基于改进差分进化算法的非线性系统模型参数辨识

Parameter estimation of nonlinear systems model based on improved differential evolution

针对非线性模型的参数估计寻优较为困难的问题,提出一种基于改进的差分进化算法的非线性系统模型参数辨识新方法。通过引入一个自适应变异率,随着迭代的进行自适应调整缩放因子,从而在初期保持种群多样性以避免早熟,并在后期逐步降低变异率,保留优良信息,避免最优解遭到破坏。交叉概率采用动态非线性增加的方法,提高了收敛速度。为了验证算法性能,针对几类典型的非线性模型参数辨识问题进行了仿真研究,并将其应用于一类发酵动力学模型参数的估计中。结果表明改进算法的参数辨识精度高,收敛速度也比较快,有效提高了模型建立的精度与效率,为解决实际系统中参数估计问题提供了一条可行的途径。

Estimation of nonlinear regression model parameters is a tough searching problem,this paper propoesd an improved differential evolution algorithm for nonlinear system model parameter identification method.It introduced an adaptive mutation rate to determine the scaling factor adaptively and avoid premature.At the other hand,it proposed a dynamic nonlinear increased crossover probability to improve convergence speed.It simulated for different model examples.The experimental results show that the improved algorithm is higher accuracy and faster convergence. At last, the new method was applied to a class of the fermentation kinetics model parameter estimation, effectively improved the accuracy and efficiency of the model, providing a feasible way to solve the problem of parameter estimation.