Pulse injection-based sensorless switched reluctance motor driver model with machine learning algorithms
- PDF / 2,022,028 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 61 Downloads / 214 Views
ORIGINAL PAPER
Pulse injection-based sensorless switched reluctance motor driver model with machine learning algorithms Ferhat Daldaban1 · Mehmet Akif Buzpinar2 Received: 4 June 2020 / Accepted: 18 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract In this study relationships of pulse injected idle phase currents are used to predict rotor position with tuned fine tree and ensemble bagged tree algorithm in MATLAB. Different classifier algorithms trained, tested, and the best accurate results are obtained via ensemble bagged tree classifier using idle phase currents. Three-phase 6/4 switched reluctance motor (SRM) with optical position sensors diagnosis pulses has been injected into idle phases and operated at constant load and speed. The measured idle phase currents were rearranged using the time series method and trained with supervised machine learning algorithms. These unprocessed idle phase currents reduce processing time and contribute to the real-time operation of the system. This study proves that SRM can be driven by predicting the active phase to be triggered by trained ensemble bagged tree and tuned fine tree machine learning algorithms from real-time measured idle phase current data. Keywords Machine learning (ML) · Ensemble bagged tree classifier · Sensorless drive · Switched reluctance motor (SRM) · Position estimation · Pulse injection
1 Introduction Switched reluctance motor (SRM) is a desirable alternative for industrial applications because of its advantages like high torque, simple and rugged structure, low production, and maintenance cost. Although, acoustic and electrical noise, torque ripple, and position sensor requirements are the drawbacks [1]. Efficient driving in SRM requires the precise estimation of the rotor position. By integrating an optical or field effectbased mechanical sensor, the rotor position is acquired. A mechanical rotor position sensor is not only decreased reliability but also increased cost, maintenance time and requires special attention for positioning of the sensor after maintenance or replacement [2]. The position sensor requires more space inside or outside of the motor case and adds to the cost.
B
Mehmet Akif Buzpinar [email protected] Ferhat Daldaban [email protected]
1
Electrical and Electronics Engineering Department, Erciyes University, Kayseri, Turkey
2
Gemerek Vocational School, Cumhuriyet University, Sivas, Turkey
[3]. Sensorless rotor position estimation algorithms overcome these difficulties. Studies on sensorless control methods have been reported by a large academic community in recent years. Traditionally, sensorless control techniques rely on the change in the magnetic behavior relative to the rotor position because of the mechanical time constant larger than the electrical time constant. For that matter, measurements are usually made from active phase. Voltages and currents are used to reach inductance or flux linkage of the active phase [4]. These values are used to determine the rotor position with di
Data Loading...
