基于PSO-ICA-BP神经网络的短期风电功率预测  

Short-Term Wind Power Forecast Based on PSO-ICA-BP Neural Network

作　　者：王帅哲 王金梅 王永奇 马文涛 WANG Shuai-zhe;WANG Jin-mei;WANG Yong-qi;MA Wen-tao(School of Physics and Electronic-Electrical Engineering,Ningxia University,Yinchuan 750021,China;Key Laboratory of Ningxia Desert Information Intelligent Perception Autonomous Region,Yinchuan 750021,China)

机构地区：[1]宁夏大学物理与电子电气工程学院,宁夏银川750021 [2]宁夏沙漠信息智能感知自治区重点实验室,宁夏银川750021

出　　处：《电工电气》2019年第2期7-11,共5页Electrotechnics Electric

基　　金：国家自然科学基金项目(51167015)

摘　　要：针对传统的BP神经网络对短期风电功率预测精度不高的缺点,提出粒子群算法改进帝国竞争算法(PSO-ICA),通过PSO算法改进殖民地同化操作提高ICA算法的全局寻优能力,输出全局最优解作为BP神经网络初始权值阈值。同时用主成分分析法降维压缩输入数据,提高网络泛化能力。利用PSO-ICA-BP预测模型对某风电场实际风电功率数据进行预测,仿真结果表明该模型预测误差更小,对短期风电功率预测更有效。In view of the shortcomings of the traditional BP neural network for short-term wind power prediction,the particle swarm optimization algorithm(PSO)was proposed to improve the Empire competition algorithm(PSO-ICA),to improve the diversity of colonial assimilation,and to optimize the initial weight threshold of the BP neural network by the output of the global optimal solution.The principal component analysis method was used to reduce dimension and to compress input data and improved the network generalization ability.The PSO-ICA-BP prediction model was used to predict the actual wind power data of certain wind farm.The simulation results show that the prediction error of this PSO-ICA-BP model is smaller and more effective for the short-term wind power prediction.

关 键 词：帝国竞争算法 粒子群算法 BP神经网络 风电功率预测 

分 类 号：TM614[电气工程—电力系统及自动化] TM715