基于样本扩展和特征标记的节假日短期负荷预测

Holiday Short-term Load Forecasting Based on Sample Expansion and Feature Extraction

针对目前节假日负荷预测中有效样本缺乏的问题,基于休息日与节假日负荷特性的相似性分析,提出一种扩展样本策略,以丰富基础样本数据量;探讨了对负荷样本节假日特征属性的标记方式,并构建了一种有效的相关因素矢量;最后结合支持向量机(support vector machine,SVM)算法,对节假日负荷进行预测,以提高其预测结果的精度。算例结果表明,与传统方法相比,所提方法能够有效提高负荷预测精度,可推广应用于实践中。

Aiming at the problem of lacking of effective samples in current holiday load forecasting, a strategy for sample expansion is proposed based on similarity analysis on load characteristics in weekends and holidays to enrich data size of basic samples. Feature extraction ways of load characteristic attributes in holidays is discussed and effective vectors of relevant factors are constructed. The support vector machine (SVM) algorithm is combined for load forecasting in holidays so as to improve accuracy of forecasting result. The example result indicates that compared with conventional methods, the proposed method can effectively improve load forecasting accuracy and be applied in practice.