Neural network maximum power point tracking for performance improvement of solar PV water pumping system

Abstract

A neural network maximum power point tracking (MPPT) controller has been proposed in order to improve the performance of a photovoltaic water pumping system. The proposed neural network MPPT controller has been trained using inputs and output data collected using the conventional P&O algorithm. The efficiency of the proposed algorithm has been studied successfully using a DC motor-pump powered by 36 PV modules via a DC-DC boost converter controller using the proposed neural network MPPT algorithm. Comparative study results between the proposed neural network MPPT controller and P&O MPPT controllers using fixed and variable step size versions as well as experimental study results using the STM32F4 board in the hardware in the loop mode prove the efficiency of the proposed controller regarding all considered performances metrics, reducing as a consequence the response time and eliminating the steady-state oscillation leading by the way to an improvement of the whole system performances.