A joint power forecasting method for hydro-wind-solar hybrid system considering extreme climate scenarios

Abstract

The power joint forecasting method of water-wind-solar complementary system considering extreme climate scenarios is used to solve the defects of the current traditional power forecasting model in the existing technology that lacks consideration of extreme climate scenario conditions and has low forecast accuracy. This method extracts floods based on super-quantitative threshold sampling methods, extracts drought events based on runoff drought index, and extracts strong wind events based on spatiotemporal factors. Then, a Transformer water-wind-solar power forecasting model considering multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios is constructed. The power historical data of extreme climate events such as floods, droughts, and strong winds and the water-wind-solar system, as well as the related meteorological data such as rainfall, evaporation, wind speed, solar radiation, and water resource data such as inflow runoff are used as inputs, and the predicted power of water-wind-solar at future moments is output to achieve high-precision joint forecasting of water-wind-solar power under extreme climate conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system power forecasting, and in particular relates to a method for jointly forecasting power of a water-wind-solar complementary system taking into account extreme climate scenarios. Background Art

[0002] Existing research has mainly used the following methods to achieve accurate short-term power forecasting: (1) Physical model. Physical models consider meteorological or hydrological processes to forecast physical quantities related to power generation, such as wind speed, wind direction, solar radiation, temperature, precipitation, evaporation and runoff, and then convert them into power generation through the power curve of wind and solar power stations or the dynamic characteristic curve of hydropower units. (2) Statistical model. Statistical models are based on historical measured meteorological, hydrological and power data, and use curve fitting, parameter estimation and other methods to construct a mapping relationship between output variables and prediction factors to perform power forecasting. (3) Combined model. Combined models forecast power by combining different models, which can make full use of the effective information and advantages of different models. Compared with a single model, a combined model can effectively reduce the frequency of large forecast errors and usually has higher accuracy.

[0003] The shortcomings of the existing technology are: in the water-wind-solar complementary system, the access of wind power and photovoltaic power has changed the original dispatching and operation strategy of the hydropower station, and the power of the renewable energy power station is affected by different meteorological and hydrological elements, showing the characteristics of volatility and randomness, which affects the accuracy of power forecasting. The current research has not yet found a specific solution to carry out high-precision joint power forecasting of the water-wind-solar complementary system while taking into account water resources and extreme climate conditions such as floods, droughts, and strong winds. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a joint power forecasting method for a water-wind-solar complementary system taking into account extreme climate scenarios, which not only improves the forecasting accuracy of the power of single-type power stations such as hydropower, wind power and photovoltaic power, but also achieves a higher-precision total power forecast while making the method more direct and simple.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] The power joint prediction method of the hydro-wind-solar hybrid system considering extreme climate scenarios includes the following steps:

[0007] S1. Extraction of extreme climate events: Extraction of extreme events including floods, droughts and strong winds is carried out based on selected data including historical runoff data, rainfall data and wind speed data of the study area;

[0008] S2. Establish a Transformer hydropower, wind-solar-solar power prediction model that considers multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios;

[0009] S3. Carry out training and evaluation of Transformer hydropower, wind-solar-solar power prediction model;

[0010] S4. After the Transformer water, wind and photovoltaic power prediction model is trained, the trained model is deployed to the production environment to predict the power output of hydropower, wind power and photovoltaic power in real time. According to the actual situation and prediction error, the model and multi-dimensional upper and lower limits and output balance constraints are adjusted in real time to ensure the accuracy and reliability of the prediction results.

[0011] Preferably, in step S1, the method for extracting extreme climate events is as follows:

[0012] 1) Extract flood events based on ultra-quantitative threshold sampling method;

[0013] 2) Extract drought events based on runoff drought indicators;

[0014] 3) Extract strong wind events based on the strong wind identification method of spatiotemporal factors.

