基于VMD与TCN的台区短期负荷预测算法研究
Research on Short-term Load Forecasting Algorithm Based on VMD and TCN
王清 1陈祉如 1李贵民 2荆臻 1张志 1王平欣 1崔琦3
作者信息
- 1. 国网山东省电力公司营销服务中心(计量中心),济南 250000
- 2. 国网山东省电力公司,济南 250000
- 3. 东北农业大学 水利与土木工程学院,哈尔滨 150006
- 折叠
摘要
针对台区短期负荷预测精度不高的问题,提出了一种基于变分模态分解(variational mode decomposi-tion,VMD)的时间卷积网络(temporal convolutional network,TCN)短期负荷预测算法.其利用VMD对负荷数据进行分解,得到规律性更强的子序列,并采用最大信息系数(maximal information coefficient,MIC)选出与负荷相关性强的天气因素,与历史负荷和分解的子序列形成新的负荷数据集,采用TCN模型完成低压台区短期负荷预测.并对TCN、LSTM、GRU预测算法进行对比分析.仿真结果表明,VMD-TCN 的预测效果最好,MAPE 和 RMSE 分别为1.65%,15.05kW,表明了采用该算法可以实现对台区负荷进行精准的短期预测,以便于台区的调度管理、优化运行以及节能减排,同时采用了另一种数据集对算法进行了验证,结果表明VMD-TCN的预测结果仍是最好的.
Abstract
Aiming at the low accuracy of short-term load forecasting in substation area,a temporal convolutional network short-term load forecasting algorithm based on variational mode decomposition is proposed in this paper.It uses VMD to decompose load data to get a more regular subsequence,and uses the maximum information coefficient to select weather factors strongly correlated with load,and forms a new load data set with the historical load and the subsequence of decomposition,using TCN model to complete short-term load forecasting in low-voltage substation areas.The prediction algorithms of TCN,LSTM and GRU are compared and analyzed.The simulation results show that the forecasting effect of VMD-TCN is the best,MAPE and RMSE are 1.65%and 15.05kW,respectively,indicating that the algorithm can be used to achieve accurate short-term forecasting of the station load,so as to facilitate the dispatch management,optimization operation,energy saving and emission reduction of the station.At the same time,another dataset was used to validate the algorithm,and the results showed that the forecasting results of VMD-TCN were still the best.
关键词
短期负荷预测/变分模态分解/时间卷积网络/最大信息系数Key words
short-term load forecasting/variational mode decomposition/temporal convolutional network/maximum information coefficient引用本文复制引用
基金项目
国家重点研发计划项目(2021YFB4001700)
国网公司科技项目(5700-202255222A-1-1-ZN)
出版年
2024
NETL
NSTL
万方数据