两类随机延迟微分方程Milstein方法的稳定性和收敛性-应用数学专业论文.docxVIP

V V 摘 要 随机延迟微分方程既可以视为确定性模型问题延迟微分方程考虑了随机因 素后的推广，也可以视为非确定性模型问题随机常微分方程考虑了时滞因素后的 推广, 所以随机延迟微分方程往往能够更加真实地模拟科学实际中的问题.因此 它已开始被广泛地应用于物理、化学、控制论、金融学、神经网络、生态学等各 个研究领域。但是对于它的研究方法，既不能等同延迟微分方程，也不能等同随 机常微分方程, 而且在具体的研究过程中必将会面临许多难以预料的困难。同 时，与延迟微分方程和随机常微分方程一样，要想办法得到随机延迟微分方程问 题本身的理论解是十分困难的。这就更加迫切地突显出随机延迟微分方程数值求 解方法研究的重要性。 本文所获得的主要结果如下: 首先，利用 Milstein 方法来求解一类极其重要的方程——Fokker-Planck 方程, 在此针对白噪声驱动随机系统的一维 Fokker-Planck 方程，证明了如果问题本身 满足适度的限制条件时，利用 Milstein 方法求解该方程时是收敛的理论结果。 其次，研究了用 Milstein 方法求解中立型随机延迟微分方程初值问题数值方 法的稳定性，给出了 Milstein 方法均方稳定的一个充分条件。 文中的数值试验进一步验证了所得理论结果的正确性。 关键词：Fokker-Planck 方程；中立型随机延迟微分方程；Milstein 方法；插值； 收敛性；均方稳定 II II Abstract If the status of a system is influenced at a certain time by numerous factors and the influence of each factor which brings to the system has great eventuality, we should consider the influence of the ambient noise in the system. When the developing trends of considered system relates to not only the current status but also the history of the past, it is usually described by stochastic delay differential equation. Nowadays, stochastic delay differential equations have been widely used in the fields of physics, chemistry, bionomics, medical, neural network and auto control and so on. Therefore more and more scholars pay close attention to how to solve the stochastic delay differential equations, however, due to the complex of the system, its analytic solutions are hard to obtain, and it makes numerical solution stochastic delay differential equations especially important. Among them, discussing the astringency and stability of the numerical methods is a significant research subject. In this paper, at first we discuss the convergence of Milstein methods for the scalar Fokker-Planck equations with a noise process. It is proved that the Milstein methods with linear interpolating process for the scalar Fokker-Planck equations are convergent. Second we discussed the numercial stability of neutral delay differential equations. It is proved that the Milstein me