Method for predicting medium-and-long-term generating capacity of wind power plant

Abstract

The invention discloses a method for predicting medium-and-long-term generating capacity of a wind power plant, and belongs to the field of electric quantity prediction. The method comprises the steps of obtaining a monthly generating capacity prediction model and an annual generating capacity prediction model; obtaining seasonal weather forecast data of a to-be-predicted month at the position of the wind power plant, and obtaining statistical characteristics of a monthly wind speed sequence; inputting statistical characteristics of the monthly wind speed sequence into the monthly generating capacity prediction model to obtain a monthly generating capacity prediction value of the wind power plant in the to-be-predicted month; when the annual energy production prediction model is used for predicting the annual energy production of a to-be-predicted wind power plant in the next one to five years, a BCMM data set of CMIP6 of the to-be-predicted year at the wind power plant position is obtained, an annual wind speed sequence is counted, and the theoretical annual energy production is calculated; and obtaining a loss coefficient of the wind power plant, and calculating an annual generating capacity prediction value of the wind power plant in the to-be-predicted year. The method provides a more accurate and effective mode for medium and long term power generation prediction, and has important practical significance for development and utilization of wind energy resources.

Description

Technical Field

[0001] The present invention relates to the field of medium- and long-term power generation prediction, and in particular to a medium- and long-term power generation prediction method for a wind farm. Background Art

[0002] As a clean, renewable energy source, wind power plays a vital role in the global energy transition. To improve wind farm operational efficiency, reduce grid scheduling challenges, and ensure stable energy supply, wind farm power forecasting technology has become a key tool in modern power systems. Wind farm power output is influenced by multiple factors, including meteorological conditions, geographical environment, and seasonal variations, resulting in significant volatility and uncertainty. While this makes wind farm power forecasting challenging, it also highlights its importance in renewable energy integration and grid scheduling.

[0003] The primary purpose of wind farm power forecasting is to predict future power generation to support the rationality and stability of grid scheduling. First, forecasting can improve the stability of grid scheduling. Due to the intermittent and random nature of wind power, wind power fluctuates dramatically, posing a challenge to grid power supply. Accurate power forecasting helps the grid make proactive scheduling adjustments, balance power supply and demand, and ensure stable grid operation. Second, power forecasting facilitates the integration of renewable energy. In the context of large-scale grid integration, the volatility of wind power can cause grid fluctuations. Accurate forecasting helps better integrate wind power resources and ensure the safe integration of a high proportion of wind power. Furthermore, scientific power forecasting can reduce wind curtailment, improve wind power utilization efficiency, and minimize resource waste. Power forecasting also provides guidance for wind farm operations and maintenance, enabling more efficient equipment maintenance schedules, thereby extending equipment life and reducing maintenance costs.

[0004] In the application of monthly power forecasts, the forecast period covers wind power fluctuations over the next month, providing support for short-term grid scheduling and wind farm operations. Monthly forecasts are of great significance in electricity market transactions and bidding. Wind farms are required to submit power supply plans for the next month. Forecasts provide a reliable reference for power generation, avoiding penalty penalties or financial losses caused by forecast deviations. Furthermore, monthly forecasts provide a reference for grid dispatch, supporting the rational allocation of reserve capacity, reducing the impact of wind power fluctuations, and ensuring power supply stability. Furthermore, monthly forecasts help wind farms maximize power generation in months with strong wind resources, rationally schedule equipment maintenance, optimize operational strategies, and improve economic efficiency.

[0005] In contrast, annual power forecasts cover a longer timeframe, primarily focusing on wind power trends within the next year and the next one to five years. They provide data support for long-term wind farm operation strategies, grid planning, and government policies. Annual forecasts provide a scientific basis for wind farm planning and expansion, helping wind farm decision-makers assess the rationale for investment and expansion. In grid planning, annual forecasts help grid companies plan transmission lines and energy storage facilities in advance, mitigate the adverse effects of wind power fluctuations, and improve the grid's adaptability to future wind power fluctuations. At the government level, annual forecasts inform policymaking on wind power development trends, supporting the formulation of appropriate incentive policies and promoting the steady development of the wind power industry. Furthermore, annual forecasts help wind farms rationally allocate equipment maintenance budgets, optimize overhaul plans, improve equipment utilization, and control operating costs. Therefore, forecasting monthly power generation within the next year and annual power generation within the next one to five years is a pressing issue. Summary of the Invention

[0006] In order to solve the problems in the prior art, the present invention provides a method for predicting the medium- and long-term power generation of a wind farm.

