Enhancing short-term probabilistic residential load forecasting with quantile long–short-term memory

被引：31

作者：

Gan, Dahua ^{[1
]}

Wang, Yi ^{[1
]}

Zhang, Ning ^{[1
]}

Zhu, Wenjun ^{[2
]}

机构：

[1] Department of Electrical Engineering, Tsinghua University, Beijing, China

[2] Electric Power Research Institute, Guangdong Power Grid Corporation, China Southern Power Grid, Guangzhou, China

来源：

Journal of Engineering | 2017年 / 2017卷 / 14期

关键词：

Deep learning - Electric power plant loads - Stochastic systems - Housing;

D O I：

10.1049/joe.2017.0833

中图分类号：

学科分类号：

摘要：

In the study of load forecasting, short-term (ST) load forecasting in the horizon of individuals is prone to manifest non-stationary and stochastic features compared with predicting the aggregated loads. Hence, better methodologies should be proposed to forecast ST residential loads more accurately, and refined representation of forecasting results should be reconsidered to make the prediction more reliable. A format of ST probabilistic forecasting results in terms of quantiles is offered, which can better describe the uncertainty of residential loads, and a deep-learning-based method, quantile long–ST memory, to implement probabilistic residential load forecasting. Experiments are conducted on an open dataset. Results show that the proposed method overrides traditional methods significantly in terms of average quantile score. © 2017 Institution of Engineering and Technology. All Rights Reserved.

引用

页码：2622 / 2627