数学杂志  2018, Vol. 38 Issue (1): 147-154   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
辛志贤
张启敏
李强
哈金才
具有年龄结构随机种群系统数值解的渐近有界性
辛志贤1, 张启敏1,2, 李强1, 哈金才1    
1. 北方民族大学数学与信息科学学院, 宁夏 银川 750021;
2. 宁夏大学数学统计学院, 宁夏 银川 750021
摘要:本文研究了一类具有年龄结构的随机种群系统的数值解问题.在线性增长条件下,利用Euler-Maruyama(EM)方法讨论了具有年龄结构的随机种群系统的数值解的p阶矩渐近有界性,并获得了渐近有界性准则.最后,通过数值算例对所得的结论进行了验证.
关键词随机种群系统    线性增长条件    Euler-Maruyama方法    p阶矩渐近有界性    
ASYMPTOTIC BOUNDEDNESS OF THE NUMERICAL SOLUTIONS OF STOCHASTIC AGE-DEPENDENT POPULATION SYSTEM
XIN Zhi-xian1, ZHANG Qi-min1,2, LI Qiang1, HA Jin-cai1    
1. School of Mathematics and Information Science, Beifang University for Nationalities, Yinchuan 750021, China;
2. College of Mathematics and Statistics, Ningxia University, Yinchan 750021, China
Abstract: A class of stochastic age-dependent population system is studied in this paper. Under linear growth condition, we discuss the p-th asymptotic boundedness of the numerical solutions for stochastic age-dependent population system and establish a criterion for the asymptotical boundedness by using Euler-Maruyama (EM) method. Finally, numerical example is presented to demonstrate the accuracy of the conclusion.
Key words: stochastic population system     linear growth condition     Euler-Maruyama method     p-th asymptotic boundedness    
1 引言

随机微分方程被广泛应用于工程、生物、金融等领域[1-6], 随机种群系统也引起了许多学者关注.本文研究如下具有年龄结构的随机种群系统模型[7]

$ \begin{equation} \label{eq:1} \left\{ \begin{aligned} &d_tP=-\displaystyle \frac{\partial P}{\partial a}dt-\mu(t, a)Pdt+f(t, P)dt+g(t, P)dB(t), &(t, a)\in Q, \\ &P(0, a)=P_0(a), &a\in[0, A], \\ &P(t, 0)=\int_0^A\beta(t, a)P(t, a)da, &t\in[0, T], \end{aligned} \right. \end{equation} $ (1.1)

其中$Q=(0, T)\times(0, A)$, $A$是种群所能达到的最大年龄, $0<T<\infty$, $d_tP=\displaystyle\frac{\partial P}{\partial t}dt$, $P(t, a)$表示$t$时刻年龄为$a$的种群密度; $\beta(t, a)$表示$t$时刻年龄为$a$的种群的出生率; $\mu(t, a)$表示$t$时刻年龄为$a$的种群的死亡率; $f(t, P)$是外界环境的干扰, 如迁移、地震、海啸等突发性灾害对种群的影响; $g(t, P)dB(t)$表示随机外界环境对系统的干扰; $B(t)$是定义在概率空间$(\Omega, \mathcal{F}_t, P)$上的Brown运动.

模型$(1.1)$考虑了随机环境对系统的扰动影响, 更符合种群模型的实际意义.由于随机种群模型的解析解很难给出, 因此数值解的计算显得尤为重要, 并且在近几年, 随机种群模型数值解的研究取得了很多成果.文献[8]研究了带跳的具有年龄结构的随机种群系统数值解的稳定性, 讨论了带有Markovian转换的与年龄相关的随机种群系统的渐近稳定性[9]; 文献[10]针对与年龄相关的随机种群扩散系统, 讨论了其数值解的指数稳定性.然而上述文献中, 均是对随机种群模型数值解的稳定性进行了研究, 但数值解的另一个性质渐近有界性同样具有重要的研究价值[11].本文利用EM方法研究具有年龄结构的随机种群系统的渐近有界性, 在线性增长条件下, 建立$p$阶矩渐近有界性准则.最后, 通过数值算例对所得的结论进行了验证.

