A wind-solar combined electricity price prediction method, device, equipment and medium

Abstract

The present application relates to the technical field of power system operation control, and discloses a wind-solar combined electricity price prediction method, device, equipment and medium, the present application divides the weather state type into extreme weather type and non-extreme weather type, and inputs the classification result into a mixed prediction model together with wind-solar weather data, so that the model can adaptively adjust its parameters according to the weather state type, thereby automatically switching the most suitable prediction strategy when extreme weather and normal weather appear alternately, avoiding the problem of uneven prediction ability of fixed model architecture under different weather scenes, improving the accuracy and stability of electricity price prediction under complex weather conditions, and providing a more reliable decision basis for power system operation control.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a method, device, equipment, and medium for predicting combined wind and solar power prices. Background Technology

[0002] Against the backdrop of continuous growth in installed capacity of renewable energy power generation, wind power and photovoltaic power generation have gradually developed into important components of the modern power system. However, the power generation characteristics of these energy sources are inherently volatile, especially under extreme weather conditions, where wind and solar power generation are significantly affected, which in turn is transmitted to electricity market prices, causing abnormal fluctuations in electricity prices.

[0003] In the context of the electricity market, electricity prices are a core technical parameter reflecting the supply-demand balance of the power system and a crucial reference for grid dispatching agencies in formulating day-ahead generation plans, allocating reserve capacity, and scheduling energy storage charging and discharging. Inaccurate electricity price forecasts under extreme weather conditions directly lead to deviations in dispatching decisions, impacting the safe and stable operation of the power system. Therefore, improving the accuracy of electricity price forecasts under extreme weather conditions is a technical requirement for ensuring the reliability of the power system under complex meteorological conditions. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and medium for wind and solar combined electricity price forecasting that takes extreme weather into account, in order to solve the technical problem that existing forecasting models lack a mechanism to adaptively adjust forecasting strategies according to weather conditions under extreme weather conditions, leading to inaccurate electricity price forecasts and thus affecting the reliability of power system dispatching decisions.

[0005] In a first aspect, the present invention provides a method for predicting wind and solar combined electricity prices, comprising: acquiring wind and solar meteorological data for a target period, and determining the weather state type based on the wind and solar meteorological data, wherein the weather state type includes extreme weather type and non-extreme weather type; inputting the wind and solar meteorological data for the target period and the weather state type into a preset hybrid prediction model to obtain a wind and solar combined electricity price prediction result, wherein the wind and solar combined electricity price prediction result is used for power system operation control; wherein the hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data and corresponding weather state type, and adjusts its own model parameters according to the input weather state type.

[0006] This invention categorizes weather conditions into extreme and non-extreme weather types and inputs the classification results along with wind, solar, and meteorological data into a hybrid prediction model. This allows the model to adaptively adjust its parameters based on the weather condition type, automatically switching to the most suitable prediction strategy when extreme and normal weather alternate. This avoids the problem of uneven prediction capabilities of a fixed model architecture under different weather scenarios, improves the accuracy and stability of electricity price prediction under complex meteorological conditions, and provides a more reliable decision-making basis for power system operation and control.

[0007] In one optional implementation, determining the weather state type based on wind, solar, and meteorological data includes: using a preset classification model to output the weather state type based on the input wind, solar, and meteorological data. The weather state type includes extreme weather types and non-extreme weather types. Extreme weather types include extreme weather types on the photovoltaic side and extreme weather types on the wind power side, while non-extreme weather types include non-extreme weather types on the photovoltaic side and non-extreme weather types on the wind power side. The classification model is trained based on historical wind, solar, and meteorological data and the corresponding weather state types. This implementation further divides the weather state type into extreme on the photovoltaic side, non-extreme on the photovoltaic side, extreme on the wind power side, and non-extreme on the wind power side by using a classification model. This allows the weather state labels to finely distinguish the weather conditions faced by photovoltaic power generation and wind power generation, providing a more targeted classification basis for subsequent hybrid prediction models. This enables the models to accurately capture the differentiated electricity price responses of different power generation sides under extreme weather conditions, thereby improving the accuracy and refinement of wind and solar joint electricity price prediction and providing more reliable decision support for power system operation and control.

[0008] In one optional implementation, the classification model is a machine learning classification model. When identifying weather state types, at least one of the following features is extracted for judgment: temporal features of wind, solar and meteorological data, spatial correlation features of regional multi-site wind, solar and meteorological data, and cross-combination features between different meteorological indicators. This implementation uses a machine learning classification model to extract at least one of the following features for weather state type identification: temporal features of wind, solar and meteorological data, spatial correlation features of regional multi-site wind, solar and meteorological data, and cross-combination features between different meteorological indicators. This enables weather classification to accurately capture the changing trends of meteorological conditions in the time dimension, regional linkages in the spatial dimension, and the coupling effects between multiple indicators. It avoids the problem of inaccurate identification in complex meteorological scenarios by traditional single-rule classification, thereby providing a more reliable and robust weather classification foundation for subsequent hybrid prediction models and improving the accuracy and stability of electricity price forecasting under extreme weather conditions.

