Study On Particle Swarm Optimization Algorithm For Multiple Application Scenarios

At present,a large number of optimization problems in practical application scenarios are NP-hard problems,which have characteristics of nonlinearity,multicriteria,discretization and uncertainty.Some features of optimization problems,such as the descriptions and characteristics,the concerned indicators and requirements,or the characteristics of optimization objectives,vary in different scenarios,which make it hard to obtain a general and standard swarm intelligence optimization algorithm to solve these problems.Thus,it is one of the feasible methods to develop efficient algorithms to solve these problems by combining the swarm intelligence optimization algorithm and the features of optimization problem in different scenarios currently.For PMSM parameter identification problem,dynamic economic load dispatch problem,dynamic economic environmental dispatch problem and flexible job shop scheduling in different application scenarios,particle swarm optimization algorithms(PSO)with different strategies are studied to solve these typical nonlinear,multi-objective,discrete optimization problems,which can improve the solving accuracy and efficiency,and can be applied to solve similar applications.The main work is as follows.(1)Aiming at PMSM parameter identification problem,a hybrid mutation PSO based on adaptive mutation probability(AMPSO)is proposed,which addressed the drawback of search stagnation caused by the decrease of population diversity in swarm intelligence optimization algorithm.In AMPSO,aggregation degree based on phenotype diversity is defined to dynamically adjust the mutation probability.Then a hybrid mutation strategy is introduced,which adopts the mechanism of hybrid mutation of Gaussian distribution and Cauchy distribution based on the global-best position and wavelet mutation based on the worst personal-best position.Simulation results show that AMPSO algorithm performs well in solving test functions,and the hybrid mutation strategy can effectively solve the application scenario of PMSM parameter identification,and has good identification accuracy.(2)Dynamic economic load dispatch problem is a complex optimization problem with high dimension,nonlinearity and nonconvexity.A PSO based on multi-information characteristics of all personal-best positions(PSO-API)is proposed by perceiving and feeding back the fitness distribution information of personal-best positions in the search space,which effectively solved the error convergence of the algorithm caused by insufficient information in the search.Centroid position,median position and personal-best position,which make use of multi-information personal-best positions,are applied to form personal cognitive experience,which is used to feed back and adjust particle velocity update.The feedback pattern of fitness distribution information in PSO-API algorithm shows excellent performance in solving test functions and multi-fuel dynamic economic load dispatch problem.(3)Dynamic economic environmental dispatch problem has characteristics of complex nonlinearity and multiple optimizing criteria.Referring to the idea of non-dominated relationship and uniform decomposition,a hybrid multi-objective PSO(MultiPSO)is proposed and multiple hybrid strategies are adopt to ensure all the convergence,diversity and uniformity of Pareto front.In MultiPSO algorithm,personal-best set and global-best set are defined to deal with the non-dominated solutions of multi-objective optimization problems,and a hybrid update mechanism based on Pareto and uniform decomposition is introduced.Then,variable neighborhood search is carrying on the global-best set to improve local search ability.Simulation results show that the approximate Pareto front obtained by MultiPSO algorithm has better convergence accuracy as multi-objective test functions and multi-fuel dynamic economic environmental dispatch problem are solved.(4)Flexible job shop problem is a NP-hard discrete optimization problem,which has characteristics of high dimensional discrete variables,high nonlinearity and multiple optimizing criteria.Due to the high relativity between discrete optimization problem and discrete operator,for the application scenarios of certainty or uncertainty flexible job shop scheduling,a novel position update framework of discrete PSO and two kinds of discrete PSO,MOPSO and EDPSO algorithm are proposed.For multi-objective flexible job shop scheduling problem,discrete operators based on flexible job shop scheduling are introduced.Then personal-best set and global-best set are defined and a non-dominated archive update strategy is designed.Variable neighborhood search based on public critical block is applied to improve local search ability.Aiming at the uncertain situation,fuzzy processing time is treated as triangular fuzzy number.Then a bybrid heuristic initialization scheme based on triangular fuzzy number is proposed and simple and effective discrete operators are defined in EDPSO.Finally,selective mechanism of multiple optimal positions is adopted to select global-best position.Simulation results show that the two algorithms achieve good performance in solving multi-objective flexible job shop scheduling problem and flexible job shop scheduling problem with fuzzy processing time respectively.