Bayesian inference based MPPT for dynamic irradiance conditions

摘要

We introduce a Bayesian inference based maximum power point tracker (BI-MPPT) and demonstrate it under static and dynamic irradiance conditions. BI-MPPT is based on a probability inference technique which uses the model of the photovoltaic (PV) module and accounts for noise in the system. Owing to the model-based approach, the tracker converges fast to the maximum power point of the PV module and is capable of tracking dynamically varying irradiance. We compare the proposed BI-MPPT to an optimized model-based P&O tracker and show that the proposed tracker consistently outperforms the latter. In experiments conducted with an in- house built solar emulator, at low illumination, the BI-MPPT achieves a static efficiency of 99.9% and a dynamic efficiency of at the least 97.4% and outperforms the optimized P&O tracker by about 10 percent point. When BI-MPPT is applied to I-V curve measurements of an outdoor PV installation, at moderate and high irradiances, the dynamic efficiency is between 98.92% and 99.61% yielding a 3 to 4 percent point improvement over the optimized P&O.