[0015] Preferably, 1) the process of extracting flood events based on the ultra-quantitative threshold sampling method is as follows:

[0016] The water inflow data of the research reservoir is sampled by over-quantitative threshold, and the threshold value interval is determined according to the over-quantitative sample mean method. The mean of the part of the over-quantitative sample exceeding the threshold value is Threshold S It is a linear function relationship, where is the over-quantitative sample mean;

[0017] Then, the flood samples in the study area are selected according to the threshold value, and the flood peak independence is judged by the following formula. The flood peak that meets the independence is the extracted flood sample:

[0018] ;

[0019] Where: D is the time interval between two consecutive flood peaks, d ; A is the basin area, km 2 ; For the i Maximum daily flow of flood water, m³ / s; It is the minimum flow between two flood peaks, m³ / s.

[0020] Preferably, 2) the process of extracting drought events based on runoff drought indicators is as follows:

[0021] Based on the hourly runoff data collected in the study area, the calculation period is selected, the runoff volume within the selected calculation period is calculated, and the runoff drought index is calculated based on the following formula:

[0022] ;

[0023] Where: Q is the runoff volume in the current calculation period, m³; Q avg is the average runoff over many years, m³; is the standard deviation of multi-year runoff, m³.

[0024] Preferably, 3) the process of extracting strong wind events based on the spatiotemporal factor strong wind identification method is as follows:

[0025] Sort the collected historical wind speed data in ascending order: After determining the sorting, a certain frequency value of the wind speed sequence is selected according to the geographical environment, climate and other indicators of the wind power station, and it is defined as the threshold of extreme "strong wind" events. When the maximum wind speed of a wind power station at a certain moment exceeds the threshold for more than two consecutive periods, it is defined as an extreme "strong wind" event.

[0026] Preferably, in step S2, a power prediction penalty equation for the water-wind-solar complementary system is first established; and then a Transformer water-wind-solar power prediction model is established that takes into account multi-dimensional upper and lower limits and output balance constraints.

[0027] Preferably, 1) establish a power prediction penalty equation for the hydro-wind-solar hybrid system:

[0028] By constructing a power prediction penalty equation for a water-wind-solar hybrid system, the model prediction results can meet the multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios; the power prediction penalty equation for a water-wind-solar hybrid system L The form is as follows:

[0029]

[0030] Where: The actual power generation capacity of hydropower, wind power and photovoltaic power; is the penalty term for the error between the simulated power generation and the actual power generation; Penalties for violating the upper and lower limits of the output of hydropower, wind and solar power stations; It is the penalty term for violating the balance constraint of the sum of the total power of the hydropower, wind-solar system and the single power of the hydropower, wind-solar system; is the weight coefficient of different penalty items, and its value is selected according to the actual prediction requirements;

[0031] In order to simulate the penalty term for the error between the generated power and the actual generated power, the error between the hydropower, wind power and photovoltaic power and the actual corresponding power is considered. The expression is as follows:

[0032] ;

[0033] Where: M is the time series length of the training period; Simulate power generation for hydropower, wind power and photovoltaic power;

[0034] It is a penalty item for violating the upper and lower limits of the output of hydropower, wind and solar power stations. The output boundary constraint includes the upper and lower boundaries of the hydropower, wind and solar power output. The upper and lower boundaries are determined according to historical power data. The expression is as follows:

[0035] ;

[0036] ;

[0037] Where: For the k The output limit of the power; For the k The lower limit of the power output;

[0038] is the penalty term for violating the balance constraint of the total power of the hydro-wind-solar system and the single power of the hydro-wind-solar system. The expression is as follows:

[0039] ;

[0040] Where: is the total power of the measured system.