2 预备知识

$V=H^1([0, A])\equiv\bigg\{\varphi|\varphi\in L^2([0, A])$, $\displaystyle\frac{\partial \varphi}{\partial a}\in L^2([0, A])\bigg\}$, 其中$\displaystyle\frac{\partial \varphi}{\partial a}$是广义函数意义下的偏导数; $V$是Sobolev空间; $H=L^2([0, A])$, 满足$V\hookrightarrow H\equiv H'\hookrightarrow V'.$$V'=H^{-1}([0, A])$$V$的对偶空间; $|\cdot|$$\|\cdot\|$分别为$V, V'$的范数; $\langle\cdot, \cdot\rangle$表示$V$$V'$空间的内积[10].

定义$(\Omega, \mathcal{F}, \{\mathcal{F}_t\}_{t\geq0}, \mathbb{P})$是带流的单调递增完备的概率空间, $\mathcal{F}_0$包含所有的$\mathbb{P}$零子集, $B(t)$是定义在概率空间$(\Omega, \mathcal{F}_t\, P)$上的Brown运动.$a\vee b$表示$a$$b$的最大值, $a\wedge b$表示$a$$b$的最小值.

定义2.1 在概率空间$(\Omega, \mathcal{F}, \mathbb{P})$上的一个带流$\mathcal{F}$的随机过程$P_t$被称为方程(1.1)的解, 如果满足下列条件

(1) $P_t\in I^2(0, T;V)\bigcap L^2(\Omega;(0, T;H))$, 其中$I^2(0, T;V)$为所有均方可测的${\left( {{P_t}} \right)_{t \in [0,T]}}$组成的空间, 满足

$ E\int_0^T\|P_t\|^2dt\leq\infty; $

(2) 对于任意的$t\in[0, T]$, 在概率空间$V'$上下列方程几乎处处成立,

$ \begin{equation} \left\{ \begin{aligned} &P(t, a)=P_0-\int_0^t\frac{\partial P(s, a)}{\partial a}ds-\int_0^t\mu(s, a)P(s, a)ds\\ &\ \ \ \ \ \ \ \ \ \ \ \ +\int_0^tf(s, P(s, a))ds+\int_0^tg(s, P(s, a))dB(s), &(0, T)\times(0, A), \\ &P(0, a)=P_0(a), &a\in[0, A], \\ &P(t, 0)=\int_0^A\beta(t, a)P(t, a)da, &t\in[0, T].\nonumber \end{aligned} \right. \end{equation} $

为了证明文中的结论, 给出以下假设条件

(A1)方程(1.1)中$f(t, P), g(t, P), \displaystyle\frac{\partial P}{\partial a}$满足线性增长条件, 即

$ \begin{eqnarray} |f(t, P)|^2\vee\|g(t, P)\|^2\vee\displaystyle\bigg|\frac{\partial P}{\partial a}\bigg|^2\leq K|P|^2+\alpha, \end{eqnarray} $ (2.1)

其中$K$$\alpha$均为正常数.

(A2)$\mu(t, a), \beta(t, a)$$Q$上是连续的且存在正常数$\mu_0, \bar{\mu}, \bar{\beta}$, 满足

$ \begin{eqnarray} 0\leq \mu_0\leq \mu(t, a)\leq \bar{\mu}<\infty, \ \ \ 0\leq \beta(t, a)\leq \bar{\beta}<\infty, \ \ \ A\bar{\beta}^2-2\mu_0\geq0. \end{eqnarray} $ (2.2)

(A3)存在正常数$D$, 使得对于任意$P\in V$,

$ \begin{aligned} \ \ \ \ \displaystyle\frac{2\langle P, f(t, P)\rangle+\|g(t, P)\|^2}{2(D+|P|^2)} -\displaystyle\frac{\langle P, g(t, P)\rangle^2}{(D+|P|^2)^2} \leq-\lambda+\displaystyle\frac{Q_1(|P|)}{D+|P|^2}+\frac{Q_3(|P|)}{(D+|P|^2)^2}, \end{aligned} $ (2.3)

其中$\lambda$是正常数, $Q_i(|P|)$$|P|$$i$次多项式.

3 EM方法

本节将利用EM方法研究具有年龄结构的随机种群系统的渐近有界性, 并建立$p$阶矩渐近有界性准则.

