Wind power prediction model method and system combined with turning weather scene

Abstract

The invention discloses a wind power prediction model method and system in combination with a turning weather scene, and the method comprises the following steps: collecting wind power data and meteorological data, and carrying out the preprocessing of the data; constructing a basic wind power prediction model; constructing a turning weather scene; constructing a wind power prediction model in a turning weather scene; and performing model fusion based on the basic wind power prediction model and the models in different turning weather scenes, outputting a final wind power prediction result, accurately extracting a data set in each scene by analyzing an influence mechanism of the turning weather scenes on wind power, constructing wind power prediction models in different scenes, and performing wind power prediction. The pertinence of wind power prediction under complex weather conditions is improved. Through model fusion, high-precision prediction can be carried out on the wind power in various scenes, the problem of large prediction errors in the prior art is effectively solved, the wind power absorption capability is improved, and safe, stable and economic operation of a power system is ensured.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation technology, and in particular to a method and system for constructing a wind power prediction model combined with transitional weather scenarios, taking into account the impact of different transitional weather factors on wind power, for accurate wind power prediction. Background Art

[0002] Fossil energy has long driven the advancement of industrial civilization and economic and social development, but it has also brought about a series of problems such as resource depletion, environmental pollution, and climate change, seriously impacting human survival and development. With increasingly severe environmental issues and growing energy demand, the development of renewable energy has become an inevitable trend. Wind power, as a key component of renewable energy generation, offers advantages such as cleanliness, short infrastructure construction cycles, flexible installation scale, and low operation and maintenance costs. It has rapidly grown globally and plays a vital role.

[0003] However, wind power is highly random and volatile. When large-scale wind power is integrated into the grid, especially during transitional weather conditions, the output of wind farms can fluctuate dramatically within a short period of time, making it difficult for the power system to maintain a balance between power supply and load in real time, affecting the security and stability of the grid. Existing wind power prediction models have many problems: The periods when extreme errors occur in wind power forecast results are almost always during transitional weather processes, which causes large deviations in the forecast results and has a significant impact on the safe and stable operation of the system and wind power absorption.

[0004] Wind power output is usually affected by multiple factors. A single prediction model cannot fully explore the mapping relationship between influencing factors and output power. Especially in the case of turning weather, a single model is difficult to fully learn, resulting in increased prediction deviation. Summary of the Invention

[0005] The present invention aims to consider the impact of different transitional weather factors on wind power and construct a wind power prediction model method and system combined with transitional weather scenarios to achieve accurate wind power prediction, improve wind power absorption capacity, and ensure the safe, stable and economical operation of the power system.

[0006] To this end, the specific technical solutions adopted in the present invention are as follows: Data collection and preprocessing Wind power data: Wind power data for a specific number of wind farms in a province was collected every 15 minutes from January 1 to August 31 of a specific year. The data included wind turbine power output, theoretical power generation, equipment parameters (rated power, cut-in and cut-out wind speeds, hub height, etc.), and location parameters (station latitude and longitude). The acquired JSON-formatted data was extracted into time series data with a standard 15-minute interval. Discontinuities in the time series were linearly interpolated and filled, and the data was stored in CSV format.

[0007] Meteorological data: We used a specific dataset provided by an authoritative organization, covering the period from January 1st to August 31st of a specific year. This dataset covers various meteorological elements, including wind speed and direction (horizontal wind speed at different altitudes), temperature and humidity (surface temperature, relative humidity, and other related indicators), air pressure, and precipitation. The spatial resolution was constant, and the temporal resolution was 1 hour. We extracted meteorological data based on wind farm location parameters, interpolated it into 15-minute time intervals, and stored it in CSV format.

[0008] Outlier detection and processing: This approach utilizes rule-based (e.g., wind speed, turbine speed, or power less than or equal to 0; wind speed less than the cut-in threshold, power greater than 0; wind speed greater than the cut-out threshold, power greater than 0; power greater than 1.2 times the rated power; actual generated power less than or equal to 30% of the theoretical power) and isolation forest algorithm-based anomaly detection. For missing power values, polynomial interpolation is used to fill in missing samples if there are no more than four consecutive missing samples within a given time interval in a day. For outliers, both zero-value replacement (e.g., data points with negative power values) and interpolation reconstruction (power values that do not conform to the wind power curve trend) strategies are employed.

