摘要
针对模拟电路直流仿真中Newton-Raphson(NR)方法存在的收敛不确定性、反复数值求导以及限于单次解等缺陷,引入进化方法以优化直流分析过程。研究了基于蚁群算法直接求解电路非线性代数方程的适应度函数构建、初始解分布、分类转移规则及信息素更新机制。鉴于蚁群算法直接求解的低精度问题,提出了将蚁群算法与NR方法相结合的新型优化方法——ACA-NR方法。实验结果表明,ACA方法具有方程求解收敛的稳定性和多解寻优能力,ACA-NR方法相比NR、ACA方法能够达到决策最优。
Aimed at the problems of convergence instability,repeated derivative operations and one-solution-one-time in DC analyses of analogue circuit simulation,an evolution method was introduced to optimize DC analyses.The issues such as fitness function,initial guess distribution,sorting and transferring rules,update mechanism for pheromone in ACA were studied.With the inexactitude of ACA,a new hybrid method-ACA-NR,which integrates ACA and NR method,was proposed.The experimental results show that ACA has the ability of stable convergence and finding multiple solutions, ACA-NR is the best choice to the optimization of solving circuit equations in DC analyses.
出处
《实验室研究与探索》
CAS
北大核心
2009年第7期53-56,160,共5页
Research and Exploration In Laboratory