首先使用EM差分方法对方程$(1.1)$进行离散, 可得

$ \begin{aligned} &P_{k+1}=P_k-\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t\\ &\ \ \ \ \ \ \ \ \ +g(t_k, P_k)\triangle B_k, \forall a\in(0, A), P(0, a)\in V.\\ \end{aligned} $ (3.1)

下面给出本文的主要定理.

定理3.1 如果条件(A1)-(A3)成立, 那么对于任意的$\varepsilon\in(0, \lambda)$, 都存在常数$p^\ast\in(0, 1)$$\triangle t^\ast\in(0, 1)$使得$\forall p\in(0, p^\ast)$$\forall\triangle t\in(0, \triangle t^\ast)$, 方程(3.1)的EM解满足

$ \begin{aligned} \lim\sup\limits_{k\rightarrow\infty}\mathbb{E}(|P_k|^p)\leq \displaystyle \frac{{C_2}''}{p(\lambda-\varepsilon)}, \ \ \ \ \forall P_0\in V, \end{aligned} $ (3.2)

其中${C_2}''$是与$K$$\alpha$$D$$p$有关, 与$P_0$无关的常正数.

  由EM差分格式(3.1)可得

$ \begin{aligned} |P_{k+1}|^2=&|P_k|^2+2\langle P_k, -\displaystyle \frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\rangle \\ &+\bigg|-\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\bigg|^2, \\ \end{aligned} $ (3.3)

则对于常数$D$, 有

$ \begin{aligned} D+|P_{k+1}|^2=&D+|P_k|^2+2\langle P_k, -\displaystyle \frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\rangle \\ &+\bigg|-\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\bigg|^2. \end{aligned} $ (3.4)

$ \begin{aligned} \xi_k=&\displaystyle\frac{1}{D+|P_k|^2}\bigg(2\langle P_k, -\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\rangle \\ &+\bigg|-\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\bigg|^2\bigg), \end{aligned} $ (3.5)

那么对于任意的$p\in(0, 1)$, 满足

$ \begin{aligned} \bigg|D+|P_{k+1}|^2\bigg|^{p/2}=\bigg|D+|P_k|^2\bigg|^{p/2}\bigg(1+\xi_k\bigg)^{p/2}. \end{aligned} $ (3.6)

由基本不等式

$ \begin{aligned} \bigg(1+u\bigg)^{p/2}\leq 1+\displaystyle\frac{p}{2}u+\frac{p(p-2)}{8}u^2+\frac{p(p-2)(p-4)}{2^3\times 3!}u^3, \ \ \ u>-1. \end{aligned} $

显然, $\xi_k>-1$, 由基本不等式可得

$ \begin{aligned} \bigg|D+|P_{k+1}|^2\bigg|^{p/2}\leq\bigg|D+|P_k|^2\bigg|^{p/2}\bigg(1+\displaystyle\frac{p}{2}\xi_k+\frac{p(p-2)}{8}\xi_k^2+\frac{p(p-2)(p-4)}{2^3\times 3!}\xi_k^3\bigg). \end{aligned} $ (3.7)

对(3.7)式两边同时取条件期望, 有

$ \begin{aligned} &\mathbb{E}\bigg(\bigg|D+|P_{k+1}|^2\bigg|^{p/2}\bigg|\mathcal{F}_{k\triangle t}\bigg)\\ \leq&\bigg|D+|P_k|^2\bigg|^{p/2}\mathbb{E}\bigg(1+\displaystyle\frac{p}{2}\xi_k+\frac{p(p-2)}{8}\xi_k^2+\frac{p(p-2)(p-4)}{2^3\times 3!}\xi_k^3\bigg). \end{aligned} $ (3.8)

因为$\triangle B_k$$\mathcal{F}_{k\triangle t}$独立, 所以$\mathbb{E}(\triangle B_k|\mathcal{F}_{k\triangle t})=\mathbb{E}(\triangle B_k)=0, $ $\mathbb{E}((\triangle B_k)^2|\mathcal{F}_{k\triangle t})=\mathbb{E}((\triangle B_k)^2)=\triangle t, $因此

