Data-driven Nonlinear MIMO ADP Method and Its Application in PMSM Control

Abstract

The power system is a nonlinear, time-varying, high-dimensional system. How to carry out effective control to ensure its safer and more stable operation has been the subject of many scholars' research, and with the continuous expansion of the power system scale and randomness. With the access of stronger new energy sources, the challenges facing the security and stability of power systems are becoming more and more severe. The conventional optimal control method has certain limitations. For example, the variational method can only solve the optimal problem that the control quantity is not constrained. The maximal/minimum value principle can only solve the optimal control problem described by ordinary differential equations. Although the plan can solve the more general optimal control problem than that described by the ordinary differential equation, it is a problem of dimensionality hazard because it is a time-backward algorithm. Adaptive dynamic programming is the product of the integration of artificial intelligence and control technology. Its essence is to approximate the solution of Hamilton-Jacobi-Bellman equation by using the approximate structure of the function of the neural network. This method does not depend on the mathematical model of the controlled object, nor does it need to define the performance index accurately, and can learn online. The introduction of this method into the power system can provide a new idea for the non-linear optimal control of the power system. Based on the traditional Adaptive Dynamic Programming (ADP) algorithm, this paper proposes a data-driven nonlinear Multi-Input and Multi-Output (MIMO) adaptive dynamic programming algorithm, and applies this algorithm to Permanent Magnet Synchronous Motor (PMSM) related control. The simulation of single objective control and under-actuated control model proves that the data-driven adaptive dynamic programming method based on least squares strategy iteration has strong robustness.