Short-Term Load Forecasting with Multi-Source Data Using Gated Recurrent Unit Neural Networks

• DOI: 10.3390/en11051138
• Publication Date: 2018-05-04T07:08:21Z
• Authors: Yixing Wang; Meiqin Liu; Zhejing Bao; Senlin Zhang
• Keywords: Electrical load

Short-term load forecasting is an important task for the planning and reliable operation of power grids. High-accuracy individual customers helps to make arrangements generation reduce electricity costs. Artificial intelligent methods have been applied short-term in past research, but most did not consider use characteristics, efficiency, more influential factors. In this paper, a method with multi-source data using gated recurrent unit neural networks proposed. The are preprocessed by clustering interference characteristics. environmental factors including date, weather temperature quantified extend input whole network so that information considered. Gated used extracting temporal features simpler architecture less convergence time hidden layers. detailed results real-world experiments shown curve mean absolute percentage error prove availability superiority proposed compared current methods.