Research On Multi-objective Integrated Optimal Control Of Dual-Fuel Engine Based On Preference Decision-driven

Natural gas is considered an ideal alternative fuel for today’s marine engines due to its advantages of abundant resources and clean combustion.However,the diesel-natural gas micro-ignition dual-fuel engine poses a greater challenge to the multi-objective performance optimization control of the engine due to its complex system and many control parameters.In particular,the complex relationship between the interactions and coupling of the control parameters brings great difficulties to the optimization of dual-fuel engine performance.At present,the intrinsic reasons for the fluctuating output performance of dual-fuel engines due to the interaction of multiple control parameters are not clear at home and abroad,especially in the study of multi-parameter matching to cope with the changes in complex operating conditions.Therefore,it is necessary to adopt a research method combining experimental tests and theoretical analysis to reveal the mechanism of the coupling effect of control parameters on engine performance,extract key control parameters,and carry out reasonable matching and optimization to realize comprehensive and optimal control of dual-fuel engines.In this paper,the research work is carried out from various aspects such as experimental research,theoretical analysis,system design,and optimization.Combining space-filling design and V-optimized design methods,the experimental condition points are reasonably designed,and the influence law of multi-control parameters on engine economy and emission under different combustion modes are mastered through dual-fuel engine bench tests,which provides accurate experimental boundary conditions for DOE experimental design.Using limited experimental data,a dual-fuel engine performance prediction model is established to describe the complex relationship between numerous control parameters and output performance.We explored the advantages and disadvantages of modeling methods,such as response surface,RBF neural network,SVM,and LSTM neural network,and compared and analyzed the prediction effect of each prediction model.Among them,the performance prediction model established based on LSTM realizes the perfect mapping between the calibration control parameters and the output performance response.The determination coefficient of the prediction model is R~2>0.99,and the average relative error MSRE<0.03,shows good prediction performance.To get a compromise between the best fuel economy and lowest emissions for dual-fuel engines,and to find the best combination of complex control parameters,must be solve a multi-objective optimization problem.It combined the LSTM neural network prediction model with the multi-objective optimization algorithm to construct a multi-objective optimization architecture.The optimization effects of the PSO and NSGA-II algorithms in the full range of operating conditions are studied respectively,and it is found that the optimization results of the PSO algorithm depend on the artificial experience.The target weight has large randomness.In many cases,when the compromise between emission and economy is achieved,the NOx emission is reduced by 20.5%compared with the original machine,which meets the requirements of IMO Tier-II regulations.The BSFC is reduced by 2.1%compared with the original machine.The experimental results of the NSGA-Ⅱalgorithm in the full range of operating conditions show that 77%reduce the NOx emission by compared with the original machine without increasing the fuel consumption.Through the marine four-condition cycle test,we found that the NOx emission is 26.6%lower than the Tier-Ⅲlimit requirement.In order to respond to the different operational requirements of decision-makers,preference information was introduced into the NSGA-Ⅱalgorithm to guide the population search,and the influence of preference parameters on the optimization results was investigated to reduce the dependence of preference parameter settings on the empirical knowledge of decision-makers by setting up an interactive optimization scheme.The experimental results show that when low emission is preferred,all optimal solutions can meet the NOx limitation requirement of IMO Tier-Ⅲ,and the average NOx emission in the full working range is 1.22g/(k W·h),which is 78.9%lower than the original machine,while the BSFC is 4.76%lower than the original machine;when low BSFC is preferred,the average BSFC is 8.44%lower than the original machine,which is lower than the When low NOx emission is preferred,the average BSFC is reduced by 8.44%compared to the original machine and 3.68%compared to the original machine,however,at this time,NOx emission is increased by 48%compared to the original machine,which shows that the lower fuel economy is at the expense of the environment,and the pursuit of the higher economy will worsen the emission drastically.The optimal solution of the preference multi-objective optimization is the optimal control parameter combination that satisfies the preference of the decision-maker.In order to ensure that it executed the got control parameters,an online optimization calibration strategy and a comprehensive optimization control system are designed,and the model reference adaptive control is studied.The effect of the algorithm on the control of rotational speed,rail pressure,and air-fuel ratio is tested,and the results show that the adaptive control strategy has better robustness and dynamic tracking performance.In the closed-loop speed control system test,the steady-state speed fluctuation rate under rated load meets the primary speed regulation index;the transient regulation rate during sudden load addition and transient regulation rate during sudden load removal meet the secondary speed regulation index;the stabilization time meets the secondary speed regulation index.In addition,the online optimization calibration function was experimentally verified on the diesel engine,and the results showed that the online optimization calibration strategy can adjust the engine operating status in time without excessive fluctuations,and the online optimization calibration is safe and reliable,and can quickly modify the engine control parameters without shutdown.