Suitability layout evaluation method and evaluation system for regional renewable energy sources

Abstract

The invention belongs to the technical field of energy informatization and geographic information systems, and relates to a regional renewable energy suitability layout evaluation method and system, and the method comprises the steps: 1, carrying out the data preprocessing, interpolating a time sequence missing value, and estimating and obtaining the meteorological and facility space distribution; 2, a renewable energy source-influence factor correlation model is constructed, and a nonlinear correlation model of energy consumption, social economy and natural environment factors is established; 3, marginal influence curve identification: determining an optimal threshold interval of a key variable by calculating a marginal effect curve of each variable, and generating a nonlinear marginal influence model; 4, suitability drawing and hot spot identification are carried out, and spatial clustering and suitability drawing are realized by using an improved algorithm; according to the invention, it can be identified that the residential facilities reach the optimal absorption threshold value under the industrial facility density of solar energy; the optimal wind speed of wind energy; the annual precipitation millimetre of water and electricity and the optimal value of the density of service facilities.

Description

Technical Field

[0001] This invention belongs to the field of energy information and geographic information system technology, specifically relating to a method and system for evaluating the suitability of regional renewable energy layout. Background Technology

[0002] Currently, research on the consumption potential of renewable energy mainly relies on the following technological approaches: (1) Statistical regression model based on panel data: Using national or provincial statistical data, the relationship between renewable energy consumption and economic, climate and other factors is analyzed through linear or multiple regression. Representative methods include Pearson correlation coefficient analysis and analytic hierarchy process (AHP).

[0003] (2) Machine learning-based prediction models, such as random forests, XGBoost, and neural networks, are used to simulate the impact of meteorological and socioeconomic factors on energy consumption. These methods can capture nonlinear relationships, but the models lack interpretability and are difficult to reveal critical thresholds and marginal effects.

[0004] (3) GIS-based spatial suitability evaluation method: Spatial mapping of renewable energy resource distribution is carried out by weighted overlay (AHP, entropy weight method, etc.), which is mainly used for suitability assessment of local areas and lacks fine modeling at the national scale.

[0005] (4) Subjective evaluation.

[0006] However, existing energy consumption assessments have the following shortcomings: (1) Weak explanatory power of variables: Existing models mostly rely on preset weights or linear assumptions, which cannot reflect the nonlinear response mechanism of natural and social factors to energy consumption.

[0007] (2) Lack of marginal effect and threshold identification: The failure to quantify the critical change pattern of renewable energy consumption with each driving factor leads to vague policy recommendations.

[0008] (3) Insufficient spatial accuracy: Existing GIS mapping is mostly at the county or provincial level, which cannot support detailed planning at the 250 m level.

[0009] (4) Poor model interpretability: Traditional machine learning models are difficult to explain the contribution of variables, lack traceability for policy and engineering applications, and cannot reveal nonlinear characteristics.

[0010] Therefore, there is a need for an energy consumption assessment method that can significantly improve accuracy, applicability, and interpretability to address the aforementioned technical issues. Summary of the Invention

[0011] This invention provides the following technical solution: a method for evaluating the suitability of regional renewable energy layout, comprising the following steps: Step S1: Data preprocessing. Use the ARIMA model to impute missing values ​​in the time series, and use Kriging interpolation and kernel density estimation to obtain the spatial distribution of meteorological and facility data.

[0012] Step S2: Construction of the Renewable Energy-Influencing Factor Correlation Model. Using CatBoost and XGBoost as the core algorithms, a nonlinear correlation model between energy consumption and socio-economic and natural environmental factors is established.

[0013] Step S3: Marginal effect curve identification. The marginal effect curves of each variable are calculated using the SHAP model to determine the optimal threshold range of key variables, generating a nonlinear marginal effect model to achieve quantitative identification and threshold determination from the causal mechanism level.

[0014] Step S4, Suitability Mapping and Hotspot Identification: Spatial clustering is achieved using the improved Getis-Ord Gi* and local Moran's I algorithms, and a GIS nonlinear weighted overlay algorithm is developed to achieve suitability mapping.

[0015] Preferably, step S1 includes the following sub-steps: Step S1-1: Input multi-source data, including renewable energy consumption statistics, meteorological data, socio-economic data, POI point data, and remote sensing raster data.

