Wind power generation power prediction method
Through multi-dimensional feature extraction and fusion, combined with autocorrelation analysis and ELM model, the problem of insufficient utilization of meteorological data and historical data in traditional wind power prediction technology is solved, and the accuracy and stability of wind power prediction is improved, and the prediction needs of different time scales are adapted.
Patent Information
- Application Number
- CN202510434732.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-08
- Publication Date
- 2025-08-08
AI Technical Summary
Traditional wind power power prediction technology relies too much on meteorological forecasting data, is susceptible to errors, and is difficult to capture the complex nonlinear characteristics and multi-scale changes of wind power systems, and fails to make full use of historical data, resulting in inconsistent prediction results on different time scales.
Through multi-dimensional feature extraction and fusion, the historical operation data of the wind turbine, including timing, seasonality and statistical characteristics, combined with autocorrelation analysis, sliding window feature analysis and standardized processing, is used to predict wind power power using a complex extreme learning machine (ELM) model.
It improves the accuracy and stability of wind power generation power prediction, adapts to forecasting needs at different time scales, and provides reliable support for wind farm management and grid scheduling.
Smart Images
Figure CN120454015A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wind turbine power prediction, and in particular to a wind power generation power prediction method. Background Art
[0002] As an important component of clean energy, wind power generation and its power forecasting are crucial for grid dispatching and energy management. Currently, traditional wind power forecasting technologies have several significant shortcomings. First, over-reliance on meteorological forecast data can easily lead to prediction results being susceptible to weather forecast errors. Second, simple time series models have difficulty capturing the complex nonlinear characteristics and multi-scale changes of wind power systems. Traditional methods often ignore the operating status and historical performance of wind turbines themselves, fail to fully utilize available historical data, and have difficulty adapting to the dynamic changes and seasonal fluctuations of wind farms. They lack comprehensive consideration of long-term trends and short-term fluctuations, resulting in inconsistent prediction results at different time scales. Based on the above problems, we designed a wind power forecasting method. Summary of the Invention
[0003] The purpose of the present invention is to provide a wind power generation prediction method, which fully utilizes historical data of wind turbines through multi-dimensional feature extraction and fusion, effectively captures temporal, seasonal and statistical characteristics, and improves the generalization ability of the model through advanced feature fusion strategies. Through standardization processing and the use of complex ELM models, the accuracy and stability of the prediction are enhanced. The application of the present invention can adapt to the wind power generation prediction needs of different time scales and provide reliable support for wind farm management and power grid scheduling.
[0004] The embodiments of the present invention are achieved through the following technical solutions:
[0005] A method for predicting wind power generation, the method comprising the steps of:
[0006] Obtain historical operating data of wind turbines, extract time series features of the historical operating data of wind turbines, and create seasonal features;
[0007] Autocorrelation analysis and sliding window feature analysis are performed on the historical operating data of wind turbines to obtain the lag characteristics and statistical characteristics of the wind turbines. The lag characteristics and time series characteristics of the wind turbines are integrated to obtain the first fusion feature. The statistical characteristics and seasonal indicators of the wind turbines are integrated to obtain the second fusion feature.
[0008] The first fusion feature and the second fusion feature are standardized and used as input of a preset wind power prediction model to complete the training of the wind power prediction model. The collected real-time operation data of the wind turbines are input into the trained wind power prediction model for calculation to complete the wind power prediction.
[0009] Optionally, the historical operating data of the wind turbine generator set specifically includes: historical environmental data of the wind turbine generator set and historical power generation data of the wind turbine generator set, wherein the historical environmental data of the wind turbine generator set includes: wind speed, wind direction, temperature and air pressure.
[0010] Optionally, the autocorrelation analysis and sliding window feature analysis are performed on the historical operating data of the wind turbine generator set respectively, wherein the autocorrelation analysis includes: performing autocorrelation function calculation on the historical operating data of the wind turbine generator set, and performing partial autocorrelation function calculation on the historical operating data of the wind turbine generator set.
[0011] Optionally, the autocorrelation function is calculated on the historical operating data of the wind turbine generator set, and the calculation formula is:
[0012]
[0013] Among them, x t is the observed value at time t, is the mean value of the entire series, N is the total number of observations, k is the time lag, R(k) is the autocorrelation coefficient of lag k, and t is the current time step.
[0014] Optionally, the partial autocorrelation function is calculated on the historical operating data of the wind turbine generator set, and the calculation formula is:
[0015]
[0016] Among them, φ kk is the partial autocorrelation coefficient at lag k.
