﻿ 基于动态贝叶斯的船舶中央冷却水系统状态推理
 舰船科学技术  2016, Vol. 38 Issue (12): 104-109 PDF

State reasoning of ship central cooling water system based on dynamic bayesian
MENG Rui, ZENG Fan-ming
College of Power Engineering, Naval University of Engineering, Wuhan 430033, China
Abstract: In order to solve the problem of the central cooling water system which becomes more and more complex and has great uncertainty, the application of dynamic Bayesian network to the detection and control of the complex pipeline system is used in this paper. Through the comprehensive analysis of the relationship between a certain type of marine central cooling water system components and evaluation parameters, the dynamic Bayesian network model is established and the BK algorithm is used for inference. Each characteristic factor of the target node and the same characteristic factor of different time slice are corrected to overcome the uncertainty, incomplete data and subjectivity. The simulation results show that the dynamic Bayesian network can take the time factor into consideration in the case of uncertain environment and incomplete data, and can effectively control state.
Key words: dynamic Bayesian network     state reasoning     central cooling water system
0 引 言

1 贝叶斯网络 1.1 贝叶斯网络定义

 $Pa({V_i}/{V_j},Pa({V_i})) = P({V_i}/Pa({V_i})) \text{，}$ (1)

 $P({V_1},{V_2},...,{V_n}) = \prod\limits_{i = 1}^n {P({V_i}/Pa({V_i}))} \text{。}$ (2)

1.2 动态贝叶斯网络

DBN 模型是把时间的因素加入到了静态贝叶斯网络模型里，将静态网络扩展到时态领域充分考虑了时间对系统运行的影响。在建立模型前先进行以下假设：

1）假设在一有限时间内条件概率变化过程对所有t 一致平稳。

2）假设动态概率过程是马氏的（Markovian），即满足：

 $\begin{array}{l} P(X[t]|X[1],X[2],...,X[t]) = P(X[t + 1]|X[t]) \end{array} \text{。}$ (3)

3）假设相邻时间的条件概率平稳，即PX[t + 1]|X[t]）与时间t 无关，可容易得到不同时间片之间的转移概率为PX[t + 1]|X[t]）。

1）先验网B0，定义在初始状态X[1] 上的联合概率分布；

2）转移网B，定义在变量X[1] 与X[2] 上的转移概率PX[t + 1]|X[t]）。

 $\begin{array}{l} P(X[1],X[2],...,X[t]) = \\ \quad \quad \quad {P_{{B_0}}}(X[1])\prod\limits_{t = 1}^T {{P_{{B_ \to }}}(X[t + 1]|X[t])} \end{array} \text{。}$ (4)
2 中央冷却水系统

 图 1 中央冷却水系统原型图 Fig. 1 Prototype of the central cooling water system
3 动态贝叶斯网络模型

 图 2 简化的中央冷却水系统 Fig. 2 Simplified central cooling water system

 图 3 单个组件的 DBN 模型 Fig. 3 DBN model of a single component

 图 4 负载子系统 1 的 DBN 模型 Fig. 4 DBN model of load subsystem 1

 图 5 负载子系统 2 的 DBN 模型 Fig. 5 Model of load subsystem 2

 图 6 资源子系统 DBN 模型 Fig. 6 DBN model of resource subsystem
4 算法分析

 $P({I_t}|{Y_{1:t}}) \approx \prod\nolimits_{c = 1}^C {P(I_t^c|{Y_{1:t}})} \text{，}$ (5)

 $P(X) \approx \prod\nolimits_{i = 1}^k {P({X^{{W_i}}})} \text{。}$ (6)

