Photovoltaic Fuzzy Logical Control MPPT Based on Adaptive Genetic Simulated Annealing Algorithm-Optimized BP Neural Network

• Published in: Processes Vol. 10; no. 7; p. 1411
• Main Authors: Zhang, Yan, Wang, Ya-Jun, Zhang, Yong, Yu, Tong
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.07.2022
• Subjects: Accuracy; Adaptive control; Algorithms; Ambient temperature; Artificial neural networks; Back propagation; Back propagation networks; Computer simulation; Control algorithms; Datasets; Efficiency; Fuzzy control; Fuzzy logic; Genetic algorithms; Irradiance; Machine learning; Mathematical models; Maximum power tracking; Neural networks; Particle swarm optimization; Photovoltaic cells; Robustness; Simulated annealing
• ISSN: 2227-9717; 2227-9717
• DOI: 10.3390/pr10071411

The P–U characteristic curve of the photovoltaic (PV) cell is a single peak curve with only one maximum power point (MPP). However, the fluctuation of the irradiance level and ambient temperature will cause the drift of MPP. In the maximum power point tracking (MPPT) algorithm of PV systems, BP neural network (BPNN) has an unstable learning rate and poor performance, while the genetic algorithm (GA) tends to fall into local optimum. Therefore, a novel PV fuzzy MPPT algorithm based on an adaptive genetic simulated annealing-optimized BP neural network (AGSA-BPNN-FLC) is proposed in this paper. First, the adaptive GA is adopted to generate the corresponding population and increase the population diversity. Second, the simulated annealing (SA) algorithm is applied to the parent and offspring with a higher fitness value to improve the convergence rate of GA, and the optimal weight threshold of BPNN are updated by GA and SA algorithm. Third, the optimized BPNN is employed to predict the MPP voltage of PV cells. Finally, the fuzzy logical control (FLC) is used to eliminate local power oscillation and improve the robustness of the PV system. The proposed algorithm is applied and compared with GA-BPNN, simulated annealing-genetic (SA-GA), particle swarm optimization (PSO), grey wolf optimization (GWO) and FLC algorithm under the condition that both the irradiance and temperature change. Simulation results indicate that the proposed MPPT algorithm is superior to the above-mentioned algorithms with efficiency, steady-state oscillation rate, tracking time and stability accuracy, and they have a good universality and robustness.