[0009] In one optional implementation, the hybrid prediction model includes a first prediction module and a second prediction module. The hybrid prediction model dynamically adjusts the fusion weights of the first prediction module and the second prediction module according to the input weather state type through a gating mechanism to adjust its own model parameters. When the weather state type is extreme weather, the fusion weight of the first prediction module is greater than the fusion weight of the second prediction module. When the weather state type is non-extreme weather, the fusion weight of the second prediction module is greater than the fusion weight of the first prediction module.

[0010] In one optional implementation, the first prediction module is an LSTM module, used to extract the time series dependencies of wind, solar and meteorological data; the second prediction module is a Transformer module, used to capture the global correlation between different locations in the wind, solar and meteorological data.

[0011] This implementation sets up a first prediction module and a second prediction module in a hybrid prediction model, and introduces a gating mechanism to dynamically adjust the fusion weight of the two according to the weather state type. In extreme weather, the first prediction module is given more weight to capture the temporal abrupt changes in the data, while in non-extreme weather, the second prediction module is given more weight to utilize the global regularity of the data. This achieves adaptive matching between the prediction strategy and the weather state, avoids the problem of uneven prediction capabilities of a fixed model architecture under different weather scenarios, and improves the overall accuracy and generalization ability of electricity price prediction under complex meteorological conditions.

[0012] In an optional implementation, the hybrid prediction model is also trained based on historical wind and solar meteorological data, historical wind and solar power data, and corresponding weather state types. The prediction results output by the hybrid prediction model also include combined wind and solar power prediction results, which are used for power system operation control. This implementation, by combining wind and solar power prediction results with electricity price predictions through the hybrid prediction model, enables the model to simultaneously provide prediction information in both electricity price and power dimensions, providing a more comprehensive decision-making basis for power system operation control. The grid dispatching department can arrange power generation plans and reserve capacity configuration based on the power prediction and electricity price prediction results, improving the reliability and economy of grid operation under complex weather conditions.

[0013] In an optional implementation, the method further includes: acquiring actual wind and solar power generation and real-time wind and solar meteorological data; when the deviation between the real-time wind and solar meteorological data and the wind and solar meteorological data for the target period exceeds a first threshold, or the deviation between the actual wind and solar power generation and the wind and solar combined power prediction result exceeds a second threshold, updating the model parameters of the hybrid prediction model online using the latest data within a preset time period before the triggering time. This implementation acquires actual wind and solar power generation and real-time wind and solar meteorological data in real time, and sets dual triggering conditions of meteorological data deviation and power deviation. When either deviation exceeds the corresponding threshold, the latest data is automatically used to update the hybrid prediction model online, enabling the model to respond quickly and adaptively when weather forecasts are wrong or power generation fluctuates abnormally, avoiding continuous prediction deviations caused by the solidification of model parameters, thereby continuously ensuring the accuracy of electricity price prediction and the reliability of power system operation control.

[0014] Secondly, the present invention provides a wind-solar combined electricity price forecasting device, comprising: a weather state classification module, used to acquire wind and solar meteorological data for a target period and determine the weather state type based on the wind and solar meteorological data, the weather state type including extreme weather type and non-extreme weather type; and a combined electricity price forecasting module, used to input the wind and solar meteorological data and weather state type for the target period into a preset hybrid forecasting model to obtain the wind-solar combined electricity price forecasting result, the wind-solar combined electricity price forecasting result being used for power system operation control; wherein, the hybrid forecasting model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data and corresponding weather state type, and adjusts its own model parameters according to the input weather state type.

[0015] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the wind and solar combined electricity price forecasting method of the first aspect or any corresponding embodiment described above.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind-solar combined electricity price forecasting method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the first step in the wind and solar combined electricity price forecasting method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the second process of the wind and solar combined electricity price forecasting method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the third process of the wind and solar combined electricity price forecasting method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the fourth process of the wind and solar combined electricity price forecasting method according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a wind and solar combined electricity price forecasting device according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0020] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0021] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0022] Most existing electricity price forecasting models are based on historical data and regular fluctuation patterns, using a single forecasting framework to directly predict overall electricity prices. These methods have the following limitations: First, existing models lack a mechanism to dynamically adjust forecasting strategies based on weather conditions. When extreme and normal weather alternate, the data distribution characteristics change significantly, and fixed model structures and parameters cannot adapt to multiple weather scenarios simultaneously, leading to a significant decrease in forecast accuracy under extreme weather conditions.

[0023] Second, existing models struggle to accurately capture the dynamic balance between temporal abrupt changes and global patterns in meteorological data under extreme weather conditions. During extreme weather events, recent data trends have a more significant impact on electricity prices; under normal weather conditions, long-term stable patterns are more valuable. Fixed-architecture forecasting models cannot flexibly adjust their emphasis on short-term abrupt changes and long-term patterns according to weather conditions.

