Constrained Multi-objective To Improve Particle Swarm Optimization Algorithm And Its Application

Lots of constrained multi-objective optimization problems exit in the practical engineering. Traditional gradient-based strategies are powerless for discontinuous, nondifferentiable and inexplicit functions. Being an intelligent evolutionary algorithm, particle swarm optimization (PSO) guides the optimal search by the cooperation and competition among the swarm population. Meanwhile, it adjusts searching strategies by dynamically tracing searching condition with its specific memory ability. Due to independent on self information of problem and powerful global optimization, particle swarm optimization (PSO) becomes a hot issue of optimum fields.The paper is to find an improved strategy for solving constrained multi-objective optimization problems, which is applied to the energy consumption for electrolytic zinc process. Based on the intensive study of the problems containing local optimum and slow late convergence, a new hybrid tabu search-particle swarm optimization algorithm (HTS-PSO) is obtained. On one hand, it prevents similarity of algorithm with choosing global extremum guided by long tabu table; On the other hand, it jump out of local extremum by adapting tabu and search mutation based on diversity of population. Convergence ratio and success ration of PSO for solving high dimensional single objective optimum problem are effectively improved by combining strong climbing and neighborhood searching ability of tabu search.Designing a reasonable constraint processed approach is the core of solving constrained-multi objective optimization problems. A constraint handling techniques which adopts constraint handling technology based on distance measures and adaptive penalty functions is proposed in this paper. According to these two values, the objective space is modified for particle swarm optimization evolution operation. The proportion of feasible solutions is used to keep the balance between objective function and constraints and improve the boundary searching capability of the algorithm. It uses an external archive to preserve the Pareto solutions that find so far and presents a new k nearest neighbors crowding density method to maintain the diversity of solutions. At the same time, it selects the global optimal particle through the combination of crowding density and roulette selection.The experiment results of simulation demonstrates that the algorithm developed is efficient for solving constrained multi-objective optimization problems.Finally, the proposed algorithms are applied to solve the energy consumption multi-objective optimization model of EZP. Based on TOPSIS, the satisfactory solution is obtained by using decision making method on multiple attribute, and the energy consumption optimization of EZP is implemented.