Adaptive neuro fuzzy inference system based decoupled control for neutral point clamped multi level inverter fed induction motor drive

Abstract

The presence of an integrator in a reference model of a rotor flux-based model reference adaptive system (RF-MRAS) and non-linearity of the inverter in the output voltage degrade the speed response of the sensorless operation of the electric drive system in terms of DC drift, initial value issues, and inaccurate voltage acquisition. To improve the speed response, a compensating voltage component is supplemented by an amending integrator. The compensating voltage is a coalition of drift and offset voltages, and reduces DC drift and initial value issues. During low-speed operation, inaccurate voltage acquisition distorts the stator voltage critically, and it becomes considerable when the stator voltage of the machine is low. Implementing a three-level neutral point clamped inverter in speed-sensorless decoupled control of an induction motor improves the performance of the drive with superior quality of inverter output voltage. Further, the performance of the induction motor drive is improved by replacing the proportional-integral (PI) controller in the adaption mechanism of RF-MRAS with an adaptive neuro-fuzzy inference system (ANFIS) controller. A prototype model of the three-level neutral point clamped inverter (3L-NPC)-fed induction motor drive is fabricated in a laboratory, and its performance for a RF-MRAS, modified RFMRAS, and modified RFMRAS using ANFIS are compared using different benchmark tests.