摘要
使化工过程同时达到经济效益最佳和废料最少是可持续发展的必然要求,此问题属于多目标优化,求解可得到一个非劣解集。应用多目标遗传算法(MOGA)求解多目标问题时,由于不能按一定精度要求收敛,求解得到的是近似非劣解集,遗传参数的取值对非劣解质量有一定影响。本文采用NSGA-Ⅱ算法对废气中丙酮回收过程的废料最少化问题进行研究,分别改变变异概率、种群规模、遗传代数和交叉概率等遗传参数的取值,得到相应的非劣解集。通过比较非劣解曲线的平滑性、非劣解集的劣解数量和端点取值,考察各个遗传参数对求解质量的影响规律。结果表明,NSGA-Ⅱ算法的求解效率较高,遗传代数在一定范围内取值对非劣解影响不大,推荐取值60代;变异概率P_m取值较小时非劣解曲线平滑性不好,劣解数量也较多,推荐P_m取值为0.3;随着种群规模增大求解质量明显改善,推荐取值80以上;交叉概率P_c增加,求解质量提高,但分布趋向不均匀,推荐P_c取值为0.8。本文推荐的遗传参数取值可为同类问题的求解提供参考。
Maximizing economic benefit and minimizing waste simultaneously for a chemical process are the request of sustainable development. In order to meet the request, a multi-objective problem need to be solved and a non-inferior solution set is obtained. Using multi-objective genetic algorithms （MOGA） to solve the problem can only achieve an approximate non-inferior solution set which quality is influenced to some extent by genetic parameters such as mutation rate, population scale, number of generations and crossover rate. How to judge the quality of solution set is also a multi-objective problem. In this paper NSGA- II is used to solve the waste minimization problem of acetone recovering process. The value of mutation rate, population scale, number of generations and crossover rate is varied respectively and achieve a series of non-inferior solution sets. The influence of genetic parameters is compared with three criteria which are the smoothness of non-inferior curve, the number of inferior solutions and the value of curve end. The results show that the solution efficiency of NSGA- II is high. The number of generations in some range, the recommended value in this paper is 60, has a little influence on non-inferior solution set. When mutation rate P~ is small（such as less than 0.1）, the curve of non-interior solution is less smooth and the number of interior solution is larger, so the recommended Pm is 0.3. Increasing population scale can improve the quality of solution set and the recommended value is 80. Increasing crossover rate Pc can also improve the quality of solution set, but the distribution of points on curve tend to uneven, so the recommended Pc value is 0.8. The recommended genetic parameters in this paper could provide reference for other similar paroblems.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2011年第12期1557-1560,共4页
Computers and Applied Chemistry
关键词
NSGA—Ⅱ
丙酮回收
非劣解
遗传参数
NSGA- II, acetone recovering, non-inferior solution, genetic parameters
