Research On Many-Objective Particle Swarm Optimization Algorithm Based On Preference

Multi-objective optimization problems always exist in practical engineering and life,such as controller design,water resources management,portfolio planning and many other optimization problems.When the objectives are more than three,the multiobjective optimization problem is called the many-objective optimization problem.The traditional optimization algorithms are not suitable for solving many-objective optimization problems due to their high computational complexity,low selection pressure and inconvenient decision making.Since the decision makers often have subjective requirements for each objective before optimization,the search process of other uninterested region will waste the computational resources.Therefore,the research on the multiobjective optimization algorithm with preference has been paid more and more attention.In order to tackle optimal difficulties of many-objective optimization problems,a new many-objective particle swarm optimization algorithm IR2-MaPSO is proposed.In this algorithm,an archive pruning strategy based on R2 indicator and objective space decomposition are proposed,and the pruning strategy is incorporated into the selection method of particle swarm optimization to solve many-objective optimization problems.There are three core parts of IR2-MaPSO:A bi-level archiving maintenance approach based on R2 indicator and objective space decomposition strategy is designed to balance convergence and diversity;Global best leader with good convergence performance,objective space decomposition leader with good distribution and personal best leaders are selected in archive,and a new velocity updated method is designed to guide the particle swarm;In order to enhance the search efficiency of the algorithm,simulated binary crossover(SBX)and polynomial.mutation(PM)operator are adopted.Aiming at the difficulties of constructing decision maker's preference model for many-objective optimization problems,the decision makers' different decision-making needs are analyzed,and the decision maker's preference framework is constructed.The framework consists of two parts:on the one hand a new weights generation strategy based on direction angle is designed,which can generate an arbitrary number of weights;on the other hand,the coordinate transformation method generating new preference in the region is used to construct corresponding preference model,with the new generation of preference weights method intergrates preference information into IR2-MaPSO,a many-objective particle swarm optimization algorithm based on preference called IR2-MaPSO-PRE is proposed to solve the multi-objective optimization problem with preference.The proposed algorithms are experimented on benchmark test functions,compared with a variety of particle swarm algorithms,evolutionary algorithms and preference multi-objective optimization algorithms,to verify the validity and rationality of algorithms.