融合RBF神经网络和集对分析的风电功率超短期预测

Ultra-short-term Wind Power Forecasting Integrated RBF Neural Network and Set Pair Analysis

风电功率的随机波动性是制约风电功率预测精度提高的关键问题之一,其中风速波动性以及风电转换不确定性是造成风电功率波动的两个主要原因.本文首先分析在风电功率预测中计及风电场状态的必要性;其次以风机运行状态充当输入变量,采用互信息理论修正外部NWP风速,引入集对分析对风电场内部状态特征参量进行匹配预测,构建计及风电场运行状态的以一种多输入-单输出的RBF神经网络为核心的风功率预测框架;最后采用吉林省某风电场的实际数据进行分析.对比多种预测算法,通过算例结果表明,所提方法可以有效地提升风电功率预测的精度.

The stochastic volatility is one of the key problems that restrict the improvement of wind power forecast precision.Wind speed volatility and non-determinacy of wind power conversion are the two principal reasons of power volatility.This paper analyzed the necessity of considering the state of the wind farm in the wind power forecast,introduced the set pair analysis to match and predict the characteristic parameters of the internal state of the wind farm,and constructed a multi input single output RBF neural network model for wind power forecast considering the running status of a wind farm.The operation state of the wind turbine was used as the input variable,and the mutual information theory was adopted to modify the external NWP wind speed.Finally,a real wind farm information in Jilin Province was analyzed,and a variety of forecast algorithms were compared.The results indicate that the proposed method can improve with effect the forecast precision.