系统工程与电子技术
繫統工程與電子技術
계통공정여전자기술
Systems Engineering and Electronics
2015年
10期
2404-2410
,共7页
混合多目标粒子群%局部理想点%约束二次逼近优化%拥挤距离%领导粒子
混閤多目標粒子群%跼部理想點%約束二次逼近優化%擁擠距離%領導粒子
혼합다목표입자군%국부이상점%약속이차핍근우화%옹제거리%령도입자
hybrid multi-objective particle swarm optimization%local optimal particles%bound optimization by quadratic approximation (BOBYQA)%crowding distance%leader particles
提出了一种结合约束二次逼近优化（bound optimization by quadratic approximation，BOBYQA）搜索算法的理想点法对非支配解进行局部优化的混合多目标粒子群方法（local search with multiobjective particle swarm optimization，LSMOPSO），以提高多目标粒子群算法的收敛性能和非支配解集的精度与多样性。LSMOPSO 算法使用拥挤距离选择领导粒子组成领导粒子集，并对其进行理想点局部搜索；分析比较了全局理想点和局部理想点对算法性能的影响，提出基于局部理想点的局部搜索策略；在粒子的设计空间的多个维度上引入均匀变异操作，降低算法陷入局部最优的可能。基本测试函数的求解结果表明，算法的收敛速度很快，而且搜索到的非支配解集的精度高、多样性好。
提齣瞭一種結閤約束二次逼近優化（bound optimization by quadratic approximation，BOBYQA）搜索算法的理想點法對非支配解進行跼部優化的混閤多目標粒子群方法（local search with multiobjective particle swarm optimization，LSMOPSO），以提高多目標粒子群算法的收斂性能和非支配解集的精度與多樣性。LSMOPSO 算法使用擁擠距離選擇領導粒子組成領導粒子集，併對其進行理想點跼部搜索；分析比較瞭全跼理想點和跼部理想點對算法性能的影響，提齣基于跼部理想點的跼部搜索策略；在粒子的設計空間的多箇維度上引入均勻變異操作，降低算法陷入跼部最優的可能。基本測試函數的求解結果錶明，算法的收斂速度很快，而且搜索到的非支配解集的精度高、多樣性好。
제출료일충결합약속이차핍근우화（bound optimization by quadratic approximation，BOBYQA）수색산법적이상점법대비지배해진행국부우화적혼합다목표입자군방법（local search with multiobjective particle swarm optimization，LSMOPSO），이제고다목표입자군산법적수렴성능화비지배해집적정도여다양성。LSMOPSO 산법사용옹제거리선택령도입자조성령도입자집，병대기진행이상점국부수색；분석비교료전국이상점화국부이상점대산법성능적영향，제출기우국부이상점적국부수색책략；재입자적설계공간적다개유도상인입균균변이조작，강저산법함입국부최우적가능。기본측시함수적구해결과표명，산법적수렴속도흔쾌，이차수색도적비지배해집적정도고、다양성호。
A new hybrid multi-objective optimizer is presented,which combines particle swarm optimization (PSO)with an innovative optimal particles local search strategy based on bound optimization by quadratic ap-proximation (BOBYQA)algorithm.The main goal of the approach is to improve the convergence performance of PSO and diversity of nondominated set.The new approach constructs the leader particles set using crowding distance to select leader particles,then makes full use of the optimal particles method to guide leader particles approach the Pareto front quickly.Meanwhile,a new local optimal particles search strategy is proposed after a-nalysis on disadvantage of the global optimal particles search method.Furthermore,the multi-dimensional uni-form mutation is introduced to prevent algorithm from being trapped into local optimum.Simulation results of benchmark functions show that our approach is highly competitive in convergence speed and generates a well dis-tributed and accurate set of nondominated solutions.