[0041] Preferably, the construction process of the Transformer hydropower, wind-solar-solar-power prediction model considering multi-dimensional upper and lower limits and output balance constraints is as follows:

[0042] 1) Collect and organize previously extracted extreme climate events such as floods, droughts, strong winds, historical hydropower, wind power, photovoltaics and total power data, and related rainfall, evaporation, wind speed, solar radiation meteorological data, and inflow water resources data, and combine historical power data, weather data, and environmental data into a multidimensional time series feature matrix as the input of the hydropower, wind power, and photovoltaic power prediction model;

[0043] 2) Generate an attention weight matrix through dot product operation to quantify the similarity between different positions, and perform weighted aggregation of the relevant features of each position in the input sequence with other positions to enhance the effect of feature extraction. Then add position encoding to the input data, integrate the position information into the input features, and input the input data containing the position encoding into the Transformer model for processing, so that the model can make full use of the position relationship in the sequence for learning and reasoning;

[0044] 3) Based on the mapping relationship between the input feature matrix and the power output, the model is used to generate preliminary hydropower, wind power, and photovoltaic power forecast values. Then, these preliminary forecast results are adjusted through the power forecast penalty equation of the hydro-wind-solar complementary system. The adjustment formula is as follows:

[0045] ;

[0046] Where: is the learning rate; is the gradient of the penalty equation with respect to the initial prediction value;

[0047] Continue to iterate the above process until the penalty equation L Convergence means that each power prediction value satisfies the upper and lower limit constraints and output balance constraints.

[0048] Preferably, S3, conduct training and evaluation of the Transformer water, wind and solar power prediction model: during the model training process, by defining the minimization penalty equation as the goal, use the goose optimization algorithm to tune the parameters, calculate the gradient of each parameter through back propagation, and dynamically adjust the learning rate and momentum to ensure stable and fast convergence, and continuously evaluate and adjust the model performance through cross-validation. In multiple training iterations, through the division and evaluation of training sets and validation sets, gradually optimize the model structure and parameter configuration, and finally ensure that the model can achieve high-precision power prediction while meeting all constraints.

[0049] A water-wind-solar complementary system power joint forecasting system considering extreme climate scenarios adopts the water-wind-solar complementary system power joint forecasting method considering extreme climate scenarios.

[0050] The present invention can achieve the following beneficial effects:

[0051] The present invention takes into account water resources and extreme climate scenarios such as floods, droughts, high temperatures, and strong winds, and constructs a joint power forecasting model for water-wind-solar systems that takes into account multi-dimensional upper and lower limits and output balance constraints. It not only improves the forecast accuracy of the power of single-type power stations such as hydropower, wind power, and photovoltaic power, but also achieves a higher-precision total power forecast while making the method more direct and simple, which can provide certain technical support for the operation planning of multi-energy complementary systems, power generation planning, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0053] Figure 1 The flood excess quantitative threshold value of the present invention is selected;

[0054] Figure 2 A schematic diagram of the construction process of the Transformer water, wind and solar power prediction model considering multi-dimensional upper and lower limits and output balance constraints of the present invention;

[0055] Figure 3 This is the power joint prediction result of the water-wind-solar complementary system of the present invention. DETAILED DESCRIPTION

[0056] Under the conditions of global climate change, extreme climates occur frequently. In the water-wind-solar complementary system, the traditional power forecast has low prediction accuracy under floods, other water resources and extreme climate conditions. In view of the above problems, the present invention proposes a method for joint power forecasting of water-wind-solar complementary systems considering extreme climate scenarios. First, floods are extracted based on the super-quantitative threshold sampling method, drought events are extracted based on the runoff drought index, and strong wind events are extracted based on the spatiotemporal factor strong wind identification method; then, a Transformer water-wind-solar power prediction model considering multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios is constructed. The power historical data of extreme climate events such as floods, droughts, strong winds and water-wind-solar systems, as well as related meteorological data such as rainfall, evaporation, wind speed, solar radiation, and water resource data such as inflow runoff are used as input, and the predicted water-wind-solar power at future moments is output to achieve high-precision joint forecasting of water-wind-solar power under extreme climate conditions. The specific scheme is as follows:

[0057] S1. Extraction of extreme climate events: Extraction of extreme events such as floods, droughts, and strong winds is carried out based on the historical runoff data, rainfall data, wind speed data, etc. of the selected study area.

