Convolutional residual network to short-term load forecasting
- PDF / 3,087,404 Bytes
- 15 Pages / 595.224 x 790.955 pts Page_size
- 0 Downloads / 225 Views
Convolutional residual network to short-term load forecasting Ziyu Sheng1 · Huiwei Wang1
· Guo Chen2 · Bo Zhou3 · Jian Sun1
Accepted: 5 September 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Since their inception, convolutional neural networks (CNNs) have been shown to have powerful feature extraction and learning capabilities, and the creation of deep residual networks (DRNs) was a milestone in the development of CNNs. However, residual networks mostly use convolution structures, which are widely applied to image recognition and classification problems. Therefore, when facing a load forecasting problem that involves nonlinear regression, will a DRN using a convolution structure still achieve great results? To answer this question, we present a network based on a DRN with a convolution structure to carry out short-term load forecasting, and we mainly focus on the effects of DRNs with different depths, widths and block structures for dealing with nonlinear regression problems. Through multiple sets of controlled experiments, we obtain the best network architecture and the corresponding hyperparameters for short-term load forecasting. The experimental results demonstrate that the model has higher prediction accuracy than existing models, and the DRN with a convolution structure can handle load forecasting while still achieving state-of-the-art results. Keywords Short-term load forecasting · Deep residual network · Convolutional neural network
1 Introduction Due to the massive penetration and utilization of renewable energy, the rapid spread of electric vehicles, and the use of smart home devices, modern power systems are becoming more complex and unpredictable. The application of various new energy sources and technologies has brought challenges to the power system as well as opportunities for development. With the development of modern energy systems especially smart grids, accurate predictions of energy loads are becoming unprecedentedly important. Accurate load forecasting can make future energy systems smarter and more powerful to cope with incoming challenges and opportunities, and the rapid development of artificial Huiwei Wang
[email protected] 1
College of Electronics and Information Engineering, Southwest University, Chongqing, 400715, People’s Republic of China
2
School of Automation, Central South University, Changsha, 410083, People’s Republic of China
3
College of Mathematics and Statistics, Chongqing Jiaotong University, Chongqing, 400074, People’s Republic of China
intelligence makes this kind of load forecasting with high precision and high reliability possible. Load forecasting can be classified into very-short-term load forecasting (VSTLF) [1], short-term load forecasting (STLF) [2], medium-term load forecasting (MTLF) [3], and long-term load forecasting (LTLF) [4] according to the time periods. VSTLF can perform load forecasting for up to one day. STLF can provide load forecasting within one to two weeks. MTLF can perform load forecasting from two
Data Loading...
