基于数值天气预报修正和气象相似日的短期风电功率区间预测

Short-term Wind Power Interval Prediction Based on Numerical Weather Forecast Correction and Meteorological Similar Days

为解决风电出力因其波动性无法准确预测的问题,提出了一种基于改进BP神经网络和Bootstrap方法的短期风电出力区间预测方法。拟合历史数值天气预报(NWP)及实际观测值,修正待预测时刻NWP数据,并根据灰色关联系数法筛选气象相似时刻。采用粒子群算法优化BP神经网络参数,引入Bootstrap法增加数据多样性,建立多个确定性预测模型。运用百分位数估计法得到了给定置信水平下各时刻的功率波动区间。以国内某风电场为例,提前24 h预测风功率,结果表明,所得区间各项评估指标均满足实际工程需要。

In order to solve the problem that wind power output cannot be accurately predicted due to its fluctuation,a short-term wind power output interval prediction method based on improved BP neural network and Bootstrap method was proposed.The historical numerical weather forecast(NWP)and the actual observation values were fitted,and the NWP data of the predicted time were corrected,and the similar meteorological time was screened according to the grey correlation coefficient method.The parameters of BP neural network were optimized by particle swarm optimization algorithm,and the Bootstrap method was introduced to increase the data diversity,and several deterministic prediction models were established.The percentile estimation method is used to obtain the power fluctuation interval at a given confidence level.Taking a domestic wind farm as an example,the wind power is predicted 24h in advance,and the results show that all the evaluation indexes in the obtained interval meet the practical engineering needs.