[0024] Therefore, when extreme weather events occur, the predictions of existing models often deviate significantly from actual electricity price trends, weakening the reliability and accuracy of the predictions.

[0025] Based on this, the present invention provides a wind-solar combined electricity price forecasting method, apparatus, equipment and medium to solve the technical problem that under extreme weather conditions, existing forecasting models lack a mechanism to adaptively adjust forecasting strategies according to weather conditions, resulting in inaccurate electricity price forecasting results and thus affecting the reliability of power system dispatching decisions.

[0026] According to an embodiment of the present invention, a wind-solar combined electricity price forecasting method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0027] This embodiment provides a method for predicting combined wind and solar power prices. Figure 1 This is a flowchart of the wind-solar combined electricity price forecasting method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain wind and solar meteorological data for the target time period, and determine the weather state type based on the wind and solar meteorological data.

[0028] The target period is the time period for which electricity prices need to be predicted, such as the next 24 hours or the next hour. Wind and solar meteorological data includes indicators directly related to wind and solar power output, such as temperature, wind speed, humidity, solar irradiance, and visibility. Weather condition types include extreme weather types and non-extreme weather types.

[0029] Step S102: Input the wind and solar meteorological data and weather condition type for the target period into the preset hybrid prediction model to obtain the wind and solar combined electricity price prediction results. The wind and solar combined electricity price prediction results are used for power system operation control.

[0030] The hybrid forecasting model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data, and corresponding weather state types. During training, the model learns the mapping relationship between meteorological data and electricity prices under different weather conditions. After training, the model can adaptively adjust its parameters according to the input weather state type: when it is extreme weather, the model focuses on capturing the temporal abrupt changes in the data; when it is non-extreme weather, the model focuses on utilizing the global regularity of the data. The final output of the wind and solar combined electricity price forecast provides a decision-making basis for the day-ahead or real-time dispatching of the power grid.

[0031] In this embodiment, the generated wind and solar combined electricity price forecast results are used for power system operation control, specifically outputting to the grid dispatching system via a data interface. Based on these forecast results and load forecast data, the dispatching system formulates power generation plans for each time period of the following day, including wind and solar power output curves, conventional unit start-up and shutdown arrangements, and energy storage system charging and discharging strategies. When the forecasted electricity price experiences peaks or troughs, the dispatching system can adjust reserve capacity configuration in advance to ensure power balance and voltage stability of the grid under extreme weather conditions.

[0032] The wind-solar combined electricity price forecasting method provided in this embodiment classifies weather conditions into extreme and non-extreme weather types and inputs the classification results along with wind and solar meteorological data into a hybrid forecasting model. This allows the model to adaptively adjust its parameters according to the weather condition type, thereby automatically switching to the most suitable forecasting strategy when extreme and normal weather alternate. This avoids the problem of uneven forecasting capabilities of a fixed model architecture under different weather scenarios, improves the accuracy and stability of electricity price forecasting under complex meteorological conditions, and provides a more reliable decision-making basis for power system operation and control.

[0033] This embodiment provides a method for predicting combined wind and solar power prices. Figure 2 This is a flowchart of the wind-solar combined electricity price forecasting method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain wind and solar meteorological data for the target time period, and determine the weather state type based on the wind and solar meteorological data. The weather state type includes extreme weather type and non-extreme weather type.

[0034] Specifically, step S201 above includes: Step S2011: Obtain wind and solar meteorological data for the target time period.

[0035] The target period refers to the time period for which electricity price forecasts are to be conducted, typically a future period. Wind, solar, and meteorological data can be obtained by connecting to a meteorological data center, including indicators such as temperature, wind speed, humidity, and solar irradiance.

[0036] Step S2012: Using a preset classification model, output the weather state type based on the input wind and solar meteorological data. The weather state type includes extreme weather type and non-extreme weather type. Extreme weather type includes extreme weather type on the photovoltaic side and extreme weather type on the wind power side. Non-extreme weather type includes non-extreme weather type on the photovoltaic side and non-extreme weather type on the wind power side. The classification model is trained based on historical wind and solar meteorological data and the corresponding weather state type.

[0037] The training data for the classification model includes historical weather data and corresponding weather condition type labels. Historical weather data is obtained by connecting to a meteorological data center and includes indicators such as temperature, wind speed, humidity, and light intensity. Weather condition type labels are generated based on preset rules or manual annotation. The preset rules include setting thresholds for meteorological elements such as wind speed, light intensity, temperature, humidity, and visibility to determine whether the solar and wind power sides experience extreme or non-extreme weather.

[0038] The weather state type label carries the state information of both sides. That is, each training data is labeled with the weather state type of the photovoltaic side and the weather state type of the wind power side. The two are independent of each other and can be combined into four cases: extreme on the photovoltaic side and non-extreme on the wind power side, non-extreme on the photovoltaic side and extreme on the wind power side, extreme on both sides, and non-extreme on both sides.

