Research On Control Performance Of Marine Ammonia/Diesel Dual Fuel Engine

Marine engine carbon emission level faces increasingly stringent greenhouse gas emission regulations,so accelerating the application research on alternative fuels has become an industry consensus.As a clean fuel without carbon,ammonia is considered to have broad application prospects in Marine engines.Among them,marine ammonia/diesel dual fuel engine can reduce carbon emissions while maintaining the original power output.In order to achieve stable operation of the dual fuel engine,it is necessary to develop a control system based on its operating characteristics and carry out research on the control performance.The main research contents of this paper are as follows:In this paper,the YC6 K high pressure common rail diesel engine is taken as the structural prototype,and a zero-dimensional simulation model of an ammonia/diesel dual-fuel engine is built,which serves as the controlled object to provide a verification platform for the control strategy and algorithm.The simulation model consists of a high pressure common rail system,a gas supply system,a cylinder system,a dynamics system and an intake and exhaust system.After the modeling is completed,the fuel injection quantity at 25 operating points in diesel mode is checked,and the average error is 3.80%,the maximum error less than 10%.The heat release law under dual-fuel mode agrees with the existing research,and the accuracy of the model meets the control requirements.Secondly,the control software requirements for the dual-fuel engine are analyzed,and the overall control strategy framework of dual-fuel engine with closed-loop speed as the core is built by using the model-based development mode.The control strategy model comprises signal processing,running state management,rail pressure control and speed closed-loop control modules.In order to overcome the drawback of the PID control parameter adjustment relies heavily on the operator’s experience in the closed-loop speed control,reinforcement learning is applied to the PID parameter self-tuning process.Deep Deterministic Gradiant Policy(DDPG)algorithm is used to train the learning process of control parameters,and the control effect of the agent is verified in the simulation environment.The simulation results show that the DDPG-based control algorithm can track the acceleration and deceleration of target speed effectively.Finally,in order to verify the actual control effect of reinforcement learning algorithm,a hardware-in-the-loop simulation platform based on NI equipment is built.The hardware-in-the-loop simulation platform is used to compare the control effects of DDPG algorithm and PID algorithm.The results show that the DDPG algorithm reduces the overshoot and settling time more efficiently in transient control than the PID algorithm.In steady-state control,the DDPG algorithm also reduces the speed fluctuation,which indicates that the control algorithm based on reinforcement learning has better control performance and robustness than the fixed parameter PID.