[0016] Step S1-2: Time series imputation. The autoregressive integral moving average (ARIMA) model is used to fill in missing values ​​in the energy and economic data of previous years to ensure the continuity of the time series.

[0017] Steps S1-3: Spatial interpolation and kernel density estimation. Kriging interpolation is used to generate a meteorological grid with spatial resolution; kernel density estimation (KDE) is used to calculate the density distribution of various facilities, and Ripley's K function is used to determine the bandwidth radius.

[0018] More preferably, in step S1-1, the meteorological data includes: solar radiation, wind speed, precipitation, and humidity; the socio-economic data includes: GDP, population, and electricity consumption; the POI data includes: industrial, residential, commercial, and service facilities; and the remote sensing raster data includes: nighttime light intensity, land use, and DEM. The output of step S1 includes a raster set of natural environmental factors and a vector set of social environmental factors under a unified geographic coordinate system.

[0019] More preferably, in step S1, the missing values ​​are imputed as follows: (1) In equation (1), Indicates time The observation time series, This indicates a shift operator that satisfies... ; This represents the difference order required to make the sequence stationary. express Autoregressive operator of order, express Order moving average operator, Indicates time The white noise error term.

[0020] Kriging interpolation is: (2) In equation (2), Indicates the target location The estimated value at that location, Indicates a known location The observed values, Indicates the assignment to the observation point The weight, This indicates the number of known sample points involved in the interpolation.

[0021] KDE is calculated as follows: (3) In equation (3), Indicates position The kernel density estimate at that location, This represents the total number of POIs. Indicates position With POI point The Euclidean distance between them Represents the kernel function. Indicates the search bandwidth.

[0022] Preferably, step S2 includes the following sub-steps: Step S2-1: Factor screening. Driver factors are screened through bivariate significance analysis and Pearson correlation coefficient. Driver factors include: sunshine hours, wind speed, precipitation, relative humidity, terrain slope, density of various facilities, and grid connection price.

[0023] Step S2-2: Nonlinear modeling. Establish a renewable energy consumption forecasting model and use CatBoost and XGBoost algorithms to achieve nonlinear fitting.

[0024] Step S2-3: Hyperparameter optimization. The Bayesian optimization algorithm is used to search for the learning rate, maximum depth, and subsampling ratio. The objective function is to minimize the relative error under the condition that R² ≥ 0.9 on the test set.

[0025] More preferably, in S2, a linear regression model is used to assess the impact of each factor on renewable energy consumption. The independent variables are screened through bivariate analysis. The linear regression model is as follows: (5) In equation (5), This indicates the projected consumption of renewable energy. Represents the intercept term. Indicates input features, Indicates model parameters or weights. Indicates the number of features.

[0026] The nonlinear regression model is as follows: (6) In equation (6), This represents the input feature vector; Indicates the first Prediction of decision trees Indicates the first A weak learner, This represents the weights of each weak learner. This indicates the number of trees in the ensemble model. This indicates increasing the number of iterations.

[0027] Preferably, step S3 includes the following sub-steps: Step S3-1: Interpretive calculation. Using the SHAP method, calculate the marginal contribution and interaction effect of each input variable to obtain the historical impact curves of different factors on solar, wind, and hydropower consumption.

[0028] Step S3-2: Threshold identification, fitting the marginal effect curve, and determining the optimal threshold range for key variables.

[0029] Step S3-3: Multi-energy complementarity analysis. Through the variable interaction matrix, the synergistic relationship between solar energy, wind energy, and hydropower is identified, revealing the complementary potential of multi-energy systems.

[0030] Preferably, step S4 includes the following sub-steps: Step S4-1, Spatial Assessment: Input the natural and social factor grids into the nonlinear correlation model and the nonlinear marginal impact model to estimate the renewable energy consumption potential of each grid.

[0031] Step S4-2: Hotspot identification. Based on the Getis-Ord Gi* and local Moran's I algorithms, identify high-value clustered hotspots and low-value clustered coldspots.

[0032] Step S4-3: Multi-energy complementary zoning. Combined with the energy suitability map, multi-energy complementary zones are divided, and finally, a renewable energy consumption potential atlas is generated.

