Method and system for estimating extreme climate risks of wind and light resources

Abstract

The application discloses a wind and light resource extreme climate risk estimation method and system, relates to the technical field of climate risk assessment and prediction, and systematically analyzes and predicts the influence of extreme climate events on the supply capacity of wind power generation and photovoltaic power generation through multiple steps. The method first acquires historical data, identifies extreme climate characteristic indexes, and performs risk grade division. Then, the influence of the extreme climate event is analyzed by using a Bayesian network model, and a risk estimation model is trained based on a deep learning algorithm. Finally, real-time data are inputted to perform future risk prediction, and specific risk estimation results are outputted. The method integrates advanced technologies such as statistical analysis, machine learning and deep learning, realizes comprehensive, accurate and dynamic estimation of the wind and light resource extreme climate risk, improves the accuracy and timeliness of risk assessment, and provides an important decision support tool for climate services and renewable energy management.

Description

Technical Field

[0001] This invention relates to the field of climate services, and in particular to climate risk assessment and prediction technologies, specifically to methods and systems for predicting extreme climate risks of wind and solar resources. Background Technology

[0002] Extreme weather events, such as heat waves, cold waves, torrential rains, droughts, and strong winds, can adversely affect renewable energy systems like wind and solar power. Traditional climate risk assessment methods rely primarily on historical data analysis and simple statistical models, which are inadequate for dealing with increasingly complex and volatile climate patterns. They often fail to adequately account for the nonlinear characteristics of climate systems and the coupling effects of multiple factors, leading to questions about the accuracy and reliability of predictions.

[0003] Furthermore, existing risk assessment methods often fail to provide sufficiently detailed and quantitative risk prediction results, which limits their application value in practical decision-making. Decision-makers need more specific and actionable risk information to develop more targeted response strategies and management measures.

[0004] In summary, the industry lacks a precise risk assessment method that can comprehensively consider historical data, real-time observations, various climate factors, and specific industry needs. This is particularly true in the wind and solar energy sector, where the high dependence of wind and solar power on meteorological conditions necessitates an urgent method to accurately assess and predict the impact of extreme weather events on wind and solar energy supply capacity. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method and system for predicting extreme climate risks of wind and solar resources, so as to solve the problem that it is difficult to accurately predict the impact of extreme climate events on wind and solar resources in the prior art.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a method for predicting extreme climate risks of wind and solar resources, the method comprising the following steps:

[0007] S10: Obtain historical climate data and historical landscape resource data within the target prediction area;

[0008] S20: Based on the historical climate data, identify extreme climate characteristic indicators of extreme climate events through statistical analysis algorithms;

[0009] S30: Based on the extreme climate characteristic indicators, and according to the climate risk classification algorithm, the extreme climate events in the target prediction area are classified into different risk levels to obtain risk level data;

[0010] S40: Based on the risk level data and the historical wind and solar resource data, the impact of various extreme weather events on the wind power and photovoltaic power supply capacity in the target prediction area is analyzed through a Bayesian network model to obtain the extreme weather risk impact results.

[0011] S50: Based on the historical climate data, the historical wind and light resource data, the extreme climate characteristic indicators, the risk level data, and the extreme climate risk impact results, train an extreme climate risk prediction model based on a deep learning algorithm;

[0012] S60: Based on the trained extreme climate risk prediction model, input the real-time climate data, real-time wind and solar resource data, real-time extreme climate characteristic indicators and real-time risk level data of the target prediction area, predict the extreme climate risks suffered by the wind and solar resources of the target prediction area within a specified future time period, and output the risk prediction results, wherein the risk prediction results represent the impact of the predicted future extreme climate events on the supply capacity of wind power and photovoltaic power generation.

[0013] Secondly, a system for predicting extreme climate risks of wind and solar resources is provided, the system comprising:

[0014] Meteorological data acquisition equipment is used to collect real-time meteorological data for the target prediction area;

[0015] Wind power generation monitoring equipment is used to monitor the real-time operating data of wind power generation equipment within the target prediction area;

[0016] Photovoltaic power generation monitoring equipment is used to monitor the real-time operating data of photovoltaic power generation equipment within the target estimated area;

[0017] A data acquisition unit, connected to the meteorological data acquisition equipment, the wind power generation monitoring equipment, and the photovoltaic power generation monitoring equipment, is used to summarize and preprocess various types of data collected;

[0018] The communication module, connected to the data acquisition unit, is used to receive external weather forecast data, historical climate data, and historical wind and solar resource data, and to transmit the risk assessment results generated by the system to external devices.

[0019] The memory is used to store historical climate data, historical wind and solar resources data, extreme climate characteristic indicators, risk level data, extreme climate risk impact results, and trained extreme climate risk prediction models.

[0020] A processor, connected to the data acquisition unit, the communication module, and the memory, is used to execute the method described in the first aspect;

[0021] The display unit, connected to the processor, is used to display risk assessment results and related data.

[0022] The above technical solution has the following beneficial effects:

[0023] The method for predicting extreme climate risks to wind and solar resources in this invention integrates multi-source data and combines advanced algorithms such as statistical analysis, machine learning, and deep learning to achieve accurate prediction of the extreme climate risks to wind and solar resources within a specified future time period.

[0024] This invention introduces a Bayesian network model to analyze the impact of extreme weather events on the supply capacity of wind and solar power. This step not only considers the impact of single factors but also simulates the interactions and complex relationships between multiple factors, thus obtaining risk impact results that more closely reflect reality. In this way, this method can more accurately quantify the potential impact of extreme weather events on the supply of renewable energy.

[0025] This invention presents an extreme climate risk prediction model based on deep learning algorithms. It comprehensively utilizes multi-dimensional information such as historical data, feature indicators, risk levels, and impact results. By capturing the complex nonlinear relationships between data through deep learning algorithms, it achieves high-precision prediction of extreme climate risks.

[0026] This invention achieves a leap from static analysis to dynamic prediction by inputting real-time data for future risk assessment. This real-time prediction capability makes risk assessment results more timely and accurate, reflecting the latest climate change trends and potential risks. The output risk prediction results directly quantify the impact of extreme weather events on the supply capacity of wind and solar power generation, providing relevant departments with specific and actionable decision-making basis for formulating response strategies, optimizing resource allocation, and conducting emergency management. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0028] Figure 1 This is a flowchart of the extreme climate risk prediction method for wind and solar resources according to an embodiment of the present invention;

[0029] Figure 2 This is a flowchart of step S10 in an embodiment of the present invention;

[0030] Figure 3 This is a flowchart of step S20 in an embodiment of the present invention;

[0031] Figure 4This is a flowchart of step S30 in an embodiment of the present invention;

[0032] Figure 5 This is a flowchart of step S40 in an embodiment of the present invention;

[0033] Figure 6 This is a flowchart of step S50 in an embodiment of the present invention;

[0034] Figure 7 This is a flowchart of step S60 in an embodiment of the present invention;

[0035] Figure 8 This is a functional block diagram of the wind and solar resource extreme climate risk prediction system according to an embodiment of the present invention;

[0036] Figure 9 This is a functional block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Example 1

[0039] like Figure 1 As shown in the figure, this embodiment provides a method for predicting extreme climate risks of wind and solar resources. The method includes the following steps:

[0040] S10: Obtain historical climate data and historical landscape resource data within the target prediction area;

[0041] S20: Based on the historical climate data, identify extreme climate characteristic indicators of extreme climate events through statistical analysis algorithms;

[0042] S30: Based on the extreme climate characteristic indicators, and according to the climate risk classification algorithm, extreme climate events in the target prediction area are classified into different risk levels to obtain risk level data;

[0043] S40: Based on the risk level data and the historical wind and solar resource data, the impact of various extreme weather events on the supply capacity of wind power and photovoltaic power generation in the target prediction area is analyzed through a Bayesian network model to obtain the results of extreme weather risk impact.