[0039] In one optional implementation, a machine learning classification model is used to replace simple rule-based classification, improving the accuracy and robustness of extreme weather identification. When identifying weather state types, at least one of the following features is extracted for judgment: temporal characteristics of wind, solar, and meteorological data; spatial correlation characteristics of wind, solar, and meteorological data from multiple stations in the region; and cross-combination characteristics between different meteorological indicators.

[0040] Among them, temporal features, such as meteorological data statistics within a sliding window; spatial features, such as the correlation between wind and solar meteorological data from multiple stations in a region; and cross features, such as temperature-humidity interaction terms and wind speed-light intensity combination features.

[0041] In one specific implementation, the classification model uses the LightGBM classifier, which is trained based on historical meteorological data and extreme weather labels. The objective function is optimized by gradient descent, and finally the weather state types of the photovoltaic side and the wind power side are output.

[0042] The objective function is: ; in, For the sample size, The loss function measures the true label. Compared with the predicted value Differences For the number of decision trees, The regularization term for the complexity of the k-th tree is given by the formula: ; in, The number of leaf nodes in the tree. The weight of the leaf node. and This is a hyperparameter that controls the strength of regularization. The loss function and the regularization term together constitute the standard objective function of LightGBM, where the regularization term controls model complexity and prevents overfitting.

[0043] The classification decision function outputs the weather state type through weighted voting, including extreme or non-extreme for the solar power side and extreme or non-extreme for the wind power side, as shown in the following formula: ; in, The total number of decision trees, i.e., the number of base learners in ensemble learning. The output value of the k-th decision tree is obtained by summing the outputs of all trees and then determining the category using a sign function.

[0044] Step S202: Input the wind and solar meteorological data and weather condition type for the target time period into the preset hybrid prediction model to obtain the wind and solar combined electricity price prediction result. The wind and solar combined electricity price prediction result is used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data, and corresponding weather condition types, and adjusts its own model parameters according to the input weather condition type.

[0045] The construction and training process of the hybrid prediction model is as follows: Step a1: Build the basic framework of the hybrid prediction model.

[0046] The hybrid prediction model includes a first prediction module and a second prediction module. The hybrid prediction model dynamically adjusts the fusion weights of the first prediction module and the second prediction module according to the input weather state type through a gating mechanism to adjust its own model parameters. When the weather state type is extreme weather, the fusion weight of the first prediction module is greater than the fusion weight of the second prediction module. When the weather state type is non-extreme weather, the fusion weight of the second prediction module is greater than the fusion weight of the first prediction module.

[0047] The input to the hybrid prediction model is represented as follows: ; Where X is the complete input vector of the model. This is a vector concatenation operation. Characteristics related to historical electricity prices and market conditions, Real-time and forecast meteorological data directly related to photovoltaic and wind power generation capacity. Based on the load demand characteristics of the power system, An embedded representation of weather category labels.

[0048] In one optional implementation, the first prediction module is an LSTM module, used to extract the time series dependencies of wind, solar and meteorological data; the second prediction module is a Transformer module, used to capture the global correlation between different locations in the wind, solar and meteorological data.

[0049] Specifically, the calculation process of the LSTM temporal coding layer is as follows: ; ; ; ; ; ; in, Let be the feature vector input to the LSTM unit at time t. This represents the hidden state of the LSTM unit output at time t-1. The cell state at time t-1. The forgetting gate control vector is output by the Sigmoid function. and Let i be the weight matrix and bias term corresponding to the forget gate. t The input gate control vector is the output of the Sigmoid function. and The weight matrix and bias terms corresponding to the input gate are... Let be a vector generated by the tanh function, representing the current input. And the previous hidden state The potential new information generated and The weight matrix and bias terms are for the candidate cell states. This represents the cell state after the update at time t. The previous state under the control of the forget gate. To input new information under gate control, o t This is the output gate control vector produced by the Sigmoid function. and The weight matrix and bias terms are for the output gate. This represents the output of the LSTM unit at time t.

[0050] The calculation process of the Transformer multi-head self-attention layer is as follows: ; ; ; Where Q is the query matrix, K is the key matrix, and V is the value matrix. Key vector The softmax function outputs the degree of attention given to different positions in the sequence, taking into account the dimension of the sequence. For the i-th attention head, , , The i-th header is used to generate the weight matrix of query, key, and value; Concat is the concatenation operation. This is the output weight matrix for linear transformation of the splicing result.

[0051] The hybrid forecasting model achieves conditional adaptation through a gating mechanism, dynamically adjusting the fusion weights of the LSTM and Transformer modules based on the weather condition type. The specific calculations are as follows: ; ; in, An embedded representation of weather category labels. and For the gated weight matrix and bias terms, It is the Sigmoid activation function. This is the gating weight, with a value ranging from 0 to 1. This is the output vector of the LSTM module. This is the output vector of the Transformer module. This represents the final output of the hybrid prediction model.

[0052] When the weather condition type is extreme weather, the gating weight The gating weights approach 1, indicating a bias towards the output of the LSTM module to capture the temporal abrupt changes in data under extreme weather conditions; when the weather condition is non-extreme, the gating weights... Approaching 0, the model prioritizes the output of the Transformer module to leverage the global regularity of data under normal weather conditions. Gating weight matrix. and bias terms The parameters are optimized together with other model parameters using gradient descent during model training, without the need for manual setting.

