Dynamic opposition learning-based rank-driven teaching learning optimizer for parameter extraction of photovoltaic models
Xuming Wang, Wen Zhang
Abstract
Due to energy depletion, Photovoltaic (PV) power has received much attention as the leading renewable solar power generation mode. How to accurately recognize PV system parameters to improve generation efficiency has become a hotspot in the field of research. A Dynamic oppositional learning strategy and Sorting Teaching-Learning-Based Optimization (DSTLBO), is proposed for accurate identification of PV models, in which TLBO is improved in the following three aspects: i) the dynamic oppositional population is generated by dynamic oppositional learning strategy in the initialization phase; ii) instead of simply choosing a learning style based on the average score, the population is divided into optimal individuals, efficient learners, and inefficient learners according to a sorting mechanism, with each group having a different learning style; iii) three individuals are randomly selected for information exchange in learner phase to strengthen the exploration ability of algorithm and avoid falling into local optimal. The performance of DSTLBO was tested against the serial benchmark function and three types of PV models, respectively. Test results indicate that DSTLBO has strong optimization capability and can efficiently extract the PV model parameters compared to other algorithms.