结合支持向量机和马尔可夫链算法的中长期电力负荷预测模型.PDFVIP

2012 年 第 6 卷 第 1 期 南 方 电 网 技 术 研究与分析 2012 ，Vol. 6 ，No. 1 SOUTHERN POWER SYSTEM TECHNOLOGY Study Analysis 文章编号：1674-0629 (2012)01-0054-04 中图分类号：TM744 文献标志码：A 结合支持向量机和马尔可夫链算法的 中长期电力负荷预测模型 1 2 陈剑勇 ，苏浩益 （1. 南宁供电局，南宁 530031 ；2. 湘潭电业局，湖南 湘潭 411104 ） 摘要：中长期电力负荷预测具有可利用的历史数据较少和受外界不确定性因素影响较大的特点，传统的单一预测模型很 难满足生产实际的需要。在简要分析了支持向量机和马尔可夫链各自优势的基础上，提出了一种基于支持向量机和马尔 可夫链的组合负荷预测模型。通过经改进的粒子群算法优化的支持向量机对历史负荷序列进行粗预测，接着借助马尔可 夫链确定负荷序列的状态转移概率矩阵，通过划分系统状态以及分析实际值与支持向量机拟合值的相对误差，得到最终 的预测结果。实际算例验证了该模型的有效性和优越性。 关键词：支持向量机；马尔可夫链；负荷预测；粒子群优化；组合模型 A Forecasting Model of Medium / Long Term Power Load in Combination of the Support Vector Machine and Markov Chain Algorithms 1 2 CHEN Jianyong , SU Haoyi (1. Nanning Power Supply Bureau, Nanning 530031, China; 2. Xiangtan Power Supply Bureau, Xiangtan, Hunan 411104, China) Abstract: In processing medium and long term power load forecasting, with less available historical data, and many uncertainties factors are influencing the results, thus the traditional single forecast model is difficult to meet actual production needs. On the basis of brief analysis of the advantages of Support Vector Machines and Markov Chain model, a new combination prediction model is put forward based on Support Vector Machines and Markov Chain theory. Support Vector Machines optimized by improved particle swarm algorithm was used to forecast the sequence of historical load; The state divert probability matrix of the load time series is gotten by Markov chain; The final results is determined by division of system state and analysis of relative error va