An efficiency optimization vector control scheme based on back propagation (BP) is presented in this paper. The controller is designed to generate d-axes currents command and q-axes currents command simultaneously. In order to achieve a robust BP efficiency optimization vector control from variation of motor parameters, an online learning algorithm is employed. Results of experiments are provided to demonstrate the effectiveness of the proposed method.
Keywords -PMSM; efficiency optimization; neural network
I INTRODUCTIONThe advantages of PMSM make them highly attractive candidates for traction and residential drive applications, such as hybrid electrical vehicles (HEV) [1][2][3]. In this application of continuous long time operation, the efficiency of the motor is one of the most important factors.Some prior studies proposed the algorithms for efficiency optimization of PMSM. Currently, all the efficiency optimization methods can be divided into two categories: based on loss model and based on search technique, which are put the minimum input power as control objectives. The loss model of PMSM was introduced in [3,4], and an optimal control of the armature current vector, which can make the electrical loss minimized, was provided. This method is suitable to do theoretical analysis, in real-time control process, there is a large amount of dependence on the parameters and computational complexity.[5] provided a search method based on genetic algorithm, but the efficiency-optimized speed of this method will slow down when a larger quantity initial species was chosen. A fuzzy logic control method based on torque compensation and the PM and iron loss was presented in [6], but fuzzy logic rules have SM mathematic model considering the effects of copper loss been influenced by anthropogenic factors, and lack of practicality. Minh and Yoichi applied the golden section technique to optimize the efficiency of induction motor firstly [7], thus shortening the optimization time greatly because of its simplicity. However, this approach is not easy to determine the direction of the initial optimization.An efficiency optimized vector control scheme for PMSM drives is designed in this paper. In the proposed strategy is based on back propagation neural network. The controller is designed to generate d-axes currents command and q-axes currents command simultaneously and an online learning algorithm is employed to achieve a robust BP efficiency optimization vector control from variation of motor parameters. At last, results of experiments are provided to demonstrate the effectiveness of the proposed method.
II LOSS MODEL OF PMSMIn the process of PMSM efficiency optimization, the motor mechanical loss and stray loss can be ignored generally, putting i d obtained by loss model, which makes the smallest electrical loss, as the efficiency of optimizing the initial value. According to three dimensional magnetic field theory [2], the application of software ANSYS with finite element method (FEM) to simulate, finite element model is built...