[0017] Optionally, a sliding window feature analysis is performed on the historical operating data of the wind turbine. The specific calculation process is as follows:
[0018]
[0019]
[0020] Among them, w is the sliding window size, x i is the observation value at time i, is the moving average in the sliding window of time t, is the maximum value in window w, is the minimum value in the window w, is the standard deviation within the window w.
[0021] Optionally, the hysteresis characteristic and the time sequence characteristic of the wind turbine generator are integrated to obtain a first fusion characteristic, which is specifically:
[0022]
[0023] in, is the first fusion feature, x t-n is the original lagged variable, H t For hours, D t For the week, M t is the month, a, b, c are parameters respectively.
[0024] Optionally, the statistical characteristics of the wind turbine generator set and the seasonal index are integrated to obtain a second integrated feature, which is specifically:
[0025]
[0026] Among them, F t s is the second fusion feature, F t is the original feature, D y For the year and day, S d is the sunshine duration, S m is the longest daylight, T c is the current temperature, T a is the average temperature, T x 、T n are the highest and lowest temperatures respectively, and α, β, and γ are parameters respectively.
[0027] Optionally, the preset wind power generation prediction model is specifically an ELM model, which consists of an input layer, a hidden layer and an output layer, and its calculation process is:
[0028] y pred =g(W in X new +b in )W out
[0029] W in =[w ij ] d×L ,b in =[b j ] L×1
[0030] H=g(W in X norm +b in )
[0031] W out =(H TH+λI) -1 H T Y
[0032] Among them, W in is the input weight matrix, b in is the input bias vector, W out is the output weight matrix, X norm is the standardized input feature, including the first fusion feature and the second fusion feature, H is the hidden layer output matrix, Y is the actual wind power matrix, g is the activation function, X new is the new standardized fusion feature, y pred is the predicted wind power generation capacity.
[0033] Optionally, the wind power generation prediction model specifically aims to maintain the prediction accuracy of the wind power generation prediction model while improving the generalization ability and stability of the wind power generation prediction. The calculation formula of the objective function is:
[0034]
[0035] Among them, J(W out ) is the objective function, is the square of the Frobenius norm to measure the difference between the predicted value and the actual value, λ1 is the L2 regularization parameter, which controls the strength of L2 regularization, λ2 is the L1 regularization parameter, which controls the strength of L1 regularization, ‖W out ‖1 is the L1 norm of the output weight matrix.
[0036] The technical solutions of the embodiments of the present invention have at least the following advantages and beneficial effects:
[0037] The present invention fully utilizes historical data of wind turbines through multi-dimensional feature extraction and fusion, effectively capturing temporal, seasonal and statistical characteristics, and improving the generalization ability of the model through advanced feature fusion strategies. The accuracy and stability of the prediction are enhanced through standardized processing and the use of complex ELM models. The application of the present invention can adapt to the wind power prediction needs of different time scales and provide reliable support for wind farm management and power grid scheduling. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 A schematic flow chart of a wind power generation power prediction method provided by the present invention. DETAILED DESCRIPTION
[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0040] like Figure 1 As shown, the present invention provides one embodiment: a method for predicting wind power generation, the method comprising the steps of:
[0041] Obtain historical operating data of wind turbines, extract time series features of the historical operating data of wind turbines, and create seasonal features;
[0042] Autocorrelation analysis and sliding window feature analysis are performed on the historical operating data of wind turbines to obtain the lag characteristics and statistical characteristics of the wind turbines. The lag characteristics and time series characteristics of the wind turbines are integrated to obtain the first fusion feature. The statistical characteristics and seasonal indicators of the wind turbines are integrated to obtain the second fusion feature.
[0043] The first fusion feature and the second fusion feature are standardized and used as input of a preset wind power prediction model to complete the training of the wind power prediction model. The collected real-time operation data of the wind turbines are input into the trained wind power prediction model for calculation to complete the wind power prediction.
[0044] During implementation, this embodiment collects historical operating data from wind turbines, including meteorological data such as wind speed, wind direction, temperature, humidity, and air pressure, as well as operating parameters such as power generation, rotational speed, and pitch angle. Time series analysis is performed on each of these historical wind turbine operating data, extracting temporal features such as trends, periodicity, and autocorrelation. Long-term trends are captured using methods such as moving averages and exponential smoothing, while periodic patterns are identified using Fourier transforms or wavelet analysis. Furthermore, seasonal features, such as month, season, and hour, are generated based on the data's timestamps. These features incorporate seasonal variations into the model, helping to capture seasonal variations in power generation.
