基于误差补偿的风速时间序列多步预测  

Multi-Step-Ahead Prediction of Wind Speed Time Series Based on Error Compensation

作　　者：韩璞[1] 王东风[1] 王富强[1] 

机构地区：[1]华北电力大学河北省发电过程仿真与优化控制重点实验室,河北保定071003

出　　处：《计算机仿真》2014年第2期206-209,385,共5页Computer Simulation

基　　金：中央高校基本科研业务费专项资金项目(11MG49)

摘　　要：传统的时间序列多步预测策略主要有:直接策略、滚动策略、直接-滚动策略、多输出策略和直接-多输出策略。由于风速具有很强的随机性和非平稳特征,使得在多步预测过程中,随着预测步数的增加,预测误差越来越大。在研究传统的多步预测策略的基础上,提出了一种误差补偿策略,以提高预测精度。采用10min采样的风速数据,进行了试验验证,实验结果表明,改进方法相比于单独多步预测策略具有更小的预测误差。Traditional multi - step - ahead prediction method for time series are mainly direct strategy, recursive strategy, direct- recursive strategy, multi -output strategy and direct -multi -output strategy. Due to the random- ness and non - stationary characteristics of wind speed, the prediction error grows bigger and bigger with the number of prediction step increases. Based on the study of traditional multi - step - ahead forecasting strategies, a new kind of multi - step - ahead forecasting strategy is put forward to improve the prediction accuracy, which is error compen- sation strategy. The wind speed data with sampling interval 10 minutes is taken to test the forecasting precision for all the multi - step - ahead forecasting methods. The experimental results show that the error compensation strategy has the smaller forecasting error compared to the prediction strategies without error compensation.

关 键 词：多步预测 误差补偿 风速 时间序列 

分 类 号：TK89[动力工程及工程热物理—流体机械及工程]