[0033] This invention also discloses a regional renewable energy suitability layout evaluation system, which employs the above-mentioned evaluation method and includes: Data preprocessing module: Uses ARIMA model to impute missing values ​​in time series data, and employs Kriging interpolation and kernel density estimation to obtain the spatial distribution of meteorological and facility data.

[0034] Renewable Energy – Impact Factor Correlation Model Construction Module: This module is used to establish a nonlinear correlation model between energy consumption and socio-economic and natural environmental factors, using CatBoost and XGBoost as the core algorithms.

[0035] Marginal effect curve identification module: used to calculate the marginal effect curves of each variable using the SHAP model, and to determine the optimal threshold range for key variables.

[0036] Suitability Mapping and Hotspot Identification Module: Used to perform spatial clustering and suitability mapping using the improved Getis-Ord Gi* and local Moran's I algorithms.

[0037] Data server: Used to store raw data from multiple sources and preprocessed results.

[0038] Modeling compute node: Used to configure GPUs and perform CatBoost / XGBoost model training and SHAP computation; Visualization terminal: Used to dynamically display results based on ArcGIS or WebGL.

[0039] Automatic output module: Used to generate output files including a national suitability map, marginal curve map, hot zone distribution map, and renewable energy influencing factor mechanism map.

[0040] The beneficial effects of this invention are: 1. This invention achieves nonlinear, threshold-based modeling: This invention can identify the threshold for solar energy absorption at industrial facility density and residential facility density; the optimal wind speed for wind energy; and the optimal values ​​for annual precipitation in millimeters and service facility density for hydropower.

[0041] 2. This invention significantly improves spatial accuracy and applicability: This invention can achieve nationwide mapping of renewable energy consumption potential with a resolution of 250 m.

[0042] 3. Enhanced model interpretability: This invention visualizes variable contributions through the SHAP method, supporting policy evaluation and regional energy layout optimization.

[0043] 4. This invention enables multi-energy complementarity analysis: This invention can simultaneously output the synergistic suitability pattern of solar energy, wind energy, and hydropower, providing a basis for the coupling of multi-energy systems. Attached Figure Description

[0044] Figure 1 This is a schematic diagram illustrating the steps of a method and system for evaluating the suitability of regional renewable energy distribution according to the present invention. Figure 2 This is a schematic diagram showing the steps of the present invention. Detailed Implementation

[0045] The related technologies of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0046] like Figures 1-2 As shown, this embodiment presents a method and system for evaluating and efficiently utilizing the suitability of renewable energy consumption based on spatiotemporal big data and interpretable machine learning. By integrating multi-source spatial data, establishing a nonlinear interpretation model, identifying the marginal impact curves and threshold ranges of key driving factors, and using geostatistical methods to achieve high-precision spatial mapping of renewable energy consumption potential and multi-energy complementary zoning, it can significantly improve the scientific nature and efficiency of regional energy system planning. The detailed steps of this embodiment are as follows: Step S1 Data Preprocessing Module: Use the ARIMA model to impute missing values ​​in the time series, and use Kriging interpolation and kernel density estimation to obtain the spatial distribution of meteorological and facility data.

[0047] Step S2 Renewable Energy – Influencing Factor Correlation Model Construction Module: Using CatBoost and XGBoost as the core algorithms, a nonlinear correlation model between energy consumption and socio-economic and natural environmental factors is established.

[0048] Step S3 Marginal Effect Curve Identification Module: Calculate the marginal effect curves of each variable using the SHAP model to determine the optimal threshold range for key variables.

[0049] Step S4 Suitability Mapping and Hotspot Identification Module: Spatial clustering and suitability mapping at 250 m resolution are achieved using the improved Getis-Ord Gi* and local Moran's I algorithms.

[0050] The four stages are executed sequentially, forming a closed-loop system of "data integration → model building → mechanism identification → spatial representation".

[0051] Detailed descriptions of each stage and module: Phase I: Data Fusion and Preprocessing (S1) Multi-source data input (S1-1): Import renewable energy consumption statistics, meteorological data (solar radiation, wind speed, precipitation, humidity, etc.), socio-economic data (GDP, population, electricity consumption), POI data (industrial, residential, commercial, service facilities), and remote sensing raster data (nighttime light intensity, land use, DEM, etc.).

