摘要
目前使用比较普遍的优化方法对BP算法改进之后,改进的BP神经网络预测过程都存在复杂程度变大、更加消耗人力资源等缺陷。针对这些缺陷,本文提出一种传递函数自我优化算法来改进神经网络。然后将改进的网络应用到风电功率预测中,以东北某风电场一段时间的风电运行数据作为实验样本,分别采用传统BP神经网络和改进的BP神经网络进行预测分析。仿真结果证明,改进之后的BP神经网络不仅有更快的收敛速度,还有更加精确的预测结果,并且不需要认为干预整个预测过程。极大提高了网络的预测能力和效率。
Using common optimization method to improve the BP algorithm will introduce shortcomings such as increased complexity and more human resource consumption in prediction process. To solve the problem, this paper presents a transfer function self-optimization algorithm for neural network improvement. The improved network is further applied to wind power prediction. Taking the operating data of a time period at a northeast wind farm as experimental samples, both traditional and the improved BP neural network are used to analyze the prediction. The results show that the improved BP neural network enhances not only convergence rate but also prediction accuracy.
出处
《电测与仪表》
北大核心
2014年第11期56-59,64,共5页
Electrical Measurement & Instrumentation
基金
国家自然科学基金资助项目(51277023)
关键词
传递函数自我优化
神经网络
风电功率预测
BP算法
transfer function self-optimization, neural network, wind power prediction, BP algorithm