$ \begin{aligned} &\ \ \ \ \mathbb{E}(\xi_k|\mathcal{F}_{k\triangle t})\\ &=\mathbb{E}\bigg(\displaystyle\frac{1}{D+|P_k|^2}\bigg(2\langle P_k, -\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+ f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_k\rangle \\ &\ \ \ \ +\bigg|-\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t+g(t_k, P_k)\triangle B_i\bigg|^2\bigg)\bigg|\mathcal{F}_{k\triangle t}\bigg)\\ &\leq\displaystyle\frac{2}{D+|P_k|^2}\langle P_k, -\displaystyle\frac{\partial P_k}{\partial a}\triangle t-\mu(t_k, a)P_k\triangle t+f(t_k, P_k)\triangle t\rangle \\ &\ \ \ \ +\displaystyle\frac{1}{D+|P_k|^2}\bigg[\bigg(\displaystyle\frac{\partial P_k\triangle t}{\partial a}\bigg)^2+(\mu(t_k, a)P_k\triangle t)^2+(f(t_k, P_k)\triangle t)^2+\|g(t_k, P_k)\|^2\triangle t\bigg]\\ &\leq\displaystyle\frac{1}{D+|P_k|^2}\bigg(A \bar{\beta}^2P_k^2-2\mu_0|P_k|^2+2\langle P_k, f(t_k, P_k)\rangle+\|g(t_k, P_k)\|^2\bigg)\triangle t\\ &\ \ \ \ +\displaystyle\frac{1}{D+|P_k|^2}\bigg(\bigg(\displaystyle\frac{\partial P_k}{\partial a}\bigg)^2+\mu^2(t_k, a)|P_k|^2+|f(t_k, P_k)|^2\bigg)\triangle t^2\\ &\leq\displaystyle\frac{1}{D+|P_k|^2}\bigg(A \bar{\beta}^2|P_k|^2-2\mu_0|P_k|^2+2\langle P_k, f(t_k, P_k)\rangle+\|g(t_k, P_k)\|^2\bigg)\triangle t\\ &\ \ \ \ +C_{11}\triangle t^2+\displaystyle\frac{C_{12}}{D+|P_k|^2}\triangle t^2. \end{aligned} $ (3.9)

类似的, 可得

$ \mathbb{E}(\xi_k^2|\mathcal{F}_{k\triangle t}) \geq\displaystyle\frac{4}{(D+|P_k|^2)^2}\langle P_k, g(t_k, P_k)\rangle^2\triangle t- C_{21}\triangle t^2-\displaystyle\frac{C_{22}}{D+|P_k|^2}\triangle t^2, $ (3.10)
$ \mathbb{E}(\xi_k^3|\mathcal{F}_{k\triangle t})\geq C_{31}\triangle t^2-\displaystyle\frac{C_{32}}{D+|P_k|^2}\triangle t^2, $ (3.11)

其中$C_{11}, C_{21}, C_{31}, C_{12}, C_{22}, C_{32}, $均为正常数; $C_{11}, C_{21}, C_{31}$$K$有关; $C_{12}, C_{22}, C_{32}$$\alpha$有关.现在考虑下面两个分数

$ \begin{aligned} \displaystyle\frac{(D+|P_k|^2)^{p/2}Q_1(|P_k|)}{D+|P_k|^2}, \displaystyle\frac{(D+|P_k|^2)^{p/2}Q_3(|P_k|)}{(D+|P_k|^2)^2}. \end{aligned} $ (3.12)

$0<p<1$时, 两个分数的分子中$|P_k|$的次数分别为$p+1$次和$p+3$次, 均小于对应分数分母中$|P_k|$的次数.所以对于任意的$|P_k|\in\mathbb{R}$, 两个分数都存在上界.同时, 当$i, j=1, 2, 3$时, 显然$\displaystyle\frac{C_{j2}}{(D+|P_{k+1}|^2)^{i-p/2}}$也有界.把(3.9)-(3.11)式代入(3.8)式, 并且根据假设(A1)-(A3)和(3.12)式, 可以得到