[0058] 1) Extract flood events based on ultra-quantitative threshold sampling method;

[0059] The water inflow data of the research reservoir is sampled by over-quantitative threshold, and the threshold value interval is determined according to the over-quantitative sample mean method. The mean of the part of the over-quantitative sample exceeding the threshold value is linearly related to the threshold value S, such as Figure 1 As shown, where is the mean of the over-quantitative samples. Then, the flood samples in the study area are selected according to the threshold value, and then the flood peak independence is judged using the following formula. The flood peak that meets the independence is the extracted flood sample:

[0060] ;

[0061] Where: Dis the time interval between two consecutive flood peaks, d ; A is the basin area, in km 2 ; For the i The maximum daily flow rate of flood in the field, in m³ / s; It is the minimum flow rate between two flood peaks, in m³ / s.

[0062] 2) Extracting drought events based on runoff drought indicators;

[0063] Based on the hourly runoff data collected in the study area, the calculation period is selected, the runoff volume within the selected calculation period is calculated, and the runoff drought index is calculated based on the following formula:

[0064] ;

[0065] Where: Q is the runoff volume in the current calculation period, in m³; Q avg is the multi-year average runoff, in m³; is the standard deviation of multi-year runoff, in m³.

[0066] According to the calculated RI As a result, it is believed that RI When it is less than or equal to -0.5, it is considered a drought event.

[0067] 3) Extract strong wind events based on the strong wind identification method of spatiotemporal factors.

[0068] Sort the collected historical wind speed data in ascending order: After determining the sorting, a certain frequency value of the wind speed sequence is selected according to the geographical environment, climate and other indicators of the wind power station, and it is defined as the threshold of extreme "strong wind" events. When the maximum wind speed of a wind power station at a certain moment exceeds the threshold for more than two consecutive periods, it is defined as an extreme "strong wind" event.

[0069] S2. Establish a Transformer hydropower, wind-solar-solar power prediction model that considers multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios;

[0070] First, the power prediction penalty equation of the hydro-wind-solar complementary system is established; then, a Transformer hydro-wind-solar power prediction model considering multi-dimensional upper and lower limits and output balance constraints is established.

[0071] By constructing a power prediction penalty equation for a water-wind-solar hybrid system, the model prediction results can meet the multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios; the power prediction penalty equation for a water-wind-solar hybrid system L The form is as follows:

[0072]

[0073] Where: is the actual power generation of hydropower, wind power and photovoltaic power, in MW; is the penalty term for the error between the simulated power generation and the actual power generation; Penalties for violating the upper and lower limits of the output of hydropower, wind and solar power stations; It is the penalty term for violating the balance constraint of the sum of the total power of the hydropower, wind-solar system and the single power of the hydropower, wind-solar system; is the weight coefficient of different penalty items, and its value is selected according to the actual prediction requirements.

[0074] In order to simulate the penalty term for the error between the generated power and the actual generated power, the error between the hydropower, wind power and photovoltaic power and the actual corresponding power is considered. The expression is as follows:

[0075] ;

[0076] Where: M is the time series length of the training period; Same meaning as above; It is the simulated power generation of hydropower, wind power and photovoltaic power, in MW.

[0077] It is a penalty item for violating the upper and lower limits of the output of hydropower, wind and solar power stations. The output boundary constraint includes the upper and lower boundaries of the hydropower, wind and solar power output. The upper and lower boundaries are determined according to historical power data. The specific expression is as follows:

[0078] ;

[0079] ;

[0080] Where: For the k The output upper limit of the power, in WM; For the k The lower limit of the power output, in WM.

[0081] is the penalty term for violating the balance constraint of the total power of the hydro-wind-solar system and the single power of the hydro-wind-solar system. The expression is as follows:

[0082] ;

[0083] Where: is the total power of the measured system.

