Position-Sensorless Control of SRM Drives Based on a Novel Super-Twisting Sliding Mode Observer With the Machine Learning Model

In this article, a novel super-twisting sliding mode observer (NSTSMO) based on the machine learning model is proposed for low-to-high speed range position-sensorless control of switched reluctance motors (SRMs). Based on the torque balance method, support vector regression (SVR) optimized by the proposed hybrid optimization algorithm (HOA) can rapidly build the accurate SRM model. Based on the model, an NSTSMO is built with three improvements. First, a torque difference estimator is added to improve the robustness of the observer without increasing chatter. Second, an advanced error function with higher accuracy is proposed by the flux linkage error, which can simplify the structure of observer. Third, with a simple structure, the dual-layer super-twisting sliding mode observer (STSMO) is designed to accurately estimate both the speed and position, which can further improve the performance of observer. Finally, the experimental results have verified that the proposed method can be applied to various operating conditions with stronger robustness, higher accuracy, and lower chatter compared to other observers. The proposed method achieves adaptive high-precision control in multiple scenarios, significantly simplifying motor structure while ensuring high reliability, which provides an innovative solution for industrial integrated applications.