Sensorless Control Of Switched Reluctance Motor Drives

Due to its simple and solid structure, the switched reluctance motor (SRM) is suitable for highspeed operation and resilience to harsh operation conditions, which has attracted great attentions overthe recent years. The research of the high performance sensorless techniques can further improve thesystem integration, and enhance the high speed capability and harsh environment adaptability, andmake full use of the strong fault-tolerant capability and high reliability, thus can expand theapplication areas of SRM, and lay a foundation for broadening the application of the SRM underextreme conditions such as high speed, high temperature and high pollution.Aiming at the starting-low speed operation, and the medium-high speed operation conditions,this paper presents several new sensorless control schemes for SRM drives. Considering thefault-tolerant operation, some exploratory study about the fault-tolerant sensorless techniques has alsobeen done.Aiming at the load-starting control, the methods such as idle phase current comparison method,idle phase current slope difference comparison method, idle phase inductance two thresholds basedmethod are proposed in this paper. The theoretical analysis and experimental results show that theturn-ON and turn-OFF angle can be adjusted in the idle phase inductance two thresholds basedmethod, which can meet the high performance starting requirements. In order to estimate the initialrotor position at standstill and rotating shaft conditions, series of algorithms based on the full-cycleinductance estimation is proposed, such as linear region inductance model based method, three-phaseinductance synthesis vector orthogonal decomposition method, and inductance vector coordinatetransformation method. These methods can estimate the initial rotor position and starting the SRMwithout reversal, and can realize the sensorless control operation under no load and light loadconditions. All these proposed starting schemes do not require any additional hardware and occupyhuge memory, and in addition, no complex computation process and no priori knowledge on magneticcharacteristics of SRM are required. Thus, they possess good versatility and portability.Aiming at the medium-high speed regions, a kind of approximation acquiring method for fluxlinkage characteristics is proposed. Based on the flux linkage and inductance characteristics, the RBFneural network based flux linkage thresholds method and the inductance dynamic thresholds methodare proposed. The former method uses the RBF neural network to modeling the reference fluxthresholds, but the latter uses the inductance mathematic model. By comparing the real flux linkage orinductance with the reference thresholds, the driving signals of each phase can be estimated, which can be used for sensorless control. The research results show that, due to low calculation speed of theRBF neural network, the applicable speed range of the RBF based flux thresholds method is limited.But the inductance dynamic thresholds based method avoided this problem. As compared to thetraditional inductance model based method and flux/current based method, this method avoids thecomplex look-up tables, complex formula calculation, as well as intelligent model implementationlinks. Thus, this method is relatively easy for implementation. And in addition, the selection of thefitting function can be varied, whether it is periodicity or not, which makes the algorithm have thecharacteristics of diversity. All these algorithms are implemented in software, thus no additionalhardware is added.In order to avoid the dependence of flux linkage and inductance characteristics, and avoid thecomplex memory, modeling and complex calculation, an inductance slope based aligned positionestimation method is proposed in medium and high speed range. As a kind of “single-cycle singlespecial position” detection method, can further improve the real-time and fast capability of theposition estimation. In the experiments, in order to avoid the error position index pulse at the turn-ONposition, two methods have been proposed in this paper. In the first method, the inductance slopebased method is combined with an inductance subregion scheme. By choosing the phase with thelargest inductance amplitudes as the estimation phase, the mutation of calculated inductance slope atthe turn-ON angle can be avoided. In the second method, the position index pulses can be restructuredby combining the phase current slope detection method and the inductance slope detection method,thus the error pulse at the turn-ON angle can be avoided, and meanwhile the aligned position is kept.Considering the fault-tolerant operation, the fault-tolerant strategies of these position estimationmethods are studied respectively, and realized the sensorless control under lacking phase faultconditions.