On the Use of DualReLU ANN for Approximating Explicit Model Predictive Control for Buck Converters

Explicit model predictive control (EMPC) has attracted extensive attention in the field of power electronics owing to its excellent dynamic performance. However, the implementation of EMPC in hardware poses considerable challenges as it requires a large amount of computing resources for online computation. In this regard, this paper proposes to use a double-rectified linear unit (DualReLU) artificial neural network (ANN) to approximate EMPC. By taking advantage of the bilaterally bounded property of the offline law distribution in power converter applications, the proposed DualReLU ANN is verified to be able to effectively approximate EMPC while significantly reducing computational load and memory usage requirements.