A Study on Performance and Robustness Improvement of Model Predictive Current Control for Interior Permanent Magnet Synchronous Motor

Abstract

In this thesis, a new robust predictive current control for Interior Permanent Magnet Synchronous Motor (IPMSM) based on discrete-time disturbance observer and symmetrical three-vector is proposed. IPMSM has been widely adopted in industrial fields requiring high performance control, from home appliances to electric vehicle applications, due to its high-power density and fast dynamic characteristics. The control methods for IPMSM include hysteresis control, proportional-integral (PI) control, sliding mode control (SMC), predictive control, and artificial intelligence control. Recently, finite-control-set model predictive control (FCS-MPC), which is one of the kinds of predictive control, has been applied to power converter and motor drive due to rapid and robust microprocessor technology development in power electronics field. FCS-MPC has some advantages of being easy to implement and easy to apply to power converter because it predicts system response by possible switching state. However, because one voltage vector is applied during one control period, it has a relatively large current ripple, the switching frequency varies, and all of the candidate voltage vectors are evaluated, which means a large computational burden. To solve these problems, various studies have been conducted. Two-vector-based MPC, a variation rates based predictive control, and a modulated model predictive control (M2PC), however, have limitations in reducing steady-state current ripple and computational burden. Meanwhile, all of the existing methods of predictive control are model-based algorithms and are therefore sensitive to motor parameter variations. The motor parameters are different from the actual values due to the measurement error, or the values are changed according to the motor drive conditions. As a result, motor parameter mismatch occurs, which causes a prediction error and makes the system unstable. Typically, the inductance mismatch causes a steady-state current error and current oscillation, which degrades the current control performance. In particular, the mismatch of the permanent magnet flux leads to an error in the back electromotive force (EMF) component. This causes a constant current error in the steady-state, and overcurrent or undercurrent in the transient state, which deteriorates the current control performance. In order to improve the robustness of these parameters, various studies have been conducted. A method of correcting the predicted values using the past values of the model, an on-line parameter identification method, and a DOB method have been proposed. The method using the past values of the model is easily affected by the current measurement error. The on-line parameter identification method can obtain accurate parameters, but the implementation is complicated and it is difficult to obtain robustness against uncertain disturbance. The DOB method is a good solution to the system disturbance including the parameter variations and model uncertainty. However, the conventional DOB method has not considered in the discrete time, and there is a limit that the permanent magnet flux is required to be measured off-line. In this thesis, a new predictive current control method that uses a concept of deadbeat control and makes the duration ratios using a relationship between the symmetrical three-vector and current ripple minimization, and a new DOB independent with the permanent magnet flux in the discrete-time domain is proposed. The proposed method obtains the predicted voltage reference value with the time delay compensation by using the characteristics of the deadbeat control. The sector of the voltage vector is obtained using the phase angle of the predicted voltage reference, and the adjacent active-vectors and the zero-vector of this sector are used. By using the relationship between the symmetrical three-vector and the duration ratios, a cost function consisting of the voltages is designed, and the duration ratios of the active-vector and zero-vector minimizing the cost function are obtained. In addition, the proposed method adds a new discrete-time DOB that is robust to parameter variations for accurate current control. The prediction error for the driving point in the IPMSM parameter mismatch was analyzed. The proposed DOB is designed in z-domain to be suitable for predictive control. The estimated disturbances are compensated to the current prediction model and the predicted voltage reference model of the symmetrical three-vector predictive current control method. The parameter variation and the model uncertainty are set as disturbances, and in particular, the term including the permanent magnet flux is included in the disturbance. Stability analysis has been performed through closed-loop pole analysis, from which gains can be obtained. To verify the effectiveness of the proposed method, various experiments using a prototype set with 600W IPMSM were performed. Based on the proposed discrete-time DOB, a new predictive current control method based on the symmetrical three-vector that is robust to the parameters and independent of the permanent magnet flux is verified.