Artificial neural networks controllers for three-phase neutral point clamped inverters

Abstract

In this study, a new current control method consisting of proportional-resonant (PR) based artificial neural networks (ANN) has been developed for a neutral point clamped (NPC) inverter. The used NPC inverter circuit is three phase three level three leg and four wire. Current control in the inverter has been carried out primarily with PR and Hysteresis current controllers for comparison purposes. Then, a dynamic ANN structure is designed and trained using the PR current controller data from the simulation. The performance of this ANN current controller is investigated by simulation and experimental studies. The proposed ANN controller has been compared with other ANN controllers used in power electronics applications in the literature, based on the size and functions of the network structures. In addition, NPC inverters have a controller whose task is to balance the voltages of direct current (DC) bus capacitors. A separate static ANN based controller has been developed for this voltage balancing controller, which usually includes two proportional-integral (PI) controllers. The performance of these two types of ANN controllers has been experimentally verified with balanced/unbalanced, linear/non-linear loads.

Keywords

• Artificial neural networks
• Capacitor voltage balancing
• Current controller
• Multi-level inverter
• Neutral point clamped inverter
• Proportional resonant