Hardware Validation for Control of Three-Phase Grid-Connected Microgrids Using Artificial Neural Networks

This paper presents a strategy for controlling inverter-interfaced DERs within a microgrid using an artificial neural network. The neural network implements a dynamic programming algorithm and is trained with a new Levenberg-Marquardt backpropagation algorithm. Hardware experiments were conducted to evaluate the performance of the neural network vector control method. They showed that the neural network control technique performs well for DER converter control if the controller output voltage is below the converter’s PWM saturation limit. If the controller’s output voltage exceeds the PWM saturation limit, the neural network controller automatically turns into a state by maintaining a constant dc-link voltage as its first priority, while meeting the reactive power control demand as soon as possible. Under variable, unbalanced, and distorted system conditions, the neural network controller is stable and reliable.