[0044] S50: Based on the historical climate data, the historical wind and light resource data, the extreme climate characteristic indicators, the risk level data, and the extreme climate risk impact results, train an extreme climate risk prediction model based on a deep learning algorithm;

[0045] S60: Based on the trained extreme climate risk prediction model, input real-time climate data, real-time wind and solar resource data, real-time extreme climate characteristic indicators and real-time risk level data of the target prediction area, predict the extreme climate risks suffered by the wind and solar resources of the target prediction area within a specified future time period, and output the risk prediction results, wherein the risk prediction results represent the impact of the predicted future extreme climate events on the supply capacity of wind power and photovoltaic power generation.

[0046] The method for predicting extreme climate risks of wind and solar resources provided in this invention has the following advantages:

[0047] First, this method employs multi-source data fusion, comprehensively utilizing historical climate data, historical wind and solar resource data, and real-time data to provide a comprehensive and rich information foundation for risk prediction. This multi-dimensional data integration not only improves the accuracy of the prediction results but also captures the complex relationship between extreme weather events and wind and solar resources, thereby achieving a more precise risk assessment.

[0048] Secondly, this method incorporates statistical analysis algorithms and climate risk classification algorithms to identify extreme climate characteristic indicators and classify risk levels. This scientific classification and grading method makes the identification of extreme climate events more objective and accurate. In this way, this method can better distinguish between extreme climate events of different degrees, improving the precision of risk assessment.

[0049] Third, this method applies a Bayesian network model to analyze the impact of extreme weather events on the supply capacity of wind and solar power. Bayesian network models can effectively handle uncertainties and interdependencies in complex systems, making the analysis of the impact of extreme weather risks more comprehensive and accurate. This method not only considers the influence of single factors but also simulates the interactions between multiple factors, thus obtaining risk impact results that are more consistent with reality.

[0050] Fourth, this method employs an extreme climate risk prediction model based on deep learning algorithms. Deep learning algorithms possess powerful nonlinear modeling and adaptive learning capabilities, enabling the extraction of valuable features and patterns from large amounts of complex historical data. By training such a model, this method can more accurately capture the complex relationship between extreme climate events and wind and solar resources, improving the accuracy and reliability of predictions.

[0051] Finally, this method represents a leap from static analysis to dynamic prediction. By inputting real-time data for future risk assessment, this method can promptly reflect the latest climate change trends and potential risks, providing more timely and accurate forecast results. This dynamic prediction capability makes risk assessment results more practical and timely, providing strong decision support for relevant departments to formulate response strategies and conduct emergency management.

[0052] like Figure 2 As shown, in a specific embodiment of the present invention, the process of obtaining historical climate data and historical wind and solar resource data within the target prediction area in step S10 can be further refined into the following sub-steps:

[0053] S11: Determine the geographical scope and time span of the target prediction area.

[0054] For example, the target prediction area can be set to a certain province or the area where a specific wind farm or photovoltaic power station is located, and the time span can be set to the past 10 years or longer to ensure the representativeness and statistical significance of the data.

[0055] S12: Extract historical climate data for the target prediction area within the defined time span from the meteorological database.

[0056] Specifically, this historical climate data can include, but is not limited to, key meteorological parameters such as temperature, precipitation, wind speed, solar radiation intensity, and humidity. These parameters directly affect the efficiency and output of wind and solar power generation. Historical climate data can be obtained from multiple sources to ensure comprehensiveness and accuracy. For example, ground-based observation data from automatic weather stations and manual observation stations can be used; satellite remote sensing data provided by meteorological satellites can be utilized; precise precipitation and wind speed data can be obtained using Doppler weather radar systems; upper-air atmospheric data can be collected using sounding balloons; wind speed and atmospheric particulate matter can be measured using ground-based lidar; and solar radiation intensity can be measured using radiometers. These diverse data sources provide comprehensive and detailed meteorological information, providing a reliable data foundation for subsequent analysis.

[0057] S13: Extract historical supply capacity data of wind power and photovoltaic power generation for the target estimated area within the defined time span from the wind and solar resource database.

[0058] Specifically, historical power generation data refers to the actual power generation data of wind farms and photovoltaic power plants over a past period, recorded hourly or daily. For example, the hourly power generation data of a 100MW wind farm on November 1, 2024. This data reflects the actual power generation of wind farms and photovoltaic power plants under different weather conditions and is crucial foundational data for assessing the impact of extreme weather events. The wind and solar resource database is housed in the data center of the power company or energy management department. Related hardware includes wind turbine monitoring systems for wind farms, inverters and data acquisition systems for photovoltaic power plants, the power company's energy management system, smart meters and power data acquisition terminals, and servers and storage devices in the data center.

[0059] S14: Extract extreme weather event records for the target prediction area within the defined time span from the historical event record database.

[0060] Specifically, these extreme weather event records can include extreme meteorological events that significantly impact wind and solar power generation, such as strong winds, torrential rains, extreme high or low temperatures, and sandstorms. These records help identify potentially high-risk meteorological conditions. The historical event record database is housed in the data centers of meteorological departments or disaster management agencies. The hardware associated with the historical event record database includes meteorological disaster monitoring systems of meteorological departments, emergency command systems of disaster management agencies, remote sensing satellite ground receiving stations, meteorological radar networks, and servers and storage devices in the data centers. For example, the National Meteorological Information Center maintains a comprehensive database of extreme weather events, recording various extreme meteorological events nationwide. This data provides crucial information for analyzing the impact of extreme weather on wind and solar power generation.

[0061] A meteorological disaster monitoring system is a system used by meteorological departments to monitor and record various meteorological phenomena in real time. It directly provides raw data to a historical event record database. For example, when extreme weather events such as strong winds or heavy rain occur, this system records detailed meteorological parameters and inputs this information into the historical event record database.

[0062] Disaster management agencies' emergency command systems are used to coordinate responses and management during extreme weather events. They record information such as the event's scope, duration, and severity, which is then integrated into a historical event database to enrich the event's description and impact assessment.

[0063] Remote sensing satellite ground receiving stations receive data from meteorological satellites, providing large-scale observations of weather systems. They are able to capture large-scale extreme weather events, such as typhoons and sandstorms, and transmit these observational data to historical event record databases, providing a macroscopic perspective for event recording.

[0064] Weather radar networks provide high-resolution local weather observations, particularly for precise measurements of precipitation and wind fields. This detailed observational data is input into historical event databases, providing crucial information for the accurate location and intensity assessment of extreme weather events.

