基于LSTM和ARIMA的负荷预测对比分析  

Comparative Analysis of Load Forecasting Based on LSTM and ARIMA

作　　者：张瀚文[1] 于长程 李依婷 宁世雄 

机构地区：[1]中国矿业大学(北京)机电与信息工程学院,北京

出　　处：《智能电网（汉斯）》2023年第3期53-62,共10页Smart Grid

摘　　要：针对负荷预测精度有待提升的问题,基于不同原理进行负荷预测研究。首先进行了数据预处理,其次形成了基于LSTM的负荷预测流程,再次形成了基于ARIMA的负荷预测流程,最后进行算例分析,验证了基于LSTM的预测方法在预测该地区综合负荷方面具有更高的准确性。进一步地,分析了两种方法的优缺点,ARIMA模型计算简单,但对于复杂的非线性负荷变化难以进行准确预测;而LSTM模型则能够更好地处理复杂的非线性负荷变化,但对计算资源和训练时间要求较高。To address the issue of improving the accuracy of load forecasting, research on load forecasting based on different principles was conducted. First, data preprocessing was performed. Secondly, a load forecasting process based on LSTM was formed, followed by a load forecasting process based on ARIMA. Finally, a case analysis was conducted, which verified that the LSTM-based prediction method has higher accuracy in predicting the comprehensive load in the region. Furthermore, the advantages and disadvantages of the two methods were analyzed. The ARIMA model has a simple calculation, but it is difficult to accurately predict complex nonlinear load changes. On the other hand, the LSTM model can better handle complex nonlinear load changes, but it requires high computing resources and training time.

关 键 词：负荷预测 LSTM模型 ARIMA模型 预测误差 

分 类 号：TM7[电气工程—电力系统及自动化]