基于数据挖掘的能源管理平台的研究

发布时间：2018-06-24 19:16

本文选题：数据挖掘 + 主成分分析　；参考：《北京建筑大学》2017年硕士论文

【摘要】：近年来,世界各国均将建筑节能减排工作视为经济发展的一项核心内容。中国的建筑能耗大约占社会总体能耗的33%,但是随着建筑物数量的增加以及人类居住洁净度舒适度的提高,建筑能耗仍将呈继续上升的趋向,因此建筑节能问题也是中国节能减排的重中之重。目前,我国进行了一系列有关建筑节能减排的工作,许多科研工作者也都在建筑能耗监测、能耗节能分析与能源故障诊断等方面展开了许多工作。本文将采用数据挖掘技术,从建筑能源管理平台中提取建筑能耗变量数据进行建模,并以综合性办公建筑为研究对象,运用建筑能耗故障诊断模型,对该建筑使用过程中的能耗异常扰动进行识别和诊断研究,及时通知运行维护人员能耗异常事件和系统故障所在,最终达到消除故障的目的。建筑能耗监控与诊断系统对于提升能源使用效率、保障系统的运行过程、保障设备和人员的安全等,具有非常重要的现实意义。首先,主成分分析是多元统计分析里面使用最广泛的方法,选定主成分分析法作为本文数据挖掘的方法,确定最优主成分个数的选择方法,确定主成分分析模型的建立规则,确定基于主成分分析的故障诊断的统计量和控制限的计算方法,基于MATLAB程序实现主成分分析;其次,利用Skyspark软件创建智慧建筑能源管理平台,把建筑中所有用能设备集中于该平台,并且该智慧建筑还建立了气象站,可以采集室外温度、室外风速、室外湿度和室外PM_(2.5)浓度。在此基础上,按照能源种类和用途对建筑系统能耗统计和监测,选取耗冷量、耗水量、空调用电量、照明用电量、景观用电量、动力用电量、生活用电量、生产用电量和商业用电量,进而分析能耗使用情况,并实现能耗统计与节能分析的展示,完善智慧能源管理平台;再次,选取连续完整的100天建筑能耗输入变量相关数据,运用MATLAB软件建立智慧建筑能耗系统故障诊断主成分模型。根据实际变量数据与主成分模型进行分析和对比,当累计方差贡献率CPV(k)=87.046%,主成分个数k=7,置信度α=99%,UCL=21.0524,Q=2.4262时,建立智慧能耗系统故障诊断模型,此时的诊断模型与实际过程最为吻合;最后,通过能源管理平台采集建筑系统全年365天的能耗数据,对已建立的智慧建筑能耗系统故障诊断模型进行应用,并建立相应的故障检测与诊断规则,发现T2统计量监控图在系统运行过程中超出其控制限UCL=21.0524的有11处,SPE统计量监控图在系统运行过程中超出其控制限Q=2.4262的较多,系统自动报警,从而判断故障产生的原因。利用上述研究成果,若建筑系统的运行能耗发生故障时,可以将系统变化特征与主成分模型进行故障匹配,结合匹配结果,最终可以达到故障检测与诊断。该结果为今后的建筑能源管理系统的故障诊断奠定了良好的基础。
[Abstract]:In recent years, countries all over the world regard building energy conservation and emission reduction as a core content of economic development. Building energy consumption in China accounts for about 33 percent of the total energy consumption in society. However, with the increase in the number of buildings and the improvement of human living cleanliness and comfort, building energy consumption will continue to rise. Therefore, the building energy-saving problem is also the top priority of energy-saving and emission reduction in China. At present, China has carried out a series of work on building energy conservation and emission reduction, and many researchers have also carried out a lot of work in building energy consumption monitoring, energy saving analysis and energy fault diagnosis. In this paper, data mining technology is used to extract building energy consumption variable data from building energy management platform for modeling. Taking comprehensive office building as research object, building energy consumption fault diagnosis model is used. The abnormal disturbance of energy consumption in the use of the building is identified and diagnosed, and the abnormal energy consumption event and the fault of the system are notified to the maintenance personnel in time, finally the purpose of eliminating the fault is achieved. The monitoring and diagnosis system of building energy consumption is of great practical significance for improving the efficiency of energy use, ensuring the operation process of the system, and ensuring the safety of equipment and personnel. First of all, principal component analysis (PCA) is the most widely used method in multivariate statistical analysis. Principal component analysis (PCA) is selected as the method of data mining in this paper. The calculation method of statistical quantity and control limit of fault diagnosis based on principal component analysis is determined, and the principal component analysis is realized based on MATLAB program. Secondly, the intelligent building energy management platform is created by using Skyspark software. All the energy-using equipments in the building are concentrated on the platform, and the intelligent building has also set up a weather station, which can collect outdoor temperature, outdoor wind speed, outdoor humidity and outdoor PM2.5 concentration. On this basis, according to the energy consumption statistics and monitoring of the building system according to the types and uses of energy, select cooling consumption, water consumption, air conditioning power consumption, lighting electricity consumption, landscape electricity consumption, power consumption, living electricity consumption, Production and commercial electricity consumption, and then analysis of energy consumption, and achieve energy statistics and energy conservation analysis display, improve the intelligent energy management platform; thirdly, select 100 days of building energy input variables related data, The principal component model for fault diagnosis of intelligent building energy consumption system is established by using MATLAB software. According to the actual variable data and principal component model, when the cumulative contribution rate of variance is 87.046, the number of principal components is 7, and the confidence degree is 21.0524Q = 2.4262, the fault diagnosis model of intelligent energy consumption system is established, and the diagnosis model is most consistent with the actual process. Finally, the energy consumption data of the building system are collected through the energy management platform, and the established fault diagnosis model of the intelligent building energy consumption system is applied, and the corresponding fault detection and diagnosis rules are established. It is found that there are 11 SPE statistic monitoring charts which exceed the control limit UCLG 21.0524 during the operation of the system. There are more SPE statistics monitoring charts than its control limit Q2.4262 during the system operation, and the system automatically alarm, thus judging the cause of the failure. Using the above research results, if the running energy consumption of the building system fails, the system change characteristics can be matched with the principal component model, and the fault detection and diagnosis can be achieved by combining the matching results. The result lays a good foundation for the fault diagnosis of building energy management system in the future.
【学位授予单位】：北京建筑大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU111.195;TP311.13

【参考文献】