IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014 1 A Fixed Switching Frequency Predictive Current Control Method for Switched Reluctance Machines Rajib Mikail, Student Member, IEEE, Iqbal Husain, Fellow, IEEE, Yilmaz Sozer, Member, IEEE, Mohammad S. Islam, Senior Member, IEEE, and Tomy Sebastian, Fellow, IEEE Abstract—The paper presents a novel fixed switching frequency predictive current control method for switched reluctance ma- chines (SRM). The proposed deadbeat predictive current con- troller accurately predicts the required duty ratio for the PWM pulse for a given reference current in each digital time step over the entire speed range of operation. The pulse width depends on the operating conditions, machine parameters and the rotor position. The controller utilizes the machine inductance profile as a function of current and rotor position to accurately predict the required voltage. The control method is studied through computer simulation and followed by experimental validation. The method is suitable for torque ripple sensitive applications requiring accurate tracking of a given current profile and mitigating the audible noise due to the switching of the inverter. Index Terms—Back EMF, deadbeat controller, fixed frequency, hysteresis control, incremental inductance, nonlinear model, pre- dictive control. I. I NTRODUCTION A switched reluctance machine produces torque with inde- pendent phase control. The current required in each phase depends on the torque demand in torque controlled applica- tions. Each phase has its own command current as a function of position depending on the control algorithm. For applications requiring smooth torque production, the torque ripple control algorithm has to perform within the specifications over the entire speed range. The primary objectives of these applications are low torque ripple, low acoustic noise and low unbalanced magnetic force that contributes to vibration, and fast posi- tion, speed and torque responses. Various control methods for smooth torque production have evolved for SRM over the past few decades. These include current-profiling-based control [1]– [3], feedback linearized and decoupled control [4]–[6], direct torque control (DTC) [7], [8], iterative-learning-based control Manuscript received September 7, 2013; revised January 20, 2014; accepted February 28, 2014. Paper 2013-EMC-633.R1, presented at the 2012 IEEE En- ergy Conversion Congress and Exposition, Raleigh, NC, USA, September 15– 20, and approved for publication in the IEEE TRANSACTIONS ON I NDUSTRY APPLICATIONS by the Electric Machines Committee of the IEEE Industry Applications Society. R. Mikail was with North Carolina State University, Raleigh, NC 27606 USA. He is now with ABB Inc., Raleigh, NC 27606 USA (e-mail: rajib. mikail@us.abb.com). I. Husain is with North Carolina State University, Raleigh, NC 27606 USA (e-mail: ihusain2@ncsu.edu). Y. Sozer is with the The University of Akron, Akron, OH 44325 USA (e-mail: ys@uakron.edu). M. S. Islam and T. Sebastian are with Halla Mechatronics, Bay City, MI 48706 USA (e-mail: mohammad.s.islam@ieee.org; t.sebastian@ieee.org). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIA.2014.2322144 (ILC) [9], fuzzy- and neural-network-based control [10], [11], control with torque-sharing functions (TSF) [12], and sliding- mode-observer-based control [13], [14]. Feedback linearizing and decoupled control method [4]–[6] introduces a transforma- tion to convert a nonlinear system into an equivalent linear one. It is suitable for high performance applications, but only when an accurate model of the machine is available. The system can become unstable with less accurate models. The digital real- time implementation is feasible only at lower speeds. DTC, ILC, fuzzy, sliding mode and neural network based control methods also suffers from instability at higher speeds. Several of the control algorithms depend on shaping the phase currents to produce a smooth torque profile. Control algorithms that compensate for torque fluctuations throughout the conduction interval precisely depend on online or offline current waveforms. Online methods work on the basis of calcu- lating a reference or command current instantaneously, while in offline methods, the current commands are fetched from a pre calculated look-up table. These command currents are established in the phases by a high bandwidth current regulator. In most of the control methods mentioned, the final block of the algorithm incorporates a current controller as a final step. The common methods for current control are the hysteresis [2], [9], [10], [15] and PI controllers [16] or a hybrid type of control [17]. A soft chopping current control method has been proposed in [18]. The dynamic performance of these controllers deteriorates at higher speeds. The hysteresis type current con- troller, in general, has better dynamic performance, but the problem of varying frequency restricts its use in many appli- cations where acoustic noise due to switching is undesirable. The PI current controller has the advantage of fixed switching frequency. Speed and position dependent PI constants result in better control at higher speeds but this increases the complexity of the controller. In the digital implementation of the hysteresis current con- troller, the voltage command of the inverter is updated at the end of each control loop time-period. The current may deviate from its reference for the entire digital time period which will increase the torque ripple especially during commutation from one phase to the next one. The switching frequency is not constant as shown in Fig. 1. The switching frequency of the inverter could fall within the audible range which would lead to acoustic noise problems. In a PI controller, several digital time steps are required to respond to a step change in command. At higher speeds, the system cannot utilize the full control bandwidth available from the converter due to the poor dynamic performance and the stability issue of the PI controller itself. 0093-9994 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.