摘要
将适用于解决大规模样本训练问题的核心向量回归(corevector regression,CVR)方法引入到电力负荷预测中,并采用粒子群优化(particle swarm optimization,PSO)方法对CVR的模型参数进行寻优,从而提出了一种基于PSO-CVR的负荷预测新模型。构造大规模负荷训练样本,研究对样本负荷产生影响的因素,从而确定样本集的构造。通过用PSO对CVR的模型参数进行优化,得到优化后的CVR预测模型,循环构造预测样本并进行连续预测。算例分析结果表明,在相同时耗下,所提出的优化CVR预测模型能够通过训练更大规模的样本得到比支持向量回归(support vector regression,SVR)方法更高的预测精度。
This paper presents a new electric load forecasting algorithm based on the core vector regression (CVR) for large scale data set. Particle swarm optimization (PSO) is applied for determining the parameters of CVR, so we get a novel scheme named PSO-CVR. The load data is analyzed for finding factors that may have great influence on load forecasting. We create several training sets in diffident size and then use CVR to train them for observing whether more accurate results could be made on a larger training data set. The proposed PSO-CVR model is verY efficient to predict one week loads. Experimental results show that the PSO-CVR model has much faster training speed and produces fewer support vectors on very large data sets compared with support vector regression (SVR).
出处
《中国电机工程学报》
EI
CSCD
北大核心
2010年第28期33-38,共6页
Proceedings of the CSEE
基金
河北省自然科学基金项目(F2007001042)~~
关键词
负荷预测
大规模样本
核心向量回归
粒子群优化
load forecasting
large scale data set
corevector regression (CVR)
particle swarm optimization (PSO)