[0084] The construction process of the Transformer hydropower prediction model considering multi-dimensional upper and lower limits and output balance constraints is as follows:

[0085] 1) Collect and organize previously extracted extreme climate events such as floods, droughts, strong winds, historical hydropower, wind power, photovoltaics and total power data, and related rainfall, evaporation, wind speed, solar radiation meteorological data, and inflow water resources data, and combine historical power data, weather data, and environmental data into a multidimensional time series feature matrix as the input of the hydropower, wind power, and photovoltaic power prediction model;

[0086] 2) Generate an attention weight matrix through dot product operation to quantify the similarity between different positions, and perform weighted aggregation of the relevant features of each position in the input sequence with other positions to enhance the effect of feature extraction. Then add position encoding to the input data, integrate the position information into the input features, and input the input data containing the position encoding into the Transformer model for processing, so that the model can make full use of the position relationship in the sequence for learning and reasoning;

[0087] 3) Based on the mapping relationship between the input feature matrix and the power output, the model is used to generate preliminary hydropower, wind power, and photovoltaic power forecast values. Then, these preliminary forecast results are adjusted through the power forecast penalty equation of the hydro-wind-solar complementary system. The adjustment formula is as follows:

[0088] ;

[0089] Where: is the learning rate; is the gradient of the penalty equation with respect to the initial prediction value;

[0090] Continue to iterate the above process until the penalty equation L Convergence means that each power prediction value satisfies the upper and lower limit constraints and output balance constraints.

[0091] S3. Carry out training and evaluation of the Transformer water, wind and solar power prediction model; Carry out training and evaluation of the Transformer water, wind and solar power prediction model: During the model training process, by defining the minimization penalty equation as the goal, use the goose optimization algorithm to tune the parameters, calculate the gradient of each parameter through back propagation, and dynamically adjust the learning rate and momentum to ensure stable and fast convergence, and continuously evaluate and adjust the model performance through cross-validation. In multiple training iterations, through the division and evaluation of training sets and validation sets, gradually optimize the model structure and parameter configuration, and finally ensure that the model can achieve high-precision power prediction while meeting all constraints.

[0092] S4. After the Transformer water, wind and photovoltaic power prediction model is trained, the trained model is deployed to the production environment to predict the power output of hydropower, wind power and photovoltaic power in real time. According to the actual situation and prediction error, the model and multi-dimensional upper and lower limits and output balance constraints are adjusted in real time to ensure the accuracy and reliability of the prediction results.

[0093] Embodiment 1:

[0094] Taking the water-wind-solar complementary system in a certain river basin as an example, the historical data from January 2021 to December 2023 are used to construct a joint prediction model for the power of the water-wind-solar complementary system considering extreme climate scenarios, with 2021-2022 as the rate period and 2023 as the test period. Taking the extreme flood event as an example, the flood events within the rate period (2021-2022) are extracted based on the ultra-quantitative threshold sampling method, and the power historical data of the water-wind-solar system in the flood event and the related meteorological data such as rainfall, evaporation, wind speed, solar radiation, and water resource data such as reservoir runoff are used as the input of the subsequent training model to output the power of the water-wind-solar system under the flood event. In order to verify the power forecast effect of the Transformer water-wind-solar power prediction model considering multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios, an independent water-wind-solar output model was set up for comparison. The forecast results of the two models for the power of each single type of power station in the water-wind-solar system during the rate period and the test period are shown in Tables 1 and 2. It can be seen from Tables 1 and 2 that the Transformer water-wind-solar power prediction model considering multi-dimensional upper and lower limits and output balance constraints proposed in the present invention can effectively improve the power forecast accuracy of each power station in the water-wind-solar system and the total power forecast accuracy.

