基于风速融合和NARX神经网络的短期风电功率预测
2020-06-19徐遵义王俊雪
徐遵义 王俊雪



摘 要: 高精度的短期风电功率预测是保证电网日常调度及运行安全的关键因素。目前,国内短期风电功率预测精度普遍低于国外水平。为了提高风机短期功率预测精度,提出一种基于风速融合和NARX神经网络的预测模型。该方法对同一地点不同数据源提供的预报风速进行融合,采用NARX神经网络进行短期风电功率预测。仿真实验结果表明,所提出的短期风电功率预测方法是可行的,预测精度可提高至87.8%,与其他风电功率预测模型相比,具有更高的预测精度和更好的适应性。
关键词: 短期风电功率预测; 预测模型; NARX神经网络; 风速融合; 数据融合; 数据处理
中图分类号: TN99?34; TP391 文献标识码: A 文章编号: 1004?373X(2020)09?0166?04
Short?term wind power prediction based on wind speed fusion
and NARX neural network
XU Zunyi, WANG Junxue
(Shandong Architecture University, Jinan 250101, China)
Abstract: High accuracy short?term wind power prediction is the key factor to ensure the daily dispatch and operation safety of power grid. At present, the accuracy of short?term wind power prediction in China is generally lower than that in foreign countries. In order to improve the short?term power prediction accuracy of fan, a prediction model based on wind speed fusion and NARX neural network is proposed in this paper. In the method, fusion of the forecasted wind speed is provided for different data sources at the same location, and NARX neural network is used for short?term wind power prediction. The simulation experiment results show that the short?term wind power prediction method proposed in this paper is feasible, and its prediction accuracy can be increased to 87.8%. In comparison with other wind power forecasting models, it has higher prediction accuracy and better adaptability.
Keywords: short?term wind power prediction; prediction model; NARX neural network; wind speed fusion; data fusion; data processing
0 引 言
随着环境污染、能源紧缺以及电能需求量大等问题的日益突出,风能作为一种具有清洁、储量丰富以及可再生等特点的新能源具有重要的发展前景[1]。由于风的不可控性、间歇性以及随机性等特点,以及国内短期风电功率预测精度普遍较低的现状[2],研究更高精确度的短期风电功率预测算法对保证风电场的日常调度、电网的平稳运行和提高国内风电利用率具有重要现实意义。短期风电功率预测方法主要有统计方法和物理方法,其中,统计方法中以建立数理统计模型的方法更为普遍。主要的统计方法有:时间序列、支持向量机、灰色关联分析、小波分析和人工神经网络等模型[3?7],预测精度一般在82%~84%,国外研究机构预测精度已达到85%以上[8]。……
