Dragonfly Optimized MLP-Controlled Cascaded Quasi-Z-Source Converter for Solar Photovoltaic Systems

Abstract

Along with the growing penetration of solar photovoltaic (SPV) systems with modern power networks, it is essential to develop advanced power conversion and intelligent control strategies capable of achieving stable operation under dynamically varying environmental conditions. Most of the conventional maximum power point tracking (MPPT) methods and conventional DC-DC converter control techniques will be having disadvantages like slow convergence, fluctuating voltages, low tracking accuracy and poor adaptability in sudden change of the irradiance and load. However, to overcome these disadvantages, this paper presents a Multi-Layer Perceptron (MLP) boost optimizers with an optimized Dragonfly Algorithm (DA-MLP) for better voltage stability in solar PV system with cascaded quasi-Z-source (C-qZS) DC—DC converter. The cascaded qZS converter offers high voltage gain, continuous input current, low switching stress and good dynamic response for renewable energy application. The developed DA-MLP controller uses the PV voltage and PV power as the inputs and is used to predict the optimal duty cycle value as required for the efficient operation of the proposed converter. The Dragonfly Algorithm is used to optimally tune the weights and biases of the MLP network for better convergence, minimum mean squared error and better adaptive learning in nonlinear operating condition. The entire system was built and tested through MATLAB/Simulink (MS) environment for various irradiance profiles varying from 200〖" W/m" 〗^2to 1000〖" W/m" 〗^2. Conventional Perturb and Observe (P&O), Incremental Conductance (INC), Artificial Neural Network (ANN) and standard MLPless MPPT techniques were compared. The simulation showed that the proposed DA-MLP framework provided better performance with a converter efficiency of 97.84%, MPPT tracking efficiency of 98.42%, smaller voltage ripple of 0.21" V" and lower total harmonic distortion (THD) of 0.82%. Additionally, the proposed controller showed shorter settling time, as well as greater voltage stabilization and dynamic adaptability, thus creating an efficient and intelligent control system for future generation photovoltaic energy converter systems.