AI-based consumption forecast to reduce energy costs for the operation of charging infrastructure in retail

Abstract

The buildup of the charging infrastructure in retail significantly changes the load profiles of these energy consumers resulting in higher costs due to power peaks. This paper proposes a new approach for energy management at supermarkets where the cooling processes are used as flexibility. The approach makes use of the time gaps between charging processes to selectively intensify the cooling processes. This energy reserve is used when new charging processes begin. Key capability is a forecast module based on deep learning. The proposed CNN-LSTM model with additional input signals for seasonality and public holidays shows good performance for a short-term prediction over two hours.