[0053] The hybrid prediction model architecture described above offers the following technical advantages: In terms of classification accuracy, the machine learning classifier can accurately identify complex weather patterns, improving the AUC by approximately 0.15 to 0.20 compared to traditional rule-based classification methods; in terms of prediction performance, the hybrid model reduces the mean absolute error by approximately 12.3% under extreme weather conditions and by approximately 8.7% under normal weather conditions; in terms of model engineering, the single architecture simplifies deployment and maintenance, and parameter sharing improves training efficiency; and in terms of adaptability, the adaptive weighting mechanism ensures that the model maintains optimal prediction performance under different weather conditions.

[0054] Step a2: Obtain training data for model training.

[0055] The training data includes historical wind and meteorological data, historical wind and solar power price data, and corresponding weather state type labels.

[0056] Historical wind and solar meteorological data were obtained by connecting to a meteorological data center, including indicators such as temperature, wind speed, humidity, and solar irradiance. Historical wind and solar power price data were obtained through the electricity market trading platform interface, including the electricity price series corresponding to wind and solar power output. The weather condition type labels are consistent with the labels used in the classification model in step S201, meaning that each training data point is labeled with both the solar-side and wind-side weather condition types. The training data covers the past three years, with a time resolution of 1 hour, and the data is aligned by timestamps.

[0057] Step a3: Train the hybrid prediction model using the training data.

[0058] Historical wind and solar meteorological data and corresponding weather condition labels are used as input features to the model, while historical wind and solar electricity price data are used as supervision signals and output features to train the hybrid prediction model. The training objective is to minimize the error between the model's predicted electricity price and the actual electricity price. The mean squared error loss function can be used. During training, the parameters of the gating mechanism are... and The parameters were optimized using gradient descent along with other model parameters, enabling the model to adaptively adjust the fusion weights of the LSTM and Transformer modules based on weather conditions. After training, the model can output a combined wind and solar power price prediction result when given new wind, solar, and meteorological data and weather conditions.

[0059] In this embodiment, the hybrid prediction model directly outputs the wind-solar combined electricity price prediction result after training. The model adaptively adjusts the gating weights based on the input weather state type labels. The solar and wind power state information carried in the labels enables the model to automatically distinguish between different extreme scenarios without the need for separate external predictions or fusion processing.

[0060] In another embodiment, the hybrid prediction model can also employ a method of separate predictions followed by fusion. Specifically: when extreme weather occurs on the photovoltaic side, the model outputs the photovoltaic-side electricity price prediction result; when extreme weather occurs on the wind power side, the model outputs the wind power-side electricity price prediction result; when extreme weather occurs on both sides, the model outputs the photovoltaic-side electricity price prediction result and the wind power-side electricity price prediction result separately, and then weights and fuses them according to the ratio of photovoltaic installed capacity to wind power installed capacity to obtain the wind-solar joint electricity price prediction result. For example, if the photovoltaic capacity is 300MW and the wind power capacity is 500MW, then the wind-solar joint electricity price prediction result = photovoltaic-side electricity price prediction value × 0.375 + wind power-side electricity price prediction value × 0.625. Non-extreme parts are automatically handled by the model, without the need to call different models separately.

[0061] The wind-solar combined electricity price forecasting method provided in this embodiment classifies weather conditions into extreme and non-extreme categories using a classification model, further subdividing them into the states of the photovoltaic and wind power sides. The classification labels, along with wind and solar meteorological data, are input into a hybrid forecasting model composed of an LSTM module and a Transformer module. The training-derived gating mechanism adaptively adjusts the fusion weights of the two modules based on the weather condition type. In extreme weather, the LSTM module is given preference to capture temporal abrupt changes, while in non-extreme weather, the Transformer module is given preference to utilize global regularities. This allows a single model to automatically switch to the most suitable forecasting strategy under different weather scenarios, thereby significantly improving the accuracy and stability of wind-solar combined electricity price forecasting under complex meteorological conditions and providing a more reliable decision-making basis for power system operation and control.

[0062] This embodiment provides a method for predicting combined wind and solar power prices. Figure 3 This is a flowchart of the wind-solar combined electricity price forecasting method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Obtain wind, solar, and meteorological data for the target time period, and determine the weather state type based on the data. Weather state types include extreme weather and non-extreme weather. See above for details. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0063] Step S302: Input the wind and solar meteorological data and weather condition type for the target time period into the preset hybrid prediction model to obtain the wind and solar combined electricity price prediction result and power prediction result. The wind and solar combined electricity price prediction result and power prediction result are used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data, historical wind and solar power data and corresponding weather condition types, and adjusts its own model parameters according to the input weather condition type.

