Research Of Artificial Immune Multi-Agent Multi-objective Optimization Algorithm And Its Application

With the develop of science and increasing demand of people, there are moreand more practical multiple targets need to optimize in the engineering design,economic management, and the natural Sciences. Nowadays, all kinds ofevolutionary algorithms which had to implement on the single objective optimizationhave been widely applied to multi-objective optimization. Some achievements havebeen acquired, but there are still many deficiencies. Artificial immune system whichsimulates the biological immune has powerful optimization mechanism. And it isconsidered the most potential intelligent calculation system in the current. However,as the complexity of the biological immune system, there are a lot of defects in theestablished artificial immune models and algorithms.Aim at the problems of multi-objective optimization algorithm and artificialimmune system. After in-depth research on artificial immune system and multi-Agentsystem, this paper puts forward an artificial immune multi-Agent system. The systemnot only has the basic characteristics of artificial immune system, but also thedistributed computing function of multi-Agent system. And the Agent in the systemupdates through clone selection operator, neighborhood competition operatorneighborhood collaboration operator and self-learning operator. It can be very goodto complete the local and global search, and have high search efficiency. Then put theartificial immune multi-Agent system in the multi-objective optimization, and theartificial immune multi-Agent multi-objective optimization algorithm (AIMAMOA)is proposed. At the same time, the algorithm combines with excellent fitnessdefinition method and external population update strategy of multi-objectiveoptimization. The simulation results show, the algorithm is effective for solving themulti-objective optimization problems, and have advantages in convergence anddistribution uniformity of Pareto front.As the shortage of resources and economic interests, chemical operationoptimization becomes particularly important. But the optimization algorithms whichused in chemical operation optimization are most based on the traditional geneticalgorithm, and there is usually just one optimization target. In this paper, AIMAMOAis applied to the multi-objective operation optimization of fractionation system. Through the AIMAMOA to search the optimal reflux ratios of fractionation system,make the minimum energy consumption and maximum product yield both reach theoptimal values, which is a pair of contradictory objectives in the fractionation system.In the optimization, the multi-objective operation optimization models offractionation system have the fractionation system Aspen Plus mechanism model andfractionation system neural network model. The experimental results show, the refluxratios that determined by this algorithm can be used to guide the operationoptimization of actual fractionation process.