摘要
针对BP神经网络容易陷入过拟合和局部极小值的缺陷,采用殖民竞争全局优化算法,将BP神经网络的权值和阈值作为变量,并将均方差作为目标函数,组成了一种新的ICA-BP神经网络算法.结合风电厂的实际数据在Matlab平台上对该方法进行了验证,并与粒子群算法、遗传算法进行比较,得出该算法可以提高风电功率预测精度的结论.
In view of the fact that BP algorithms are fast but they tend to be trapped in local minimums, ICA is employed as a global optimum search algorithm to overcome BP neural network adversities, ANN connection weights are formed as variables of ICA and the Mean Square Error is used as a cost function in ICA, composing the new ICA-BP algorithm. Combined with the actual data of wind power plants on the MATLAB platform to validate the method, and a conclusion is made that this algorithm can improve the precision of wind power forecasting.
出处
《上海电力学院学报》
CAS
2014年第3期203-207,222,共6页
Journal of Shanghai University of Electric Power
基金
国家自然科学基金(60801056)
上海市青年科技启明星计划基金(11QA1402800)
上海教育委员会科研创新重点项目(11ZZ170)
关键词
殖民竞争算法
权值阈值优化
BP神经网络
风电功率预测
imperialist competition algorithm
weight value optimization
BP nerve network
wind power forecast
