电力综合数据网运行态势评估与预测方法研究

发布时间：2018-02-11 01:16

本文关键词： 电力综合数据网态势评估态势预测 SNMP 支持向量机　出处：《电子科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：态势感知可以细分为态势评估及态势预测,是网络的结构配置情况,网络资源的使用情况,网络运行情况以及业务因素综合呈现出的网络整体状态以及可能的发展趋势。通过评估当前网络态势以及预测网络未来运行趋势,可以为网络管理员提供深入理解网络及用户行为的途径。电力综合数据网,是电网信息化的支撑网络,承担电网企业内部高速数据、话音以及多媒体等业务传输,呈现网络规模庞大,网络拓扑复杂、业务种类繁多、网络协议丰富的特点。为了保证电力综合数据网的正常高效运行,准确把握网络运行态势的现状及发展趋势是网络管理员关心的重中之重,因此建立针对电力综合数据网的运行态势评估与预测系统是一种行之有效的手段。本文针对电力综合数据网,通过采集骨干网络的SNMP协议信息,提取网络运行态势因子,对态势因子进行分类评估及时间序列预测,以达到评估整个网络当前运行态势及预测未来运行态势的目的。本文的具体工作如下:1.研究网络态势感知系统模型,提出一种基于骨干路由器运行状态的态势评估与预测模型,以骨干路由器SNMP协议信息为数据来源,建立态势因子指标体系,并实现了电力综合数据网运行态势评估与预测系统功能模块。2.研究态势评估方法,针对电力综合数据网骨干网提出一种基于K-means聚类预标签及支持向量机的态势评估方法,利用结合人工经验的K-means聚类为态势因子加上状态标签,解决了利用有监督分类评估时,大量态势因子样本无标签的问题。3.研究态势预测方法,针对时间序列预测的缺点,提出一种基于累加误差修正及支持向量机的态势预测方法,同步对态势因子及预测误差值通过累加处理后进行预测,利用误差预测值对态势因子初步预测值进行误差修正,减少了支持向量机回归预测滞后性及样本突变波动产生的误差,提高了预测准确度。本文在现有态势感知研究基础上,针对电力综合数据网提出了一套运行态势评估与预测模型,在评估与预测的过程中分别使用本文提出的基于K-means聚类预标签及支持向量机的态势评估方法及基于累加误差修正及支持向量机的态势预测方法,经过实验分析和系统实现,本文的模型和方法具有一定实用价值。
[Abstract]:Situational awareness can be subdivided into situation assessment and situation prediction, which is the configuration of the network structure, the use of network resources, The overall state of the network and the possible development trend presented by the network operation and service factors. By evaluating the current network situation and predicting the future network running trend, It can provide a way for network administrators to understand the network and the behavior of users in depth. Power integrated data network is the supporting network of power network informatization, which undertakes the transmission of high-speed data, voice, multimedia and other services within power grid enterprises. In order to ensure the normal and efficient operation of the power integrated data network, the network has the characteristics of large scale, complex network topology, various types of services and abundant network protocols. Accurately grasping the current situation and developing trend of network operation situation is the most important concern of network administrator. Therefore, it is an effective means to set up an operational situation assessment and prediction system for power integrated data network. In this paper, the SNMP protocol information of backbone network is collected to extract the running situation factor of the network. In order to evaluate the current situation of the whole network and predict the future situation of the whole network, the classification evaluation and time series prediction of the situation factor are carried out. The specific work of this paper is as follows: 1. The model of network situation awareness system is studied. This paper presents a situation assessment and prediction model based on the running state of backbone routers. Based on the SNMP protocol information of backbone routers, a situation factor index system is established. The function module of the power integrated data network running situation assessment and forecasting system is realized. 2. The situation assessment method is studied, and a situation assessment method based on K-means clustering pre-label and support vector machine is proposed for the power integrated data network backbone network. K-means clustering combined with artificial experience is used to add state label to situation factor, which solves the problem that a large number of situation factor samples are not tagged when using supervised classification and evaluation. 3. Study the method of situation prediction, aiming at the shortcomings of time series prediction. A situation prediction method based on accumulative error correction and support vector machine is proposed. The situation factor and prediction error are predicted by accumulative processing synchronously, and the initial prediction value of situation factor is corrected by error prediction value. It reduces the error of prediction lag and sample mutation fluctuation by using support vector machine regression, and improves the accuracy of prediction. A set of operational situation assessment and prediction model for power integrated data network is proposed. In the process of evaluation and prediction, the situation assessment method based on K-means clustering prelabel and support vector machine and the situation prediction method based on accumulative error correction and support vector machine are used in this paper. The model and method in this paper have certain practical value.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TM73

【参考文献】