In practical industrial applications, the control performance in a wide speed range is hard to ensure, especially under the low-speed condition with a low-cost incremental encoder, while the unknown structure parameters may also degrade the tracking performance. This paper proposes a low-order adaptive instantaneous speed estimator (AISE) and a self-tuning control strategy to promote the speed control performance in a wide speed range with unknown inertia parameters. Together with the adaptive-Kalman-filter-based AISE, a novel measurement noise variance updating scheme, which allows more appropriate compensation in the different speed range than fixed error variance, is introduced through the theoretical analysis based on probability and stochastic process. Moreover, an easy-to-implement self-tuning law, integrated with an online recursive-least-square-based parameters identification method, is developed to tune the speed controller, while the AISE is also adjusted online to ensure the control performance with a considerable variation of load inertia.All strategies were implemented in a TMS320F28335-based permanent magnet synchronous motor (PMSM) control system with a low-cost 2500-line incremental encoder, and the results demonstrated the effectiveness of the proposed techniques. KEYWORDSadaptive Kalman Filter, instantaneous speed estimator, measurement noise adaptation, motor control, parameter self-tuning Recently, high-performance PMSM speed control system has been more and more popular in many industrial applications, such as robots, manufacturing instruments, hybrid or pure electric vehicles, due to the demand of high precision, high reliability, and user-friendliness. However, there are many obstacles to achieve high-performance PMSM speed control [1][2][3][4][5][6]. Among them, how to acquire the speed information accurately and rapidly is one of the principal challenges. In traditional PMSM applications, the methods by direct detection from the position sensors, including frequency measurement and period measurement [7,8], is widely applied. These methods are usually reliable owing to the development and progress of the sensor technology in the past decades, while there are

No abstract