Research on the Optimization of Dual-Fuel Engines Based on the Non-Dominated Sorting Whale Optimization Algorithm

To address the complex calibration parameters and low optimization efficiency of dual-fuel engines, this paper innovatively proposes an optimization calibration method based on a simulation model and the Non-Dominated Sorting Whale Optimization Algorithm (NSWOA). Taking the YC6K dual-fuel engine as the research object, a high-precision simulation model was constructed within the GT-Power environment, and its reliability was confirmed through the external characteristic curve (the maximum deviation of torque and specific fuel consumption rate is less than 5%). A total of 260 parameter samples were generated using a Sobol sequence space-filling experimental design, and a performance prediction model was established by combining the Crested Porcupine Optimization algorithm and the Back-Propagation Neural Network (CPO-BP). The experimental results show that the CPO-BP model exhibits excellent predictive capability, with the coefficient of determination (R2) of nitrogen oxides (NOx) and brake-specific fuel consumption rate (BSFC) reaching 0.98964 and 0.99501 respectively. Based on this, the NSWOA algorithm was introduced to optimize key parameters such as speed, torque, main injection timing, and rail pressure, with the optimization objectives being NOx emissions and BSFC. The optimization results show that under 100% load conditions, the reduction in BSFC ranges from 1.5% to 4.3%, and NOx emissions are reduced by 48.6% to 67.1%. The effectiveness of the optimized parameters was also verified through bench tests, providing an efficient solution for complex engineering optimization problems.