【作者】 周晖；

【导师】 王毅；

【作者基本信息】 北京交通大学 ， 电力系统及其自动化， 2010， 博士

【摘要】 随着电力工业改革进程的深入,电力公司所面临的电力客户的欠费问题,变得愈加严峻。为了保证电网公司资金的正常运转以及效益,电力客户的信用管理显得十分必要。首先,为了解决电力公司在电费交费问题上所处的被动局面的问题,借助于电力公司的内部信息和公开的社会信息,设计了适于电力企业使用的信用评价体系。其中,基于量化度量电力客户信用差异的需要,利用电费数据库的交费纪录对其进行数据挖掘,设计了变现动态信用评估算法。它客观性强,能较好地区分电力客户欠费、拖费、正常交费以及还款与不还款、还款及时性等各种复杂的缴费行为所表现出来的信用差异。考虑到电力企业信用评估量大,且要求每月一次,评估频繁的特点,以及企业内部信用管理成本的控制问题,设计了由抄表员负责的信息收集卡,运用电力公司的人力资源,有效地解决了电力客户信用信息收集难的问题。其次,利用主成分分析方法或突变理论等综合评价方法,完成了对众多用户的信用评估从而实现客户细分管理的要求。采用信用风险分析方法,通过对历史数据的训练,建立了多个用户信用风险判别模型(如Beyes, ANN, ANFIS等),它们能根据及时掌握的最新用户信用特征信息,有效地诊断用户的信用类别,从而达到有效防范信用风险的目的。在实际运行中,模型判断的正确率达85%以上；此外,还建立了Logistic判别模型,实现对电力客户的违约概率进行估计与判断。最后,在对用户拖欠费历史纪录及其原因进行深入分析的基础上,设计了统计预警方式下的欠费预警指标及预警标准。它具有指标实用、可观测的特点,且可根据实际管理工作的需要,进行调节,运用对用户预警指标下的观察值,来确定其是否列入预警黑名单。此外,运用灰色系统理论,建立了估计每个用户的信用变化趋势的模型,应用聚类方法确定出欠费类别判断准则,从而断定客户下一阶段可能位于的预警区域,为加强电费管理提供了可靠的技术支持。考虑到欠费控制手段需相互配合使用,着重从理论上探讨了预付费这种创新的制度,对于欠费风险控制的作用。更多还原

【Abstract】 With deeper reform in electricity industry, electric power companies have to face with the problem, i.e. great amount of account receivable is dued by clients. In order to ensure electric power enterprise have reasonable cash flow as well as gain corresponding profit, it is necessary to enhance clients’credit management.First of all, considering that electric power company is at disadvantage situation on the issue of electricity fees payment, we design a credit assessment systems which are suitable for electric power company. The data used in evaluating is conveniencely collected from MIS (managemant information system) or from public bulletin. In addition, to measure the difference of clients’credit, we mined the database of payment record, and design a new algorithm which is able to dynamically calculate their credit. The proposed algorithm could distinguish complicated paying behaviors objectively. For example, whether electrivity fee is paid on time or not, is in debt or not, if accumulative arrears existed for long time or short time etc. Considering that lots of clients are needed to be evaluated monthly, in order to save the management cost and make the investigation being accessible, we design a survey questionnair, which is assigned to staffs who are responsible for meter-reading. Therefore, we got credit information efficiencelly by our staffs in the electric power company.Then, with principal component approach and catastrophe theory, we accomplished the comprehensive credit evaluation for the purpose of subdivision clients and classification management. By means of credit risk analysis technique, we trained historical data and constructed credit discriminent models based on Beyes, ANN, RBF, ANFIS etc., then we could use the lastest credit information to judge the clients claasification. In the case we studied,correct-judging rate of the model achieved to 85% above. If the logistic discriminent model is employed, to estimate the default probability is available.Finally, we analyzed the reasons of arrears, designed default forewarning indecies and defined corresponding forewarning range. The indecies are simple, observable, and also adjustable according to the management requirement. Compared to the forewarning benchmark,it is easily for us to know if the client must be listed in the warning sheet. Besides, with grey system theory, we constructed a model which is able to forecast the credit variation of a client. With the determined rules, then we got to know if the client would likely enter the forewarning range. These results from model forewarning and statistical forewarning are an important reference in electricity fee management. We also pointed out that multimal control means in controlling default risk are very necessary, and analyzed the validity of pre-paying management system.更多还原