[0065] The servers and storage devices in a data center are the physical carriers of historical event record databases. They are responsible for storing, processing, and managing massive amounts of data collected from various monitoring systems and devices. Servers run database management systems, performing data storage, retrieval, and analysis operations, while storage devices provide large-capacity, highly reliable data storage space.

[0066] S15: Time-matching extreme weather event records with historical supply capacity data to obtain supply change data related to historical extreme weather events.

[0067] Specifically, this step establishes a direct link between extreme weather events and changes in power generation capacity, providing crucial information for subsequent risk analysis. This matching clearly shows how the supply capacity of wind and solar power changes during specific extreme weather events. This correlation analysis helps identify which types of extreme weather events have the greatest impact on wind and solar power, as well as the specific extent and duration of that impact.

[0068] Specifically, this process can be achieved through the following steps:

[0069] First, determine the precision and range of the time matching. Hourly or daily precision can be chosen, as this better balances the level of data detail with the complexity of processing. The time range should include a period before, during, and after the extreme weather event to comprehensively observe its impact. For example, for a strong typhoon event lasting 3 days, select to analyze data from 1 day before the event, 3 days during the event, and 2 days after the event, for a total of 6 days.

[0070] Secondly, extreme weather event records and historical power supply capacity data are aligned using timestamps. This involves database join operations, using timestamps as key fields. For example, if there is a record of a severe typhoon event that occurred on August 1, 2024 at 12:00, this embodiment will find the corresponding power generation data for that time point and the time periods before and after it in the historical power supply capacity data. This process can be implemented using database query languages ​​(such as SQL) or data processing tools (such as Python's pandas library).

[0071] Then, time series analysis is performed on the matched data to calculate the changes in supply capacity. This includes calculating the average supply capacity before, during, and after the event, as well as the rate of change in supply capacity. For example, this embodiment found that during a strong typhoon event, the supply capacity of wind power generation decreased significantly, by as much as 90% or more, while the supply capacity of photovoltaic power generation decreased by about 80%.

[0072] S16: Perform data cleaning on the data extracted in steps S12 to S15. This process includes removing outliers, filling in missing values, and standardizing the data.

[0073] Specifically, outlier removal can employ statistical methods, such as the 3σ criterion or interquartile range; missing value imputation can utilize interpolation or machine learning methods; and data standardization can employ methods such as z-score standardization or min-max standardization. This step helps improve data quality and ensures the accuracy of subsequent analyses. Through these processes, noise and inconsistencies in the data can be eliminated, making the dataset more reliable and consistent, and providing high-quality input for subsequent model training and risk analysis.

[0074] S17: Integrate the cleaned data into a unified dataset as historical climate data and historical landscape resource data.

[0075] Specifically, this unified dataset includes historical supply capacity data for wind and solar power within the target prediction area, as well as data on supply changes related to historical extreme weather events. This comprehensive dataset provides a robust and reliable data foundation for subsequent extreme climate characteristic identification, risk level classification, and model training. By integrating these data from different sources and types, it is possible to comprehensively reflect the climate characteristics, wind and solar resource status, and the impact of extreme weather events on power generation capacity in the target prediction area.

[0076] like Figure 3 As shown, in a specific embodiment of the present invention, step S20 specifically includes the following sub-steps:

[0077] S21: Perform time series analysis on historical climate data to identify the abnormal fluctuation patterns corresponding to temperature, precipitation, wind speed and solar radiation intensity, respectively.

[0078] This step begins by collecting and organizing historical climate data from multiple years, including key indicators such as temperature, precipitation, wind speed, and solar radiation intensity. Then, time series analysis techniques are used to identify anomalous fluctuations in these indicators. For example, for temperature data, moving averages can be used to identify abnormally high or low temperature events. Specifically, the daily average temperature over the past 30 years can be calculated as a long-term average, and then the actual temperature of each day is compared to this long-term average. If the temperature on a given day exceeds the long-term average plus two standard deviations, it is marked as an abnormally high temperature event. This method effectively captures anomalous fluctuations in climate data.

[0079] Specifically, time series analysis methods can include moving averages, autoregressive models, autoregressive integral moving average (ARIMA) models, and anomaly detection algorithms. Moving averages calculate an average value by setting a window (e.g., 30 days) to smooth data, thereby identifying long-term trends or outliers. Autoregression (AR) and moving average (MA) models utilize the autocorrelation of historical data to predict future values ​​and capture specific fluctuation patterns. ARIMA models are extensions of AR and MA, suitable for time series data with trends and seasonality, and can effectively detect long-term trends and anomalous changes in climate data such as temperature and precipitation. Furthermore, anomaly detection algorithms, such as control charts and standard deviation analysis, identify anomalies by defining the range of fluctuations within normal values. For example, when temperature data deviates significantly from the average (e.g., more than two standard deviations), it is marked as an anomalous fluctuation.

[0080] S22: Based on multiple predefined thresholds, potential extreme climate events are selected from each of the aforementioned abnormal fluctuation patterns, wherein each abnormal fluctuation pattern corresponds to one or more predefined thresholds;

[0081] This step involves setting appropriate thresholds to define extreme events. Taking precipitation as an example, the following thresholds can be set: 24-hour precipitation exceeding 50 mm is defined as heavy rain, exceeding 100 mm as torrential rain, and exceeding 250 mm as extremely heavy rain. By applying these thresholds, genuine extreme precipitation events can be filtered out from the anomalous fluctuations identified by S21. This method is not only applicable to precipitation but can also be applied to other climate indicators, such as wind speed and temperature.

[0082] S23: Perform cluster analysis on the screened potential extreme climate events to obtain multiple extreme climate event categories;

[0083] Specifically, clustering analysis techniques, such as the K-means algorithm, can be used to group the selected extreme events. Taking wind speed data as an example, the following clustering results are obtained: Category 1 consists of short-duration events with high maximum wind speeds (corresponding to tornadoes); Category 2 consists of long-duration events with moderate wind speeds (corresponding to tropical storms); and Category 3 consists of long-duration events with extremely high maximum wind speeds (corresponding to typhoons). This clustering analysis helps identify different types of extreme weather events, providing a classification foundation for subsequent feature extraction and model building.

[0084] S24: For each category of extreme weather events, extract its characteristic parameters, including event duration, maximum intensity, average intensity, and rate of change;

[0085] Specifically, for each type of extreme event, key features are extracted. Taking heatwave events as an example, the following feature parameters can be extracted: duration (number of consecutive high-temperature days), maximum intensity (highest temperature during the period), average intensity (average temperature during the heatwave), and rate of change (the rate of temperature increase, such as the number of degrees per day).

[0086] S25: Using principal component analysis algorithm, select the most representative index from the extracted feature parameters as the extreme climate characteristic index;

[0087] Principal Component Analysis (PCA) can reduce the dimensionality of data and identify the features that best explain data variability. For example, in the case of heavy rainfall events, PCA analysis shows that the first two principal components explain 85% of the data variation; these two principal components mainly consist of maximum precipitation intensity and total precipitation. Therefore, these two indicators can be selected as the main characteristic indicators of heavy rainfall events. This method not only simplifies the subsequent analysis process but also ensures that the selected indicators reflect the essential characteristics of extreme events to the greatest extent possible.

