基于小波分解技术和随机森林算法的负荷概率预测

Probabilistic Load Forecasting Based on Wavelet Decomposition Technology and Random Forest Algorithm

当前负荷概率预测受到越来越多研究人员的关注,其中多阶段预测系统已经证明了其在提高负荷概率预测整体性能方面的有效性。在使用分位数回归森林和随机森林建立概率预测之前,采用4种基于小波分解的方法预处理负荷时间序列,通过不同的模型对变换得到的负荷分量进行预测,以提高预测精度并减少计算工作量。以2014年全球能源预测竞赛期间公布的实际数据为例进行了数值仿真分析,并与多个基准进行了比较,证明了所提方法的有效性。

Currently,more and more researchers pay attention to probabilistic load prediction forecasting,and the multistage prediction system has proven its effectiveness in improving the overall performance of load prediction forecasting.In this paper,before the use of the quantile regression forest and random forest to establish probabilistic forcasting,four wavelet-based decomposition methods are used to preprocess the load time series,and the load components obtained by these transformations are predicted through different models to improve accuracy and reduce calculation workload.Finally,a numerical application was proposed based on actual data released during the 2014 Global Energy Forecasting Contest,and its effectiveness was demonstrated by comparing with multiple benchmarks.