Optimal Policy Gradient-Based Type-2 Fuzzy Control for Multi-DC Terminal PEC Converter in 5G-Based Commercial Buildings

Among various technologies being integrated into smart cities electric vehicles (EVs) and advanced communication systems emerge as critical components. They require extensive distributed infrastructure for effective operation. Given their substantial energy footprint and widespread distribution, commercial buildings offer an ideal framework. They are well-suited for accommodating these technologies. This paper introduces a DC power supply unit utilizing a Nine-Level Packed E-Cell (PEC9) converter, specifically designed for commercial buildings to efficiently support both EV charging stations and 5G communication infrastructure. A structured Interval Type-2 Fuzzy Proportional Derivative Plus Integral (IT2F-PD+I) controller has been developed to stabilize this multi-DC terminal Active Power Factor Correction (APFC) multilevel rectifier. The primary objective of the proposed controller is to maintain the supplied DC voltages of PEC9 at their nominal values under unbalanced and variable dc loads. To optimize the input/output scaling factors of the IT2F-PD+I controller and intelligently regulate DC voltages, a deep Deterministic Policy Gradient (DDPG) algorithm is employed. Two Deep Neural Networks (DNNs), namely Actor and Critic, are trained to determine the optimal policy by maximizing return signals received from PEC9 converter. Experimental implementation using DS1202 is conducted to evaluate the feasibility of the IT2F-PD+I controller-based DDPG approach in stabilizing the DC output voltages of the proposed multi-DC terminal multilevel-based power supply unit.