[0052] Time series interpolation (S1-2): The Autoregressive Integral Moving Average (ARIMA) model was used to impute missing values ​​in energy and economic data from 2000 to 2024 to ensure the continuity of the time series.

[0053] Spatial interpolation and kernel density estimation (S1-3): A meteorological grid with a spatial resolution of 250 m was generated using Kriging interpolation; the density distribution of various facilities was calculated using kernel density estimation (KDE), and the bandwidth radius was determined using Ripley's K function.

[0054] The output of this stage includes a raster set of natural environmental factors and a vector set of social environmental factors under a unified geographic coordinate system, providing the input data foundation for subsequent modeling.

[0055] Phase II Nonlinear Correlation Modeling (S2) Factor screening (S2-1): Driving factors were screened using bivariate significance analysis (p<0.01) and Pearson correlation coefficient (PCC<0.3), including sunshine hours (SSD), wind speed (WIN), precipitation (PRE), relative humidity (RHU), topographic slope, density of various facilities, and feed-in tariff (FIT).

[0056] Nonlinear modeling (S2-2): A renewable energy consumption forecasting model was established, employing CatBoost and XGBoost algorithms for nonlinear fitting, applicable to both photovoltaic and wind (hydropower) data, respectively. The model presented here can be replaced; see the alternative solutions section.

[0057] Hyperparameter optimization (S2-3): The Bayesian optimization algorithm is used to search for parameters such as learning rate, maximum depth, and subsampling ratio. The objective function is to minimize the relative error under the condition that R² ≥ 0.9 on the test set.

[0058] Through Phase II modeling, a nonlinear correlation model (Model 1) between renewable energy and its driving factors was constructed, providing a foundation for interpretive analysis.

[0059] Phase III: Marginal Effect Analysis and Threshold Identification (S3) Interpretive computation (S3-1): The SHapley Additive exPlanations (SHAP) method was used to calculate the marginal contribution and interaction effect of each input variable, and to obtain the historical impact curves of different factors on solar, wind and hydropower consumption.

[0060] Threshold recognition (S3-2): By fitting marginal effect curves, the optimal threshold ranges for key variables are determined. For example, solar energy consumption increases significantly when the density of industrial facilities is >2.26 facilities / km²; wind power utilization is optimal when the wind speed is 6.21 m / s; and hydropower benefits are maximized when the annual precipitation is 898 mm.

[0061] Multi-energy complementarity analysis (S3-3): By using a variable interaction matrix, we can identify the synergistic relationship between solar, wind, and hydropower, and reveal the complementary potential of multi-energy systems.

[0062] This stage generates a nonlinear marginal impact model (Model 2), enabling quantitative identification from the perspective of causal mechanisms.

[0063] Phase IV: Suitability Mapping and Spatial Clustering (S4) Spatial assessment (S4-1): By inputting natural and social factor raster data into Model 1 and Model 2, the renewable energy consumption potential of each 250 m grid can be estimated.

[0064] Hotspot identification (S4-2): Based on the Getis-Ord Gi* and local Moran's I algorithms, high-value clusters (Hotspot) and low-value clusters (Coldspot) regions are identified.

[0065] Multi-energy complementary partition (S4-3): By combining the three types of energy suitability maps, multi-energy complementary zones are divided, such as "solar-wind coupling zone" and "wind-hydro complementary zone".

[0066] The final result is a nationwide atlas of renewable energy consumption potential with a resolution of 250 m, providing spatial decision-making support for planning departments.

[0067] This embodiment identifies that solar energy reaches its optimal absorption threshold when the density of industrial facilities is 2.26 units / km² and that of residential facilities is 21.95 units / km²; wind energy is optimal at a wind speed of 6.21 m / s; and hydropower is most suitable when the annual precipitation is 898 mm and the density of service facilities is 10.41 units / km².

[0068] Example The data used in this embodiment includes panel data and geospatial data (vector data and raster data). Data types, time ranges, and data sources are shown in Table 1. (1) Interpolation and preprocessing For panel data such as grid-connected electricity price, population, GDP, and renewable energy consumption, the Autoregressive Integrated Moving Average (ARIMA) model is used to impute missing values, as shown in equation (1): (1) in, For a moment The observation time series, For the shift operator, satisfying ; The order of difference required to make the sequence stationary; for Autoregressive (AR) operator; for Moving average (MA) operator; For a moment The white noise error term.

