Deep learning based digital twins augmented reality: Model predictive control for battery and storage optimization in renewable energy prosumers districts

Abstract

In modern energy systems, integrating renewable energy sources has accelerated the development of prosumer networks, where entities produce and consume energy. The inherent variability of renewable generation challenges maintaining a balance between supply and demand. Optimizing battery and storage systems is required to overcome this volatile behavior of renewable generation to ensure energy and cost efficiency. This paper proposes a Deep Learning-Driven Model Predictive Control (DL-MPC) framework that uses Digital Twin (DT) and Augmented Reality (AR) technologies to optimize energy storage in prosumer districts. The framework utilises the DL model for energy and consumption forecasting for MPC to adjust energy storage and distribution in real-time dynamically. The DT technology emulates the real-time prosumer district to generate feedback for continuous improvement of the control decisions. Moreover, the AR interface provides an intuitive view of the real-time energy flow, storage levels, and usage statistics to improve decision-making and inform immediate adjustments. The proposed DL-MPC framework is simulated using real-world data. The framework improves 15 % energy efficiency and reduces 20 % operational cost in comparison to traditional methods. This framework presents a robust, scalable, adaptive tool to optimize the energy flows in modern power grids. • Integrating renewables accelerates prosumer networks, enabling energy production and consumption. • Effective optimization of batteries ensures balance amid renewable energy variability. • The DL-MPC framework leverages Digital Twin and AR for real-time energy optimization. • Real-world simulations show 15 % energy efficiency gain and 20 % cost reduction. • A scalable, adaptive tool, DL-MPC optimizes energy flow in modern power grids.

Topics & Concepts