[0045] In this embodiment, the autocorrelation analysis and sliding window feature analysis are performed on the historical operating data of the wind turbine generator set respectively, wherein the autocorrelation analysis includes: performing autocorrelation function calculation on the historical operating data of the wind turbine generator set, and performing partial autocorrelation function calculation on the historical operating data of the wind turbine generator set.
[0046] The autocorrelation function calculation is performed on the historical operation data of the wind turbine generator set, and the calculation formula is:
[0047]
[0048] Among them, x t is the observed value at time t, is the mean value of the entire series, N is the total number of observations, k is the time lag, R(k) is the autocorrelation coefficient of lag k, and t is the current time step.
[0049] The partial autocorrelation function is calculated for the historical operating data of the wind turbine generator set, and the calculation formula is:
[0050]
[0051] Among them, φ kk is the partial autocorrelation coefficient at lag k.
[0052] During implementation, this embodiment calculates the autocorrelation function (ACF) and partial autocorrelation function (PACF) to identify important lag features, so as to identify repetitive patterns in historical data and help the ELM model understand temporal relationships.
[0053] The sliding window feature analysis of the historical operating data of wind turbines is performed, and the specific calculation process is as follows:
[0054]
[0055] Among them, w is the sliding window size, x i is the observation value at time i, is the moving average in the sliding window of time t, is the maximum value in window w, is the minimum value in the window w, is the standard deviation within the window w.
[0056] This embodiment uses a sliding window technique to calculate statistical features such as mean, variance, maximum, minimum, etc., so as to capture short-term fluctuations and local patterns of the data.
[0057] In a specific application of this embodiment, the hysteresis characteristics and the timing characteristics of the wind turbine generator are integrated to obtain a first fusion characteristic, which is specifically:
[0058]
[0059] in, is the first fusion feature, x t-n is the original lagged variable, H t For hours, D t For the week, M t is the month, a, b, c are parameters respectively.
[0060] The statistical characteristics of the wind turbine generator set and the seasonal index are integrated to obtain the second integrated characteristics, which are specifically:
[0061]
[0062] Among them, F t s is the second fusion feature, F t is the original feature, D y For the year and day, S d is the sunshine duration, S m is the longest daylight, T c is the current temperature, T a is the average temperature, T x 、T n are the highest and lowest temperatures respectively, and α, β, and γ are parameters respectively.
[0063] This embodiment integrates the lagged characteristics of wind turbines with their time series characteristics. By using trigonometric functions, it can capture complex nonlinear periodic patterns. It combines lagged variables with intra-day, intra-week, and monthly periodic variations, comprehensively capturing the multi-scale influence of time. The parameters a, b, and c allow for adjustment of the weights of influence at different time scales, enabling the ELM model to adapt to the unique patterns of different regions or seasons. Furthermore, by integrating the statistical characteristics of wind turbines with seasonal indicators and comparing current conditions with average or extreme conditions, it not only enables adaptive adjustment of feature strength but also integrates information from multiple dimensions, such as annual seasonality, sunshine duration, and temperature variations, providing a more comprehensive feature representation.
[0064] In the application of this embodiment, the preset wind power prediction model is specifically an ELM model, which consists of an input layer, a hidden layer, and an output layer. The training process of the ELM model is to randomly initialize the input weights and biases, calculate the hidden layer output, and calculate the output weights. The calculation process is as follows:
[0065] W in =[w ij ] d×L ,b in =[b j ] L×1
[0066] H=g(W in X norm +b in )
[0067] W out =(H T H+λI) -1 H T Y
[0068] Among them, Win is the input weight matrix, b in is the input bias vector, W out is the output weight matrix, X norm is the standardized input feature, including the first fusion feature and the second fusion feature, H is the hidden layer output matrix, Y is the actual wind power matrix, g is the activation function, X new is the new standardized fusion feature, y pred is the predicted wind power generation capacity.
[0069] For new input data: y pred =g(W in X new +b in )W out .
[0070] Furthermore, the wind power generation prediction model specifically aims to maintain the prediction accuracy of the wind power generation prediction model while improving the generalization ability and stability of the wind power generation prediction. The calculation formula of its objective function is:
[0071]
[0072] Among them, J(W out ) is the objective function, is the square of the Frobenius norm to measure the difference between the predicted value and the actual value, λ1 is the L2 regularization parameter, which controls the strength of L2 regularization, λ2 is the L1 regularization parameter, which controls the strength of L1 regularization, ‖W out ‖1 is the L1 norm of the output weight matrix.
[0073] This embodiment establishes an objective function so that the ELM model can achieve better generalization and stability while maintaining prediction accuracy. The regularization effect of the ELM model can be fine-tuned by adjusting λ1 and λ2, thereby achieving optimal performance in different application scenarios.
