返回
类型 应用研究 预答辩日期 2018-03-11
开始(开题)日期 2015-06-11 论文结束日期 2017-12-19
地点 动力楼#102 论文选题来源 省(自治区、直辖市)项目    论文字数 7.64 (万字)
题目 智能电网环境下电力数据挖掘研究
主题词 数据挖掘,故障诊断,负荷区间预测,负荷调度,需求响应
摘要 针对日益严峻的全球能源危机,各个国家相继在智能电网上开展研究。若要实现智能电网运行的安全性与可靠性,需要对遍布于整个智能电网的数据进行全方位的实时采集、传输和存储,并加以解析。其中,数据解析是保证智能电网正常运作的一个重要技术支撑,是对电力数据分析处理的理论方法;通过挖掘隐藏在电网中的未知信息、深层次剖析电力数据等方式,有效地推动智能电网的建设。数据挖掘是数据解析在智能电网中的主要应用形式,可用于负荷预测、需求响应、电网灾难预警、电气设备状态监测等领域。 由于参加博士联合培养项目,本人在东南大学和北卡罗莱纳州立大学分别针对电气设备和商业楼宇负荷进行研究。电气设备是智能电网的硬件支撑,常见的数据解析形式是状态评估或故障诊断。变压器和高压断路器是本文在电气设备方面主要的研究对象,前者承担电网能量的转换与传输,后者承担电力系统的运行维护。这两类设备的故障会严重影响智能电网的运行,因此本文选择对其智能化故障诊断进行研究;通过人工智能技术对故障样本的自动分析和学习,可有效地提高电气设备运行的可靠性,同时促进电网智能化发展。 商业楼宇拥有大量室内电器,并且是楼宇型分布式能源的主要安装对象,其负荷解析有助于提高电力系统能效与经济性,常见的数据解析形式是负荷预测、负荷调度等。负荷区间预测、需求响应和负荷调度是本文在商业楼宇方面的主要研究对象。负荷区间预测描述预测结果的可能范围,帮助决策人员在负荷调度时了解未来的不确定性;需求响应的操作是:根据市场价格信号或者激励机制,电力用户调整自己原有的用电模式;负荷调度通过控制能源终端,可以加强电网安全运行和电力资源配置的能力。本文的具体工作如下: 为了解决三比值法在变压器故障诊断中出现的问题,本文提出一种基于SVM多分类概率输出与证据理论融合的变压器综合诊断方法。证据理论通过证据融合将多种判据结果归纳为统一的结论,适用于不确定性分析。SVM多分类概率输出可将SVM特有的硬判决输出转化为软判决输出,为证据理论的数据融合工作提供客观的基本概率分配函数。油中溶解气体的实例分析结果证明该方法具有较高的可靠性、普适性和识别率。 当前高压断路器的故障诊断方法少有未知故障识别和模型分类边界实时更新的研究,本文结合粒子群算法(PSO)、支持向量域(SVDD)和核模糊聚类算法(KFCM),提出一种基于P-SVDD和P-KFCM的高压断路器自适应故障诊断方法。PSO-SVDD检测样本是否属于未知故障;属于已知故障的样本由PSO-KFCM判断其故障类型;PSO-KFCM联立改进划分系数聚类有效性分析(MPC)学习新故障。本文采用线圈电流样本进行故障模拟实验,通过与几种故障诊断算法的比较证明本文方法的有效性。 当前关于短期负荷预测的方法多数属于点预测范畴,给出确定性数值预测结果,缺乏对预测负荷波动范围的描述;同时关于各类小区域型电力用户(居民、商业、工业等)的负荷预测研究相对偏少。针对上述问题,本文提出一种基于相似日和核密度估计的日前商业负荷区间预测。相似日单元将负荷点预测过程分解为日平均负荷预测与标幺曲线搜索,最终每个时间节点获得多个预测值;核密度估计单元将各时间节点的预测值集合成一个负荷预测区间。负荷点预测过程的分解保证两个分解模块在求解过程中互不影响;核密度估计的选择原因是其函数形式自由及不依赖样本分布。仿真数据来源于北卡罗莱纳州立大学百年校区,仿真结果表明该算法具有普适性、强鲁棒性等优点。 商业楼宇月度电费通常按峰谷分时(TOU)电价结算,楼宇管理员可以通过需求响应和负荷调度等方式优化楼宇月度电费,然而国内外在这方面的研究偏少。因此,在分布式能源和楼宇用户终端的基础上,本文提出一种商业楼宇日前负荷调度方法,并嵌入一种即插即用型需求响应(P&P-DR)算法。在日前负荷调度中,三种日优化模式的设计目的是最小化楼宇月负荷峰值,同时兼顾优化用量电费。当楼宇在某月实施这个调度方法时,楼宇每天需要选择一种合适的日优化模式,三种优化模式的切换和配合可以有效地削减楼宇当月的需量电费和用量电费。另外,P&P-DR算法凭借惩罚函数决定转移负荷的操作。本文提出的DR算法无需考虑楼宇的可控负荷配置,凭借其互通性和易实施性,可轻松地适用于任何商业楼宇。实验仿真以北卡罗莱纳州立大学百年校区为例(其电价模式来自Duke Energy公司的商业楼宇TOU机制),实验结果证明该算法的可行性和有效性。
英文题目 Research on power data mining in the smart grid
英文主题词 data mining, fault diagnosis, load interval forecast, load scheduling,demand response
英文摘要 Because of the severe global energy problems, every country has been putting forward to the construction of smart grid. To keep the smart grid operation under safety and reliability, it is necessary to real-time acquire, transfer, storage and analyze the data flow in the power grid. Data analysis that is defined as the theory method of data processing and analyzing is an important technical support for smart grid. It pushes forward the development of smart grid by the ways of digging out the information hidden in the power grid and the deep-level analysis of power data. The most common analytic methods include data mining, data analysis and optimization theory. These analytic methods can be used in the fields, such as load forecast, demand response, power grid disaster warning, electrical equipment condition monitoring, etc.. Because I used to particpate in the PhD joint program, electrical equipment and commercial building load are the research points at the Southeast University and North Carolina State University (NCSU), respecrtively. Electrical equipment is