CC BY
32
Applied Data Analysis in Energy Monitoring System
Kychkin А.V., Mikriukov G.P.
Perm National Research Polytechnic University Perm, Russia
Abstract. Software and hardware system organization is presented as an example for building energy monitoring of multi-sectional lighting and climate control / conditioning needs. System key feature is applied office energy data analysis that allows to provide each type of hardware localized work mode recognition. It is based on general energy consumption profile with following energy consumption and workload evaluation. Applied data analysis includes primary data processing block, smoothing filter, time stamp identification block, clusterization and classification blocks, state change detection block, statistical data calculation block. Time slot consumed energy value and slot time stamp are taken as work mode classification main parameters. Energy data applied analysis with HIL and OpenJEVis visualization system usage experimental research results for chosen time period has been provided. Energy consumption, workload calculation and eight different states identification has been executed for two lighting sections and one climate control / conditioning emulating system by integral energy consumption profile. Research has been supported by university internal grant №2016/PI-2 «Methodology development of monitoring and heat flow utilization as low potential company energy sources».
Keywords: energy data, energy monitoring, energy data visualization, applied data analysis, pattern recognition, OpenJEVis, Janitza.
Analiza aplicativa a datelor in sistemul de energomonitorizare Cacichin A.V., Micriucov G.P.
Universitatea Nationala de Cercetare Politehnica din Perm Perm, Federatia Rusa
Rezumat. Se studiaza un exemplu de organizare software §i hardware a sistemului de monitorizare a consumului de energie electrica in oficii la necesitati de iluminat multi-sectional §i de sistemul de climatizare/ventilare. Particularitatea siustemului consta in analiza aplicativa a datelor oficiului, ce permite recunoa§terea regimurilor localizate de functionare a fiecarui tip de utilaj in baza profilului energetic consolidat cu estimarea ulterioara a consumului de energie §i evolutii. Analiza aplicativa a datelor cu caracter energetic include blocul de prelucrare primara, filtru trece jos, blocul de identificare a marcii intervalului de timp, blocurile de clasterizare §i de clasificare a datelor, blocul de estimare a consumului netipic, blocul de calcul al consumului netipic, blocul de calcul al indicilor statistice de functionare in clase §i blocul de calcul al consumului de energie. in calitate de paramentri de baza a clasificarii regimului de functionare sunt utilizate valorile de energie consumata pentru un anumit interval de timp §i marca de timp a slotului. Se ofera estimarea rezultatelor cercetarilor experimentale ale analizei aplicative a datelor cu utilizarea echipamentului pentru modelarea seminaturala §i a sistemului de vizualizare OpenJEVis pentru perioada aleasa de observari. Pentru doua sectii de iluminat studiate §i a unei instalatii, care imiteaza sistemul de dirijare a ventilatiei, utilizand curba de sarcina integrala a fost efectuata identificarea de opt stari posibile de punere in functiune, a fost calculat consumul de energie §i evolutia. Lucrarea a fost efectuata cu suportul grantului Universitatii PNIPU №2016/PI-2 «Elaborararea metodologiei de monitorizare §i utilizare a fluxurilor termice, ca resursului de potential termic scazut al intreprinderii». Cuvinte-cheie: datele de consum de energie, monitorizare energetica, vizualizarea datelor, analiza aplicativa a datelor, recunoa§terea imaginilor, OpenJEVis, Janitza.
Прикладной анализ данных в системе энергомониторинга Кычкин А.В., Микрюков Г.П.
Пермский национальный исследовательский политехнический университет
Пермь, Россия
Аннотация. Рассматривается пример программно-аппаратной организации системы мониторинга потребления электрической энергии офисными помещениями на нужды многосекционного освещения и управления климатом/вентиляцией. Особенностью системы является прикладной анализ энергетических данных офиса, позволяющий произвести распознавание локализованных режимов работы каждого типа оборудования на основе сводного энергетического профиля с последующей оценкой энергопотребления и наработки. Прикладной анализ энергоданных включает блок первичной обработки, сглаживающий
фильтр, блок идентификации временной метки, блоки кластеризации и классификации данных, блок оценки нетипичного потребления, блок расчета статистических показателей наработки в классах и энергопотребления. В качестве основных параметров классификации режима работы использованы величины затраченной электроэнергии за один временной слот и временная метка слота. Приводится оценка результатов экспериментальных исследований прикладного анализа энергоданных с использованием оборудования для полунатурного моделирования и системы визуализации ОрепШУ18 за выбранный период наблюдения. Для исследуемых двух секций освещения и одной установки, имитирующей систему управления вентиляцией, по интегральному профилю нагрузки проведена идентификация восьми возможных состояний включения, проведен расчет энергопотребления и наработки. Работа выполнена при поддержке внутривузовского гранта ПНИПУ №2016/ПИ-2 «Разработка методологии мониторинга и утилизации тепловых потоков, как низкопотенциального ресурса предприятия».
