| Calibration technology is an important means to reduce emissions and improve economy of engines,and its essence is a process of seeking the optimal combination of control parameters at each operating point of the engine.However,there is an irreconcilable conflict between engine emission targets and fuel consumption,which often requires multi-objective optimization.The current research status of multi-objective performance optimization calibration technology is mainly focused on off-line optimization calibration,which cannot respond to external environmental factors and internal device consumption once the calibration is completed.This paper combines model-based calibration method and online calibration technology to carry out online multi-objective performance optimization calibration of marine 6K micro-injected diesel/natural gas dual-fuel engine to control the engine in good operating condition for a long time.In order to shorten the calibration period and reduce the experimental cost,firstly,the model-based offline calibration technique is adopted to initially optimize the performance of the dual-fuel engine in the full range of operating conditions.It mainly includes: determining the optimization targets,and the control parameters need to be optimized and the boundary range.A hybrid experimental design scheme combining space-filling and V-optimal experimental design is used to reduce the experimental cost.Establishing a dual-fuel engine performance prediction model based on an LSTM neural network that can effectively remember the key information in the time-series data.Based on the NSGA-Ⅱ algorithm to find the optimal combination of control parameters for each operating condition.Combined with the demand of online fast optimization of dual-fuel engines,the original time-consuming non-dominated ranking method is replaced by the simple and efficient dominance matrix ranking method.And a multi-objective optimization strategy based on the improved NSGA-Ⅱ optimization algorithm coupled with LSTM neural network is designed for more efficient multi-objective performance optimization of the dual-fuel engine.In order to improve the performance optimization of the full range of operating conditions of the dual-fuel engine,online optimization is carried out for the operating conditions points with unsatisfactory offline optimization results.It mainly includes: the sliding time window strategy is introduced on the basis of LSTM predictive modeling,and the data of online modeling is continuously updated to more accurately predict the performance status of the actual operation of the dual-fuel engine;Based on the MLIB/MTRACE function to calibrate the optimization results into the RCP control system without stopping the engine to complete the online calibration.Based on the proposed multi-objective optimization strategy,an online multi-objective optimization platform was built.Firstly,the optimization of the dual-fuel engine was verified offline at full working conditions,and the experimental results showed that the NOx emission and fuel consumption rate were improved compared with the original engine,and the offline calibrated MAP could improve the engine performance overall according to the optimization requirements.Then online optimization was carried out for five operating points with unsatisfactory offline optimization results.The experimental results showed that the emission of NOx decreased by an average of 20.186% and the consumption of BSFC increased by an average of 2.738% after online optimization,and the calculation speed of online multi-objective optimization was feasible.After experimental verification,the online multi-objective optimization strategy not only retains the advantages of the model-based calibration method with low calculation amount and high calibration efficiency,but also combines the online optimization calibration technology to solve the MAP offset problem. |
PDF Full Text Request