Summary
Today, grid‐connected inverters are widely used in renewable energy systems, distributed generation systems, and microgrids. Traditionally, control of a grid‐connected inverter is developed using proportional‐integral controllers in the dq reference frame. However, in many applications, a connection from the dc neutral point to the ground is needed, and in some applications, the inverters are operated in parallel over the DC bus such as the bidirectional converters in hybrid dc and ac microgrids. For such a situation, circulating current will appear, and control design in dq0 reference frame would be needed to suppress the circulating current. This article presents a novel neural network control design in dq0 reference frame and compared it with conventional control methods. The proposed neural network control is developed based on the full state‐space equation of the grid‐connected inverter system and is trained to implement optimal control based upon approximate dynamic programming. Compared to the conventional control, the neural network control responds faster, is able to suppress the circulating current, and preserves adequate power quality.