Transient fluctuation prediction and control of tracked vehicle electric drive engine-generator set based on LSSVM

To address the challenges of delayed control responses and suboptimal performance due to the absence of predictive capabilities for pre-power chain speed fluctuations in the electromechanical composite transmission system of armored vehicles, a transient fluctuation prediction and control method based on the Least Squares Support Vector Machine (LSSVM) is proposed for the engine-generator set within the system. This approach leverages real-world generator data collected from actual vehicles as the training dataset to establish a data-driven model. A specific LSSVM training model is developed, with experimental data serving as the test set. Within the model's predictive framework, transient fluctuations of critical engine-generator parameters are generated in real-time under test conditions. Simulations are conducted on a test platform for the electromechanical composite transmission system, evaluating both single-generator operation and a variety of driving conditions. Comparative analysis is performed to assess the operational factors influencing system performance under single and multiple conditions, as well as the control effects of transient power chain fluctuation prediction. Under multiple-condition scenarios, the system demonstrates faster dynamic recovery in response to significant load disturbances, with voltage peak fluctuations remaining within 5 %, which meets engineering application standards. This validates the model's adaptability and generalization capability for broader use cases.