Community energy storage system: Deep learning based optimal energy management solution for residential community

The concept of community energy storage system (CESS) is required for the efficient and reliable utilization of renewable energy and flexible energy sharing among consumers. This paper proposes a novel approach to assess the practical benefits of CESS deployment in a residential community by decreasing the daily electricity cost and maximizing the self-consumption of PV energy. To this end, a deep-learning-based forecasting model, namely a bi-directional long short-term memory model, is implemented to predict the operational constraints and dependency. Furthermore, a hybrid optimization technique that comprises a clustering and optimization algorithm is developed in which the clustering algorithm ensures appropriate combinations of user groups to develop optimal control policies. Finally, the forecasting model is integrated with the hybrid optimization algorithm to find the optimal solution involving PV-CESS energy utilization. Numerical analyses are performed using real historic data of the energy demand and PV generation for three consecutive days considering different scenarios. The results demonstrate that the electricity costs and self-consumption associated with the CESS are lower and greater than those of an individual ESS system, respectively, with the daily electricity cost decreasing by 21.89%, 13.81%, and 7.66% in the three analyzed scenarios. •Deep Learning based day-ahead energy generation and consumption prediction modeling.•Integration of optimization and clustering algorithm with Bi-LSTM model.•Daily electricity cost minimization through optimal control policies.•Economic feasibility of CESS integration in the residential community.•The proposed approach for CESS outperformed individual ESS.