[0095] Table 1 Average absolute error and root mean square error of periodic power forecast for each model rate

[0096]

[0097] Table 2 Average absolute error and root mean square error of power forecast for each model during the validation period

[0098]

[0099] The forecast error complementarity rates of the two models under different power combinations under different extreme climate scenarios are shown in Table 3. From analysis of Table 3, it can be seen that the forecast error complementarity rates obtained based on the model proposed in the present invention are greater than those of the independent water-wind-solar output model. The annual average forecast error complementarity rate of the water-wind-solar power combination is 28% higher than that of the independent model, indicating that this model can significantly improve the forecast error complementarity rate of the system power and improve the forecast accuracy of the total power by conducting joint power forecasting for the water-wind-solar system under extreme climate scenarios.

[0100] Table 3 Forecast error complementarity of each model in different power combinations under different extreme climate scenarios

[0101]

[0102] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A power joint forecasting method for a water-wind-solar hybrid system considering extreme climate scenarios, characterized by: S1. Extraction of extreme climate events: Extraction of extreme events including floods, droughts and strong winds is carried out based on selected data including historical runoff data, rainfall data and wind speed data of the study area; S2. Establish a Transformer hydropower, wind-solar-solar power prediction model that considers multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios; S3. Carry out training and evaluation of Transformer hydropower, wind-solar-solar power prediction model; S4. After the Transformer hydropower, wind-power, and photovoltaic power prediction model is trained, the trained model is deployed to the production environment to predict the power output of hydropower, wind power, and photovoltaic power in real time. According to the actual situation and prediction error, the model and multi-dimensional upper and lower limits and output balance constraints are adjusted in real time to ensure the accuracy and reliability of the prediction results. In step S2, a power prediction penalty equation for the water-wind-solar complementary system is first established; then a Transformer water-wind-solar power prediction model is established that considers multi-dimensional upper and lower limits and output balance constraints; 1) Establish the power prediction penalty equation for the hydro-wind-solar hybrid system: By constructing a power prediction penalty equation for a water-wind-solar hybrid system, the model prediction results can meet the multi-dimensional upper and lower limits and output balance constraints under extreme climate scenarios; the power prediction penalty equation for a water-wind-solar hybrid system L The form is as follows: ； Where: The actual power generation capacity of hydropower, wind power and photovoltaic power; is the penalty term for the error between the simulated power generation and the actual power generation; Penalties for violating the upper and lower limits of the output of hydropower, wind and solar power stations; It is the penalty term for violating the balance constraint of the sum of the total power of the hydropower, wind-solar system and the single power of the hydropower, wind-solar system; is the weight coefficient of different penalty items, and its value is selected according to the actual prediction requirements; In order to simulate the penalty term for the error between the generated power and the actual generated power, the error between the hydropower, wind power and photovoltaic power and the actual corresponding power is considered. The expression is as follows: ； Where: M is the time series length of the training period; Simulate power generation for hydropower, wind power and photovoltaic power; It is a penalty item for violating the upper and lower limits of the output of hydropower, wind and solar power stations. The output boundary constraint includes the upper and lower boundaries of the hydropower, wind and solar power output. The upper and lower boundaries are determined according to historical power data. The expression is as follows: ； ； Where: For the k The output limit of the power; For the k The lower limit of the power output; is the penalty term for violating the balance constraint of the total power of the hydro-wind-solar system and the single power of the hydro-wind-solar system. The expression is as follows: ； Where: is the total power of the measured system; The construction process of the Transformer hydropower prediction model considering multi-dimensional upper and lower limits and output balance constraints is as follows: 1) Collect and organize previously extracted extreme climate events such as floods, droughts, strong winds, historical hydropower, wind power, photovoltaics and total power data, and related rainfall, evaporation, wind speed, solar radiation meteorological data, and inflow water resources data, and combine historical power data, weather data, and environmental data into a multidimensional time series feature matrix as the input of the hydropower, wind power, and photovoltaic power prediction model; 2) Generate an attention weight matrix through dot product operation to quantify the similarity between different positions, and perform weighted aggregation of the relevant features of each position in the input sequence with other positions to enhance the effect of feature extraction. Then add position encoding to the input data, integrate the position information into the input features, and input the input data containing the position encoding into the Transformer model for processing, so that the model can make full use of the position relationship in the sequence for learning and reasoning; 3) Based on the mapping relationship between the input feature matrix and the power output, the model is used to generate preliminary hydropower, wind power, and photovoltaic power forecast values. Then, these preliminary forecast results are adjusted through the power forecast penalty equation of the hydro-wind-solar complementary system. The adjustment formula is as follows: ； Where: is the learning rate; is the gradient of the penalty equation with respect to the initial prediction value; Continue to iterate the above process until the penalty equation L Convergence means that each power prediction value satisfies the upper and lower limit constraints and output balance constraints.

