Fault Tolerant Sliding Mode Controller Design Subject to Sensor Faults for Output Voltage Regulation of a Self-Excited Induction Generator

• Published in: Electric power components and systems Vol. 49; no. 6-7; pp. 681 - 695
• Main Authors: Calgan, Haris, Demirtas, Metin
• Format: Journal Article
• Language: English
• Published: Philadelphia Taylor & Francis 09.02.2022; Taylor & Francis Ltd
• Subjects: Control systems design; Controllers; Electric potential; Fault detection; Fault tolerance; fault tolerant control; Faults; Induction generators; Neural networks; Performance enhancement; Radial basis function; radial basis function neural network; self-excited induction generator; Sensors; sliding mode; Sliding mode control; Thyristors; Tracking; Voltage
• ISSN: 1532-5008; 1532-5016
• DOI: 10.1080/15325008.2021.2011483

This paper aims to improve the performance of the output voltage regulation of the Self-Excited Induction Generator (SEIG) with fixed capacitor-thyristor controlled reactor (FC-TCR) structure in case of sensor faults. A sliding mode controller (SMC) is designed to generate the triggering angle of the FC-TCR to regulate the output voltage. The performance of proposed SMC is firstly tested without sensor faults and high success in tracking the reference is obtained. Since an additive fault occurs on the sensor, SMC performs well tracking dynamic but is on the incorrect reference. Therefore, Radial Basis Function Neural Network (RBFNN) is trained accurately with the 2% mean squared error (MSE) by taking the stator currents, triggering angle and generator speed as inputs, whereas the output voltages of the SEIG are outputs. Further, fault detection architecture is well designed with a voltage error index calculated by subtracting the sensor output from the estimated output. The threshold based selector reconfigures the controller well. The simulation results show that in case of sensor faults and various loading conditions, proposed fault tolerant SMC remains the actual output voltage of the SEIG at real reference with maximum 2.2% steady state error in a finite time.