基于PSO-BSNN的短期风速预测

发布时间：2018-03-11 14:08

  本文选题：PSO　切入点：BSNN　出处：《电力系统保护与控制》2015年15期 　论文类型：期刊论文

【摘要】：考虑到风的随机性和波动性,提出一种基于粒子群(PSO)优化B样条神经网络(BSNN)的短期风速预测方法。利用相空间重构方法确定BSNN的输入空间向量,BSNN可以灵活地改变对输入空间的划分和对隐层基函数的定义,对任意的网络输入,隐层基函数的输出只有少数非零,使网络输出简单,收敛速度快。但在传统的BSNN中,对输入空间节点位置的均匀划分是粗糙的,预测结果容易陷入局部极小而影响预测精度。粒子群优化算法是一种智能搜索方法,它具有较强的搜索能力并且容易实现,利用PSO优化BSNN输入空间的节点位置划分,可避免BSNN陷入局部极小并提高预测精度。仿真结果表明,基于PSO-BSNN的预测模型比传统的BSNN和BPNN预测模型具有更高的预测精度。
[Abstract]:Considering the randomness and volatility of the wind, A method of short-term wind speed prediction based on particle swarm optimization B-spline neural network (BSNN) is proposed. The method of phase space reconstruction is used to determine the input space vector of BSNN, which can flexibly change the partition of input space and the definition of hidden layer basis function. For arbitrary network input, the output of hidden layer basis function is only a few non-zero, which makes the network output simple and convergent speed. But in traditional BSNN, the uniform partition of node position in input space is rough. Particle Swarm Optimization (PSO) is an intelligent search method, which has strong searching ability and is easy to realize. PSO is used to optimize the node location partition of BSNN input space. The simulation results show that the prediction model based on PSO-BSNN has higher prediction accuracy than the traditional BSNN and BPNN models.
【作者单位】： 燕山大学电气工程学院;
【基金】：河北省自然科学基金项目(F2012203088)~~
【分类号】：TP18;TM614
，

本文编号：1598497