$ \begin{aligned} &\ \ \ \ \mathbb{E}((D+|P_{k+1}|^2)^{p/2}|\mathcal{F}_{k\triangle t})\\ &\leq(D+|P_{k+1}|^2)^{p/2} \bigg(1+\displaystyle\frac{p}{2(D+|P_k|^2)}\bigg(A{\bar\beta}^2|P_k|^2 -2\mu_0|P_k|^2\\ &\ \ \ \ +2\langle P_k, f(t_k, P_k)\rangle+\|g(t_k, P_k)\|^2\bigg)\triangle t\\ &\ \ \ \ +\displaystyle\frac{p(p-2)}{2(D+|P_k|^2)^2}\langle P_k, g(t_k, P_k)\rangle^2\triangle t+{C_1}'\triangle t^2\bigg)+{C_2}'\triangle t\\ &=(D+|P_{k+1}|^2)^{p/2}\bigg[1+p\triangle t\bigg(\displaystyle\frac{(A{\bar\beta}^2-2\mu_0)|P_k|^2+2\langle P_k, f(t_k, P_k)\rangle}{2(D+|P_k|^2)}\\ &\ \ \ \ +\displaystyle\frac{\|g(t_k, P_k)\|^2}{2(D+|P_k|^2)} -\displaystyle\frac{\langle P_k, g(t_k, P_k)\rangle^2}{(D+|P_k|^2)^2}\bigg)+\displaystyle\frac{p^2\triangle t\langle P_k, g(t_k, P_k)\rangle^2}{2(D+|P_k|^2)^2} +{C_1}'\triangle t^2\bigg]\\ &\ \ \ \ +{C_2}'\triangle t\\ &\leq(D+|P_{k+1}|^2)^{p/2}\bigg[1+p\triangle t\bigg(\displaystyle\frac{2\langle P_k, f(t_k, P_k)\rangle+\|g(t_k, P_k)\|^2}{2(D+|P_k|^2)} -\displaystyle\frac{\langle P_k, g(t_k, P_k)\rangle^2}{(D+|P_k|^2)^2}\bigg)\\ &\ \ \ \ +\displaystyle\frac{p^2\triangle t\langle P_k, g(t_k, P_k)\rangle^2}{2(D+|P_k|^2)^2} +{C_1}'\triangle t^2\bigg]+{C_2}'\triangle t\\ &\leq(D+|P_{k+1}|^2)^{p/2}\bigg(1-p\lambda \triangle t+\displaystyle\frac{p^2\triangle tK}{2}+{C_1}''\triangle t^2\bigg)+{C_2}''\triangle t, \end{aligned} $ (3.13)

其中${C_1}', {C_1}''$是与$K$$p$有关的正常数, ${C_2}', {C_2}''$是与$K$, $\alpha$, $D$$p$有关的正常数.对于任意给定的$\varepsilon\in(0, \lambda)$, 取充分小的$p^\ast\in(0, 1)$使得$p^\ast K<\varepsilon$, 同时取充分小的$\triangle t^\ast\in(0, 1)$使得$p^\ast\lambda\triangle t^\ast\leq1$, ${C_1}''\triangle t^\ast\leq\displaystyle\frac{1}{2}p\varepsilon$.则对于任意的$p\in(0, p^\ast)$$\triangle t\in(0, \triangle t^\ast)$, 有

$ \begin{aligned} \mathbb{E}((D+|P_{k+1}|^2)^{p/2}|\mathcal{F}_{k\triangle t})\leq(D+|P_k|^2)^{p/2}(1-p(\lambda-\varepsilon)\triangle t)+{C_2}''\triangle t. \end{aligned} $ (3.14)

两边同时取期望, 可得

$ \begin{aligned} \mathbb{E}((D+|P_{k+1}|^2)^{p/2}\leq\mathbb{E}((D+|P_k|^2)^{p/2})(1-p(\lambda-\varepsilon)\triangle t)+{C_2}''\triangle t. \end{aligned} $ (3.15)

通过迭代, 易知

$ \begin{aligned} &\ \ \ \ \mathbb{E}((D+|P_{k+1}|^2)^{p/2}\\ &\leq\mathbb{E}((D+|P_0|^2)^{p/2})(1-p(\lambda-\varepsilon)\triangle t)^k+ \displaystyle\frac{1-(1-p(\lambda-\varepsilon)\triangle t)^{k-1}}{p(\lambda-\varepsilon)}{C_2}''. \end{aligned} $ (3.16)