[0007] The technical solutions of the present invention are as follows:

[0008] The present invention discloses a method for predicting medium- and long-term power generation of a wind farm, comprising the following steps:

[0009] Obtaining a prediction model, wherein the prediction model includes a monthly power generation prediction model for predicting the monthly power generation of the wind farm to be predicted within the next year, and an annual power generation prediction model for predicting the annual power generation of the wind farm to be predicted within the next one to five years;

[0010] When predicting the monthly power generation of a wind farm to be predicted within the next year, seasonal weather forecast data for the predicted month at the wind farm location is obtained to obtain the statistical characteristics of the monthly wind speed series for the predicted month; the statistical characteristics of the monthly wind speed series for the predicted month are input into the monthly power generation prediction model to obtain the predicted value of the monthly power generation of the wind farm in the predicted month;

[0011] When using an annual power generation prediction model to predict the annual power generation of a wind farm to be predicted in the next one to five years, the prediction method includes: obtaining a BCMM data set of CMIP6 for the year to be predicted at the location of the wind farm, downloading the east-west wind speed component and the north-south wind speed component for the year to be predicted, and interpolating the east-west wind speed component and the north-south wind speed component to the hub height of the wind farm, calculating the wind speed sequence at the hub height of the wind farm for the year to be predicted, counting the annual wind speed sequence for the year to be predicted, and then obtaining the theoretical power curve and the number of wind turbines of the wind turbines to calculate the theoretical annual power generation of the wind farm for the year to be predicted; obtaining the loss coefficient of the wind farm, multiplying the theoretical annual power generation for the year to be predicted by the loss coefficient, and obtaining a predicted value of the annual power generation of the wind farm in the year to be predicted.

[0012] Furthermore, the training process of the monthly power generation prediction model includes:

[0013] Obtain the historical monthly power generation of the wind farm to be predicted; obtain historical seasonal weather forecast data at the location of the wind farm to be predicted, obtain the historical monthly wind speed series at the hub height of the wind farm in units of months, and calculate the statistical characteristics of the historical monthly wind speed series; construct a data set together with the statistical characteristics of the historical monthly wind speed series and the historical monthly power generation for all months; wherein the historical monthly power generation is a label; use the data set to train the monthly power generation prediction model to be trained to obtain a trained monthly power generation prediction model, that is, obtain a monthly power generation prediction model for predicting the monthly power generation of the wind farm to be predicted in the next year.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] The present invention provides an innovative framework. In the method for predicting monthly power generation in the next year, the present invention establishes the statistical characteristics of wind speed at the hub height of the wind farm by statistically analyzing the monthly statistical characteristics of wind speed corresponding to seasonal weather forecast data. The relationship between the monthly statistical characteristics of wind speed and the monthly power generation of the wind farm is predicted through artificial neural network modeling, thereby realizing the prediction of the monthly power generation of the wind farm. In the method for predicting annual power generation in the next one to five years, the present invention uses an annual power generation prediction model to obtain the annual wind speed sequence corresponding to the BCMM data set of CMIP6 to the hub height of the wind farm, calculate the theoretical annual power generation of the wind farm in the year to be predicted, obtain the loss coefficient of the wind farm, calculate the predicted value of the annual power generation of the wind farm in the year to be predicted, and realize the accurate estimation of the annual power generation of the wind farm. The present invention not only solves the prediction problem in the prior art, but also provides strong support for wind farm managers to evaluate future benefits and risks, and has significant practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the flow of the method for predicting medium- and long-term power generation of a wind farm according to the present invention;

[0017] Figure 2 This is a graph showing the application test results of the monthly power generation prediction method within one year of the medium- and long-term power generation prediction method for a wind farm according to the present invention;

[0018] Figure 3 This is a graph showing the application test results of the invented method for predicting annual power generation within one to five years for the medium- and long-term power generation forecasting method for wind farms. DETAILED DESCRIPTION

[0019] In order to facilitate ordinary technicians in this field to understand and implement the present invention, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the implementation examples described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0020] The present invention provides a medium- and long-term wind farm power generation forecasting method for forecasting the monthly power generation of a wind farm within the next year and the annual power generation within the next one to five years, so as to solve the problems existing in the prior art.