[0009] Model building Basic Wind Power Prediction Model: The XGBoost model was selected to construct a basic wind power prediction model. The input sample data included wind speed at 100 meters, wind direction at 100 meters, wind speed at 10 meters, wind direction at 10 meters, surface temperature, relative humidity, air pressure, precipitation, and other major meteorological factors affecting wind power, as well as wind power. The data was divided into training and test sets with a ratio of 8:2. A grid search algorithm was used to optimize the XGBoost model's parameters, including n_estimators, max_depth, learning_rate, and colsample_bytree. The optimal parameters were selected to construct the model.

[0010] Construction of transitional weather scenarios: Taking three representative transitional weather processes—cold waves, heavy rainfall, and icing—as examples, this study analyzes the relationship between wind power forecast errors and weather processes, as well as the mechanisms that cause these errors. The two transitional weather processes, cold waves and precipitation, are divided into three scenarios: high winds, icing, and precipitation.

[0011] Wind power prediction model for transitional weather scenarios: The Pearson correlation coefficient method was used to analyze the correlation between various meteorological factors and wind power, optimizing meteorological sensitive features for each scenario. For high wind scenarios, wind speed was selected as the model input feature; for icing scenarios, precipitation, air pressure, temperature, and wind speed were selected; for heavy precipitation scenarios, wind speed, air pressure, and large-scale precipitation were selected as model inputs. Based on this, the XGBoost model was used to construct wind power prediction models for different transitional weather scenarios.

[0012] Model fusion: This system integrates the basic wind power forecast model with models for different transitional weather scenarios. During real-time wind power forecasting, the system determines whether the current input weather data represents transitional weather and its type, and outputs the duration of the current transitional weather. The system then replaces the basic model's forecast results with the time period of the transitional weather event, and outputs the final forecast results. Beneficial effects of the present invention By analyzing the impact mechanism of transitional weather scenarios on wind power, accurately extracting data sets under each scenario, and constructing wind power prediction models under different scenarios, the targetedness of wind power prediction under complex weather conditions is improved.

[0013] Through the fusion of models, wind power can be predicted with high precision in various scenarios, effectively solving the problem of large prediction errors in existing technologies, improving wind power absorption capacity, and ensuring the safe, stable and economical operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is a schematic diagram of the present invention showing that when the wind speed is too high or even exceeds the cut-out wind speed, wind turbines in a wind farm may experience climbing, high output, partial cut-out, or full cut-out; Figure 2 This is a schematic diagram showing that when ice accumulates on a wind turbine blade at the same wind speed, its geometric shape changes, causing blade resistance to increase and lift to decrease, thereby reducing output power; Figure 3 This is a schematic diagram showing that, before heavy rainfall occurs, wind power output varies accordingly with wind speed fluctuations, and the power and wind speed fluctuation trends are consistent; Figure 4 This is a flow chart of a wind power prediction model method combined with transitional weather scenarios according to the present invention; DETAILED DESCRIPTION

[0015] According to an embodiment of the present invention, a method and system for a wind power prediction model incorporating transitional weather scenarios are provided.

[0016] Data collection and preprocessing Wind power data collection: Establish data connections with a certain number of wind farms in a province, obtain real-time operating data of wind turbines, record data according to the specified time range and collection frequency, and store it in the original JSON format.

[0017] Wind power data preprocessing: Develop a data processing program to extract JSON-formatted data into time series data with 15-minute intervals, mapping time to relevant elements. Linear interpolation is used to fill in gaps in the time series data. The processed data is then stored as a CSV file for subsequent analysis and model training.

[0018] Meteorological data collection: The ERA5 dataset was obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) through a network interface, and the corresponding meteorological data were filtered out according to the time range and geographical location of the wind farm.

[0019] Meteorological data preprocessing: Using geographic information processing tools, we extracted the downloaded gridded meteorological data based on the longitude and latitude of the wind farm stations to form a meteorological element time series dataset. We then used an interpolation algorithm to adjust the meteorological element time resolution from 1 hour to 15 minutes and stored it in CSV format.

[0020] Outlier detection and processing: Develop an outlier detection program to detect wind power data using both rule-based and isolation forest algorithms. Detected missing and outlier values are handled according to established filling and processing strategies.

