Artificial neural network-based power smoothing control of lithium-ion battery-supercapacitor hybrid energy storage systems for renewable energy applications

Abstract

Grid stability, power quality, and the lifespan of energy storage devices are all severely hampered by the intrinsic intermittency and quick power fluctuations of Renewable Energy Sources (RES). A battery–supercapacitor Hybrid Energy Storage System (HESS) is an effective solution because it combines the high energy density of batteries with the high-power density and fast dynamic response of supercapacitors. But to get the best power sharing between these storage parts, a smart and adaptable control strategy is required. This paper proposes an improved Artificial Neural Network (ANN) based power smoothing control for a battery-supercapacitor HES integrated with RES. The ANN controller in the suggested method lowers battery stress and makes the system more reliable by smartly sending high-frequency transient components to the supercapacitor and low-frequency power components to the battery. Three bio-inspired optimization techniques, namely, Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), and Ant Colony Optimization (ACO) are also used to improve the performance of the ANN controller by optimizing the network parameters for better control accuracy and dynamic response. We use the MATLAB/Simulink environment to build and test the proposed system with different loads and RES. When compared to non-optimized controllers, optimized ANN controllers greatly reduce power fluctuations, improve overall energy management performance, and make DC bus voltage regulation better. Among the evaluated methods, ACO-ANN achieves the best DC voltage regulation (0.6% overshoot, 0.3 s settling time) and power smoothing (±25 W ripple), while PSO-ANN minimizes battery current stress (4.135 A RMS). This reveals a trade-off: controllers that aggressively smooth power demand higher transient currents from the battery. Therefore, the choice of optimizer depends on the application priority; ACO-ANN for grid stability, PSO-ANN for battery longevity. To strengthen the validation of the proposed controller, quantitative performance of the controllers is evaluated using voltage RMSE, voltage ripple, battery current RMS, power ripple, efficiency, settling time, and composite performance score. The study also includes multiple operating condition analysis, covering load variation, renewable power fluctuation, combined disturbances, parameter variation, and measurement-noise conditions. Furthermore, a statistical validation section based on repeated simulation trials and ANOVA significance testing is added to confirm that the observed performance improvements are statistically supported rather than being based only on individual simulation waveforms. The study also discusses real-time implementation feasibility and identifies the limitations of the present simulation-based validation. The results support the potential of optimization-assisted ANN control for HESS power smoothing, while further experimental and hardware-in-the-loop validation is required before practical deployment.