Machine Learning Approach for Mode Selection and Control of Multiport DC-DC Converters In PV-Battery Applications

This paper analyzes the performance of a multi-port DC-DC converter in hybrid Renewable Energy Source (RES) applications. In the hybrid source integrations, the challenging role for the controller is to choose the appropriate source based on its availability; thus, performing a promising job for the energy management system. In that way, the mode selection still remains a complicated concept for the design engineers due to the fact that involves in detailed learning and prediction algorithm. This paper is one among the kind in designing a Machine Learning (ML) algorithm for mode selection and control of hybrid RES system. In this article, an extendable DC-DC converter is considered with multi-port capability on the source side with battery back-up. The operation of the converter in different modes will be decided by the ML algorithm. The performance analysis of the proposed technique is conducted in Hardware-In-Loop fashion, where the ML logic is developed in Matlab and integrated with experimental hardware using dSPACE-MicroLabBox. The mode selection and operational features with the proposed technique are validated with simulation and experimental results as well.