[0088] S26: Perform statistical analysis on the selected extreme climate characteristic indicators to obtain statistical analysis characteristics, including the distribution characteristics, frequency of occurrence, and trend of change of the indicators;

[0089] In this step, the distribution characteristics of the indicators describe how the data's central values ​​are distributed. This includes the central tendency (mean, median), dispersion (standard deviation, variance), and the shape of the distribution (skewness, kurtosis). In extreme weather event analysis, understanding the distribution characteristics helps in understanding the typical values ​​and variability of event intensity. For example, for extreme heat events, a right-skewed distribution of maximum temperatures means that most extreme heat events have temperatures concentrated near lower values, but a few events reach very high temperatures, causing the distribution to stretch to the right. A mean of 35°C represents the average maximum temperature of all extreme heat events, while a standard deviation of 2°C reflects the dispersion of temperature values ​​around the mean. This information helps in assessing the typical intensity and variability of extreme heat events.

[0090] Frequency refers to the number of times a specific event occurs within a certain period. It is expressed as the average number of occurrences per unit of time. Understanding the frequency of extreme weather events is crucial for risk assessment and disaster prevention and mitigation planning. In the given example, extreme heat events occur an average of 3 times per year. This information can help policymakers and the public understand the prevalence of extreme heat events and provide a basis for developing response strategies.

[0091] A trend describes the direction and rate of change of an indicator over time. In climate analysis, identifying long-term trends is crucial for understanding the impacts of climate change and predicting future events. Trends can be linear or non-linear, and are determined using statistical methods such as linear regression.

[0092] Taking extreme heat events as an example, the following statistical characteristics can be obtained: distribution characteristics (the maximum temperature shows a right-skewed distribution, with an average of 35℃ and a standard deviation of 2℃), frequency of occurrence (an average of 3 extreme heat events occur per year), and trend (the frequency of extreme heat events has increased by 1 per decade over the past 20 years). These statistical characteristics provide important information for understanding the overall pattern of extreme climate events and help identify long-term trends and potential climate change signals.

[0093] S27: Based on the aforementioned statistical analysis characteristics, establish a mathematical model for extreme climate characteristic indicators, which is used to describe the characteristics of different categories of extreme climate events;

[0094] Specifically, mathematical models are constructed using the obtained statistical characteristics to describe various extreme climate events. For example, for extreme precipitation events, the Generalized Extreme Value Distribution (GEV) can be used to simulate the annual maximum daily precipitation.

[0095]

[0096] Where μ is the location parameter, controlling the location of the extreme value distribution; σ is the scale parameter, controlling the scale (width or extent) of the extreme value distribution; ξ is the shape parameter, determining the shape of the extreme value distribution, i.e., the frequency and shape of the distribution of extreme events; and x has a physical meaning that depends on the type of extreme climate event being analyzed, and can be an observed or measured value of the extreme climate event, such as maximum precipitation or maximum temperature. F(x) represents the cumulative distribution function of the generalized extreme value distribution (GEV distribution) used in extreme event analysis, which describes the probability distribution of extreme events and is used to estimate the probability of a specific extreme value (e.g., extremely high temperature, extremely high precipitation) occurring within a certain range. These parameters can be estimated from historical data using the maximum likelihood estimation method. This mathematical model can accurately describe the probability distribution of extreme events, providing a theoretical basis for risk assessment and prediction.

[0097] S28: Apply the mathematical model of the extreme climate characteristic index to historical climate data for simulation to verify the mathematical model and obtain the verification results;

[0098] For example, data from 2000–2015 can be used to train a GEV model, and then data from 2016–2020 can be used to test the model. The model's predicted frequency and intensity of extreme precipitation events from 2016–2020 are compared with actual observations, and the root mean square error (RMSE) is calculated to evaluate model performance. This validation process helps identify the model's strengths and limitations, providing a basis for further optimization.

[0099] S29: Based on the verification results, the extreme climate characteristic indicators are quantitatively adjusted and optimized to form the final extreme climate characteristic indicator set.

[0100] Based on the validation results, the model is adjusted or the selected feature indices are re-evaluated. For example, if the Generalized Extreme Value Distribution (GEV) model is found to systematically underestimate the intensity of extreme precipitation events, it is necessary to consider introducing additional covariates (such as sea surface temperature) to improve the model. Through this iterative optimization process, an optimized set of feature indices is ultimately formed, which can most accurately describe and predict various extreme climate events.

[0101] like Figure 4 As shown, in one embodiment of the present invention, step S30 specifically includes the following sub-steps:

[0102] S31: Based on the extreme climate characteristic indicators, establish a climate risk classification model, wherein the extreme climate characteristic indicators include any multiple of extreme temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation; the high-intensity solar radiation refers to solar radiation intensity exceeding a predetermined threshold.

[0103] Specifically, step S31 establishes a climate risk classification model based on extreme climate characteristic indicators. In this embodiment, extreme temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation are selected as input indicators for the model. High-intensity solar radiation is defined as daily cumulative solar radiation exceeding 25 MJ / m². 2 The criteria for extreme temperatures include daily high temperatures exceeding 35°C and daily low temperatures below -10°C. Extreme wind speed is defined as a daily maximum wind speed exceeding 17.2 m / s (Force 8), while extreme precipitation is defined as 24-hour precipitation exceeding 50 mm. These indicators and thresholds were chosen based on their potential impact on wind and solar power systems.

[0104] S32: The climate risk classification model is trained using machine learning methods. The climate risk classification model takes extreme climate characteristic indicators as input and outputs extreme climate event types and their corresponding risk levels, where the risk levels include low risk, medium risk and high risk.

[0105] Specifically, step S32 uses machine learning methods to train a climate risk classification model; in this embodiment, the random forest algorithm is employed. The model's input consists of the aforementioned extreme climate characteristic indicators, and its output is the type of extreme climate event and its corresponding risk level (low, medium, and high risk). The training dataset includes daily meteorological data from the past 30 years and corresponding extreme event records. Each data point includes the date, location, meteorological index value, known extreme event type, and the degree of impact on wind and solar power systems. Risk levels are classified based on historical impact levels. For example, for extreme wind speed events, short-term shutdowns of wind turbines are classified as low risk, shutdowns of large-scale wind turbines for several hours are classified as medium risk, and large-scale damage to wind farms is classified as high risk. The model training process includes data preprocessing, feature engineering, model parameter optimization, and cross-validation. The hyperparameters of the random forest are optimized using a grid search method, and 10-fold cross-validation is used to evaluate model performance, ensuring the model's generalization ability.

[0106] Data preprocessing is the first step in model training, transforming raw data into a format suitable for machine learning models. Preprocessing includes steps such as handling missing values, normalization or standardization, and encoding. Missing value handling can involve imputation using the mean or median, or directly deleting data points with a large number of missing values ​​to ensure data integrity. Normalization or standardization scales feature values ​​to a similar range, avoiding model bias caused by different numerical scales. For categorical variables (such as event types), one-hot encoding can be used to convert categorical data into numerical values, making it suitable for the model's input requirements.

