grid mapping, multi-source meteorological data with different spatial resolutions are mapped one by one to the corresponding meteorological basic grid cells. Multi-source meteorological data falling into the same meteorological basic grid cell are subjected to weighted averaging, optimal interpolation, or statistical fusion and aggregation to eliminate spatial scale differences between different data sources and generate meteorological basic grid data with unified spatial dimensions and consistent accuracy.

[0051] Based on the unified meteorological basic grid data, feature extraction is carried out on meteorological elements in both spatial and temporal dimensions within each meteorological basic grid unit. Spatial statistical features are obtained through mean calculation, extreme value statistics, variance analysis, and spatial gradient solving in the spatial dimension. Temporal variation features are obtained through time series difference, trend fitting, rate of change calculation, and periodic component extraction in the temporal dimension. The spatial statistical features and temporal variation features corresponding to the same grid unit are spliced ​​and fused to form a meteorological spatiotemporal grid feature set with spatiotemporal joint representation capabilities, thereby completing the construction of the meteorological spatiotemporal grid.

[0052] In one possible implementation, step S140 further includes: The multi-source meteorological data includes numerical weather prediction data, satellite cloud image data, ground observation data, and all-sky imager data.

[0053] Specifically, the multi-source meteorological data includes numerical weather prediction data, satellite cloud image data, ground meteorological station observation data, and all-sky imager data. Numerical weather prediction data is used to provide regional meteorological forecast information, satellite cloud image data is used to obtain information on cloud cover and cloud distribution over a large area, ground observation data is used to collect meteorological elements such as measured temperature, irradiance, wind speed, and wind direction, and all-sky imager data is used to achieve high spatiotemporal resolution real-time monitoring of local cloud type and cloud cover. These multiple types of data complement each other, providing a comprehensive and accurate meteorological data source for photovoltaic power generation prediction.

[0054] In one possible implementation, step S200 further includes: Step S210: Divide the target area into regular grids according to a preset spatial resolution to establish initial photovoltaic spatial grid units; map each photovoltaic power station to the corresponding initial photovoltaic spatial grid unit according to the geographical coordinates of the photovoltaic power station to construct the photovoltaic spatial grid.

[0055] In this process, photovoltaic power plants are grouped based on the correlation analysis of historical power time series, and the photovoltaic spatial grid is calibrated and reconstructed based on the photovoltaic grouping relationship, so that photovoltaic power plants in the same photovoltaic spatial grid are in the same group, and photovoltaic power plants in the same group have consistent power response characteristics under similar meteorological conditions.

[0056] Specifically, the target area is divided into regular grids according to a preset spatial resolution consistent with the meteorological base grid, and initial photovoltaic spatial grid units with uniform size and regular arrangement are established. Based on the actual geographical coordinate information of each photovoltaic power station, each photovoltaic power station is accurately mapped to the corresponding initial photovoltaic spatial grid unit through spatial matching and positioning. The grid units containing photovoltaic power stations are marked and integrated, thereby constructing a photovoltaic spatial grid that is spatially matched with the meteorological spatiotemporal grid and is suitable for photovoltaic power generation prediction.

[0057] First, Pearson correlation coefficients are calculated for the historical power time series of each photovoltaic power station to construct a power correlation matrix between power stations. Then, a clustering algorithm is used to divide photovoltaic power stations with correlations higher than a set threshold into the same power characteristic group. Next, the original photovoltaic spatial grid is calibrated and reconstructed using the grouping results as constraints. Each initial photovoltaic spatial grid is traversed. If a grid contains photovoltaic power stations in different groups, the grid boundaries are adjusted, divided, or merged to unify the photovoltaic power stations in the same group into the same photovoltaic spatial grid. Finally, each photovoltaic spatial grid corresponds to only one power characteristic group, and photovoltaic power stations in the same grid have consistent power response characteristics under similar meteorological conditions.

[0058] In one possible implementation, step S400 further includes: Step S410: Based on the power grid mapping relationship, the meteorological driving characteristics, installed capacity characteristics and historical power time series characteristics corresponding to the photovoltaic spatial grid unit are used as inputs to input the pre-constructed multi-scale power prediction model. The multi-scale power prediction model includes an ultra-short-term prediction model, a short-term prediction model and a medium- and long-term prediction model.

[0059] Step S420: Obtain the predicted power values ​​at different time scales output by the multi-scale power prediction model, perform continuity correction or smoothing on the predicted power values ​​at different time scales to ensure consistency of prediction results between adjacent time scales, and determine the predicted power values ​​of each grid cell at different time scales.

