This paper presents the design of Unscented Kalman Filter (UKF) for estimation of state space variables of permanent magnet synchronous machine (PMSM). The UKF is shown together with the field oriented speed control. At first, the position and the speed of PMSM are measured, and UKF is used only for a load torque estimation. It is indicated how differences in sampling time of the speed and the current loop affects overall estimation performance. Subsequently, speed sensorless performance of the UKF with the same parameters is shown for comparison. Designed filter is verified only by Matlab simulation.

This paper presents an estimator-based speed sensorless field-oriented control (FOC) method for induction machines, where the state estimator is based on a self-contained, non-linear model. This model characterises both the electrical and the mechanical behaviours of the machine and describes them with seven state variables. The state variables are estimated from the measured stator currents and from the known stator voltages by using an estimator algorithm. An important aspect is that one of the state variables is the load torque and, hence, it is also estimated by the estimator. Using this feature, the applied estimator-based speed sensorless control algorithm may be operated adequately besides varying load torque. In this work, two different variants of the control algorithm are developed based on the extended and the unscented Kalman filters (EKF, UKF) as state estimators. The dynamic performance of these variants is tested and compared using experiments and simulations. Results show that the variants have comparable performance in general, but the UKF-based control provides better performance if a stochastically varying load disturbance is present.

This paper presents a new control algorithm development approach for induction machines by using model-based design and a systematically built model architecture implemented in MATLAB/Simulink. The model architecture follows a three-layer structure, and it is developed according to the principle of functional decomposition and the needs of reusability and expandability. The first model layer consists of elementary model and algorithm components, the second contains a machine simulation model and a field-oriented control (FOC) algorithm, built upon the first layer’s components, and the third realises the executable models by connecting the models and algorithms defined in the second layer. Furthermore, rapid control prototyping (RCP) is discussed as an experimental validation method, and an experimental setup with RCP is also introduced. The application of the presented methods is demonstrated by simulations as well as by experiments, and by using a control algorithm based on FOC as an example.

Development of control algorithms for electrical machines may be a difficult procedure if functional safety, software quality, reusability and expandability come into scope. These properties might be required in both of research and industrial development projects. State of the art methods and tools like model-based design (MBD) and automated code generation may help to meet these requirements. In this paper, MBD methods and a modular, reusable model architecture are presented for implementation of field oriented control (FOC)-based controller software for induction motors.

State estimation of induction machines may be a difficult problem, due to the non-linear behavior of theirs. For non-linear state estimation, the unscented Kalman filter (UKF) is a well-known extension of the linear Kalman filter. Operation of the UKF algorithm strongly depends on the process and measurement noise covariance parameters of the estimator. Determination of these parameters is not straightforward and can be difficult, especially if the number of state variables and hence the system complexity is relatively high. In this paper, the UKF algorithm is applied for speed sensorless state estimation of induction machines in such a way that seven state variables are estimated from the measured stator currents and from the known excitation voltages. In order to tune the noise parameters of the UKF, a new, optimization-based method is presented. This tuning method provides adequate behavior for the observer beside difficult operating conditions as it has been shown by simulation experiment.

In this study, three feasible speed sensorless estimators of induction machines are presented by using extended, unscented and cubature Kalman filter algorithms. The estimators are based on an augmented non-linear state-space model of these machines, which describes the dynamics in stationary reference frame with six state variables. As an important part of the estimator design, an observability study is provided for the nonlinear model and an observability condition is formulated as well. The estimators are compared experimentally around zero stator frequency with respect to the speed estimation performance. However, the estimators are investigated only in open-loop and without external load disturbance.