[0021] In the method for predicting monthly power generation in the next year, the present invention establishes the statistical characteristics of wind speed at the hub height of the wind farm location by statistically analyzing the monthly statistical characteristics of wind speed corresponding to seasonal weather forecast data, and then establishes a prediction model for monthly power generation, thereby achieving an accurate estimation of the monthly power generation of the wind farm.

[0022] In the method for predicting annual power generation in the next one to five years, the present invention uses the wind speed sequence corresponding to the predicted year in the BCMM dataset of CMIP6 to obtain the annual wind speed sequence at the hub height of the wind farm location, and then calculates the theoretical annual power generation of the wind farm in the predicted year; obtains the loss coefficient of the wind farm, and calculates the predicted annual power generation value of the wind farm in the predicted year, thereby achieving an accurate estimation of the wind farm's annual power generation.

[0023] like Figure 1 As shown, the method for predicting medium- and long-term wind farm power generation of the present invention includes the following steps:

[0024] S1: Obtain prediction models, including a monthly power generation prediction model for predicting the monthly power generation of the wind farm to be predicted in the next year, and an annual power generation prediction model for predicting the annual power generation of the wind farm to be predicted in the next one to five years.

[0025] S2: When predicting the monthly power generation of a wind farm to be predicted within the next year, obtain the seasonal weather forecast data for the month to be predicted at the location of the wind farm to obtain the statistical characteristics of the monthly wind speed series for the month to be predicted; input the statistical characteristics of the monthly wind speed series for the month to be predicted into the monthly power generation prediction model to obtain the predicted value of the monthly power generation of the wind farm in the month to be predicted; wherein, the statistical characteristics of the monthly wind speed series include the maximum wind speed, the minimum wind speed, the average wind speed and the wind speed variance.

[0026] S3: When using the annual power generation prediction model to predict the annual power generation of the wind farm to be predicted in the next one to five years, the prediction method includes: obtaining the BCMM data set of CMIP6 for the year to be predicted at the location of the wind farm, downloading the east-west wind speed component and the north-south wind speed component for the year to be predicted, and interpolating the east-west wind speed component and the north-south wind speed component to the hub height of the wind farm, calculating the wind speed sequence at the hub height of the wind farm for the year to be predicted, counting the annual wind speed sequence for the year to be predicted, and then obtaining the theoretical power curve and the number of wind turbines of the wind turbines to calculate the theoretical annual power generation of the wind farm for the year to be predicted; obtaining the loss coefficient of the wind farm, multiplying the theoretical annual power generation for the year to be predicted by the loss coefficient, and obtaining the annual power generation forecast value of the wind farm in the year to be predicted.

[0027] In S1, the monthly power generation prediction model of the present invention is an artificial neural network model built using Pytorch. The monthly power generation prediction model is obtained by training the monthly power generation prediction model to be trained. The training process of the monthly power generation prediction model in S1 includes:

[0028] Obtain the historical monthly power generation of the wind farm to be predicted; obtain historical seasonal weather forecast data at the location of the wind farm to be predicted, obtain the historical monthly wind speed series at the hub height of the wind farm on a monthly basis, and calculate the statistical characteristics of the historical monthly wind speed series; construct a data set by combining the statistical characteristics of the historical monthly wind speed series and the historical monthly power generation for all months; wherein the historical monthly power generation is a label; use the data set to train the monthly power generation prediction model to obtain a trained monthly power generation prediction model, that is, obtain an annual power generation prediction model for predicting the annual power generation of the wind farm to be predicted in the next one to five years. The historical monthly power generation data collected for the wind farm to be predicted should be at least five years long and can be a monthly power generation report of the wind farm, containing monthly power generation information, which serves as a label for the training model.

[0029] The dataset is divided into a training set and a test set. The training set is used for learning, and the test set is used to evaluate the algorithm model. During the training process, hyperparameter tuning is performed. By trying different parameter values, the corresponding loss function values ​​are calculated, and the parameter combination that minimizes the loss function is selected as the final model parameter setting.

[0030] In the present invention, when training the monthly power generation prediction model to be trained, the historical monthly wind speed sequence at the hub height of the wind farm is obtained, including:

[0031] After obtaining the historical seasonal weather forecast data for the location of the wind farm to be predicted, the east-west wind speed component, north-south wind speed component, and wind potential height at all timestamps of each month are downloaded. Based on the potential height, the east-west wind speed component and the north-south wind speed component are interpolated to the hub height of the wind farm. The historical monthly wind speed series at the hub height of the wind farm is calculated in units of months.

