ニューラルネットワークを用いたPWMインバータのディジタル電流制御の安定性

抄録

There have been growing demands for high performance in current control of PWM inverters. Digital current controls are becoming dominant in the current controllers of the PWM inverters due to the rapid progress of microprocessor technologies. But the performances of the digital current controllers are degraded by changes of load impedances and delay necessary for calculation of the controllers. The authors proposed a digital current controller for a PWM inverter using a neural network. The new controller could compensate the delay necessary for the calculation and tune the gains of the controller to optimal ones. Stability of the learning of the neural network, however, was not guaranteed in case where the load impedances of the inverter were not exactly known.
This paper presents a method to guarantee the stability of the learning of the neural network. The Lyapunov method is used and the stability condition for the learning rate of the neural network is clarified. Simulations and experiments are done to verify the analytic results.