Heuristic dynamic programming for neural network vector control of a grid-connected converter

This paper analyzes optimal control of a grid-connected converter (GCC) based on the adaptive critic designs (ACDs), especially on heuristic dynamic programming (HDP). Instead of using a trained model neural network to identify the dynamics of the plant, the paper uses exact GCC plant mathematical model to reflect the system dynamics accurately. Thus, the HDP for GCC only contains a critic neural network and an action neural network, which simplifies the control design. The training cycles of the critic and action networks in our HDP design are combined into one training cycle, which reduces the calculation time and makes it more suitable to online training. Results show that the RNN controller based on the designed HDP exhibits good tracking ability. The weights of the critic and action networks are adjusted online to adapt the changing GCC system automatically.