[0107] Feature engineering is the process of improving model prediction performance through feature selection, feature creation, and other techniques. Feature selection identifies and retains important features closely related to the target variable while removing irrelevant or redundant features, thereby improving the efficiency and accuracy of the model. Feature creation generates new features from existing data, such as calculating the fluctuation characteristics of daily average temperature or extreme wind speed, thus helping the model better capture the impact of climate change on risk. These new features often reveal information hidden in the original data.

[0108] Model parameter optimization involves adjusting the model's hyperparameters to achieve optimal training results. Hyperparameters in a random forest model, such as the number of trees (n_estimators), the maximum tree depth (max_depth), and the minimum number of sample splits per node (min_samples_split), significantly impact model performance. Proper parameter optimization can effectively prevent overfitting (performing well only on training data) or underfitting (performing poorly even on training data), allowing the model to achieve better performance on test data. By optimizing hyperparameters, the model's predictive accuracy can be improved, enabling it to more accurately identify the risk levels of extreme weather events.

[0109] Grid search is a hyperparameter optimization method that automatically tries all possible combinations of a set of hyperparameters to find the optimal parameter configuration. For example, to optimize `n_estimators` and `max_depth`, multiple alternative values ​​can be specified, and grid search will try each combination in turn, ultimately selecting the best-performing parameter combination. To improve the reliability of hyperparameter selection, grid search is combined with cross-validation. The model is evaluated using cross-validation on each set of parameters to arrive at the optimal hyperparameter combination, ensuring the model has good generalization ability on test data.

[0110] Cross-validation is a technique for evaluating model performance and generalization ability, with 10-fold cross-validation being a common method. Specifically, the dataset is divided into 10 parts. Each time, one part is used as the validation set, and the remaining 9 parts are used as the training set. The model is trained, and its performance on the validation set is evaluated. This process is repeated 10 times, each time using a different validation set. The final result is the average of the 10 validation results to obtain the overall performance evaluation of the model. This method effectively assesses the model's stability and generalization ability, ensuring consistent performance across different datasets.

[0111] S33: Using a trained climate risk classification model, analyze the extreme climate characteristic indicators of the target prediction area to obtain the analysis results, which include the types of extreme climate events in the target prediction area and their corresponding risk levels.

[0112] Specifically, in step S33, the trained climate risk classification model is used to analyze the extreme climate characteristic indicators of the target prediction area. Taking a wind power and photovoltaic power generation base as an example, the meteorological forecast data for the region for the next week is input, including daily maximum temperature, minimum temperature, expected precipitation, expected maximum wind speed, and expected solar radiation intensity. The model analysis results show that: an extreme wind speed event may occur on the second day, with the maximum wind speed expected to reach 20 m / s, and the risk level is medium risk; an extreme low temperature event may occur on the fourth day, with the minimum temperature expected to drop to -15℃, and the risk level is high risk; sustained high-intensity solar radiation may occur on the sixth and seventh days, with the daily cumulative radiation exceeding 28 MJ / m². 2 The risk level is low.

[0113] S34: Based on the analysis results, generate risk level data, which includes the types of extreme climate events in the target predicted area, the corresponding extreme climate characteristic indicators, and the risk level.

[0114] Specifically, in step S34, risk level data is generated based on the analysis results. For the analysis results of the aforementioned wind power and photovoltaic power generation bases, the generated risk level data includes, for example:

[0115] Extreme wind speed event (November 12, 2024, maximum wind speed 20m / s, medium risk);

[0116] Extreme low temperature event (November 14, 2024, minimum temperature -15℃, high risk);

[0117] High-intensity solar radiation event (November 16-17, 2024, with a daily cumulative solar radiation of 28 MJ / m²). 2 (Low risk).

[0118] These risk level data provide crucial decision support information for power generation base managers. For example, for medium-risk extreme wind speed events, the operating parameters of wind turbine generators can be adjusted in advance to ensure safety; for high-risk extreme low temperature events, enhanced equipment anti-freezing measures can be implemented to prevent potential equipment failures; and for low-risk high-intensity solar radiation events, the angle of photovoltaic panels can be optimized to maximize power generation efficiency. In this way, the climate risk classification model of this invention can provide a scientific basis for the safe operation and efficiency optimization of wind and photovoltaic power generation systems, effectively reducing the potential losses and impacts caused by extreme weather events.

[0119] like Figure 5 As shown, in one embodiment of the present invention, step S40 specifically includes the following sub-steps:

[0120] S41: Construct a Bayesian network model, using extreme climate event types, risk levels, wind power supply capacity, and photovoltaic power supply capacity as network nodes of the Bayesian network model;

[0121] In one specific embodiment of the present invention, step S41 constructs a Bayesian network model, using extreme weather event types, risk levels, wind power supply capacity, and photovoltaic power supply capacity as network nodes. This Bayesian network model includes four main nodes: extreme weather event types (extreme high temperature, extreme low temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation), risk levels (low, medium, and high), wind power supply capacity, and photovoltaic power supply capacity. The connections between nodes reflect their causal relationships; for example, extreme weather event types affect risk levels, and risk levels, in turn, affect wind and photovoltaic power supply capacities.

[0122] S42: Based on the historical scenic resource data and the risk level data, calculate the conditional probability distribution among each network node;

[0123] Step S42 calculates the conditional probability distribution among network nodes based on historical wind and solar resource data and risk level data. For example, it calculates the probability that wind power supply capacity will decrease by more than 50% under given extreme high-temperature events and high-risk levels. This step involves statistical analysis of a large amount of historical data, including wind speed, solar radiation intensity, and power generation under different extreme weather events. Using this data, a conditional probability such as 0.3 for the probability of wind power supply capacity decreasing by more than 50% under extreme high-temperature and high-risk conditions can be derived.

[0124] S43: Using the historical landscape resource data, the risk level data, and the conditional probability distribution among the network nodes, train the Bayesian network model to obtain a trained Bayesian network model.

[0125] In step S43, a Bayesian network model is trained using historical wind and solar resource data, risk level data, and the calculated conditional probability distribution. The training process includes two aspects: parameter learning and structure learning. Parameter learning primarily involves accurately calculating the conditional probability table for each node using methods such as maximum likelihood estimation or Bayesian estimation. Structure learning involves adjusting the network structure, such as adding, deleting, or reversing connections between nodes, to better fit the observed data. For example, it might be found that extreme wind speed events have a more direct impact on wind power generation than expected; therefore, a connection directly linking the extreme wind speed event to the wind power generation capacity could be added to the model.

[0126] S44: Using a trained Bayesian network model, reason about different types and risk levels of extreme climate events to obtain data on expected supply loss of wind and solar resources, which serves as the result of the impact of extreme climate risks.

[0127] In step S44, a trained Bayesian network model is used to infer the expected supply loss data of wind and solar resources for different extreme weather event types and risk levels. For example, given an extreme low temperature event (-20℃) and a high risk level, the model can infer that the probability of a 30% decrease in wind power supply capacity and a 45% decrease in photovoltaic power supply capacity is the highest. This inference can be performed for various extreme weather scenarios, such as extreme high temperature (40℃), medium risk; extreme wind speed (25m / s), high risk, etc. In this way, the Bayesian network model can provide detailed predictions of wind and solar resource supply losses for different extreme weather scenarios, helping grid operators and energy planners to develop more accurate response strategies, such as adjusting power generation plans, preparing backup power sources, or implementing demand-side management measures.

