Optimizing Solar PV System Performance Using Feedforward Neural Network Control Algorithm

Artificial Neural Networks (ANNs) are important in solar energy applications as a way to improve the efficiency of solar energy systems by overcoming the limitations of traditional Maximum Power Point Tracking (MPPT) techniques. This paper will introduce Feedforward Neural Network (FNN) as the control algorithm to make the solar PV system in a specific residential village at Universiti Teknologi PETRONAS (UTP) more effective. The algorithm will aim at optimizing energy generation and improving system efficiency through a dynamic adjustment of boost converter pulse width modulation (PWM) duty cycle. The MATLAB Simulink software platform is used to develop and analyze a FNN control algorithm. Using MATLAB, a simulation of the photovoltaic (PV) system is created and simulated to supply sufficient power for the residential village at UTP. The result of this study shows that FNN control algorithm is effective in maximizing output power of PV systems by adjusting the PWM signal controlling the boost converter and is comparable to common MPPT methods such as Perturb and Observe (P&O) algorithm. Future work includes fine-tuning the hyperparameters of FNN, incorporation of additional features through feature engineering and the application of regularization techniques.