Itineraries for charging and discharging a BESS using energy predictions based on a CNN-LSTM neural network model in BCS, Mexico

Abstract

Renewable energy generation (REG) is irrupting throughout the globe and it points to be the path for a sustainable energy future. Nevertheless, due to their volatile nature, they present a challenge for integrating these intermittent sources into the grid. In this work we present itineraries for charging and discharging two ideal Battery Energy Storage Systems (BESS), one powered with a solar PV generation system and the other one powered with wind energy. Using predictions for REG and electric demand (ED), based on a hybrid Convolutional Long-Short Time Memory (CNN-LSTM) neural network, we propose accurate itineraries to know when to charge and when to discharge variable REG, in the area of Baja California Sur (BCS) in Mexico, pursuing to reduce the ED in peak hours. The convolution net will extract local features and the LSTM the temporal ones. The proposed itineraries of charge and discharge based on predictions with the hybrid CNN-LSTM model, are compared with itineraries made with a well known benchmark and itineraries based on true observations points of REG. The results show that the integration of two BESS with charging and discharging itineraries based con a CNN-LSTM model, can effectively mitigate two important peaks of the electric demand profile in the studied location.