Load-Sharing Optimization of Parallel Marine Diesel Generators Using Differential Evolution Algorithm

Efficient load sharing among parallel marine diesel generators enables the minimization of fuel consumption and improves the operational efficiency of shipboard power systems under variable loading conditions. This study presents a Differential Evolution (DE)-based optimization framework for load distribution in a marine AC power system consisting of six MAN 8L21/31 diesel generators while operating in parallel. Polynomial regression models of specific fuel consumption were developed using manufacturer performance data and testbed measurements, enabling accurate fuel-consumption prediction over the entire operating range. Two operational strategies were investigated: continuous manual operation with six generators online and automatic generator activation controlled by the Power Management System (PMS) according to predefined loading thresholds. The optimization objective was to determine the optimal generator load distribution that minimizes total fuel consumption while satisfying operational constraints. The results show that asymmetric load sharing can provide measurable fuel savings compared with conventional equal load distribution. In manual-operation mode, the proposed optimization achieved fuel-consumption reductions, particularly within certain system load ranges. Under PMS-controlled automatic activation, the optimization produced segmented operating regions with peak fuel savings exceeding relative reductions at medium load levels. The optimized operating strategy consistently shifted generator loading toward higher system efficiency operating regions and reduced inefficient system load operation. The results obtained from this research confirm the effectiveness of evolutionary optimization methods for improving the energy efficiency of marine diesel-electric power systems. The proposed DE-based optimization framework represents a practical approach for implementation in advanced shipboard power management systems and provides a basis for further multi-objective optimization.