[0128] like Figure 6 As shown, in one embodiment of the present invention, step S50 specifically includes the following sub-steps:

[0129] S51: Construct a deep learning neural network model, which includes:

[0130] The input layer is used to receive preprocessed input features;

[0131] Multiple hidden layers are used to perform nonlinear transformations and feature extraction on the preprocessed input features. Adjacent hidden layers are connected by weight matrices and activation functions.

[0132] The output layer, used to generate extreme climate risk prediction results, is connected to the last hidden layer via a weight matrix;

[0133] In this embodiment, the deep learning neural network model employs a multilayer perceptron (MLP) structure. Specifically, the input layer contains 100 neurons, corresponding to the dimensions of the preprocessed input features. The hidden layers employ a three-layer structure, with each layer containing 64, 32, and 16 neurons respectively, using the ReLU activation function for nonlinear transformation. The output layer contains one neuron, used to predict the degree of impact of extreme climate risks, and employs a linear activation function.

[0134] S52: Take the historical climate data, the historical wind and light resource data, the extreme climate characteristic indicators and the risk level data as input features, take the extreme climate risk impact results as output labels, and preprocess the input features to obtain preprocessed input features;

[0135] In this step, input features can include daily average temperature, wind speed, precipitation, solar radiation intensity, and corresponding wind and solar power generation data for the past 10 years. Extreme climate characteristic indicators and risk level data are derived from the analysis results of previous steps. Preprocessing includes data standardization (scaling each feature to a distribution with a mean of 0 and a variance of 1) and missing value handling (filling with the average values ​​of consecutive time points). The output label is the percentage of actual loss caused to wind and solar power systems by extreme climate events.

[0136] S53: Input the preprocessed input features into the deep learning neural network model;

[0137] The preprocessed input features are organized into batches (batch size set to 64) and sequentially fed into the neural network model. Each sample contains 100 feature values, corresponding to 100 neurons in the input layer.

[0138] S54: The deep learning neural network model is trained using the backpropagation algorithm and optimizer. By iteratively adjusting the network parameters, the error between the prediction results and the actual extreme climate risk impact results is minimized, and the trained deep learning neural network model is obtained.

[0139] This embodiment employs the Adaptive Moment Estimation (Adam) optimizer with a learning rate of 0.001 and a Mean Squared Error (MSE) loss function. The backpropagation algorithm calculates the gradient of the loss function with respect to the weights of each layer, and then updates the network parameters using gradient descent. After each epoch, the performance of the deep learning neural network model is evaluated on a validation set to prevent overfitting.

[0140] S55: Repeat steps S53 to S54 until the performance of the deep learning neural network model reaches a preset threshold or the number of iterations reaches a predetermined value, to obtain the trained extreme climate risk prediction model.

[0141] The model training process was set to a maximum of 1000 epochs, with an early stopping strategy: training stopped if the loss on the validation set did not improve for 10 consecutive epochs. Ultimately, the model achieved an R-value of 0.95 on the validation set. 2 The score and root mean square error (RMSE) of 0.05 meet the preset performance threshold requirements. The trained model can accurately predict the loss of wind and solar resource supply under different extreme weather scenarios, providing strong support for power grid dispatch and risk management.

[0142] like Figure 7 As shown, in one embodiment of the present invention, step S60 specifically includes the following sub-steps:

[0143] S61: Obtain real-time climate data and real-time wind and solar resource data for the target prediction area;

[0144] In this embodiment, the target estimation area is a provincial power grid coverage area. Climate data such as temperature, wind speed, precipitation, and solar radiation intensity, as well as wind power generation and photovoltaic power generation data, are acquired in real time within this area through a meteorological station network and a power monitoring system. Data is collected every 5 minutes to ensure real-time data accuracy.

[0145] S62: Based on the real-time climate data, the same statistical analysis algorithm as in step S20 is used to calculate the real-time extreme climate characteristic index;

[0146] The extreme climate characteristic index system established in step S20 is used to analyze real-time climate data. For example, the percentile of the current temperature relative to the same period in history is calculated to determine whether the extreme high or low temperature standard has been reached; the wind speed is analyzed to determine whether extreme wind conditions exist, etc.

[0147] S63: Based on the real-time extreme climate characteristic indicators, the same climate risk classification algorithm as in step S30 is used to calculate the real-time risk level data.

[0148] The real-time extreme climate characteristic indicators obtained in step S62 are input into the climate risk classification model trained in step S30. The model outputs the risk level at the current moment, for example, dividing the risk level into three levels: low, medium, and high, and giving a specific risk score (e.g., 0-100 points).

[0149] S64: Input the real-time climate data, the real-time wind and solar resource data, the real-time extreme climate characteristic indicators, and the real-time risk level data into the extreme climate risk prediction model trained in step S50.

[0150] S65: Using the extreme climate risk prediction model, calculate the probability distribution of extreme climate risks to the wind and solar resources of the target prediction area within a specified future time period;

[0151] Based on the input data, the model predicts the risk probability distribution for the next 24, 48, and 72 hours. For example, the model might output probability distribution information such as: "The probability of wind power generation in this area decreasing by more than 30% in the next 24 hours is 0.2, and the probability of decreasing by more than 50% is 0.05".

[0152] S66: Based on a preset risk threshold, the probability distribution is converted into a specific risk level and degree of impact;

[0153] Risk thresholds are set; for example, a probability greater than 0.3 indicates high risk, 0.1-0.3 indicates medium risk, and less than 0.1 indicates low risk. The probability distribution is then converted into risk levels. Simultaneously, the degree of impact is determined based on the predicted supply loss; for example, a supply loss of less than 30% indicates a mild impact, 30%-50% indicates a moderate impact, and more than 50% indicates a severe impact.

[0154] S67: Generate and output risk prediction results including risk level, impact degree and possible extreme climate event types. The risk prediction results include: the impact of future extreme climate on the supply capacity of wind and solar resources in the target prediction area, the numerical range of predicted supply loss and the predicted supply interruption time.

[0155] The final result is a detailed risk assessment report, such as: "Within the next 48 hours, the target area may face a high-risk extreme high-temperature event, which is expected to have a moderate impact on photovoltaic power generation. Specifically, the photovoltaic power generation supply capacity may decrease by 35%-45%, with an expected duration of 6-8 hours. It is recommended to take corresponding emergency measures." Such risk assessment results provide specific and actionable guidance for grid dispatch and risk management.

[0156] In a specific embodiment of the present invention, the method may further include step S70: generating a corresponding risk warning based on the risk prediction results and providing suggestions on coping strategies to reduce the impact of extreme weather on the supply of wind and solar resources.

[0157] Specifically, the system first automatically generates risk warnings of corresponding levels based on the risk assessment results obtained in step S60. Warning levels are divided into four levels: blue, yellow, orange, and red, corresponding to mild, moderate, moderate, and severe risk levels, respectively. Warning information includes elements such as warning level, warning time, warning area, warning content, and warning impact. For example, when the assessment results indicate that photovoltaic power generation may face a 35%-45% supply loss within the next 48 hours, the system will issue a yellow warning, which might read: "Issued at 14:00 on November 12, 2024, Yellow Warning for District B, City A, Province: Sustained high temperatures are expected in this area within the next 48 hours, affecting photovoltaic power generation equipment. Photovoltaic power generation supply capacity may decrease by 35%-45%, with an expected duration of 6-8 hours."