[0069] For spatial interpolation of meteorological data, the Kriging interpolation method was adopted, as shown in equation (2). Based on measured data from 1,947 meteorological observation stations in China, a spatial gridded dataset of meteorological parameters was generated.

[0070] (2) in, For target location The estimated value at that location; For known location Observed values; To be assigned to observation points The weights; The number of known sample points used in the interpolation.

[0071] For POI data of different types of facilities, the kernel density function is used to estimate their spatial distribution probability, as shown in equation (3). This method is used to calculate the spatial density distribution of residential, industrial and commercial facilities.

[0072] (3) in, For position The kernel density estimate at that location; This represents the total number of POIs. For position With POI point The Euclidean distance between them; For kernel functions; The search bandwidth (radius) is determined by Ripley's K function, as shown in equation (4): (4) in, Indicates the distance of any point. The expected number of other points within the range, after point strength normalization; The area of ​​the study region; This represents the number of POIs. For point With point The distance between them; This is an indicator function; it takes the value 1 if the condition is met, and 0 otherwise. Boundary correction weights are introduced to compensate for boundary effects.

[0073] (2) Model selection and interpretability analysis First, a linear regression model was used to assess the impact of each factor on renewable energy consumption. The independent variables were screened through bivariate analysis, and their mathematical form is shown in equation (5): (5) in, Forecasted renewable energy consumption; For the intercept term; Input features (influencing factors); These are model parameters or weights; The number of features.

[0074] After identifying the main influencing factors of various renewable energy consumption, a nonlinear model was further employed to test whether these factors exhibit nonlinear effects. Due to the large variance and skewed distribution characteristics of the data, a tree-based model was considered a potentially suitable choice, as shown in equation (6): (6) in, Forecasted renewable energy consumption; The input feature vector; For the first Prediction using a decision tree; For the first A weak learner (decision tree); The weights of each weak learner; The number of trees in the ensemble model, To increase the number of iterations.

[0075] Based on the selected baseline model, the Bayesian hyperparameter optimization algorithm is used to determine the optimal combination of hyperparameters for the nonlinear model. The optimization process can be expressed as equation (7): (7) in, Represents a hyperparameter vector; The objective function (here, the objective function is defined as: on the test set R) 2 Minimize the relative error when it exceeds 0.9. This represents the set of hyperparameter configurations that have been evaluated.

[0076] For the constructed nonlinear regression model, the SHapley Additive exPlanations (SHAP) model is used to reveal the marginal effect of each influencing factor on the consumption of different types of renewable energy, and its form is shown in equation (8): (8) in, Features Given a predicted value The SHAP value below; For the entire feature set; for A subset of (i.e., excluding features) (All features other than those mentioned above); Indicates using only subsets The model prediction results; Indicates using only subsets The model prediction results at that time.

[0077] (3) Nonlinear evaluation algorithm for geographic rasters Based on the calculation results of SHAP, the marginal effect characteristic curves of each variable in the consumption of different types of renewable energy are fitted, and their form is shown in equation (9).

[0078] (9) in, Indicates energy type The projected consumption of renewable energy; Representation of features For renewable energy types The marginal effect curve.

[0079] Based on the fitting results and geographical neighborhood analysis, equation (10) can be used to calculate the future potential and suitability assessment of regional renewable energy consumption.

[0080] (10) in, Representing geographical units Energy type The projected consumption of renewable energy; Features In unit Above energy types The marginal effect; Representation unit The set of neighborhood grids; Indicates neighboring elements to elements Spatial influence weight.

[0081] Equation (10) can be used to estimate and assess the suitability of regional solar, wind, and hydropower consumption at a spatial resolution of 250 m. For county-level results, the Getis-Ord Gi* statistic is used to calculate spatial distribution characteristics to identify hot and cold areas, as shown in equation (11): (11) in, Features In distance Getis-Ord Gi* statistics at the location; For position The values ​​of the variables under study; The average of the variables studied across all locations; Indicates distance Location within range The neighborhood set; For position With position Spatial weights between them; This represents the total number of spatial units.

