Neural Network Control Based on Nonlinear Autoregressive Exogenous Model and Adaptive Dynamic Programming Algorithm for High-Gain DC-DC Converter

DC-DC converters are extensively deployed across a range of new energy applications. However, traditional control methods that rely on accurate state space models encounter limitations in dc-dc converters due to the presence of parasitic parameters, component nonlinearities, and various uncertainties. To address these challenges, this article introduces a neural network control (NNC) for dc-dc converters, which is based on a nonlinear autoregressive exogenous (NARX) model and adaptive dynamic programming (ADP). This approach only requires the input-output data of the converter. The process begins with the NARX model capturing the dynamic behavior of the actual dc-dc converter, supplying the gradient information and an equivalent model necessary for the ADP algorithm. Subsequently, the NNC weight parameters are adjusted by the ADP algorithm. The proposed NNC inputs include the error, the integral of the error, and the derivative of the error, offering greater flexibility than a traditional PID controller due to the neural network’s nonlinear characteristics. Most notably, the NNC developed in the simulation environment can be directly implemented in the actual circuit. Finally, simulations and experiments conducted on a high-gain dc-dc converter demonstrate that the proposed NNC outperforms traditional controllers in terms of control performance.