Control Optimization Research Of PMSM Base On Neural Network Inverse System

With the rapid development of AC serve control technology in industrialmodernization, vector control method for AC motor has greatly improved the controlperformance, which makes AC serve control as a main stream of electric driving.Moreover, with many advantages in control, permanent magnet synchronous motor(PMSM) has increasingly become a research focus, meeting the needs of thedevelopment of modern industrial society. As for a complex nonlinear object, PMSMcontrolled by traditional method cannot get ideal decoupling performance. As a result,its applications and development has been limited to some extent, for the sake of suchfactors as dynamic changes of the parameters amd the internal state of the coupledeffects. Based on this, neural network inverse system decoupling methodology is analternative to the intelligent decoupling control and provide an effective solution to thethe intelligent motor control problems.This thesis aims to the critical problems in PMSM control and applications, appliesthe neural network inverse system theory to nonlinear systems decoupling process,builds up simulation model to verify the real system and thoroughly analysis the theoryand principles of the neural network inverse system of permanent magnetic synchronousmotor control system in response to this problem.In detail, firstly this thesis mainly studies the inverse system theory and deduceswith the reversible motor system mathematical expressions by means of analyzing themathematical model of PMSM. Considering the expression type can be used to establishmodel with inverse system method, the simulation results of this model shows that it iseffective in improving the decoupling performance compared with the traditionalcontrol strategy in a certain degree. Secondly, this thesis investigates the main principleof neural network inverse system and provides the steps for establishing the controlsystem based on neural network inverse system. Next, the thesis studies the structureand parameters of BP and RBF neural network, with aims to obtaining the ideal resultsof neural network module through large amounts of training test. After establishing theneural network inverse system control system of PMSM for decoupling and optimize the RBF structure parameters using the particle swarm intelligent algorithm, thedecoupling effect then reveals the method’s effectiveness. In the end, experiments basedon RBF neural network inverse system decoupling control system of PMSM shows thatthe fast response and excellent static and dynamic decoupling characteristics can beused to overcome the disadvantages of parameters’ and conditions’ variation. The paperbuilds the complete dSPACE experiment platform through connected the DS1104single board with the IPM intelligent modules and signal detection module and thenfinishes constructing control system model based on the dSPACE platform anddownloading program. The run-time effect of PMSM on the dSPACE experimentplatform is so ideal that it can prove the effectiveness of the control method.