Research And Application Of Many-objective Evolutionry Algorithm Based On Preference Vector Guidance

Many-objective optimization problems is the hotspot and difficulty in multi-objective optimization field,which is widely used in practical engineering,mechanical engineering and communication engineering,and has important research value.With the increase of the target dimension,the defects of traditional multi-objective optimization algorithms appear gradually: firstly,the individual selection pressure decreases sharply;secondly,population size shows exponential growth;thirdly,the performance of the algorithm is difficult to balance;fourthly,the uniformity of the solution set on the complex Pareto front is insufficient.To solve the above difficulties,the relevant scholars have considered to use the preference information of decision-makers to deal with Many-objective optimization problems.In the actual situation,what the decision maker is really interested in is not the whole Pareto front,but a local region based on the preference information.It is effective to fouse the attention of searching non-dominated solution sets on the preference region,through preference information given by decision makers,which can not only focus on searching resources,but also help decision makers select satisfactory programmes from a large number of solutions.This paper proposed the preference vector generation strategy based on the preference guidance.The uniformly distributed weight vectors in the target space are proportionally scaled to generate the preference vector,through the given reference point g and radius parameter of preference region ? given by decision makers,to reduce influence of the reference point position to algorithm convergence.At the same time,an angle punishment distance mechanism based elite selection strategy is proposed to allocate computational resource more adaptively,which aiming at balancing the convergence and diversity.Then,this paper proposed a weight vectoradjustment strategy based on the micro decomposition.In the proposed method,the approximated Pareto front is simulated using current population shape,and the distribution of weight vectors are adjusted based on the cooperative relationship between population and weight vectors,which improves the distribution uniformity of the solution set on the complex Pareto front.Finally,the proposed algorithm is applied to the problem of green supply chain partner selection,to provide as best as possible for decision makers to choose suitable supply chain partners,and to provide new ideas and methods for solving multi-constrained and nonlinear Many-objective optimization problems with real logistics background.