Modeling of diesel generators in DC microgrids is fundamental for simulating large-scale power systems, but the difficulty in obtaining electromagnetic parameters of diesel generators hinders the accuracy of electrical characteristic modeling. To address the modeling issue of diesel generators in DC microgrids, a data-driven equivalent modeling approach is proposed. By taking the voltage, current, and output speed of the diesel generator after AC-DC-DC conversion as inputs, and the effective values of phase voltage and phase current of the diesel generator as outputs, a diesel generator voltage prediction model and a current prediction model based on Long Short-Term Memory (LSTM) neural networks are constructed. The temporal logic and internal non-linear mapping properties of LSTM neural networks are utilized to describe the electrical characteristics of the diesel generator output. By comparing with actual operational data of diesel generators in DC microgrids, the rationality and accuracy of the modeling approach are validated, demonstrating the practical value of the proposed method.