| With the continued development of modern industry, permanent magnet synchronous machines (PMSM) are now widely employed from high performance accurate servo drives, aerospace/avigation, wind power generators, to electric/hybrid electric vehicles, due to their high efficiency, high starting torque and high power/torque density, etc. In a PMSM based control system, accurate machine parameters are usually necessary for assistance of controller design such as sensorless control, optimal PI controller constants for vector control and online identification/compensation of voltage source inverter (VSI) nonlinearlity. However, the parameter values of PMSM such as winding inductance, resistance and rotor flux linkage will change with the variation of motor temperature, load and magnetic saturation and be away from their nominal values at normal temperature. Especially, the impact from temperature on PMSM parameter values is most significant and occurs most frequently, especially for stator winding resistance and rotor flux linkage values. For stator winding, the winding resistance value increases with the winding temperature rise. For rotor magnet, the rotor flux linkage value decreases with the rotor magnet temperature rise. The performance of designed control system is easy to suffer from the machine parameter variation, especially when the actual parameter values are far away from their nominal values at normal temperature, which even causes the system cannot work. For example, with the temperature rise in stator and rotor, the drive system will suffer from the low power efficiency, magnet demagnetization, power density drop and even irreversible demagnetization due to high temperature. Therefore, the existing literatures start to focus on estimating the stator winding resistance and rotor flux linkage values through the system identification theory associated with the measured motor terminal signals such as stator currents, voltages and rotor speed, which will be further applied for online adjusting the controller constants and estimating the temperatures of stator winding and rotor magnet. A deep and comprehensive research is described in this dissertation, which shows the key technology of PMSM parameter identification is to solve ’two problems’. Further, based on the research on the ’two problems’, three schemes for multi-parameter estimation of PMSM are proposed and verified on a vector control based surface-mounted PMSM test rig finally. The’two problems’are shown as follows:(1) The rank number of employed reference/variable model for identification algorithm. The rank deficient PMSM steady state equation is analyzed and the merits/drawbacks of existing rank deficient state equation based parameter estimation strategies are investigated by theory and experimental analysis, which show that the identification algorithm can converge to the actual machine parameter values only if the applied reference/variable model is full rank. Based on the steady state rank deficient PMSM models in existing literatures, the online estimation strategies based on model reference adaptive system (MRAS) and Adaline Neural Network (NN) are developed for investigation.(2) The accuracy of employed PMSM parameter identification algorithm suffers from the accuracy of VSI (voltage source inverter) nonlinearity compensation. The influence of VSI nonlinearity compensation on different PMSM parameter estimation is investigated by theories and experiments. Compared with the estimation of stator winding resistance and inductance values, it is verified by theories and experiments that the influence of VSI nonlinearity on rotor flux linkage estimation is most insignificant at rated rotor speed. In addition, the state equation, including both the specified PI regulator output voltage and the VSI nonlinearity in PMSM synchronously rotating reference frame, is derived and a strategy based on this state equation is proposed for online estimating the VSI nonlinearity under jd=0control.From the foregoing theory and experimental analysis of the’two problems’, it is evident that the most crucial content of PMSM parameter identification technology includes two points:the design of a full rank reference/variable model for identification and a proper scheme of VSI nonlinearity compensation. Based on the two points, the second part of this research is performed and three PMSM parameter estimation schemes are proposed. The rank of applied reference/variable models is taken into account in these three schemes. Further, the influence of VSI nonlinearity is also taken into account in the three schemes, and proper VSI nonlinearity compensation schemes are synthesized in estimator design as well. The detail of proposed three schemes is shown as follows:(1) A simplified full rank model based on error analysis is proposed for non-salient pole PMSM parameter estimation. This strategy needs only to inject a pule of flux weakening current in d-axis and the PMSM stator inductance, resistance and rotor flux linkage can be online synchronously estimated. Using the proposed strategy, the online PMSM parameter estimators based on Adaline and ICA (immune clonal algorithm) are proposed and verified by experiments, respectively.(2) Under constant rotor speed and load torque, a full rank SPMSM estimation strategy is proposed. This strategy needs only to inject a positive pulse in d-axis current to obtain the full rank reference/variable model. Without measuring the load torque, the dq-axis inductances and rotor flux linkage variation due to injected current pulse are cancelled by the constant torque equation and can be effective in online estimating rotor flux linkage and stator winding resistance synchronously. Using this strategy, the estimators based on Adaline and ICA are proposed and verified by experiments, respectively.(3) The scheme based on embedding thermocouple in stator winding for online estimating the variation of stator winding resistance is investigated for reducing the estimated parameter number in PMSM steady-state model. Under id=0control, this scheme can be employed for designing a full rank reference/variable model for online estimating the variation of rotor flux linkage. This scheme associated with the VSI nonlinearity compensation method which is proposed in this dissertation and fitted for id=0control, can be used for online accurately estimating the variation of rotor flux linkage corresponding to the variation of rotor temperature.In all, the two most crucial factors of PMSM parameter identification technology are analyzed by theory and experiments and three schemes for PMSM parameter identification are proposed and verified by experiments, which are aiming at the foregoing two factors. The proposed three parameter identification schemes are not only fitted for different application fields but also of great theoretic and practical significance. |
PDF Full Text Request