[0082] To verify the significance of the spatial hot and cold regions obtained by equation (11) and to distinguish the spatial autocorrelation, the Anselin Local Moran's I (ALMI) method was used for testing, and its calculation form is shown in equation (12).

[0083] (12) in, For position The local Moran's I value; For position The value that the variable can take; The average of the values ​​taken by all positional variables; For position With position Spatial weights between them; This represents the total number of locations (sample points).

[0084] The results of the potential suitability assessment for renewable energy development in this embodiment are shown in Table 2:

[0085] The threshold calculation results in Table 2 show that when a characteristic variable reaches a specific threshold, it will have a positive marginal impact on renewable energy consumption. "No significant correlation" indicates that the variable's historical contribution to the corresponding renewable energy consumption is less than 5%. "Optimal value" indicates the maximum marginal impact of the variable on renewable energy consumption (if applicable). The symbols "+" and "-" indicate a significant association between the influencing factor and the corresponding renewable energy consumption, but this association cannot be fully characterized by conventional linear or nonlinear models, although an overall positive or negative trend can still be observed.

[0086] Based on the marginal impact curves of key influencing factors, the suitability thresholds for renewable energy development were further assessed, as shown in Table 2. For meteorological factors, solar energy consumption exhibits a positive marginal impact when the annual sunshine duration (SSD) exceeds 1645.69 h. Conversely, the marginal impact of solar energy consumption turns negative when the annual average wind speed (WIN) exceeds 0.79 m / s or the annual precipitation (PRE) exceeds 418.15 mm. Regarding energy-consuming facilities, a residential facility density of 21.95 facilities / km² is most conducive to solar energy absorption. No significant saturation effect was found for other types of facilities, indicating that the widespread distribution of industrial, service, and commercial facilities all contribute to solar energy utilization.

[0087] In summary, this invention can identify the threshold for optimal absorption of solar energy at industrial facility density and residential facility density; the optimal wind speed for wind energy; and the optimal values ​​for annual precipitation in millimeters and service facility density for hydropower.

[0088] It should be emphasized that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A method for evaluating the suitability of a layout of a regional renewable energy source, characterized by, The method comprises the following steps: Step S1, data preprocessing, using ARIMA model to interpolate missing values of time series, using Kriging interpolation and kernel density estimation to obtain spatial distribution of weather and facilities; Step S2, renewable energy-influencing factor correlation model construction, taking CatBoost and XGBoost as core algorithms, establishing a nonlinear correlation model of energy consumption and social and economic and natural environmental factors; Step S3, marginal effect curve identification, calculating the marginal effect curve of each variable through the SHAP model, determining the optimal threshold interval of the key variable, generating a nonlinear marginal effect model, and realizing quantitative identification and threshold determination from the causal mechanism level; Step S4, suitability mapping and hotspot identification, using improved Getis-Ord Gi* and local Moran's I algorithm to realize spatial clustering, and developing GIS nonlinear weighted superposition algorithm to realize suitability mapping.

2. The method for evaluating the suitability of layout of regional renewable energy sources according to claim 1, characterized in that, The step S1 comprises the following steps: Step S1-1, multi-source data input, importing renewable energy consumption statistical data, weather data, social and economic data, POI point data, and remote sensing raster data; Step S1-2, time series interpolation, using autoregressive integrated moving average model ARIMA to fill in missing values of energy and economic data in previous years to ensure time series continuity; Step S1-3, spatial interpolation and kernel density estimation, using Kriging interpolation to generate spatial resolution weather grid; using kernel density estimation KDE to calculate the density distribution of various facilities, and using Ripley's K function to determine the bandwidth radius.

3. The method for evaluating the suitability of layout of regional renewable energy sources according to claim 2, characterized in that, In the step S1-1, the weather data includes solar radiation, wind speed, precipitation, and humidity; the social and economic data includes GDP, population, and electricity consumption; the POI point data includes industrial, residential, commercial, and service facilities; and the remote sensing raster data includes night light intensity, land use, and DEM. The output result of the step S1 includes a natural environmental factor grid set and a social environmental factor vector set in a unified geographic coordinate system.

