Research On Neural Network DTC Algorithm And Special Control Chip Of Asynchronous Motor

At present,Electric Vehicle/Hybrid Electric Vehicle(EV/HEV)is gradually becom-ing the mainstream trend in the transportation industry.Asynchronous Motor(AM)has the advantages of simple structure,low cost and stable performance,which is widely used in BMW,Tesla and other high-end coupe brands,but it is strict with the Motor control performance.Vector Control Method(VCM)is widely used for better dynamic and static performance,including Direct Torque Control(DTC)and Field Oriented Control(FOC).There are following disadvantages about the traditional DTC scheme:the torque ring introduces a switching regulator,which leads to high ripple ripple;The PWM wave is generated directly by the switch table without fixing its period.The accuracy of flux ob-servation is greatly affected by motor parameters and operating conditions.Therefore,how to improve the flux observation model to obtain higher observation accuracy is the primary task.Based the principle of induction motor,this paper analyzes in detail the different in-fluences of selecting different boundary basic vectors in DTC on the control effect of flux and torque,and determines the appropriate switch table logic.The motor speed is obtained by differentiating the stator side flux phase.Using u-i model as a reference,the flux ob-servation model based on BP-ANN was established by off-line training,and the DTC-IP transformation was completed after algorithm verification on SG platform.Finally,IP integration is carried out on Vivado platform and the solidification of DTC algorithm is completed.At the same time,the performance of DTC-ASIC is comprehensively evalu?ated.In addition,the physical verification of the algorithm is completed based on ARTIX platform.There are a few work points as follows:(1)Completed the modeling and simulation of DTC hardware algorithm based on SG platform,realized the core conversion of DTC-IP,and laid the foundation for DTC-ASIC solidification;(2)The phase of the stator flux linkage is differentiated by time to realize the non-inductive prediction of the speed.The BP algorithm is combined with the traditional DTC,and the speed sensorless DTC algorithm based on the BP-ANN flux model is designed.The transient response of the speed loop is 0.25s,the relative error of the observed flux linkage does not exceed 2%,and the relative error of the electromagnetic torque does not exceed 2.5%,which improves the control performance of asynchronous motors while reducing hardware overhead;(3)Completed the design,timing constraint,performance and power consumption evalu-ation of DTC-ASIC based on Vivado platform.At the same time,the peripheral hardware circuit of DTC system was designed and the ARTIX-FPGA hardware platform was built.Finally,The solidification and prototype verification of DTC-ASIC were completed.