基于改进双并联过程神经网络的光伏功率预测  

Photovoltaic Power Prediction Based on Improved Dual Parallel Process Neural Network

作　　者：何祖军[1] 周光辉 杨奕飞[1] HE Zujun;ZHOU Guanghui;YANG Yifei(College of Electronics and Information,Jiangsu University of Science and technology,Zhenjiang 212001)

机构地区：[1]江苏科技大学电子信息学院

出　　处：《计算机与数字工程》2019年第8期1890-1894,共5页Computer & Digital Engineering

摘　　要：在光伏发电领域,面对复杂多变的天气,实现光伏发电量的准确预测是一个难点问题。光伏发电量的准确预测能为电网平稳调度提供参考依据,但传统方法进行光伏发电输出功率预测误差大、响应时间长,难以满足电网需求调度的问题。提出一种改进双并联过程神经网络的光伏功率预测方法,对神经元的聚合运算机制和激励方式向时间域进行扩展,利用动态增量来更新双并联过程神经网络的权值,避免因误差变化小而使网络陷入局部极小值。仿真结果表明,使用改进后的模型预测光伏发电功率能得到更高的预测精度。In the field of photovoltaic power generation,the accurate prediction of photovoltaic power generation is a difficult problem in the face of complex and changeable weather.The accurate prediction of photovoltaic power generation can provide a reference for the smooth dispatching of power grid.However,the traditional method of photovoltaic power output has large prediction error and long response time,which is difficult to meet the demand scheduling problem of power grid.A prediction method for photovoltaic power improved double parallel process neural network is put forward,neuron aggregation operation mechanism and incentive mode to the time domain are extended by the dynamic incremental update double parallel process neural network weights,in order to avoid the error due to small changes in the network into local minima.The simulation results show that the predicted PV power can be predicted by the improved model.

关 键 词：调度 过程神经网络 聚合 动态增量 

分 类 号：TM615[电气工程—电力系统及自动化]