•  
  •  
 

Keywords

Battery-supercapacitor integration, energy management, PI controller comparison, neural network controller, renewable energy sources

Document Type

Research Article

Abstract

Standalone photovoltaic (PV) systems are essential for off-grid applications but suffer from solar intermittency, resulting in an unreliable power supply. Batteries are commonly integrated into PV systems to provide energy storage; however, they are less effective in handling rapid power fluctuations. To overcome this limitation, hybrid energy storage systems (HESS) combining batteries and supercapacitors are employed to utilise their complementary characteristics. This study proposes a neural network-based energy management system (EMS) for standalone PV systems using a hybrid battery-supercapacitor HESS. The objectives are to develop a neural network-based controller for optimal power sharing between battery and supercapacitor, and to evaluate its performance through a comparative simulation study against a conventional proportional–integral (PI) controller. An existing Simulink/MATLAB model was utilised, with a conventional PI controller initially implemented as a performance benchmark. A feedforward neural network controller was then trained using datasets generated from a custom Simulink model using PV power, load power, load voltage, and load resistance as inputs and optimised switching duty cycle signals for the battery and supercapacitor as outputs. The model achieved a high regression accuracy (R = 0.99948), indicating successful learning of the input–output relationship. Results showed that the neural network controller significantly improved system performance and voltage stability, reducing overshoot by 73% and steady-state ripple by 95%. In terms of power-sharing coordination, battery power ripple was reduced by 82%, while supercapacitor ripple was nearly eliminated, showing a 99.9% reduction along with a 25% improvement in settling time. These results demonstrate that the neural network controller offers more effective energy management compared to the conventional PI controller.

html

Publication Date

31-3-2026

First Page

27

Last Page

34

Download

Share

COinS