基于优化稀疏编码的超短期负荷滚动多步预测  

Multi-Step Rolling Ultra-Short-Term Load Forecasting Based on the Optimized Sparse Coding

作　　者：储晨阳 秦川[1] 鞠平[1] 赵静波 赵健 Chu Chenyang;Qin Chuan;Ju Ping;Zhao Jingbo;Zhao Jian(Energy and Electrical Engineering College Hohai University,Nanjing 211100 China;Electric Power Research Institute of State Grid Jiangsu Electric Power Co.Ltd,Nanjing 210008 China;State Grid Nanjing Power Supply Company,Nanjing 210013 China)

机构地区：[1]河海大学能源与电气学院,南京211100 [2]国网江苏省电力有限公司电力科学研究院,南京210008 [3]国网南京供电公司,南京210013

出　　处：《电工技术学报》2021年第19期4050-4059,共10页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(51837004);111引智计划(B14022)资助项目。

摘　　要：小时级的超短期负荷预测是调度部门制定日内滚动计划的基础。该文提出了一种基于优化稀疏编码的超短期负荷多步预测算法,对未来4h的负荷进行滚动预测。首先,将历史负荷功率序列组成具有时延的特征字典和目标字典,并利用稀疏编码算法建立多步预测模型;然后,基于负荷曲线的相似性特性,借助实时的负荷功率特征向量与特征字典中原子的拓展符号化距离对特征字典原子进行筛选,提高了稀疏编码的预测精度;最后,对预测结果进行误差分析。针对负荷爬坡时段误差偏大的问题,通过基于负荷增量预测值的修正模型进行优化,进一步提高了预测精度。采用实际负荷数据进行分析,验证了所提方法的有效性。Ultra-short-term load forecasting is the basis of intra-day rolling scheduling for the dispatching departments.An optimized sparse coding based multi-step load forecasting method is proposed to make rolling forecast of the load power in the next 4 hours.Firstly,the historical load power time series data are used to create the predictor/response dictionaries pair with time-lag,then the multi�step load forecasting model can be built via sparse coding.Secondly,considering the similarity of the load power time series data,the atoms of the dictionaries pair are filtered according to the extended symbolic aggregate approximation distance between the real-time load power data and the vectors of the dictionaries,which improves the load forecasting accuracy.Finally,error analysis is performed.It is found that the forecast errors during the period of load ramping up are always larger than that of the other periods.Therefore,the error correction model based on the load increment forecasting is proposed to further improve the prediction accuracy.The effectiveness of the proposed method is verified by the case of real-world load power dataset.

关 键 词：超短期负荷预测 滚动预测 优化稀疏编码 拓展符号化距离 负荷增量预测 

分 类 号：TM715[电气工程—电力系统及自动化]