Data Science-Based Maximum Power Point Tracking Control for DC-DC Power Converters

Abstract

Recent trends in renewable power conversion practices highlight maximum power point tracking (MPPT) for photovoltaic (PV)-fed Direct Current-Direct Current (DC-DC) converters, considering that variations in irradiance, temperature, and load fluctuation as well as partial shading generate diminution in energy extraction and converter stability. This work presents a boost converter with an MPPT controller developed using a DS-based approach, in which PV voltage, PV current, power, irradiance, temperature and the duty ratio inputs on day duty variations and weekly power variations are used as impact variables. The goal is to achieve short-term predictions of the duty cycle, eliminate oscillation in dynamic and static states, and successfully locate the global maximum power point (MPP) of the current direction connected with stable DC link voltage. As mentioned previously, the research gap is that conventional fixed-step and rule-based MPPT methods have a limited ability to track the global MPP under the changing circumstances of multi-peak PV characteristics. The issues of slow convergence, being trapped in local peaks, converter loss, voltage ripple, and reduced accuracy due to partial shading are significant problems. These are several techniques that have been adopted, such as data science (DS)-based prediction, pulse width modulation (PWM) gate control-based approach, boost converter topology, PV singlediode modeling, converter dynamic modelling and multiobjective loss minimization. Based on the simulation results, the proposed model using intelligent control algorithms achieves a tracking efficiency of 98.4%, a tracking accuracy of 97.1%, a convergence time with oscillation of 0.31 seconds, a DC-link stability of 5.8 W, and a partial shading ability of 94.8%. Applications such as rooftop PV systems, EV charging stations, battery-supported converters and DC microgrids.