又因为$\mathbb{E}(|P_k|^p)\leq\mathbb{E}((D+|P_k|^2)^{p/2}), $所以

$ \begin{aligned} \mathbb{E}(|P_k|^p)\leq\mathbb{E}((D+|P_0|^2)^{p/2})(1-p(\lambda-\varepsilon)\triangle t)^k+ \displaystyle\frac{1-(1-p(\lambda-\varepsilon)\triangle t)^{k-1}}{p(\lambda-\varepsilon)}{C_2}''. \end{aligned} $ (3.17)

$k\rightarrow\infty$, 则

$ \begin{aligned} \lim\sup\limits_{k\rightarrow\infty}\mathbb{E}(|P_k|^p)\leq\displaystyle\frac{{C_2}''}{p(\lambda-\varepsilon)}, \ \ \ \ \forall P_0\in V. \end{aligned} $ (3.18)

定理得证.

4 数值算例

本节通过以下例子对定理进行验证

$ \begin{equation} \left\{ \begin{aligned} &d_tP=\bigg[-\displaystyle \frac{\partial P}{\partial a}-\frac{1}{4(1-a)^2}P dt+1.5-4P\bigg]dt\\ &\ \ \ \ \ \ \ \ \ +(0.125+0.25P)dB(t),&(t, a)\in Q, \\ &P(0, a)=e^{\frac{1}{1-a}},&a\in[0, 0.8], \\ &P(t, 0)=\int_0^A\frac{1}{(1-a)^2}P(t, a)da,&t\in[0, T], \end{aligned} \right. \end{equation} $ (4.1)

其中$Q=[0, T]\times[0, 0.8]$, $A=0.8$, $\mu(t, a)=\displaystyle\frac{1}{4(1-a)^2}$, $\beta(t, a)=\displaystyle\frac{1}{(1-a)^2}$, $f(t, P)=1.5-4P$, $B(t)$是标准布朗运动, $g(t, P)=0.125+0.25P$.显然$f(t, P), g(t, P)$均满足线性增长条件, 同时,

$ \begin{eqnarray*}&&0\leq 0.25=\mu_0\leq \mu(t, a)\leq \bar{\alpha}=6.25<\infty, \\ &&0\leq \beta(t, a)\leq \bar{\beta}=1<\infty, A\bar{\beta}^2-2\mu_0>0, \end{eqnarray*} $

所以假设条件(A2)也成立.

下面证明假设条件(A3)中$D$的存在性并给出$D$的选取过程.

$f(t_k, P_k)=\theta_1+\theta_2P_k, g(t_k, P_k)=\sigma_1+\sigma_2P_k$, 则对于假设条件(A3)有

$ \frac{{2\langle {P_k},f({t_k},{P_k})\rangle + {{\left\| {g({t_k},{P_k})} \right\|}^2}}}{{2(D + |{P_k}{|^2})}} = \frac{{P_k^2(2{\theta _2} + \sigma _2^2)}}{{2(D + |{P_k}{|^2})}} + \frac{{\sigma _1^2 + (2{\theta _1} + 2{\sigma _1}{\sigma _2}){P_k}}}{{2(D + |{P_k}{|^2})}}, $ (4.2)
$ \begin{array}{l} {\langle {P_k},g({t_k},{P_k})\rangle ^2} = {({\sigma _1}{P_k} + {\sigma _2}P_k^2)^2} = \sigma _1^2P_k^2 + \sigma _2^2P_k^4 + 2{\sigma _1}{\sigma _2}P_k^3\\ \;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\; = \sigma _2^2{(P_k^2 + \frac{{\sigma _1^2}}{{2\sigma _2^2}})^2} - \frac{{\sigma _1^4}}{{4\sigma _2^2}} + 2{\sigma _1}{\sigma _2}P_k^3. \end{array} $ (4.3)

