【作者】 曾晶；

【导师】 李江涛；

【作者基本信息】 重庆大学 ， 统计学， 2019， 硕士

【摘要】 电能是维持电力公司正常运行和管理的一种特殊商品,同时也是推动电力企业持续发展的重要经济来源之一。电力公司通过向广大用电客户提供电力和电能的服务进而收取电费。收取电费量的多少体现了电力公司的工作效率与能力。电力公司通过打造先进的电力服务网络来为用电客户提供既优质又安全的电力电能。为用户提供持续安全的电力服务是电力公司的责任,也是电力公司拥有资金来源的基础保障。所以说,只有确保电力公司全部电费的如数按时顺利回收,才能保障电力公司健康发展。在电力服务网络不断发展的进程中,广大用电客户拖欠电费的现象日益激增。电力公司面临着严重的电费回收风险。拖欠的电费金额不断增加的问题会严重制约电力公司的生产与发展。致使电力公司无法为社会提供稳定持续的电力资源,影响社会生产节奏和人们的生活质量。目前国内电网公司采取的经营模式是先用电后支付。这就会出现电费拖欠、拖欠的电费金额不断增加的问题。在电力公司庞大的用电量数据中,高压用电客户的耗电量占比相当惊人,其每月产生的电费就占据电力公司收取电费总额的70%以上,这就存在着非常大的潜在电费回收风险。一旦出现高压用电客户缴费违约、电费拖欠、拖欠的电费金额不断增加的现象,将会给电力公司带来难以弥补的损失,严重制约电力公司的生产与发展。所以实现高压用电客户电费拖欠风险的预警迫在眉睫。为建立高压用电客户电费拖欠风险的预警模型,本课题尝试从基本的电力数据入手,具体的研究工作有:首先,详细分析本课题的选题背景及意义,对电力客户电费拖欠风险预警研究现状和现阶段主要的一些研究成果进行说明。分析现有研究成果存在的问题和不足,阐述了建立高压用电客户电费拖欠风险的预警的必要性及现实意义。采取皮尔逊相关系数方法筛选影响电费拖欠的7个主要因素,并通过数据标准化消除不同变量间的量纲影响。然后根据这七个主要因素采用神经网络预测高压客户的风险得分值,再用风险得分值与7个主要影响因素建立逻辑回归预测高压客户的风险,将风险管理模式由劳动密集型向技术型转变,及时预警高风险用电客户。实验证明本文建立的神经网络与逻辑回归的混合预测模型有较高的预测精度,与实际情况具有良好的一致性,具有一定的现实意义,可以较为准确地预测高压风险用电客户,避免高压用户因违约给电力公司带来难以补救的损失。更多还原

【Abstract】 Electric energy is a special commodity to maintain the normal operation and management of power companies,and it is also one of the important economic sources to promote the sustainable development of power enterprises.Power companies charge electricity fees by providing electricity and power services to customers.The amount of electricity charged reflects the efficiency and ability of the power company.Electric power companies provide high quality and safe electric power for customers by building advanced electric power service network.It is the responsibility of power companies to provide continuous and safe power services for users,and it is also the basic guarantee for power companies to have sources of funds.Therefore,only by ensuring that all the electricity charges of the power company are recovered on time and smoothly,can the healthy development of the power company be guaranteed.In the process of the continuous development of power service network,the phenomenon that the majority of electricity users are in arrears of electricity charges is increasing sharply.Electric power companies are facing serious risk of electricity tariff recovery.The increasing amount of electricity arrears will seriously restrict the production and development of power companies.As a result,power companies can not provide stable and sustainable power resources for society,affecting the pace of social production and people’s quality of life.At present,the operation mode adopted by domestic power grid companies is to use electricity first and then pay for it.This will lead to the problem of electricity arrears and the increasing amount of electricity arrears.In the huge electricity consumption data of power companies,the power consumption of high-voltage customers accounts for an astonishing proportion,and their monthly electricity charges account for more than 70%of the total electricity charges collected by power companies,which has a very large potential risk of electricity tariff recovery.Once the phenomenon of default of payment,arrears and increasing amount of arrears occurs,it will bring irreparable losses to power companies and seriously restrict the production and development of power companies.Therefore,it is imminent to realize the early warning of the risk of high-voltage electricity customers’ electricity bill arrears.In order to establish an early warning model for the risk of high-voltage electricity customers’ electricity bill arrears,this paper attempts to start with the basic electricity data.Specific research work is as follows:Firstly,the background and significance of this topic are analyzed in detail,and the research status and main research results of electricity customer default risk early warning are explained.This paper analyses the existing problems and shortcomings of the research results,and expounds the necessity and practical significance of establishing the early warning system for the risk of high-voltage electricity customers’ electricity arrears.Pearson correlation coefficient method was used to screen the seven main factors affecting electricity bill arrears,and the dimension effects among different variables were eliminated by data standardization.Then,according to these seven main factors,we use neural network to predict the risk score value of high-voltage customers,and then use the risk score value and seven main factors to establish logistic regression to predict the risk of high-voltage customers,change the risk management model from labor-intensive to technical,and timely warn high-risk electricity customers.Experiments show that the hybrid forecasting model of neural network and logistic regression established in this paper has high forecasting accuracy and good consistency with the actual situation.It has certain practical significance.It can accurately predict the high-voltage risky customers and avoid the irreparable losses caused by default of high-voltage users to power companies.更多还原