Machine Learning-Driven Digital Twin for Real-Time Optimization of Hybrid Energy Storage Systems

Abstract

The growing adoption of renewable energy sources and the electric mobility has heightened the requirement of intelligent energy storage control that can act under dynamic and uncertain conditions. In this paper, we suggest a machine-learning-based algorithmic digital twin of real-time optimization of hybrid energy storage systems (HESS) consisting of batteries and supercapacitors. The created digital twin creates a high-fidelity cyber-physical model of the physical HESS by keeping system conditions in parity with real-time measurements of voltage, current, power, and state-of-charge variables. The digital twin uses machine learning models to forecast the demand of the load, availability of renewable power, and the dynamics of energy storage systems to allow it to anticipate and react to changes in energy management. The proposed framework is the best method to divide power between the battery and the supercapacitor because it separates low-frequency and high-frequency power elements and, therefore, allows decreasing the current load in the battery, which will increase the efficiency of the entire system. The machine learning predictions are combined with control constraints via a real-time optimization layer, and the resulting reference signals are optimum ones to the power electronic converters. The simulation results at different operating conditions such as the rapid loading transient and renewable power variations have shown that the proposed digital twin-based solution is considerably better than the existing rule-based energy management solutions in terms of power tracking performance, battery current ripple, and battery life. The findings prove the usefulness of the proposed framework as a scalable and smart proposal to next generation hybrid energy storage systems in renewable energy and electric mobility applications.