[0021] (2) Model construction Building a basic wind power prediction model: Using programming languages such as Python, we call the XGBoost library. We divide the preprocessed sample data into training and test sets in an 8:2 ratio. We use a grid search algorithm to optimize the XGBoost model's parameters, including n_estimators, max_depth, learning_rate, and colsample_bytree. After finding the optimal parameter combination, we build a basic wind power prediction model and evaluate and validate the model using the test set.

[0022] Construction of transitional weather scenarios: Based on the meteorological-wind power dataset, the relationship between wind power forecast error and two types of weather processes, namely cold waves, heavy rainfall, and icing, is analyzed by taking three representative transitional weather processes as examples, and the mechanism of forecast error under transitional weather is analyzed.

[0023] A cold wave is a widespread, intense cold air movement characterized by a significant drop in temperature, strong winds, and frequent rain, snow, rime, and frost. These cold waves are always accompanied by high-impact weather conditions, which can be divided into five categories based on the dryness and wetness of the air: high winds with cooling temperatures, rain and snow with cooling temperatures, rain and snow followed by high winds with cooling temperatures, high winds and rain and snow occurring on the same day with cooling temperatures, and high winds followed by rain and snow with cooling temperatures. High winds and blade icing are the primary factors affecting wind power output.

[0024] When the wind speed is too high or even exceeds the cut-out wind speed, the wind turbines in the wind farm will experience climbing, high output, partial cut-out or full cut-out, as shown in Figure (1). If the scale of wind turbine cut-out is large, it will have a significant impact on the voltage and frequency adjustment of the region or even the entire province. At this time, if the wind power prediction system fails to predict the high wind cut-out situation, the power prediction accuracy of the entire province and the station will be significantly insufficient.

[0025] In addition to the strong wind conditions mentioned above, blade icing under low temperature and high humidity conditions will also affect the output of the wind turbine. At the same wind speed, when ice accumulates on the wind turbine blades, their geometric shape changes, causing the blade resistance to increase and the lift to decrease, thereby reducing the output power. As shown in Figure (2), when a variable-speed variable-pitch wind turbine has a small amount of ice accumulation, if the wind speed is large enough, it can also reach the rated power. However, when a variable-speed variable-pitch wind turbine has severe ice accumulation and a stalled wind turbine has blade icing, it cannot reach the rated power, which seriously affects the wind turbine output. Therefore, if only wind speed and wind direction data are selected as the input of the prediction model in the case of blade ice accumulation, it will inevitably cause large errors.

[0026] As shown in Figure (3), before heavy rainfall, wind power output will change accordingly with wind speed fluctuations, and the power and wind speed fluctuation trends are consistent. Near the peak of rainfall, power no longer changes with wind speed, and the power decreases as wind speed increases. After the rainfall, the power and wind speed fluctuation trends return to a state close to synchronization. Therefore, when rainfall occurs, wind speed will no longer be the dominant factor affecting wind power output.

[0027] Based on the above analysis, the two types of transitional weather, cold waves and precipitation, are divided into three scenarios: strong winds, icing and precipitation.

[0028] 4. Construction of wind power prediction model under transitional weather scenarios The meteorological sensitive features in each scenario were optimized, and the Pearson correlation coefficient method was used to analyze the correlation between individual meteorological elements and wind point processing.

[0029] High Wind Scenario: We extracted meteorological data and corresponding output power data for high wind scenarios without ice and precipitation, and conducted calculations and analysis. We concluded that wind speed is the primary meteorological factor causing fluctuations in wind power output in high wind scenarios, and wind speed serves as the model input feature for these scenarios. Icing scenario: We extracted meteorological data and actual output data under icing conditions from the wind farm and conducted calculations and analysis. The optimal feature selection results for the icing scenario were precipitation, air pressure, temperature, and wind speed. Heavy precipitation scenarios: When total daily precipitation exceeds 10 mm and the temperature is above 0°C, precipitation is considered to occur, and a precipitation dataset is extracted from this. Using the Pearson correlation coefficient, it is found that wind power output in precipitation scenarios has a high correlation with wind speed, air pressure, and large-scale precipitation. Therefore, these three meteorological factors are selected as model inputs.

[0030] Based on the above analysis, a regression model is used to construct a wind power prediction model under different transition weather scenarios.