[0032] In a preferred embodiment of the present invention, the dataset used to train the monthly power generation prediction model also includes monthly features, and the monthly features are obtained by encoding the month information using a sin function. After the monthly power generation prediction model is trained using the dataset including the monthly features, when the trained monthly power generation prediction model is applied, the monthly features of the month to be predicted need to be input into the trained monthly power generation prediction model.

[0033] In S3, the theoretical annual power generation of the wind farm in the predicted year is calculated, including: finding the wind turbine output power corresponding to the wind speed data in the wind speed sequence from the theoretical power curve of the wind turbine, then multiplying the wind turbine output power by the duration of the corresponding wind speed to obtain the annual power generation of the next wind turbine at the wind speed, and then multiplying the annual power generation of one wind turbine by the number of wind turbines to obtain the annual power generation of the wind farm, that is, the theoretical annual power generation.

[0034] In a specific embodiment of the present invention, in S3, the BCMM dataset of CMIP6 is a dataset of the ssp245 path in the BCMM data of CMIP6.

[0035] Furthermore, because actual wind farm power generation is often affected by factors such as power curtailment, terrain losses, and wind speed uncertainty, after calculating the theoretical annual power generation, a loss coefficient must be introduced to correct the theoretical annual power generation to obtain the annual power generation forecast. The selection of the loss coefficient depends on factors such as the specific wind farm characteristics, geographical environment, climate conditions, and wind turbine layout. It is usually determined through experience and historical data and is generally preset.

[0036] An embodiment of the present invention provides a method for predicting medium- and long-term wind farm power generation based on numerical weather forecasts. The specific steps for predicting monthly power generation within a year are as follows:

[0037] Wind farms at sites A and B have historical monthly power generation data for nearly five years, from January 2020 to October 2024. To model the relationship between climatological wind speed data and monthly power generation, we first downloaded the SeasonalForecast monthly wind speed data for the corresponding historical months. EC Seasonal Forecast is a high-precision weather forecast product that provides forecasts for key meteorological variables such as wind speed and temperature for the coming months. Wind speed information was then interpolated to the site and hub height. Based on this data, the maximum, minimum, mean, and variance of the monthly wind speed series were calculated. These statistical features, along with the month code, were used as input features for artificial neural network training, with monthly power generation serving as the model's prediction label.

[0038] Because the data is divided into monthly units, the data volume is relatively small, making it difficult to support efficient neural network training. In this context, the goal of this model is to use climatological wind speed data to establish a correlation with monthly power generation. Since wind farms A and B share the same turbine model and number, the historical data from both wind farms can be combined to form a unified training set. This approach not only effectively expands the dataset size but also increases the number of training samples for the model, thereby improving the model's prediction accuracy and generalization ability for monthly power generation.

[0039] Therefore, the model training set includes the monthly power generation data of wind farms at sites A and B from January 2020 to December 2022, and the model effect is verified by the monthly power generation data of the wind farm at site A in 2023.

[0040] The test results are as follows Figure 2 As shown in Figure 2, the average annual error in the monthly power generation forecast for the wind farm at Site A from January to December 2023 is 14.9%. This model can well reflect the power generation trend of the wind farm at Site A.

[0041] To predict the annual power generation of the wind farm at site C within the next one to five years, the specific steps are as follows:

[0042] The wind farm at Site C, which has six years of actual annual power generation data from 2017 to 2022, was selected as a test wind farm. Meteorological data for the corresponding time period at the site of Site C, including all timestamps for the east-west and north-south wind speed components, was downloaded from the CMIP6-BCMM model. Spatial interpolation was used to convert the gridded wind speed data to the specific location and hub height of the wind farm site. The theoretical annual power generation of the entire wind farm was calculated by multiplying the theoretical power generation of a single wind turbine by the number of wind turbines in the wind farm. Multiplying the power generation of a single wind turbine by the number of wind turbines in the wind farm yields the theoretical annual power generation of the wind farm under zero power losses. This theoretical value is calculated based on meteorological data and wind turbine power curves and does not take into account other factors affecting actual operation. The actual power generation of a wind farm is often affected by various factors, such as power curtailment, terrain losses, and wind speed uncertainty. To further align the theoretical annual power generation, a loss factor was introduced to correct the theoretical power generation. This value is set based on experience.