 图 7 BK 算法更新投影过程 Fig. 7 Updated projection process of BK algorithm

5 仿真实验

 图 8 负载的温度变化情况 Fig. 8 The temperature change of the load

 图 9 负载 1 的流量概率值变化趋势 Fig. 9 Change trend of flow probability of load 1

 图 10 负载 2 的流量概率值变化趋势 Fig. 10 Change trend of flow probability of load 2

6 结 语

 [1] 刘富斌, 田志定, 向征涛. 混流式中央冷却系统设计探讨[J]. 船海工程 , 2002 (2) :49–52. LIU Fu-bin, TIAN Zhi-ding, XIANG Zheng-tao. Design study of mixed flow central cooling system[J]. Ship&Boat , 2002 (2) :49–52. [2] 付尧, 曾凡明, 陈于涛, 等. 舰用柴油机冷却水系统贝叶斯状态推理方法[J]. 舰船科学技术 , 2014, 36 (8) :42–45. FU Yao, ZENG Fan-ming, CHEN Yu-tao, et al. Research on the Bayesian state reasoning methods for marine diesel cooling water system[J]. Ship Science and Technology , 2014, 36 (8) :42–45. [3] XIANG Yang, POOLE D, BEDDOES M P. Multiply sectioned Bayesian networks and junction forests for large knowledge-based systems[J]. Computational Intelligence , 1993, 9 (2) :171–220. DOI:10.1111/coin.1993.9.issue-2 [4] 王三民, 王宝树. 贝叶斯网络在战术态势评估中的应用[J]. 系统工程与电子技术 , 2004, 26 (11) :1620–1623. WANG San-min, WANG Bao-shu. Application of Bayesian networks in tactical situation assessment[J]. Systems Engineering and Electronics , 2004, 26 (11) :1620–1623. [5] 奚海荣, 马文丽, 梁斌. 基于贝叶斯网络SP算法的改进研究[J]. 计算机技术与发展 , 2009, 19 (3) :155–157. XI Hai-rong, MA Wen-li, LIANG Bin. Improvement of SP algorithm based on Bayesian networks[J]. Computer Technology and Development , 2009, 19 (3) :155–157. [6] 赵晓辉, 姚佩阳, 张鹏. 动态贝叶斯网络在战场态势估计中的应用[J]. 光电与控制 , 2010, 17 (1) :44–47. ZHAO Xiao-hui, YAO Pei-yang, ZHANG Peng. Application of dynamic Bayesian network in battlefield situation assessment[J]. Electronics Optics&Control , 2010, 17 (1) :44–47. [7] 宫义山, 高媛媛. 基于信息融合的诊断贝叶斯网络研究[J]. 计算机技术与发展 , 2009, 19 (6) :106–108. GONG Yi-shan, GAO Yuan-yuan. Diagnostic Bayesian networks research based on information fusion[J]. Computer Technology and Development , 2009, 19 (6) :106–108. [8] MIHAJLOVIC V, PETKOVIC M. Dynamic Bayesian networks:a state of the art[R]. CTIT technical reports series, TR-CTIT-34, University of Twente Document Repository, 2001. [9] COOPER G F. The Computational complexity of probabilitic inference using Bayesian belief networks[J]. Artificial Intelligence , 1990, 42 (2/3) :393–405. [10] MURPHY K P. Dynamic Bayesian networks:representation, inference and learning[D]. Berkeley, Fall:University of California, 2002. [11] 胡大伟. 动态贝叶斯网络的近似推理算法研究[D]. 合肥:合肥工业大学, 2009. HU Da-wei. The research on approximate reasoning algorithm for dynamic Bayesian networks[D]. Hefei:Hefei Polytechnic University, 2009. [12] 俞奎. 贝叶斯网络建模及推理算法研究[D]. 合肥:合肥工业大学, 2007. YU Kui. The research of learning and inference algorithm for bayesian networks[D]. Hefei:Hefei Polytechnic University, 2007. [13] 刘俊娜. 贝叶斯网络推理算法研究[D]. 合肥:合肥工业大学, 2007. LIU Jun-na. Research on the Bayesian networks inference[D]. Hefei:Hefei Polytechnic University, 2007. [14] 董晨. 贝叶斯网络在电力系统故障中的应用研究[D]. 沈阳:沈阳工业大学, 2010. DONG Chen. Applications and research of the Bayesian network in power system fault diagnosis[D]. Shenyang:Shenyang University of Technology, 2010.