Adaptive Critic Based Optimal Neurocontrol of a Distributed Microwave Heating System Using Diagonal Recurrent Network

The dynamic capture ability of diagonal recurrent neural network (DRNN) makes them a suitable candidate for implementing real-time nonlinear adaptive controllers to handle the nonlinearity and uncertainty of high-power distributed microwave heating system (HPDMHS). In conventional DRNN-based adaptive control, the diagonal recurrent neural controller (DRNC) is trained online with one step cost and control law are not always optimal. To improve this, this paper couples a simple direct adaptive neural control with adaptive critic design (ACD) technique to achieve the optimal temperature tracking in HPDMHS. After transforming the original optimal temperature tracking control problem into an error regulation problem, the desired control is obtained by a regular DRNC, while the error regulation control is solve by ACD technique using DRNN. Simulation results demonstrate the superiority of the proposed DRNC-ACD method over conventional adaptive control in temperature tracking for HPDMHS.