Abstract: In order to fully play the excellent physicochemical characteristics of F-T coal‑to‑oil  and improve the trade‑off relationship between SOOT and NOX of diesel engineeffectively，a four‑cylinder diesel engine was taken as the research object, and SOOT and NOX was taken as the optimization target. The test bench was used to obtain performance data with the variation of injection parameters. According to these test samples, a numerical simulation model for SOOT and NOX prediction under different injection parameters was established by SVM, which was used as fitness function of NSGA-Ⅱ(Non‑dominated Sorting Genetic Algorithms-Ⅱ), and then the Pareto solutions of SOOT and NOX were found in the sample spaces. On this basis, SOOT and NOX targets were endowed with weights by means of Topsis analysis method, and then the principles of Pareto decisions were described. As a matter of fact, Pareto solutions decided by Topsis method showed a marked drop compared to the original points, in which SOOT declined 3.7%-7.1% and NOX declined 1.2%-2.6%. At last, the validity of the results was demonstrated by the experiments, which showed that the  maximum relative error of SOOT simulations was 8%, while it was 5% for NOX simulations, and R2 under various operating conditions were more than 0.95. In this paper, analysis shows that the methods make a significant improvement in the trade‑off relationship between SOOT and NOX emissions, which can improve emission performance of direct injection diesel engine.

Key words: F-T coal?to?liquid, support vector machine, genetic algorithm, emission performances, multi?objective optimization

摘要： 为充分发挥F-T煤制油的优良理化特性，改善柴油机SOOT和NOX之间的“Trade‑off”关系，以一台四缸柴油机为研究对象，将SOOT和NOX作为优化目标。通过台架试验获得柴油机燃用F-T煤制油时SOOT和NOX随喷油参数变化的性能试验数据，采用支持向量机SVM建立喷油参数与SOOT、NOX间的预测模型，并将其作为带精英策略的非支配排序遗传算法的适应度评价模型，在样本空间内获得SOOT和NOX的Pareto前沿。赋予SOOT和NOX不同的权重，通过Topsis对Pareto前沿进行决策分析，与原机工况点相比，SOOT下降幅度为3.7%~7.1%，NOX下降幅度为1.2%~2.6%。经试验验证，SOOT的Pareto前沿与试验验证值之间的相对误差不超过8%，NOX不超过5%，各工况下的R2均大于0.95。该方法改善了柴油机SOOT和NOX之间的“Trade‑off”关系，使柴油机排放性能得到提高。

关键词: F-T煤制油, 支持向量机, 遗传算法, 排放性能, 多目标优化