Research on output power modeling of all-electric marine diesel generator set based on data-driven

• Proposed a data-driven modeling method for diesel generator sets based on information entropy and Elastic Net regression. • Validated the algorithm using actual operational data collected during vessel operation. • The developed model reveals the underlying correlation mechanisms among the operational parameters of the diesel generator set. Multi-energy integrated all-electric vessels, centred on generator sets and lithium-ion battery packs, represent a pivotal direction for the development of green intelligent vessels. Their intelligent operation and maintenance heavily rely on digital twin technology. As the core power source, shipboard diesel generator sets exhibit strong multi-physics coupling characteristics, making the construction of mechanism-based models challenging and difficult to simultaneously meet the accuracy and real-time requirements of digital twin systems. Considering that most operational conditions during vessel navigation are steady-state or quasi-steady-state, high-precision steady-state models provide a reliable foundation for conducting energy efficiency analysis, load distribution optimisation, and performance degradation assessment. To this end, this paper proposes a data-driven steady-state modelling method for diesel generators. This approach uses prime mover operational data as input and generator power as output, employing information entropy for feature selection and an elastic net regression algorithm for model construction. A dataset derived from actual vessel operational data was used for model training and validation. After normalisation, the model achieved a Mean Absolute Error (MAE) of 0.0110, a Root Mean Square Error (RMSE) of 0.0162, and a Coefficient of Determination (R²) of 0.9974. This model reveals underlying correlation mechanisms among unit characteristics, providing a critical steady-state simulation module for the digital twin system of all-electric vessels. It also lays a robust foundation for subsequent energy efficiency optimisation and condition monitoring efforts.