基于ARIMA-SVM模型的微电网短期负荷组合预测研究  

Study on Microgrid Short-term Load Combination Forecasting Based on ARIMA-SVM Model

作　　者：王友春 文闪闪 秦跃进 范黎 杨再鹤[2] 郑丹[2] 

机构地区：[1]湖北省电力勘测设计院,湖北武汉430024 [2]武汉大学电气工程学院,湖北武汉430072

出　　处：《陕西电力》2014年第3期19-23,共5页Shanxi Electric Power

基　　金：国家科技支撑计划项目资助(2013BAA02B00)

摘　　要：在对比分析微电网负荷特性与传统电力负荷特性的基础上,针对微电网负荷不确定性和波动性强的特点,提出了一种基于预测误差指标的最优组合预测方法进行微电网短期负荷预测。方法中结合了自回归积分移动平均法和支持向量机模型的优点,以两种预测方法误差绝对值和最小为目标,分析确定预测方法在组合模型中的权重,进而得到组合预测中的最优权重组合。研究表明,与方差倒数法组合预测模型所得结果相比,该组合预测方法具有更高的预测精度,能满足实际要求。The load of microgrid has more uncertainty and high frequency changes than the conventional power system＇s loads.According to the volatility features of microgrid load,the combination forecasting model based on Auto-regressive Integrated Moving Average Model（ARIMA） and Support Vector Machines（SVM） is proposed,in which the weight is solved by reaching the sum of absolute value of forecasting error to each individual method minimum.And then the optimal weights are drawn by the proposed method.A load forecast empirical example has shown that compared with the combination model based on the variance reciprocal weighting method,the proposed method can achieve higher prediction accuracy,which is more suitable for the demand of microgrid.

关 键 词：微电网 自回归积分移动平均模型 支持向量机 组合预测 

分 类 号：TM715[电气工程—电力系统及自动化]