[0064] This step S302 is the same as the above. Figure 2 Compared to step S202 in the illustrated embodiment, the difference lies in that the hybrid prediction model is also configured to simultaneously output power prediction results. That is, the hybrid prediction model is also trained based on historical wind and solar meteorological data, historical wind and solar power data, and corresponding weather state types; the prediction results output by the hybrid prediction model also include wind and solar combined power prediction results, which are used for power system operation control.

[0065] Specifically, during the model training phase, training data also includes historical wind and solar power data. Historical wind and solar meteorological data and their corresponding weather condition type labels are used as model input, while historical wind and solar electricity price data and historical wind and solar power data are used as supervision signals. The hybrid prediction model is jointly trained, enabling the model to simultaneously learn the mapping relationship between meteorological conditions, weather condition types, and electricity prices and power. The training objective is to minimize the joint loss of electricity price prediction errors and power prediction errors.

[0066] During the forecasting phase, wind, solar, and meteorological data for the target period, along with weather condition types, are input into the trained hybrid forecasting model. The model adaptively adjusts the gating weights based on the weather condition type and outputs both the wind-solar combined electricity price forecast and the wind-solar combined power forecast. Both are used for economic dispatch decisions and power balance dispatch, providing a more comprehensive decision-making basis for power system operation and control.

[0067] This embodiment provides a method for predicting combined wind and solar power prices. Figure 4 This is a flowchart of the wind-solar combined electricity price forecasting method according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps: Step S401: Obtain wind, solar, and meteorological data for the target time period, and determine the weather state type based on the data. Weather state types include extreme weather and non-extreme weather. See above for details. Figure 3 Step S301 of the illustrated embodiment will not be described again here.

[0068] Step S402 involves inputting the wind and solar meteorological data and weather condition type for the target time period into a preset hybrid prediction model to obtain the combined wind and solar power price prediction result and power prediction result. These results are used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar power price data, historical wind and solar power data, and corresponding weather condition types, and adjusts its model parameters according to the input weather condition type. See the above for details. Figure 3 Step S302 of the illustrated embodiment will not be described again here.

[0069] Step S403: Obtain actual wind and solar power generation and real-time wind and solar meteorological data; when the deviation between the real-time wind and solar meteorological data and the wind and solar meteorological data for the target period exceeds the first threshold, or the deviation between the actual wind and solar power generation and the wind and solar combined power prediction result exceeds the second threshold, use the latest data within the preset time period before the trigger time to update the model parameters of the hybrid prediction model online.

[0070] In actual operation, weather forecasts may be inaccurate, and wind and solar power generation may fluctuate abnormally. To address this, this embodiment introduces a dynamic verification and online update mechanism, namely step S403 above, which specifically includes three steps: deviation calculation, trigger condition judgment, and model fine-tuning.

[0071] First, deviation calculation is performed, and the actual measured wind and solar power values ​​are continuously obtained. And real-time meteorological data, calculate the deviation between real-time meteorological data and forecast meteorological data for the target period, and Compared with the wind and solar power values ​​predicted by the model The deviation between them.

[0072] Next, trigger conditions are determined. When the deviation between real-time meteorological data and forecast meteorological data is significant, or... and Deviation exceeds threshold When this occurs, the model re-optimization process is triggered. Here, ε is a preset error threshold, which can be determined through historical data analysis and expert experience.

[0073] Finally, model fine-tuning is performed. After the above steps are triggered, the hybrid prediction model is fine-tuned online using the latest short-term data. This involves using data from the most recent period before the trigger time, including the latest wind, solar, and meteorological data, wind and solar power prices, wind and solar power data, and corresponding weather condition types, to iteratively update the model parameters of the trained model with a small learning rate. This allows the model to quickly adapt to the new data distribution without destroying existing knowledge. After the update is complete, the updated model is used to re-execute predictions.

[0074] Through the aforementioned dynamic verification and online update mechanism, a rapid response can be made when weather forecasts are wrong or power generation fluctuates abnormally, continuously ensuring the accuracy and reliability of electricity price and power forecasts, thereby ensuring the effectiveness of power system operation and control.

[0075] In retrospective testing using historical data containing multiple extreme weather events, the wind-solar combined electricity price prediction method provided in this embodiment significantly reduces the average absolute error between the predicted and actual electricity prices compared to traditional single models. Under extreme weather conditions, the stability of the prediction results is improved, and the range of error fluctuations is reduced. The model maintains high prediction accuracy and reliability in various scenarios, including extreme weather, non-extreme weather, and mixed weather.

[0076] In this embodiment, the classification module identifies the weather state type for the target time period based on real-time wind, solar, and meteorological data and historical training data. The hybrid prediction model adaptively adjusts its parameters according to the weather state type and outputs the electricity price prediction result. The system compares the predicted electricity price with the actual electricity price in real time, calculates the mean absolute error, and verifies the effectiveness of the prediction. When the error value is within a preset threshold, the current prediction model is deemed effective, and the existing prediction process is maintained. When the error value exceeds the preset threshold, the current prediction model is deemed to have failed under the current conditions, triggering a model re-optimization process. The prediction model is fine-tuned online using the latest data before the triggering time, the model parameters are updated, and the prediction is re-executed to continuously ensure prediction accuracy.

