Research On Imperialist Competitive Algorithms For Solving Constrained Optimization Problems

Optimization problems are frequently encountered in many scientific research and engineering technology fields.Such problems always possess complex and diverse constants,namely constrained optimization problems.Owing to the addition of constants,efficient and reasonable constraint handling strategies should be drawn into solving this kind of problems,only then the optimal solution of such problems can be accepted;if conflicting objectives are considered simultaneously,multi-objective processing is also needed.Therefore,how to apply constraint handling strategies and multi-objective processing methods are extremely important for constrained optimization problems.Constrained single optimization problems and constrained multi-objective optimization problems are deeply discussed in this thesis.Then mathematic models based on those two problems are constructed,and two imperialist competitive algorithms are proposed considering corresponding optimization strategies.Finally,advantages of constraint process and algorithms are proved by computational experiment.The main work of the thesis is listed as follow:(1)The constrained optimization problems are induced by introducing research background and significance,then the related works and analyses on constrained optimization problems are given,finally constrain handling strategies are depicted and three related intelligent optimization algorithms are described.(2)To solve constrained optimization problems,the lexicographical method is used to simultaneously optimize objective function and the degree of violation and a novel imperialist competitive algorithm is presented,In which,cost and normalized cost are redefined to guarantee the power of all imperialists exceeds zero in the procedure of constructing initial imperialists,then the global search of colonies is given in assimilation,however the local search of the best colony is applied in revolution,and the differential evolution of imperialists is added,finally a new imperialist competition is developed,all of these strategies are to improve solution quality.Many computational experiments are conducted and imperialist competitive algorithm is compared with global best artificial bee algorithm and tree and seed algorithm from literature.The results show that the novel imperialist competitive algorithm has promising advantages for solving constrained optimization problems.(3)To solve constrained multi-objective optimization problem,we propose an effective strategy for constraint handling by using the degree of constraint violation and Pareto dominance,and design a novel multi-objective imperialist competitive algorithm.We present a simple process of generating initial empires in the novel multi-objective imperialist competitive algorithm.We adopt a learning mechanism of colonies from non-dominated solutions in external archive in assimilation and use a new way to imperialist competition based on the new definition of empire power to generate the good solution.Finally,we use many computational simulations to conduct experiments,and compare the novel multi-objective imperialist competitive algorithm with NSGA-? and the modified differential evolution algorithm.The computational results indicate that the algorithm has good advantages both in convergence and the distribution of solution.