Enhanced Bayesian Based MPPT Controller for PV Systems

Abstract

We tackle the problem of a photovoltaic (PV) controller for maximum power point tracking (MPPT) under varying insolation and shading conditions. A general-purpose adaptive maximum power controller is tailored to maintain operation of the PV system at the maximum power point while constantly avoiding local maxima for changing environmental conditions. While a variety of conventional MPPT algorithms have been designed for ideal operating situations, very few were able to deliver true maximum power under abrupt changes in sun shading. Under these dynamic changes, most MPPT techniques fail to rapidly locate the global maximum power point and are stuck at global maxima, leading therefore to inconsistent power generation and low system efficiency. In this paper, we apply Bayesian fusion, a machine learning technique otherwise used for unsupervised classification, curve detection, and image segmentation, in order to achieve global MPPT in record time. Simulation results validated with real-life experimental studies demonstrated the ameliorations of the proposed technique compared to state-of-the-art methods. Using this algorithm, the total output power of the solar system is maximized while minimizing the steady-state oscillations and the tracking time.