Research Based On Adversarial Decomposition And Neighborhood Evolution For Dynamic Multi-objective Particle Swarm Optimization Algorithm

Multi-objective optimization problems(MOPs)have the characteristics of simultaneously optimizing multiple conflicting targets.This type of problem is widely used in practical applications such as industrial dispatching,power systems,controller design,etc.Multi-objective evolutionary algorithms and multi-objective particle swarm optimization algorithms are often applied to solve such problems.They perform heuristic searches in the decision space to find a set of non-dominated solutions to meet the needs of decision-makers.In recent years,a special type of multi-objective optimization problem has become a research hotspot and is widely used in the engineering field.In addition to the characteristics of conflicting objectives,it also has the characteristics of objective functions,constraints,and parameters changing over time.They are called dynamic multi-objective optimization problems(DMOPs).Compared with static multi-objective problems,DMOPs have higher requirements on algorithm design because of their time variability.To effectively solve DMOP,the following two points are generally considered:(1)When the environment changes,the algorithm should provide an effective dynamic response mechanism to track the latest optimal solution and adapt to the new environment.(2)Between two environmental changes,the environment is equivalent to a static optimization problem.Therefore,the algorithm should ensure that the current Pareto solution set(PS)can be quickly found in less algebra.This fast convergence ability is important for tracking performance.It is also crucial.Based on the above two points,this paper proposes a dynamic multi-objective particle swarm algorithm(ADNEPSO)based on adversarial decomposition and neighborhood evolution.The proposed algorithm mainly has the following three contributions: First,by combining adversarial decomposition and particle swarm optimization,a novel dynamic multi-objective optimization framework is proposed.By maintaining the alternating update and information interaction of two opposing populations,The purpose of maintaining population diversity and adapting to more complex Pareto Optimum Faces(PF).Secondly,a novel neighborhood evolutionary particle swarm update strategy is proposed.Each vector is assigned a globally optimal particle.The neighborhood vector and the current vector are selected by the priority selection strategy as the parent guide particle evolution trend.Finally,this paper proposes a hybrid forecasting strategy based on particle swarm and population center points.Through population center point prediction,global optimal information development,and individual retention operations,historical information is used to effectively achieve rapid response and diversity to the environment.And the generation of astringent individuals.The algorithm was tested on a series of benchmark problems and compared with several recent algorithms.The results show that ADNEPSO has an excellent performance in terms of convergence and distribution,and has strong competitiveness in dealing with dynamic problems.