2. The method for joint forecasting of power of a water-wind-solar hybrid system considering extreme climate scenarios according to claim 1 is characterized by: In step S1, the method for extracting extreme climate events is as follows: 1) Extract flood events based on ultra-quantitative threshold sampling method; 2) Extracting drought events based on runoff drought indicators; 3) Extract strong wind events based on the strong wind identification method of spatiotemporal factors.

3. The power joint forecasting method of the water-wind-solar complementary system considering extreme climate scenarios according to claim 2 is characterized by: 1) The process of extracting flood events based on the ultra-quantitative threshold sampling method is as follows: The water inflow data of the research reservoir is sampled by over-quantitative threshold, and the threshold value interval is determined according to the over-quantitative sample mean method. The mean of the part of the over-quantitative sample exceeding the threshold value is Threshold S It is a linear function relationship, where is the over-quantitative sample mean; Then, the flood samples in the study area are selected according to the threshold value, and the flood peak independence is judged by the following formula. The flood peak that meets the independence is the extracted flood sample: ； Where: D is the time interval between two consecutive flood peaks, d ; A is the watershed area; For the i Maximum daily flow of flood water; It is the minimum flow between two flood peaks.

4. The power joint forecasting method of the water-wind-solar hybrid system considering extreme climate scenarios according to claim 2 is characterized by: 2) The process of extracting drought events based on runoff drought indicators is as follows: Based on the hourly runoff data collected in the study area, the calculation period is selected, the runoff volume within the selected calculation period is calculated, and the runoff drought index is calculated based on the following formula: ； Where: Q is the runoff volume in the current calculation period; Q avg is the average runoff over many years; is the standard deviation of multi-year runoff.

5. The method for joint forecasting of power of a water-wind-solar hybrid system considering extreme climate scenarios according to claim 2 is characterized by: 3) The process of extracting strong wind events based on the strong wind identification method based on spatiotemporal factors is as follows: Sort the collected historical wind speed data in ascending order: After determining the sorting, a certain frequency value of the wind speed sequence is selected according to the geographical environment and climate indicators of the wind power station, and it is defined as the threshold of extreme "strong wind" events. When the maximum wind speed of a wind power station at a certain moment exceeds the threshold for more than two consecutive periods, it is defined as an extreme "strong wind" event.

6. The method for joint forecasting of power of a water-wind-solar hybrid system considering extreme climate scenarios according to claim 1 is characterized by: S3. Conduct training and evaluation of the Transformer hydropower, wind-solar-power prediction model: During the model training process, the goal is to minimize the penalty equation, and the goose optimization algorithm is used for parameter tuning. The gradient of each parameter is calculated through back propagation, and the learning rate and momentum are dynamically adjusted to ensure stable and fast convergence. The model performance is continuously evaluated and adjusted through cross-validation. In multiple training iterations, the model structure and parameter configuration are gradually optimized through the division and evaluation of training sets and validation sets, ultimately ensuring that the model can achieve high-precision power prediction while meeting all constraints.

7. A power joint forecasting system for water-wind-solar hybrid systems considering extreme climate scenarios, characterized by: A method for joint power forecasting of a water-wind-solar complementary system considering extreme climate scenarios according to any one of claims 1 to 6 is adopted.

Images