中文摘要：

针对遗传算法在谐波平衡仿真中经常出现随机性大、迭代过程慢、局部搜索能力弱等问题,提出一种改进的混合遗传算法。该算法将Volterra级数、拟牛顿算法与遗传算法相结合,利用Volterra级数的记忆特性估算频域初始值,通过遗传算法进行全局优化,最后利用拟牛顿算法进行局部优化。基于对MRF281的谐波平衡仿真结果表明,该算法与遗传算法相比,迭代次数减少了40%左右,同时仿真数据与实测数据拟合较好。改进算法兼备了全局优化和局部优化的特点,明显提高了精度和收敛速度,并克服了遗传算法随机性大、局部搜索能力弱等缺点,对非线性电路分析具有较大的参考价值。

英文摘要：

Genetic algorithm for harmonic balance simulation often appear randomness,the iterative process is slow,weak local search capabilities and other issues,this paper proposed an improved hybrid genetic algorithm. The algorithm combined with Volterra series method,genetic algorithm and the quasi-Newton algorithm. Firstly,it used the characteristics of the memory of Volterra series to estimate the initial value of frequency domain,and then by means of genetic algorithm for global optimization,and finally used the quasi-Newton algorithm for local optimization. Based on the MRF281 harmonic balance simulation results show that the algorithm and genetic algorithm,the number of iterations reduced by 40%,while the simulation data and measured data fitting better. Algorithm combines the global optimization and local optimization features,significantly improves the harmonic balance equations accuracy and convergence rate,and to overcome the randomness of genetic algorithms,local search ability is weak and other shortcomings,the nonlinear circuit analysis has great reference value.