[0077] The following uses an electricity market that includes both wind and solar power as an example to illustrate the solution of this invention.

[0078] In a certain regional electricity market, there are 500MW of wind farm capacity, 300MW of photovoltaic power plant capacity, and 10 meteorological monitoring stations with a data sampling interval of 1 hour. When executing day-ahead electricity price forecasts, the following steps are performed: Step b1: Collaborative acquisition of multi-source data. Temperature, humidity, wind speed, light intensity, and visibility data are acquired through the meteorological data platform interface. Historical power output and nodal electricity price data for wind and solar power plants are acquired through the electricity market operation system. The data range covers the past three years, with a time resolution of 1 hour.

[0079] Specifically, historical meteorological data for wind and solar power includes temperature, humidity, wind speed, light intensity, and visibility sequences; historical power data for wind and solar power includes wind farm output curves and photovoltaic power plant output curves; and historical price data for wind and solar power includes market-clearing price sequences corresponding to wind and solar power output.

[0080] Step b2: Based on the meteorological characteristic sequence in the historical wind and solar meteorological data, classify the historical data into historical data of extreme weather for photovoltaic power, historical data of extreme weather for wind power, historical data of non-extreme weather for photovoltaic power, and historical data of non-extreme weather for wind power.

[0081] The classification rules are defined as follows: Ice accumulation event, with temperatures consistently below 0°C and relative humidity exceeding 85%; Dust storm events, with visibility less than 1km and wind speeds greater than 8m / s; Strong wind event, with wind speeds consistently exceeding 12 m / s; A windless event is defined as a wind speed consistently below 2 m / s; all other situations are considered non-extreme weather.

[0082] The classification module has a built-in feature extraction algorithm that performs sliding window detection on the input meteorological sequence to identify meteorological event segments that meet the above rules and label their type and time interval.

[0083] Step b3: Construct extreme weather electricity price prediction models and non-extreme weather electricity price prediction models.

[0084] The extreme weather electricity price prediction model uses an LSTM neural network structure. The input is the past 24-hour meteorological sequence, power sequence, and electricity price sequence. The output is the predicted wind and solar power and the corresponding electricity price under extreme weather conditions. The non-extreme weather electricity price prediction model uses a Transformer model. The input is the 24-hour historical sequence. The output is the predicted power and electricity price under non-extreme weather conditions.

[0085] During model training, the extreme weather model is trained using the historical extreme weather data classified in step b2, while the non-extreme weather model is trained using the historical non-extreme weather data. The training objective is to minimize the mean square error of the electricity price forecast.

[0086] Step b4: Perform day-ahead or real-time electricity price forecasting.

[0087] Obtain the meteorological forecast data and the actual data for the most recent 24 hours required for forecasting, and determine the future weather conditions according to the classification rules in step b2; if extreme weather is identified for photovoltaic power, the extreme weather model is used to forecast the electricity price for the photovoltaic portion; if extreme weather is identified for wind power, the extreme weather model is used to forecast the electricity price for the wind power portion; for non-extreme weather portions, the non-extreme weather model is used for forecasting; finally, the electricity price forecast results for photovoltaic and wind power are combined by capacity weighting. For example, if the photovoltaic capacity is 300MW and the wind power capacity is 500MW, then the comprehensive electricity price = photovoltaic electricity price × 0.375 + wind power electricity price × 0.625.

[0088] Step b5: Output the day-ahead or real-time electricity price forecast curve and confidence interval.

[0089] The prediction results include point prediction values ​​and probability ranges, with the probability ranges achieved through quantile regression or Monte Carlo sampling. The system updates the prediction results every 15 minutes and supports visualization and API interface output.

[0090] Furthermore, it also includes human-computer interaction and visualization steps, receiving user instructions, displaying data classification results, model training status, real-time prediction process, and final electricity price prediction results.

[0091] In a test cycle that included multiple extreme weather events, the system provided in this embodiment showed a reduction in prediction error compared to traditional single-model systems. In particular, the accuracy of electricity price prediction was improved in icing events accompanied by a sudden drop in temperature and high humidity, verifying the effectiveness of multi-model collaboration and meteorological feature classification mechanisms.

[0092] This embodiment also provides a wind-solar combined electricity price forecasting device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0093] This embodiment provides a wind-solar combined electricity price forecasting device, such as... Figure 5 As shown, it includes: The weather state classification module 501 is used to acquire wind and solar meteorological data for the target period and determine the weather state type based on the wind and solar meteorological data. The weather state type includes extreme weather type and non-extreme weather type. The joint electricity price prediction module 502 is used to input wind and solar meteorological data and weather state type for the target period into a preset hybrid prediction model to obtain the wind and solar joint electricity price prediction result. The wind and solar joint electricity price prediction result is used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data and corresponding weather state type, and adjusts its own model parameters according to the input weather state type.

