Abstract
This paper presents an adaptive implementation of model predictive control (MPC) for a variable speed diesel generator operated as a back-up energy source in an ac ship power system. The variable-speed operation is based on a diesel motor driving a synchronous machine (SM) with a diode rectifier and a boost converter as the interface to the dc-link of a voltage source converter (VSC) connected to the ac bus. The VSC is operated as a Virtual Synchronous Machine (VSM) for ensuring flexibility in supporting islanded operation of the ship power system at low load. The MPC strategy is designed for controlling the diesel generator torque and the excitation of the SM, and for regulating the dc-link voltage by providing a current reference for the boost converter. Adaptive operation of the MPC implementation is introduced by using a linearized prediction model updated at the operating conditions of each time-step. Simulation results demonstrate how the proposed implementation can ensure a more robust performance and a wider range of stability in response to large load variations in the ac bus than a conventional approach based on independent PI-controllers.This paper presents an adaptive implementation of model predictive control (MPC) for a variable speed diesel generator operated as a back-up energy source in an ac ship power system. The variable-speed operation is based on a diesel motor driving a synchronous machine (SM) with a diode rectifier and a boost converter as the interface to the dc-link of a voltage source converter (VSC) connected to the ac bus. The VSC is operated as a Virtual Synchronous Machine (VSM) for ensuring flexibility in supporting islanded operation of the ship power system at low load. The MPC strategy is designed for controlling the diesel generator torque and the excitation of the SM, and for regulating the dc-link voltage by providing a current reference for the boost converter. Adaptive operation of the MPC implementation is introduced by using a linearized prediction model updated at the operating conditions of each time-step. Simulation results demonstrate how the proposed implementation can ensure a more robust performance and a wider range of stability in response to large load variations in the ac bus than a conventional approach based on independent PI-controllers.