4. The method for evaluating the suitability of layout of regional renewable energy sources according to claim 2, characterized in that, In the step S1, the missing values are interpolated as follows: (1) In formula (1), denotes the observation time series at time , denotes the backshift operator satisfying ; denotes the order of differencing required to make the series stationary, denotes the order of the autoregressive operator, denotes the order of the moving average operator, denotes the white noise error term at time ; The Kriging interpolation is as follows: (2) In formula (2), denotes an estimated value at a target position, denotes an observation value at a known position, denotes a weight assigned to an observation point, denotes a number of known sample points participating in interpolation;​​​ The KDE calculation is as follows: (3) In formula (3), denotes the kernel density estimate at the location denotes the total number of POI points, denotes the location the Euclidean distance between the location and the POI point denotes the kernel function, denotes the search bandwidth.​ 5. The method of claim 1, wherein the method further comprises: The S2 comprises the following steps: Step S2-1, factor screening, screening driving factors through bivariate significance analysis and Pearson correlation coefficient, the driving factors including sunshine hours, wind speed, precipitation, relative humidity, terrain slope, facility density, and on-grid electricity price; Step S2-2, nonlinear modeling, establishing a renewable energy consumption prediction model, and using CatBoost and XGBoost algorithms to realize nonlinear fitting; Step S2-3, hyperparameter optimization, using Bayesian optimization algorithm to search learning rate, maximum depth, and subsampling ratio, and the objective function is to minimize the relative error under the condition that R²≥0.9 in the test set.

6. The method for evaluating the suitability of layout of regional renewable energy sources according to claim 5, characterized in that, In the S2, the influence of each factor on renewable energy consumption is evaluated by using a linear regression model, the independent variables are selected by bivariate analysis, and the linear regression model is as follows: (5) In formula (5), represents the predicted renewable energy consumption, represents the intercept term, represents the input features, represents the model parameters or weights, represents the number of features; The nonlinear regression model is as follows: (6) In formula (6), represents an input feature vector; represents a prediction of the decision tree, represents the weak learner, represents the weight of each weak learner, represents the number of trees in the ensemble model, represents the number of boosting iterations.

7. The method of claim 1, wherein the method further comprises: The S3 comprises the following steps: Step S3-1, explanatory calculation, using the SHAP method, to calculate the marginal contribution degree and interaction effect of each input variable, and to obtain the historical influence curve of different factors on solar energy, wind energy and hydropower consumption; Step S3-2, threshold identification, fitting the marginal effect curve to determine the optimal threshold interval of the key variable; Step S3-3, multi-energy complementary analysis, through the variable interaction matrix, to identify the synergistic relationship of solar energy, wind energy and hydropower, and to reveal the complementary potential of multi-energy system.

8. The method of claim 1, wherein, S4 includes the following steps: Step S4-1, spatial evaluation, inputting natural and social factors grid into the nonlinear correlation model and the nonlinear marginal influence model to estimate the renewable energy consumption potential of each grid; Step S4-2, hotspot identification, based on Getis-Ord Gi* and local Moran's I algorithm, to identify high-value aggregation Hotspot and low-value aggregation Coldspot area; Step S4-3, multi-energy complementary partition, combined with energy suitability map, to divide multi-energy complementary area, and finally to generate renewable energy consumption potential map set.

9. A system for evaluating the suitability of a layout of a regional renewable energy source, characterized by, The evaluation system adopts the evaluation method of any one of claims 1 to 8, and the evaluation system comprises: A data preprocessing module: using ARIMA model to interpolate time series missing values, using Kriging interpolation and kernel density estimation to obtain meteorological and facility spatial distribution; A renewable energy-influencing factor correlation model construction module: used to establish a nonlinear correlation model of energy consumption and social economic and natural environmental factors based on CatBoost and XGBoost as core algorithms; A marginal influence curve identification module: used to calculate the marginal effect curve of each variable through the SHAP model, and to determine the optimal threshold interval of the key variable; A suitability mapping and hotspot identification module: used to realize spatial clustering and suitability mapping by using improved Getis-Ord Gi* and local Moran's I algorithm; A data server: used to store multi-source original data and preprocessing results; A modeling calculation node: used to configure GPU to execute CatBoost / XGBoost model training and SHAP calculation; A visualization terminal: used to realize dynamic display of results based on ArcGIS or WebGL; An automatic output module: used to generate result files including national suitability map, marginal curve map, hot area distribution map and renewable energy influencing factor mechanism map.

Images