[0094] In some optional implementations, the weather state classification module 501 includes: The target data acquisition unit is used to acquire wind, solar and meteorological data for the target time period; The weather state determination unit is used to output the weather state type based on the input wind and solar meteorological data using a preset classification model. The weather state type includes extreme weather type and non-extreme weather type. Extreme weather type includes extreme weather type on the photovoltaic side and extreme weather type on the wind power side. Non-extreme weather type includes non-extreme weather type on the photovoltaic side and non-extreme weather type on the wind power side. The classification model is trained based on historical wind and solar meteorological data and the corresponding weather state type.

[0095] The wind-solar combined electricity price forecasting device provided in this embodiment of the invention can execute the wind-solar combined electricity price forecasting method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0096] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0097] The following is a detailed reference. Figure 6This diagram illustrates a suitable structural design for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 601, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 602 or a program loaded from memory 608 into random access memory (RAM) 603. RAM 603 also stores various programs and data required for the operation of the electronic device. The processor 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0098] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0099] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a memory 608, or installed from a ROM 602. When the computer program is executed by the processor 601, it performs the functions defined in the wind-solar combined electricity price forecasting method of the embodiments of the present invention.

[0100] Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0101] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the wind-solar combined electricity price forecasting method shown in the above embodiments is implemented.

[0102] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0103] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting combined wind and solar power prices, characterized in that, The method includes: Acquire wind, solar and meteorological data for the target time period, and determine the weather state type based on the wind, solar and meteorological data. The weather state type includes extreme weather type and non-extreme weather type. The wind and solar meteorological data for the target period and the weather state type are input into a preset hybrid prediction model to obtain the wind and solar combined electricity price prediction result, which is used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data, and corresponding weather state types, and adjusts its own model parameters according to the input weather state type.

2. The wind-solar combined electricity price forecasting method according to claim 1, characterized in that, The step of determining the weather state type based on the wind, solar and meteorological data includes: using a preset classification model to output the weather state type based on the input wind, solar and meteorological data. The weather state type includes extreme weather type and non-extreme weather type. The extreme weather type includes extreme weather type on the photovoltaic side and extreme weather type on the wind power side. The non-extreme weather type includes non-extreme weather type on the photovoltaic side and non-extreme weather type on the wind power side. The classification model is trained based on historical wind, solar and meteorological data and the corresponding weather state type.

3. The wind-solar combined electricity price forecasting method according to claim 2, characterized in that, The classification model is a machine learning classification model. When identifying weather state types, it extracts at least one of the following features for judgment: the temporal features of wind, solar and meteorological data, the spatial correlation features of wind, solar and meteorological data from multiple stations in the region, and the cross-combination features between different meteorological indicators.

4. The wind-solar combined electricity price forecasting method according to claim 1, characterized in that, The hybrid prediction model includes a first prediction module and a second prediction module. The hybrid prediction model dynamically adjusts the fusion weights of the first prediction module and the second prediction module according to the input weather state type through a gating mechanism to adjust its own model parameters. When the weather state type is extreme weather, the fusion weight of the first prediction module is greater than the fusion weight of the second prediction module. When the weather state type is non-extreme weather, the fusion weight of the second prediction module is greater than the fusion weight of the first prediction module.

5. The wind-solar combined electricity price forecasting method according to claim 4, characterized in that, The first prediction module is an LSTM module, used to extract the time series dependencies of wind, solar and meteorological data; the second prediction module is a Transformer module, used to capture the global correlation between different locations in the wind, solar and meteorological data.

6. The wind-solar combined electricity price forecasting method according to claim 1, characterized in that, The hybrid prediction model is also trained based on historical wind and solar meteorological data, historical wind and solar power data, and corresponding weather state types; the prediction results output by the hybrid prediction model also include wind and solar combined power prediction results, which are used for power system operation control.

7. The wind-solar combined electricity price forecasting method according to claim 6, characterized in that, Also includes: Acquire actual wind and solar power generation capacity and real-time wind and solar meteorological data; When the deviation between the real-time wind and solar meteorological data and the wind and solar meteorological data for the target time period exceeds a first threshold, or the deviation between the actual wind and solar power generation and the wind and solar combined power prediction result exceeds a second threshold, the model parameters of the hybrid prediction model are updated online using the latest data within a preset time period before the trigger time.

8. A wind-solar combined electricity price forecasting device, characterized in that, The device includes: The weather state classification module is used to acquire wind and solar meteorological data for a target time period and determine the weather state type based on the wind and solar meteorological data. The weather state type includes extreme weather type and non-extreme weather type. The combined electricity price forecasting module is used to input wind and solar meteorological data for the target period and the weather state type into a preset hybrid forecasting model to obtain the wind and solar combined electricity price forecasting result, which is used for power system operation control. The hybrid prediction model is trained based on historical wind and solar meteorological data, historical wind and solar electricity price data, and corresponding weather state types, and adjusts its own model parameters according to the input weather state type.

9. An electronic device, characterized in that, include: The system includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the wind and solar combined electricity price forecasting method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the wind-solar combined electricity price forecasting method according to any one of claims 1 to 7.

Images