Deep learning application: Load forecasting in big data of smart grids

Abstract

Load forecasting in smart grids is still exploratory; despite the increase of smart grids technologies and energy conservation research, many challenges remain for accurate load forecasting using big data or large-scale datasets. This chapter addresses the problem of how to improve the forecasting results of loads in smart grids, using deep learning methods that have shown significant progress in various disciplines in recent years. The deep learning methods have the potential ability to extract problem-relevant features and capture complex large-scale data distributions. Existing research in load forecasting tends to focus on finding predicted loads using small historical datasets and the behavior of the load’s consumers in smart grids. Moreover, current research which applies the conventional deep learning methods for load forecasting has shown better performance than conventional load forecasting methods. However, there is little evidence that researchers have addressed the issue of hybridizing different deep learning methods for complex large-scale load forecasting in smart grids, with the intent of building a robust predictive model in smart grids and understanding the relationships that exist between different predictive models and deep learning methods. Consequently, the purpose of this chapter is to provide an overview of how the load forecasting performances using deep learning methods in smart grids can be improved.