the hardware support of smart grid, of which the major forms of data analysis are condition assessment and fault diagnosis. This paper selects transformer and high voltage circuit breaker (HVCB) to be discussed. The former transforms voltages and transmits power energy while the latter plays an important role in control and protection of power transmission system. These two equipments are essential to the operation of smart grid so that it is necessary to do research on intelligent fault diagnosis. Through automatic analysis and self-study of electrical equipment data, intelligent fault diagnosis could improve the reliability of electrical equipments and promote the development of smart grid. There exists lots of electrical appliances and distributed energy resources on the commercial buildings so that data analysis of commercial building, such as load forecast, load dispatch, can enhance the energy efficiency and economy of power system. This paper selects load interval forecast, demand response and load dispatch to be discussed. Load interval forecast describes the fluctuating range of predicted results, helping managers master the load uncertainty in future. Demand response refers to a tariff or program coordinated with power market conditions for motivating changes in electricity consumptions by end-use customers. With the centralized control of energy terminal, load dispatch strengthens the stability of smart grid and the ability of power resources allocation. The followings are the content of this paper In order to solve the problems, which arise in three ratio method applied in transformer fault diagnosis, this paper proposes a comprehensive fault diagnosis based on SVM multi-class probability output and evidence theory. Evidence theory is suitable for uncertainty analysis, fusing various criterions into one uniform conclusion by means of evidence fusion formula. With the help of SVM multi-class probabilistic output, SVM output turns from hard decision to soft decision, providing evidential theory with basic probability assignment. Case study based on dissolved gas validates the reliability, universality and accuracy of the proposed method. The current research on high voltage circuit break (HVCB) fault diagnosis lacks of unknown fault detection and real-time model classification updating. This paper proposes an adaptive fault diagnosis based on particle swarm optimization (PSO), support vector domain (SVDD) and fuzzy kernel clustering algorithm (KFCM). In the proposed method, PSO-SVDD can detect the unknown fault sample while P-KFCM and modified partition coefficient (MPC) cluster validity analysis are used in known sample category recognition and new fault learning. The simulation results based on HVCB closing coil current demonstrate the effectiveness of the proposed method, compared with the existing algorithms. At present, the most of short term load forecast (STLF) belong to point load forecast and cannot give the fluctuation range of predicted load. To solve the above problems, this paper proposes a day-ahead commercial load interval forecast based on similar day and kernel density estimation. Similar day unit decomposes point load forecast into