$D=\displaystyle\frac{\sigma_1^2}{2\sigma_2^2}$, 则有

$ \begin{aligned} &\ \ \ \ \displaystyle\frac{2\langle P_k, f(t_k, P_k)\rangle+\|g(t_k, P_k)\|^2}{2(D+|P_k|^2)} -\displaystyle\frac{\langle P_k, g(t_k, P_k)\rangle^2}{(D+|P_k|^2)^2}\\ &\leq\bigg(\theta_2-\displaystyle\frac{1}{2}\sigma_2^2\bigg)+ \displaystyle\frac{\sigma_1^2+(2\theta_1+2\sigma_1\sigma_2)P_k}{2(D+|P_k|^2)} +\displaystyle\frac{1}{(D+|P_k|^2)^2}\bigg(\frac{\sigma_1^4}{4\sigma_2^2}-2\sigma_1\sigma_2P_k^3\bigg), \end{aligned} $ (4.4)

因此取

$ -\lambda=\theta_2-\displaystyle\frac{1}{2}\sigma_2^2 =-4.03125, $
$ Q_1(|P_k|)=\sigma_1^2+(2\theta_1+2\sigma_1\sigma_2)P_k=0.015125+3.0625P_k, $
$ Q_3(|P_k|)=\displaystyle\frac{\sigma_1^4}{4\sigma_2^2}-2\sigma_1\sigma_2P_k^3=\displaystyle\frac{(0.0125)^4}{0.25}-0.00625P_k^3. $

所以假设条件(A3)成立, 即常数$D$存在并且$D=\displaystyle\frac{\sigma_1^2}{2\sigma_2^2}=0.121$.

最后根据定理1的证明过程(3.4)-(3.18)式得到$P(t, a)$渐近有界.

取时间步长$\Delta t=0.01$, 空间步长$h=0.05$, 采用EM方法对方程进行差分离散, 分别作出该算例数值解的二维图(图 1)和三维图(图 2).由图可知当时间趋于无穷时, 该种群系统的种群密度上下波动, 并存在一个上界, 显然, 此种群系统渐近有界.

图 1 $P(t, a)$数值解的三维轨迹

图 2 $P(t, a)$数值解的二维轨迹
5 结论

本文基于EM方法研究了一类具有年龄结构的随机种群系统的数值解的$p$阶矩渐近有界性.在线性增长条件下, 利用基本不等式建立了渐近有界性准则.所得结论为种群的最优控制提供了有效的工具.

参考文献
[1] Jean J, Albert N, Shiriaev. Limit theorems for stochastic process[M]. New York: Springer Verlag, 2002.
[2] Luo Hui, Ma Xuemin, Ma Zhiwei, Zhou Xuan. The statistics analysis of randomly trimmed means and their bootstrap[J]. J. Math., 2014, 34(1): 1–15.
[3] 杨洪福, 张启敏. 与年龄相关的随机时分数阶种群系统温和解的存在性、唯一性[J]. 数学杂志, 2016, 36(5): 1083–1090.
[4] 李荣华, 戴永红, 孟红兵. 与年龄相关的随机时滞种群方程的指数稳定性[J]. 数学年刊, 2006, 27A(1): 39–52.
[5] Mao Xuerong. Stochastic differential equations and applcations[M]. Chichester: Horwood Pulishing, 2007.
[6] Mao Xuerong, Yuan Chenggui. Stochastic differential equations with Markovian switching[M]. London: Imperial College Press, 2006.
[7] Zhang Qimin, Liu Wenan, Nie Zankan. Existence, uniqueness and exponential stability for stochastic age-dependent population[J]. Appl. Math. Comp., 2004, 154: 183–201. DOI:10.1016/S0096-3003(03)00702-1
[8] Mao Wei. Exponential stability of numerical solutions to stochastic age-dependent population equations with Poisson jumps[J]. Engin. Tech., 2011, 55: 1103–1108.
[9] Ma Weijun, Zhang Qimin, Wang Zhanping. Asymptotic stability of stochastic age-dependent population equations with Markovian switching[J]. Appl. Math. Comp., 2014, 227: 309–319. DOI:10.1016/j.amc.2013.11.006
[10] Zhang Qimin. Expontial stability of numerical solutions to a stochastic age-structured population system with diffusion[J]. J. Comp. Appl. Math., 2008, 220(1-2): 22–33. DOI:10.1016/j.cam.2007.08.026
[11] Luo Qi, Mao Xuerong, Shen Yi. Generalised theory on asymptotic stability and boundedness of stochastic functional differential equations[J]. Automatica, 2011, 47(9): 2075–2081. DOI:10.1016/j.automatica.2011.06.014