Artificial neural network with modified rider optimization for design and control of PV-integrated quasi Z-source cascaded multilevel inverter system

Abstract

The modeling of the controller for quasi Z-Source Cascaded Multilevel Inverter (qZS-CMI)-dependent 3-phase grid-tie Photovoltaic (PV) power system is considered in this paper. The state-of-the-art controller requires precise conceptual models and sophisticated optimization principles based on the derived models. However, such processes are limited to known system models, which are uncertain in future systems. Here, the controller for 3-phase qZS-CMI is modeled based on two phases, and the source PV voltage and output grid current are controlled. In Phase I, optimized Proportional Integral (PI) controller is used for finding out the total PV voltage, and Phase II utilizes the optimized Proportional Resonant (PR) controller enabled with the Artificial Neural Network (ANN) for controlling the grid current. For two phases, the modified optimization algorithm called Fitness Enabled-Rider Optimization Algorithm (FE-ROA) is used. Moreover, in Phase II, ANN is trained in an offline mode with the exact dataset arranged by the proposed FE-ROA, and it guarantees the control of grid current. The two phases plan to optimize the gain of both PI and PR controllers respectively using the same proposed algorithm. The main objective of phase I is to lessen the error among the reference PV voltage, and measured voltage, and phase II is to lessen the error among the reference and measured grid current. Hence, the grid-tie current injection is achieved by the developed module, and system-level control offers independent Maximum Power Point Tracking (MPPT). Lastly, the performance of the proposed controller for qZS-CMI is compared over the other controllers and substantiates the efficacy of the proposed one.