[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.
Claims
1. A method for predicting wind power generation, characterized in that: The steps of the method include: Obtain historical operating data of wind turbines, extract time series features of the historical operating data of wind turbines, and create seasonal features; Autocorrelation analysis and sliding window feature analysis are performed on the historical operating data of wind turbines to obtain the lag characteristics and statistical characteristics of the wind turbines. The lag characteristics and time series characteristics of the wind turbines are integrated to obtain the first fusion feature. The statistical characteristics and seasonal indicators of the wind turbines are integrated to obtain the second fusion feature. The first fusion feature and the second fusion feature are standardized and used as input of a preset wind power prediction model to complete the training of the wind power prediction model. The collected real-time operation data of the wind turbines are input into the trained wind power prediction model for calculation to complete the wind power prediction.
2. The wind power generation power prediction method according to claim 1, characterized in that: The historical operating data of the wind turbine generator set specifically includes historical environmental data of the wind turbine generator set and historical power generation data of the wind turbine generator set, wherein the historical environmental data of the wind turbine generator set includes wind speed, wind direction, temperature and air pressure.
3. The wind power generation power prediction method according to claim 2, characterized in that: The autocorrelation analysis and sliding window feature analysis are performed on the historical operation data of the wind turbine generator set respectively, wherein the autocorrelation analysis includes: performing autocorrelation function calculation on the historical operation data of the wind turbine generator set, and performing partial autocorrelation function calculation on the historical operation data of the wind turbine generator set.
4. The wind power generation power prediction method according to claim 3, characterized in that: The autocorrelation function calculation is performed on the historical operation data of the wind turbine generator set, and the calculation formula is: Among them, x t is the observed value at time t, is the average value of the entire series, N is the total number of observations, k is the time lag, R(k) is the autocorrelation coefficient of lag k, and t is the current time step.
5. The wind power generation power prediction method according to claim 4, characterized in that: The partial autocorrelation function is calculated for the historical operating data of the wind turbine generator set, and the calculation formula is: Among them, φ kk is the partial autocorrelation coefficient at lag k.
6. The wind power generation power prediction method according to claim 5, characterized in that: The sliding window feature analysis of the historical operating data of wind turbines is performed, and the specific calculation process is as follows: Among them, w is the sliding window size, x i is the observation value at time i, is the moving average in the sliding window of time t, is the maximum value in window w, is the minimum value in the window w, is the standard deviation within the window w.
7. The wind power generation power prediction method according to claim 6, characterized in that: The hysteresis characteristics and the time sequence characteristics of the wind turbine generator system are integrated to obtain the first fusion characteristics, which are specifically: in, is the first fusion feature, x t-n is the original lagged variable, H t For hours, D t For the week, M t is the month, a, b, c are parameters respectively.
8. The wind power generation power prediction method according to claim 7, characterized in that: The statistical characteristics of the wind turbine generator set and the seasonal index are integrated to obtain the second integrated characteristics, which are specifically: in, is the second fusion feature, F t is the original feature, D y For the year and day, S d is the sunshine duration, S m is the longest daylight, T c is the current temperature, T a is the average temperature, T x 、T n are the highest and lowest temperatures respectively, and α, β, and γ are parameters respectively.
9. The wind power generation power prediction method according to claim 8, characterized in that: The preset wind power generation prediction model is specifically an ELM model, which consists of an input layer, a hidden layer and an output layer. Its calculation process is: y pred =g(W in X new +b in )W out W in =[w ij ] d×L ,b in =[b j ] L×1 H=g(W in X norm +b in ) W out =(H T H+λI) -1 H T Y Among them, W in is the input weight matrix, b in is the input bias vector, W out is the output weight matrix, X norm is the standardized input feature, including the first fusion feature and the second fusion feature, H is the hidden layer output matrix, Y is the actual wind power matrix, g is the activation function, X new is the new standardized fusion feature, and ypred is the predicted wind power.
10. The wind power generation power prediction method according to claim 9, characterized in that: The wind power generation prediction model specifically aims to maintain the prediction accuracy of the wind power generation prediction model while improving the generalization ability and stability of the wind power generation prediction. The calculation formula of its objective function is: Among them, J(W out ) is the objective function, is the square of the Frobenius norm to measure the difference between the predicted value and the actual value, λ1 is the L2 regularization parameter, which controls the strength of L2 regularization, λ2 is the L1 regularization parameter, which controls the strength of L1 regularization, ‖W out ‖1 is the L1 norm of the output weight matrix.