EMD-ISMO算法在电力负荷预测中的应用  

Load Forecasting Based on EMD-ISMO Algorithm

作　　者：翟永杰[1] 周倩[2] 韩璞[1] 

机构地区：[1]华北电力大学自动化系,保定071003 [2]华北电力科学研究院有限责任公司,北京100045

出　　处：《系统仿真学报》2010年第12期2858-2861,共4页Journal of System Simulation

基　　金：华北电力大学"校内科研基金"(200814003;200712015)

摘　　要：电力负荷是受多种因素影响的复杂非线性系统,具有明显的周期波动性和趋势性。利用集平稳化和层次化处理能力于一体的经验模态分解(EMD)方法处理非线性非平稳信号的有效性,对电力负荷数据进行平稳化处理,分离出12组IMF数据,包含若干个不同频率的平稳分量,能更明显地看出原负荷序列的周期项、随机项和趋势项;结合对负荷数据具有很好预测能力的改进SMO算法(ISMO),对IMF数据进行分别预测和组合预测,提出了一种EMD-ISMO算法。实验结果表明,该方法无论在预测精度还是收敛速度上都比单纯的SMO算法有了很大改进,取得了很好的预测效果。Electrical load is a complex nonlinear system which is affected by many factors.It has obvious volatile,cyclical,and tendency.Empirical mode decomposition （EMD） algorithm has smoothing and hierarchical processing ability.It can process the nonlinear and non-stationary digtal signal effectively.EMD algorithm was used to process the electrical load data.12 groups IMF data were decomposed,including a number of smooth components with different frequency.The cyclical term,random term and tendency term could be observed clearly.EMD-ISMO algorithm was proposed combining with the improved SMO algorithm（ISMO） which had perfect forecasting ability.Forecasting model was established with IMF data to forecast separately and jointly.The experimental results show that EMD-SMO algorithm can greatly improve the forcast accuracy and computation speed.It achieve very good forcast results.

关 键 词：经验模态分解 支持向量机 序列最小优化 负荷预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]