With the increasing maturity of digital twin technology, the traditional simulation model has a long running time and poor interpretability of simulation model parameters, which makes it difficult to develop into a digital twin simulation model. Aiming at this problem, this paper proposes an agent simulation model of marine diesel engine with high simulation efficiency and high accuracy. In this study, the traditional mean value theory of diesel engine is improved, and the performance parameter update factor is introduced to improve the interpretability of the simulation model. Using the new factor of performance parameters combined with the existing high-precision diesel engine simulation model, the response value of diesel engine monitoring parameters corresponding to the performance parameter update factor is obtained. The Kriging surrogate model is constructed by using the obtained high-precision data set training, and the trained Kriging model is used as the lightweight simulation model of marine diesel engine. The experimental results show that the simulation results of the marine diesel engine surrogate simulation model are more accurate, and the evaluation index R2 of the training set and the test set is close to 1. The simulation results are more reliable and can be used as the basic simulation model of digital twin.