Research Of PMSM Adaptive Speed Regulation Based On BPNN

Owing to its high power density,high efficiency,strong overload ability,excellent control performance, Permanent Magnet Synchronous Motors (PMSM) develops rapidly in small-middle capacity speed control system and precision control occasions, and have the trend to be extended gradually. However, the PMSM magnetic field has the unique features of cross-coupling and cross-saturation and the system is a strongly nonlinear, time-varying and multivariable one, in order to achieve high-precision speed control result, the traditional linear control methods should be altered.In the PMSM speed control system, the sensor's existence makes the system reliability worse, costs increased and maintenance complex. So research of sensorless PMSM speed control becomes a new hot spot. In the speed control system, we have recognized that the neural network has good approximation ability, and this ability is also available in sensorless system.Artificial neural network (ANN) approach has the ability of strong self-adaptive & self-learning and capable approach to arbitrary complex nonlinear function, so it is perfect to be utilized in complex nonlinear control occasions. BP neural network (BPNN), among all the ANN structures, is better known and most frequently used currently. BPNN has simple structure; few parameters need to determine in advance, strong generalization, sufficient approximation precision and good real-time response. It's quite practical that BP neural network to be implemented in the PMSM speed control.The paper presents a adaptive speed regulation strategy for PMSM, based on BPNN, completes a duel-neural network control system, which include an identification-NN and a controlling-NN. The first NN identifies system parameters and adjusts the speed controller's parameters, the later NN achieve the adaptive speed regulation with a PI controller. The simulation result validates the quick response, perfect real time cooperation, better precise characteristics of the control system.The paper presents a BPNN position senseless control strategy for PMSM, based on Mixed Training Algorithm (MTA), which consists of a chaos optimal and gradient descent algorithm. After offline training, the BPNN could be utilized for rotor position estimation. The simulation result demonstrates that MTA could efficiently fast the NN's convergence speed, as well as better rotor position estimation precision and low errors.The paper put forwards a PMSM speed regulation system based on TI TMS320F2812, as well as relevant software and hardware implements, which will provide a platform for PMSM control strategy application .The DSP control system prepares samples for NN, which is the precondition for PMSM self-adaptive speed regulation and rotor position estimation.