[0060] Specifically, based on the power grid mapping relationship, the meteorological driving characteristics corresponding to each photovoltaic spatial grid unit, such as irradiance, temperature, and wind speed, as well as the installed capacity characteristics, such as the total photovoltaic installed capacity within the grid and module type parameters, and the historical power time series characteristics, such as historical power time series fluctuations and change rates, are standardized and then used as unified inputs into a pre-constructed multi-scale power prediction model. This multi-scale power prediction model is pre-constructed using a structure of independent modeling at different scales and unified integration and calling. Adaptive algorithm structures are designed and trained for different prediction periods. The ultra-short-term prediction model uses the LSTM (Long Short-Term Memory) network algorithm, whose structure includes an input layer, a hidden layer, a forget gate, an input gate, an output gate, and an output layer. The input layer receives standardized feature data, filters out invalid historical information through the forget gate, controls the input weights of current feature information through the input gate, and regulates the output of information from the hidden layer through the output gate. The hidden layer stores time-series dependencies through memory units. The system uses features to ultimately output ultra-short-term forecasts at the minute or hour level. The short-term forecast model employs the Transformer algorithm, whose core structure consists of a self-attention mechanism and a feedforward neural network. The self-attention mechanism captures the temporal correlation and feature interaction relationships of the input features, accurately extracting the long-term and short-term dependencies of meteorological driving features and historical power features. The feedforward neural network performs non-linear transformations on the features output by the attention mechanism, ultimately outputting short-term forecasts for the day or day. The medium- and long-term forecast model uses the XGBoost algorithm, whose structure is based on Gradient Boosting Decision Tree (GBDT). Multiple weak decision trees are integrated to form a strong learner. The splitting nodes of each decision tree are iteratively optimized using gradient descent, and a regularization term is introduced to suppress overfitting. At the same time, combined with time series trend features, weekly or monthly power changes of grid cells are fitted and predicted. The integration of these three models forms a complete multi-scale power forecast model, which can call the model of the corresponding scale to complete the forecast according to the forecast requirements.

[0061] The original predicted power values ​​of each photovoltaic spatial grid cell at three time scales—ultra-short-term (minute / hour), short-term (day / day), and medium-to-long-term (week / month)—output by the multi-scale power prediction model were obtained. Using the short-term prediction results as the baseline time series reference, numerical consistency constraints were constructed at the time series transition points between ultra-short-term and short-term, and between short-term and medium-to-long-term. Weighted fusion and linear smoothing fitting algorithms were used to connect the predicted power sequences of adjacent time scales, correcting time series jumps and numerical abrupt changes to ensure that the predicted power at each time scale transitioned continuously in time and remained consistent in value. The results after continuous correction and smoothing were then used as the final predicted power values ​​of each photovoltaic spatial grid cell at the corresponding time scale.

[0062] Example 2, based on the same inventive concept as the multi-source meteorological-driven regional photovoltaic power generation prediction method in the foregoing examples, such as... Figure 2 As shown, this application provides a multi-source meteorological-driven regional photovoltaic power generation prediction system. The system and method embodiments in this application are based on the same inventive concept. The system includes: The meteorological spatiotemporal grid construction module 10 is used to acquire multi-source meteorological data of the target area, extract and aggregate spatiotemporal scale features of the multi-source meteorological data, and construct a meteorological spatiotemporal grid.

[0063] The photovoltaic spatial grid construction module 20 is used to construct a photovoltaic spatial grid based on the structural distribution and operation data of photovoltaic power plants, and to determine the capacity weight according to the photovoltaic installed capacity in each grid.

[0064] The power grid mapping relationship establishment module 30 is used to establish the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid.

[0065] The predicted power value acquisition module 40 is used to predict the power at each time scale according to the power grid mapping relationship and through a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid cell.

[0066] The power generation prediction result acquisition module 50 is used to perform weighted aggregation of the predicted power values ​​of each grid unit according to the capacity weight to obtain the power generation prediction result of the regional photovoltaic.

[0067] Furthermore, the system also includes: Based on the spatial relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid, a mapping weight coefficient between the grids is determined. This mapping weight coefficient is obtained based on one or more of the following: spatial intersection ratio, spatial distance attenuation relationship, or upwind sampling area determined by real-time wind direction. Based on the mapping weight coefficient, meteorological features in the meteorological spatiotemporal grid are mapped to the corresponding photovoltaic spatial grid, generating a meteorological driving feature set corresponding to the photovoltaic spatial grid, and establishing the power grid mapping relationship.

[0068] Furthermore, the system also includes: Based on the proportion of spatial overlap between meteorological spatiotemporal grid units in the meteorological spatiotemporal grid and photovoltaic spatial grid units in the photovoltaic spatial grid, a spatial intersection weight is obtained; based on the spatial distance between the center points of the meteorological spatiotemporal grid units and the photovoltaic spatial grid units, a spatial distance attenuation weight is calculated using a preset distance attenuation function; based on the real-time wind direction, the upstream direction of the photovoltaic spatial grid units is determined, and a sampling corridor of preset length and width is constructed in the upstream direction, assigning upwind dynamic influence weights to the meteorological spatiotemporal grid units located in the sampling corridor; using one or more of the spatial intersection weights, spatial distance attenuation weights, and upwind dynamic influence weights, the mapping weight coefficient is determined.