[0158] Secondly, based on early warning information and historical response experience, the system automatically generates targeted response strategy recommendations. These recommendations help relevant departments and grid operators take timely measures to minimize the impact of extreme weather on wind and solar resource supply. Response strategy recommendations include aspects such as power generation equipment protection, grid dispatch optimization, demand-side management, emergency resource allocation, and enhanced monitoring and early warning. For example, the system may recommend: cooling photovoltaic panels and, if necessary, partial shading; increasing the reserve capacity of conventional energy generation to ensure grid supply and demand balance; implementing peak-shifting electricity plans to guide large users to consume electricity during non-high-temperature periods; deploying mobile generators to key areas to address potential power shortages; and strengthening real-time monitoring of photovoltaic power generation equipment to promptly detect and handle abnormal situations.

[0159] Finally, the system will send the generated risk warnings and response strategy recommendations to relevant responsible persons and decision-makers through various channels. These channels include, but are not limited to, real-time displays at the power dispatch and control center, mobile application push notifications, SMS and email alerts, and automated voice call notifications. Through timely and accurate risk warnings and targeted response strategy recommendations, grid operators and relevant departments can prepare in advance, take necessary preventative and responsive measures, effectively reduce the adverse impact of extreme weather on wind and solar resource supply, and ensure the safe and stable operation of the power grid.

[0160] Example 2

[0161] like Figure 8 As shown, this embodiment provides a wind and solar resource extreme climate risk prediction system 200, the system 200 including:

[0162] Meteorological data acquisition equipment 210 is used to collect real-time meteorological data of the target prediction area;

[0163] Wind power generation monitoring equipment 220 is used to monitor the real-time operating data of wind power generation equipment within the target prediction area;

[0164] Photovoltaic power generation monitoring equipment 230 is used to monitor the real-time operating data of photovoltaic power generation equipment within the target prediction area;

[0165] The data acquisition unit 240 is connected to the meteorological data acquisition device 210, the wind power generation monitoring device 220 and the photovoltaic power generation monitoring device 230, and is used to summarize and preprocess the various types of data collected;

[0166] The communication module 250 is connected to the data acquisition unit 240 and is used to receive external weather forecast data, historical climate data and historical wind and light resource data, and transmit the risk prediction results generated by the system to external devices.

[0167] The memory 260 is used to store historical climate data, historical wind and light resource data, extreme climate characteristic indicators, risk level data, extreme climate risk impact results, and trained extreme climate risk prediction models.

[0168] The processor 270, connected to the data acquisition unit 240, the communication module 250 and the memory 260, is used to execute the extreme climate risk prediction method for wind and solar resources in Embodiment 1;

[0169] Display unit 280, connected to processor 270, is used to display risk assessment results and related data.

[0170] Specifically, the historical climate data includes: temperature, precipitation, wind speed, solar radiation intensity, and humidity;

[0171] The historical wind and solar resource data includes: historical supply capacity data of wind power and photovoltaic power generation in the target estimated area, as well as supply change data related to historical extreme weather events;

[0172] The extreme climate characteristic indicators include any multiple of the following: extreme temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation; the high-intensity solar radiation refers to solar radiation intensity exceeding a predetermined threshold.

[0173] The extreme climate risk impact results are data on the expected supply loss of wind and solar resources under different risk levels;

[0174] The risk assessment results include: the impact of future extreme weather on the supply capacity of wind and solar resources in the target area, the range of predicted supply losses, and the predicted time of supply interruption.

[0175] The meteorological data acquisition device 210 includes a temperature sensor, a humidity sensor, an anemometer, a wind vane, a rain gauge, a barometer, and a radiometer, used to measure ambient temperature, relative humidity, wind speed, wind direction, precipitation, atmospheric pressure, and solar radiation intensity, respectively. In wind power generation monitoring, a power sensor, a speed sensor, and a vibration sensor are configured to measure the wind turbine's output power, rotor speed, and monitor its operating status. In photovoltaic power generation monitoring, a voltage sensor, a current sensor, and a temperature sensor are provided to measure the photovoltaic panel's output voltage, current, and surface temperature. Furthermore, the data acquisition unit 240 also includes an analog-to-digital converter and signal conditioning circuitry to convert analog signals into digital signals and perform preliminary processing to ensure data accuracy and consistency.

[0176] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0177] Example 3

[0178] In one embodiment, a computer device is provided, which can be a terminal or a server. When the computer device is a terminal, its internal structure diagram can be as follows: Figure 9 As shown. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method. Those skilled in the art will understand that... Figure 9The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0179] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method.

[0180] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), etc.

[0181] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0182] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for predicting extreme climate risks of wind and solar resources, characterized in that, The method includes the following steps: S10: Obtain historical climate data and historical landscape resource data within the target prediction area; S20: Based on the historical climate data, identify extreme climate characteristic indicators of extreme climate events through statistical analysis algorithms; S30: Based on the extreme climate characteristic indicators, and according to the climate risk classification algorithm, the extreme climate events in the target prediction area are classified into different risk levels to obtain risk level data; S40: Based on the risk level data and the historical wind and solar resource data, the impact of various extreme weather events on the wind power and photovoltaic power supply capacity in the target prediction area is analyzed through a Bayesian network model to obtain the extreme weather risk impact results. S50: Based on the historical climate data, the historical wind and light resource data, the extreme climate characteristic indicators, the risk level data, and the extreme climate risk impact results, train an extreme climate risk prediction model based on a deep learning algorithm; S60: Based on the trained extreme climate risk prediction model, input the real-time climate data, real-time wind and solar resource data, real-time extreme climate characteristic indicators and real-time risk level data of the target prediction area, predict the extreme climate risk suffered by the wind and solar resources of the target prediction area within a specified future time period, and output the risk prediction result, wherein the risk prediction result represents the impact of the predicted future extreme climate events on the supply capacity of wind power and photovoltaic power generation. Specifically, step S30 includes the following sub-steps: S31: Based on the extreme climate characteristic indicators, establish a climate risk classification model, wherein the extreme climate characteristic indicators include any multiple of extreme temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation; the high-intensity solar radiation refers to solar radiation intensity exceeding a predetermined threshold. S32: The climate risk classification model is trained using machine learning methods. The climate risk classification model takes extreme climate characteristic indicators as input and outputs extreme climate event types and their corresponding risk levels, where the risk levels include low risk, medium risk and high risk. S33: Using a trained climate risk classification model, analyze the extreme climate characteristic indicators of the target prediction area to obtain the analysis results, which include the types of extreme climate events in the target prediction area and their corresponding risk levels. S34: Based on the analysis results, generate risk level data, which includes the types of extreme climate events in the target prediction area, the corresponding extreme climate characteristic indicators, and the risk level; Specifically, step S40 includes the following sub-steps: S41: Construct a Bayesian network model, using extreme climate event types, risk levels, wind power supply capacity, and photovoltaic power supply capacity as network nodes of the Bayesian network model; S42: Based on the historical scenic resource data and the risk level data, calculate the conditional probability distribution among each network node; S43: Using the historical landscape resource data, the risk level data, and the conditional probability distribution among the network nodes, train the Bayesian network model to obtain a trained Bayesian network model. S44: Using a trained Bayesian network model, reason about different types and risk levels of extreme climate events to obtain data on expected supply loss of wind and solar resources, which serves as the result of the impact of extreme climate risks.