daily average load prediction and pre-unit curve search and provides each time node with multiple predictive values. Then, these multiple predictive values are synthesized into load interval by kernel density estimation unit. The decomposition of point load forecast ensures that the two decomposition modules will not interfere with each other. The reason for selecting kernel density estimation is due to its free-form and independence of the sample distribution. The simulation results collected from centennial campus at NCSU demonstrate the universality and strong robustness of the proposed method. Few load scheduling strategies focus on minimizing the monthly electrical bill under the time-of-use (TOU) rate, and this has received wide attentions from commercial building managers. We present a plug-and-play demand response algorithm for day-ahead commercial building load scheduling on the basis of distributed energy resources and building end-uses. Three daily optimal modes are designed for minimizing the monthly peak load during on-peak hours with consideration of optimizing the daily energy charge. When implementing this scheduling method for everyday in one month, one suitable mode is selected for each day. The use of switching and cooperation among modes can save money both in terms of the monthly charge of energy and demand. Moreover, plug-and-play DR uses a penalty function to adjust the amount of building end-use loads to be shifted before or after the targeted hours. The proposed DR, which is resource agnostic, can be as a generic control method for enhancing building energy management systems. Data collected from the Centennial Campus at North Carolina State University and the Duke Energy commercial TOU rate are used for the case studies. The simulation results show that the proposed method is satisfactory and robust.
学术讨论
主办单位时间地点报告人报告主题
课题组讨论会 2013.09.07 动力楼303 朱克东 高压开关柜在线监测与故障诊断装置研究报告
课题组讨论会 2013.09.16 动力楼303 朱克东 断路器实验组项目汇报
东南大学电气工程学院 2013.12.20 动力楼318 Costinett Analysis and Design of High Efficiency, High Conversion Ratio DC-DC Power Converters
课题组讨论会 2014.03.05 动力楼303 朱克东 基于支持向量回归模型的多参量变压器固体绝缘与剩余寿命预测技术研究
东南大学电气工程学院 2014.05.14 逸夫科技馆一楼多功能厅 Sir Michael Sterling Towards a Smart Grid
东南大学电气工程学院 2014.05.22 动力楼316 黄少聪 A Series of Exponential Step-Down SC Converters and Their Applications in Two-Stage Converters
东南大学电气工程学院 2014.06.25 动力楼316 Leon M. Tolbert CURRENT’s Reconfigurable Grid Emulator and Power Electronics Research at The University of Tennessee
课题组讨论会 2015.01.06 动力楼303 朱克东 高压断路器故障诊断及算法鲁棒性研究
     
学术会议
会议名称时间地点本人报告本人报告题目
东南大学 2015.04.14 前工院107 基于后验概率支持向量机与证据理论融合的变压器故障诊断
ICEEP组委会 2016.09.17-2016.09.18 深圳 Optimization method of hybrid energy storage capacity of wind farm
美国北卡罗莱纳州立大学 2015.09-2017.09 美国,北卡州,罗利市 Design of a Plug-and-Play Demand Response Algorithm for Day-ahead Commercial Building Load Scheduling Considering the Minimum Monthly Peak Load
江苏省国家电网 2017.09.10 南京 国家电网江苏省检项目验收会
     
代表作
论文名称
Adaptive fault diagnosis of HVCBs based on P-SVDD and P-KFCM
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
鞠平 正高 教授 博导 河海大学
姚建国 正高 教授级高工 博导 中国电力科学研究院
赵剑锋 正高 教授 博导 东南大学
金龙 正高 教授 博导 东南大学
吴在军 正高 教授 博导 东南大学
      
答辩秘书信息
姓名职称工作单位备注
杨燕 其他 讲师 东南大学