[0031] (3) Model Fusion A model fusion program was developed. During the real-time wind power forecast process, the program first determined whether the currently input weather data was a transitional weather event, identified the type of transitional weather event, and determined the duration of the transitional weather event. Based on this determination, the predicted values for the transitional weather event period in the base model's forecast were replaced with the model's predictions for the corresponding transitional weather scenario. Finally, a fused, high-precision wind power forecast was output.

[0032] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A wind power prediction method combined with transitional weather scenarios, characterized in that: The following steps are involved: Collecting wind power data and meteorological data, and preprocessing the data; Construct a basic wind power prediction model; Construct transitional weather scenarios; Construct a wind power prediction model under transitional weather scenarios; Based on the basic wind power prediction model and the models under different transition weather scenarios, the model is integrated to output the final wind power prediction results.

2. The wind power prediction method combined with transitional weather scenarios according to claim 1, characterized in that: The wind power data is collected from multiple wind farms, including wind turbine power output, theoretical power generation, equipment parameters and location parameters. The meteorological data includes wind speed, wind direction, temperature, humidity, air pressure and precipitation. The data preprocessing includes data format conversion, time series alignment and outlier processing.

3. The wind power prediction method combined with transitional weather scenarios according to claim 1, characterized in that: The construction of the basic wind power prediction model includes selecting a regression model, preprocessing sample data and dividing it into a training set and a test set, and using a grid search algorithm to tune the model parameters.

4. The wind power prediction method combined with transitional weather scenarios according to claim 1, characterized in that: The construction of the transitional weather scenario is based on three representative transitional weather processes: cold wave, heavy rainfall and icing. The relationship between wind power prediction error and weather process and the error generation mechanism are analyzed, and the transitional weather is divided into three scenarios: strong wind, icing and precipitation.

5. The wind power prediction method combined with transitional weather scenarios according to claim 1, characterized in that: The construction of the wind power prediction model under the transitional weather scenario includes using the Pearson correlation coefficient method to analyze the correlation between various meteorological elements and wind power, optimizing the meteorological sensitive characteristics under each scenario, and constructing the wind power prediction model under different scenarios based on the regression model.

6. The wind power prediction method combined with transitional weather scenarios according to claim 1, characterized in that: The model fusion is to determine whether the current input weather data is turning weather and which type of turning weather it belongs to during the real-time prediction of wind power, output the time period of the current turning weather, and replace the prediction result of the basic model in the time period when the turning weather occurs.

7. A wind power prediction system combined with transitional weather scenarios, characterized in that: include: A data collection and preprocessing module, used to collect wind power data and meteorological data, and preprocess the data; Basic model building module, used to build a basic wind power prediction model; A transitional weather scenario construction module, used to construct transitional weather scenarios; Scenario model building module, used to build a wind power prediction model under transitional weather scenarios; The model fusion module is used to perform model fusion based on the basic wind power prediction model and models under different transitional weather scenarios, and output the final wind power prediction results.

8. The wind power prediction system combined with transitional weather scenarios according to claim 7, characterized in that: The wind power data collected by the data collection and preprocessing module include wind turbine power output, theoretical power generation, equipment parameters and location parameters; meteorological data include wind speed, wind direction, temperature, humidity, air pressure and precipitation; preprocessing includes data format conversion, time series alignment and outlier processing.

9. The wind power prediction system combined with transitional weather scenarios according to claim 7, characterized in that: The basic model building module selects a regression model, preprocesses the sample data and divides them into training sets and test sets, and uses a grid search algorithm to tune the model parameters.

10. The wind power prediction system combined with transitional weather scenarios according to claim 7, characterized in that: The transitional weather scenario construction module takes three representative transitional weather processes, namely cold waves, heavy rainfall, and icing, as examples to analyze the relationship between wind power prediction errors and weather processes and the error generation mechanism, and divides transitional weather into three scenarios: strong winds, icing, and precipitation; the scenario model construction module uses the Pearson correlation coefficient method to analyze the correlation between various meteorological elements and wind power, optimizes the meteorological sensitive features under each scenario, and constructs wind power prediction models under different scenarios based on regression models; when predicting wind power in real time, the model fusion module determines whether the current input weather data is transitional weather and which type of transitional weather it belongs to, outputs the time period of the current transitional weather, and replaces the prediction results of the basic model in the time period when the transitional weather occurs.

Images