[0043] The test results are as follows Figure 3 As shown, CMIP6 data performs significantly more stably in annual power generation forecasts. The results show that the error between the predicted and actual annual power generation from 2017 to 2022 is small, with the forecast error remaining within 10% for most years and an average annual error of 7.01%. The annual data is highly accurate. Although CMIP6 data have limitations in capturing monthly climate fluctuations, they can well reflect the changing characteristics of the annual climate state in terms of long-term trends.

[0044] The embodiment described above is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Persons skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent substitution or equivalent transformation falls within the scope of protection of the present invention.

Claims

1. A method for predicting medium- and long-term power generation of a wind farm, characterized in that: The following steps are involved: Obtaining a prediction model, wherein the prediction model includes a monthly power generation prediction model for predicting the monthly power generation of the wind farm to be predicted within the next year, and an annual power generation prediction model for predicting the annual power generation of the wind farm to be predicted within the next one to five years; When predicting the monthly power generation of a wind farm to be predicted within the next year, seasonal weather forecast data for the predicted month at the wind farm location is obtained to obtain the statistical characteristics of the monthly wind speed series for the predicted month; the statistical characteristics of the monthly wind speed series for the predicted month are input into the monthly power generation prediction model to obtain the predicted value of the monthly power generation of the wind farm in the predicted month; When using an annual power generation prediction model to predict the annual power generation of a wind farm to be predicted in the next one to five years, the prediction method includes: obtaining a BCMM data set of CMIP6 for the year to be predicted at the location of the wind farm, downloading the east-west wind speed component and the north-south wind speed component for the year to be predicted, and interpolating the east-west wind speed component and the north-south wind speed component to the hub height of the wind farm, calculating the wind speed sequence at the hub height of the wind farm for the year to be predicted, counting the annual wind speed sequence for the year to be predicted, and then obtaining the theoretical power curve and the number of wind turbines of the wind turbines to calculate the theoretical annual power generation of the wind farm for the year to be predicted; obtaining the loss coefficient of the wind farm, multiplying the theoretical annual power generation for the year to be predicted by the loss coefficient, and obtaining a predicted value of the annual power generation of the wind farm in the year to be predicted.

2. A method for predicting medium- and long-term wind farm power generation according to claim 1, characterized in that: The training process of the monthly power generation prediction model includes: Obtain the historical monthly power generation of the wind farm to be predicted; obtain historical seasonal weather forecast data at the location of the wind farm to be predicted, obtain the historical monthly wind speed series at the hub height of the wind farm in units of months, and calculate the statistical characteristics of the historical monthly wind speed series; construct a data set together with the statistical characteristics of the historical monthly wind speed series and the historical monthly power generation for all months; wherein the historical monthly power generation is a label; use the data set to train the monthly power generation prediction model to be trained to obtain a trained monthly power generation prediction model, that is, obtain a monthly power generation prediction model for predicting the monthly power generation of the wind farm to be predicted in the next year.

3. A method for predicting medium- and long-term wind farm power generation according to claim 2, characterized in that: The step of obtaining the historical monthly wind speed sequence at the hub height of the wind farm includes: After obtaining the historical seasonal weather forecast data for the location of the wind farm to be predicted, the east-west wind speed component, north-south wind speed component, and wind potential height at all timestamps of each month are downloaded. Based on the potential height, the east-west wind speed component and the north-south wind speed component are interpolated to the hub height of the wind farm. The historical monthly wind speed series at the hub height of the wind farm is calculated in units of months.

4. A method for predicting medium- and long-term wind farm power generation according to claim 1, characterized in that: The monthly power generation prediction model is an artificial neural network model.

5. The method for predicting medium- and long-term wind farm power generation according to claim 1, characterized in that: The statistical characteristics of the monthly wind speed series include the maximum wind speed, the minimum wind speed, the average wind speed and the wind speed variance.

6. A method for predicting medium- and long-term wind farm power generation according to claim 2, characterized in that: The data set also includes a month feature, which is obtained by encoding the month information using a sin function.

7. A method for predicting medium- and long-term wind farm power generation according to claim 6, characterized in that: After the monthly power generation prediction model to be trained is trained using the data set including the monthly features, when the trained monthly power generation prediction model is applied, the monthly features of the month to be predicted also need to be input into the trained monthly power generation prediction model.

8. The method for predicting medium- and long-term wind farm power generation according to claim 1, characterized in that: The CMIP6 BCMM dataset is a dataset of the ssp245 pathway in the CMIP6 BCMM data.

Images