基于多重数据筛选的短期风电功率区间优化预测OA

To solve the problem of inaccurate prediction of wind power output due to its volatility,a short-term wind power output interval prediction scheme based on multiple data filtering was proposed by combining the dependence of meta heuristic algorithms on training data.Firstly,the wind speed in weather forecasts was corrected by using curve fitting,and historical data similar to the predicted daily weather types was obtained by conducting initial wind speed screening based on meteorological types.A secondary screening was conducted on the correlation of historical data using grey correlation degree,and the screening results were given that were similar to the trend of wind speed changes on the day to be predicted.Then,the empirical mode decomposition method was used to decompose the secondary screening results and obtain the intrinsic mode functions(IMF)of different frequencies.Moreover,the war strategy optimization algorithm(WSO)was used to optimize the parameters of long short term memory(LSTM)networks and establish deterministic prediction models for different components.By converting the relationship to obtain the point prediction results of wind power,the Bootstrap sampling percentile method was introduced to provide the fluctuation range of power at a certain confidence level.Finally,an analysis was conducted using measured data from an actual wind farm.Results confirm the effectiveness and accuracy of the proposed multiple data screening method,showcasing obtained intervals with excellent coverage and narrower bandwidth in wind power interval forecasting.