2. The method according to claim 1, characterized in that, Step S10 specifically includes the following sub-steps: S11: Determine the geographical scope and time span of the target prediction area; S12: Extract historical climate data of the target prediction area within the time span from the meteorological database; The historical climate data includes: temperature, precipitation, wind speed, solar radiation intensity, and humidity; S13: Extract historical supply capacity data of wind power and photovoltaic power generation for the target estimated area within the time span from the wind and solar resource database; S14: Extract extreme weather event records for the target prediction area within the time span from the historical event record database; S15: Time-match the extreme weather event records with the historical supply capacity data to obtain supply change data related to historical extreme weather events; S16: Perform data cleaning processing on the data extracted in steps S12 to S15, which includes removing outliers, filling in missing values, and standardizing data. S17: Integrate the cleaned data into a unified dataset as the historical climate data and historical wind and solar resource data; the historical wind and solar resource data includes: historical supply capacity data of wind power and photovoltaic power generation in the target prediction area, as well as supply change data related to historical extreme weather events.

3. The method according to claim 1, characterized in that, Step S20 specifically includes the following sub-steps: S21: Perform time series analysis on historical climate data to identify the abnormal fluctuation patterns corresponding to temperature, precipitation, wind speed and solar radiation intensity, respectively. S22: Based on multiple predefined thresholds, potential extreme climate events are selected from each of the aforementioned abnormal fluctuation patterns, wherein each abnormal fluctuation pattern corresponds to one or more predefined thresholds; S23: Perform cluster analysis on the screened potential extreme climate events to obtain multiple extreme climate event categories; S24: For each category of extreme weather events, extract its characteristic parameters. The characteristic parameters include event duration, maximum intensity, average intensity, and rate of change; S25: Using principal component analysis algorithm, select the most representative index from the extracted feature parameters as the extreme climate characteristic index; S26: Perform statistical analysis on the selected extreme climate characteristic indicators to obtain statistical analysis characteristics, including the distribution characteristics, frequency of occurrence, and trend of change of the indicators; S27: Based on the aforementioned statistical analysis characteristics, establish a mathematical model for extreme climate characteristic indicators, which is used to describe the characteristics of different categories of extreme climate events; S28: Apply the mathematical model of the extreme climate characteristic index to historical climate data for simulation to verify the mathematical model and obtain the verification results; S29: Based on the verification results, the extreme climate characteristic indicators are quantitatively adjusted and optimized to form the final extreme climate characteristic indicator set.

4. The method according to claim 1, characterized in that, Step S50 specifically includes the following sub-steps: S51: Construct a deep learning neural network model, which includes: The input layer is used to receive preprocessed input features; Multiple hidden layers are used to perform nonlinear transformations and feature extraction on the preprocessed input features. Adjacent hidden layers are connected by weight matrices and activation functions. The output layer, used to generate extreme climate risk prediction results, is connected to the last hidden layer via a weight matrix; S52: Take the historical climate data, the historical wind and light resource data, the extreme climate characteristic indicators and the risk level data as input features, take the extreme climate risk impact results as output labels, and preprocess the input features to obtain preprocessed input features; S53: Input the preprocessed input features into the deep learning neural network model; S54: The deep learning neural network model is trained using the backpropagation algorithm and optimizer. By iteratively adjusting the network parameters, the error between the prediction results and the actual extreme climate risk impact results is minimized, and the trained deep learning neural network model is obtained. S55: Repeat steps S53 to S54 until the performance of the deep learning neural network model reaches a preset threshold or the number of iterations reaches a predetermined value, to obtain the trained extreme climate risk prediction model.

5. The method according to claim 1, characterized in that, Step S60 specifically includes the following sub-steps: S61: Obtain real-time climate data and real-time wind and solar resource data for the target prediction area; S62: Based on the real-time climate data, the same statistical analysis algorithm as in step S20 is used to calculate the real-time extreme climate characteristic index; S63: Based on the real-time extreme climate characteristic indicators, the same climate risk classification algorithm as in step S30 is used to calculate the real-time risk level data. S64: Input the real-time climate data, the real-time wind and solar resource data, the real-time extreme climate characteristic indicators, and the real-time risk level data into the extreme climate risk prediction model trained in step S50. S65: Using the extreme climate risk prediction model, calculate the probability distribution of extreme climate risks to the wind and solar resources of the target prediction area within a specified future time period; S66: Based on a preset risk threshold, the probability distribution is converted into a specific risk level and degree of impact; S67: Generate and output risk prediction results including risk level, impact degree and extreme climate event type. The risk prediction results include: the impact of future extreme climate on the supply capacity of wind and solar resources in the target prediction area, the numerical range of predicted supply loss and the predicted supply interruption time.

6. The method according to claim 1, characterized in that, Also includes: S70: Based on the risk forecast results, generate corresponding risk warnings and provide suggestions on coping strategies to reduce the impact of extreme weather on the supply of wind and solar resources.

7. A system for predicting extreme climate risks of wind and solar resources, characterized in that, The system includes: Meteorological data acquisition equipment is used to collect real-time meteorological data for the target prediction area; Wind power generation monitoring equipment is used to monitor the real-time operating data of wind power generation equipment within the target prediction area; Photovoltaic power generation monitoring equipment is used to monitor the real-time operating data of photovoltaic power generation equipment within the target estimated area; A data acquisition unit, connected to the meteorological data acquisition equipment, the wind power generation monitoring equipment, and the photovoltaic power generation monitoring equipment, is used to summarize and preprocess various types of data collected; The communication module, connected to the data acquisition unit, is used to receive external weather forecast data, historical climate data, and historical wind and solar resource data, and to transmit the risk assessment results generated by the system to external devices. The storage device is used to store historical climate data, historical wind and solar resources data, extreme climate characteristic indicators, risk level data, extreme climate risk impact results, and trained extreme climate risk prediction models. A processor, connected to the data acquisition unit, the communication module, and the memory, is used to execute the method according to any one of claims 1-6; The display unit, connected to the processor, is used to display risk assessment results and related data.

8. The system according to claim 7, characterized in that, The historical climate data includes: temperature, precipitation, wind speed, solar radiation intensity, and humidity; The historical wind and solar resource data includes: historical supply capacity data of wind power and photovoltaic power in the target estimated area, as well as supply change data related to historical extreme weather events; The extreme climate characteristic indicators include any multiple of the following: extreme temperature, extreme wind speed, extreme precipitation, and high-intensity solar radiation; the high-intensity solar radiation refers to solar radiation intensity exceeding a predetermined threshold. The extreme climate risk impact results are data on the expected supply loss of wind and solar resources under different risk levels; The risk assessment results include: the impact of future extreme weather on the supply capacity of wind and solar resources in the target area, the range of predicted supply losses, and the predicted time of supply interruption.

Images