Artificial bee colony optimization–based enhancement of output power generation in grid-connected photovoltaic systems

Abstract

The integration of photovoltaic systems to utility grids becomes a durable alternative for power generation in many countries. Given the recent advances in artificial intelligence, the performance in output power maximization can be further improved. This paper presents an intelligent control strategy of a grid-connected photovoltaic (GPV) plant based on an efficient artificial bee colony (ABC) metaheuristic. The motivation behind the use of such a design tool is to formulate and systematically solve the GPV output power maximization issue in comparison to empirical and traditional techniques. Such a competing ABC algorithm efficiently contributes to design cascaded proportional–integral (PI) controllers for the DC-link voltage and grid currents dynamics under rapidly changing conditions of irradiance and temperature. The tuning of PI controllers is formulated as a non-analytical optimization problem subject to the operational constraints of output power generation. An offline tuning approach is performed for all PI controllers operating across the entire modeled GPV system. Performance in terms of reproducibility capabilities, algorithmic convergence, and quality of solutions are compared to the most commonly used state-of-the-art algorithms. Analysis of variance (ANOVA) study based on the Friedman ranking and post hoc Bonferroni–Dunn tests is made and discussed. Demonstrative results in terms of output power maximization, harmonics mitigation, and robustness against irradiance and temperature changes are presented. Compared to traditional methods, the ABC-based tuning presents more advantages in system performance with 94.50% of power efficiency and 1.10% of harmonics mitigation, in addition to a benefit in terms of allocated design time and systematization of the control procedure.