[0069] Furthermore, the system also includes: When one of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight is used, the used weight is used as the mapping weight coefficient; when multiple of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight are used, the multiple weights are weighted and combined or multiplied and normalized to construct a mapping weight matrix from meteorological spatiotemporal grid unit to photovoltaic spatial grid unit, wherein the element values ​​of the mapping weight matrix are used as the mapping weight coefficient.

[0070] Furthermore, the system also includes: The mapping weight matrix is ​​dynamically updated over time based on the impact of real-time wind direction changes on meteorological propagation paths and cloud movement directions.

[0071] Furthermore, the system also includes: Acquire multi-source meteorological data for the target area, including meteorological data with different spatial and temporal resolutions; perform time-scale unification processing on the multi-source meteorological data, including resampling according to a preset time step, missing value imputation, and outlier removal; divide the target area into spatial grids according to the preset spatial resolution to establish meteorological basic grid units; map meteorological data with different spatial resolutions to the meteorological basic grid units, and fuse and aggregate multi-source meteorological data mapped to the same meteorological basic grid unit to generate unified meteorological basic grid data; extract spatial statistical features and temporal variation features based on the meteorological basic grid data, combine and aggregate the spatial statistical features and temporal variation features to generate a meteorological spatiotemporal grid feature set, and construct the meteorological spatiotemporal grid.

[0072] Furthermore, the system also includes: The multi-source meteorological data includes numerical weather prediction data, satellite cloud image data, ground observation data, and all-sky imager data.

[0073] Furthermore, the system also includes: The target area is divided into regular grids according to a preset spatial resolution to establish initial photovoltaic spatial grid units. Each photovoltaic power station is mapped to its corresponding initial photovoltaic spatial grid unit according to its geographical coordinates to construct the photovoltaic spatial grid. The photovoltaic power stations are grouped according to the correlation analysis of historical power time series, and the photovoltaic spatial grid is calibrated and reconstructed based on the photovoltaic grouping relationship so that the photovoltaic power stations in the same photovoltaic spatial grid are in the same group, and the photovoltaic power stations in the same group have consistent power response characteristics under similar meteorological conditions.

[0074] Furthermore, the system also includes: Based on the power grid mapping relationship, the meteorological driving characteristics, installed capacity characteristics, and historical power time series characteristics corresponding to the photovoltaic spatial grid unit are used as inputs to a pre-constructed multi-scale power prediction model. The multi-scale power prediction model includes an ultra-short-term prediction model, a short-term prediction model, and a medium-to-long-term prediction model. The predicted power values ​​at different time scales output by the multi-scale power prediction model are obtained. The predicted power values ​​at different time scales are continuously corrected or smoothed to ensure consistency between prediction results at adjacent time scales, and the predicted power values ​​of each grid unit at different time scales are determined.

[0075] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Specific embodiments of this specification have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0076] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0077] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for predicting regional photovoltaic power generation driven by multiple meteorological sources, characterized in that, include: Acquire multi-source meteorological data of the target area, extract and aggregate spatiotemporal scale features from the multi-source meteorological data, and construct a meteorological spatiotemporal grid; Based on the structural distribution and operation data of photovoltaic power plants, a photovoltaic spatial grid is constructed, and the capacity weight is determined according to the photovoltaic installed capacity in each grid. Establish the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid; Based on the power lattice mapping relationship, power prediction at each time scale is performed using a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid cell. The predicted power values ​​of each grid cell are weighted and aggregated according to the capacity weight to obtain the predicted power generation results of the regional photovoltaic system.

2. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 1, characterized in that, Establishing the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid includes: Based on the spatial relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid, the mapping weight coefficient between the grids is determined. The mapping weight coefficient is obtained based on one or more of the following: spatial intersection ratio, spatial distance attenuation relationship, or upwind sampling area determined by real-time wind direction. Based on the mapping weight coefficient, meteorological features in the meteorological spatiotemporal grid are mapped to the corresponding photovoltaic spatial grid, generating a set of meteorological driving features corresponding to the photovoltaic spatial grid, and establishing the power grid mapping relationship.

3. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 2, characterized in that, Determine the mapping weight coefficients between grids, including: The spatial intersection weight is obtained based on the proportion of spatial overlap area between meteorological spatiotemporal grid units in the meteorological spatiotemporal grid and photovoltaic spatial grid units in the photovoltaic spatial grid. Based on the spatial distance between the center points of the meteorological spatiotemporal grid unit and the photovoltaic spatial grid unit, the spatial distance attenuation weight is calculated based on a preset distance attenuation function. The upstream direction of the photovoltaic spatial grid unit is determined based on the real-time wind direction, and a sampling corridor with a preset length and width is constructed in the upstream direction. The meteorological spatiotemporal grid units located in the sampling corridor are assigned upwind dynamic influence weights. The mapping weight coefficient is determined by using one or more of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight.

4. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 3, characterized in that, The mapping weight coefficients are determined using one or more of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight, including: When one of the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight is adopted, the adopted weight is used as the mapping weight coefficient. When multiple weights among the spatial intersection weight, spatial distance attenuation weight, and upwind dynamic influence weight are used, the multiple weights are weighted and combined or multiplied and normalized to construct a mapping weight matrix from meteorological spatiotemporal grid unit to photovoltaic spatial grid unit, wherein the element values ​​of the mapping weight matrix are used as the mapping weight coefficients.

5. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 4, characterized in that, The mapping weight matrix is ​​dynamically updated over time based on the impact of real-time wind direction changes on meteorological propagation paths and cloud movement directions.

6. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 1, characterized in that, Acquire multi-source meteorological data for the target area, extract and aggregate spatiotemporal scale features from the multi-source meteorological data, and construct a meteorological spatiotemporal grid, including: Acquire multi-source meteorological data for the target area, wherein the multi-source meteorological data includes meteorological data with different spatial and temporal resolutions; The multi-source meteorological data is subjected to time-scale uniform processing, including resampling according to a preset time step, filling in missing values, and removing outliers. The target area is divided into spatial grids according to the preset spatial resolution to establish meteorological basic grid units; Meteorological data with different spatial resolutions are mapped to the meteorological basic grid unit, and multi-source meteorological data mapped to the same meteorological basic grid unit are fused and aggregated to generate unified meteorological basic grid data. Based on the meteorological basic grid data, spatial statistical features and temporal variation features are extracted, and the spatial statistical features and temporal variation features are combined and aggregated to generate a meteorological spatiotemporal grid feature set, thus constructing the meteorological spatiotemporal grid.

7. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 6, characterized in that, The multi-source meteorological data includes numerical weather prediction data, satellite cloud image data, ground observation data, and all-sky imager data.

8. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 1, characterized in that, Based on the structural distribution and operational data of photovoltaic power plants, a photovoltaic spatial grid is constructed, including: The target area is divided into regular grids according to a preset spatial resolution to establish initial photovoltaic spatial grid units; each photovoltaic power station is mapped to its corresponding initial photovoltaic spatial grid unit according to its geographical coordinates to construct the photovoltaic spatial grid. In this process, photovoltaic power plants are grouped based on the correlation analysis of historical power time series, and the photovoltaic spatial grid is calibrated and reconstructed based on the photovoltaic grouping relationship, so that photovoltaic power plants in the same photovoltaic spatial grid are in the same group, and photovoltaic power plants in the same group have consistent power response characteristics under similar meteorological conditions.

9. The regional photovoltaic power generation prediction method driven by multi-source meteorological conditions according to claim 8, characterized in that, Based on the power lattice mapping relationship, power prediction at each time scale is performed using a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid cell, including: Based on the power grid mapping relationship, the meteorological driving characteristics, installed capacity characteristics and historical power time series characteristics corresponding to the photovoltaic spatial grid unit are used as inputs to input a pre-constructed multi-scale power prediction model. The multi-scale power prediction model includes an ultra-short-term prediction model, a short-term prediction model and a medium- and long-term prediction model. The predicted power values ​​at different time scales output by the multi-scale power prediction model are obtained. The predicted power values ​​at different time scales are continuously corrected or smoothed to ensure that the prediction results between adjacent time scales are consistent, and the predicted power values ​​of each grid cell at different time scales are determined.

10. A multi-source meteorological-driven regional photovoltaic power generation prediction system, characterized in that, The system is used to implement the multi-source meteorological-driven regional photovoltaic power generation prediction method according to any one of claims 1-9, the system comprising: The meteorological spatiotemporal grid construction module is used to acquire multi-source meteorological data of the target area, extract and aggregate spatiotemporal scale features of the multi-source meteorological data, and construct a meteorological spatiotemporal grid. The photovoltaic spatial grid construction module is used to construct a photovoltaic spatial grid based on the structural distribution and operation data of photovoltaic power plants, and determine the capacity weight according to the photovoltaic installed capacity in each grid. A power grid mapping relationship establishment module is used to establish the power grid mapping relationship between the meteorological spatiotemporal grid and the photovoltaic spatial grid; The predicted power value acquisition module is used to predict the power at each time scale according to the power grid mapping relationship and through a pre-constructed multi-scale power prediction model to obtain the predicted power value of each grid cell. The power generation prediction result acquisition module is used to perform weighted aggregation of the predicted power values ​​of each grid unit according to the capacity weight to obtain the power generation prediction result of the regional photovoltaic.

Images