Ключевые слова: энергоданные, энергетический мониторинг, визуализация энергоданных, прикладной анализ данных, распознавание образов, ОрепШУ1з, Тап^а.
Introduction
Energy monitoring systems (EMS) are widely used in manufacturing, housing, transport, and other spheres in order to automate data acquisition, transmission, storage and analysis, energy loss estimation.
Building energy management system potential is in possibility of main consumers groups pattern recognition, such as lighting and climate control/conditioning systems [1].
The main component of building EMS is data analysis system (DAS), which is implemented on the «client-server» architecture with usage of available information analysis platforms. Mathematical module of DAS is usually presented by fixed set of standard statistical functions, which are used by energy services staff.
Lighting system separate sections workload evaluation task is being solved rarely. This caused by irregular exploitation of separate sections by staff. Such sections usage irregularity caused by change of sunlight fall direction during the day and human factor. Moreover, climate control/conditioning management system can be interesting for current research among other office consumers. Thus, for specified groups of consumers recognition implementation DAS functionality should be expanded for complex automated energy data consumption analysis implementation, of decision making on energy consumption performance increasing [2]. This could be achieved by adding new analytical blocks for classifying and clustering [3, 4].
1. Methodology
The main objective of the research is to create DAS analytical apparatus for office
P P11 P12
P2 P21 P22
> = <
Pn. Pn1 Pn2,
buildings aimed at numerical values obtaining, statistics calculation and additional information forming on lighting and climate control/conditioning system energy consumption of single consumer or group of consumers detected by general consumption profile [5].
Common office energy data can be characterized by the following set:
P = \ f = \ f - set of energy data
p°ints Pi = {Pi1 Pi2 } ,
where, n — number of measures, pn — time stamp (time values ranged from 0:00 to 23:45 with 15 minutes step, pi1 e{l..96}), pi2 —
energy consumption for period i.
The main tasks of DAS are to identify monitoring consumer workload states, and to assign appropriate class for each energy data value:
S = {S1, S 2,..., Sn} — full set of lighting and climate control/conditioning work modes,
Si ^ K,, j = 1..k
i j ^ ,
K = {K1, K2,..., Kk } — reduced work modes set,
where k — typical work modes combinations number.
The implementation of the tasks mentioned above is provided by following DAS functions [6,7]:
- primary data processing: filtering and
smoothing for state change detection;
- time stamps definition, clusterization
and classification;
- statistical parameters calculation for
result evaluation.
DAS functioning algorithm is shown at Figure 3.
Initially, lighting and climate control/conditioning system energy data is being filtered (Fj) by threshold criteria (Eq. 1). This step is needed to exclude errors of measuring, reading, recording, CRC etc. Further, filtered data xi goes to the other units of DAS for the analysis.
F1: Xm,n < X < Xm
(1)
where x, - filtered energy data point, xmin and xmax - system defined minimal and maximal values.
Data smoothing (F2) is implemented by the moving average method (Eq. 2) as having a low number of smoothing points results in the graph with the closest form to the original one. On the contrary, large number of smoothing points heavily distorts the form of the graph [8].
F2: xs, =■
1
2N +1
i+k
(2)
k=-N
where 2N+1 - the number of smoothing energy data points, xsi - is smoothed energy data point.
State change (F3) of monitored system (for instance, change of working energy consumers combination) is detected by smoothed and filtered data comparing (Eq. 3). Received data used for classification accuracy evaluation and work mode of lighting and climate control/conditioning system definition [9, 10].
F3: Ax, = x, - xs,
(3)
Time stamps definition block (F4) converts time into numerical value (Eq. 4). Time values ranged from 0:00 to 23:45 with 15 minutes step (96 values per day) [11, 12].
F4: tt = getTime(xt), tt e {1 ..96} (4)
Clusterization (F5) is carried out according to the k-means method (Eq. 6).
Number of required lighting and climate control/conditioning system work mode clusters is determined by Eq. 5. K-means aimed to minimize standard deviation of cluster points from cluster centroids:
k = 2m
F5: V = £ -c,)2 ^
j=1 XiGSj
min
(5)
(6)
where m - number of real consumers, V -standard deviation, k - is the number of lighting and climate control/conditioning system working combinations, Sj - cluster j, Cj - is centroid of xi e Sj [13].
Consumers classification (F6) is carried out on the basis of clusterization (Eq. 7). Figure 4 shows obtaining and visualization of information on energy consumption.
Total consumption is determined by the consumption of particular consumers
n
E = ^ Ej . Every work mode consumer class
j=1
fj can be represented through the simplified model in the following way:
F6: f, =
0,12 < t < tj Cj , t! < t < 12
(7)
where Cj - is amount of energy consumption (centre of cluster). Time intervals ti and t2 are determined by values of cluster edge elements.
Statistical data calculation block (F7) required for object workload computation for different lighting and climate control/conditioning system work modes (Eq. 8).
t,
F7: wl, =
15-11
, =1 t.
t
100%
(8)
total
where tt : xi c Sj, ttotai - duration of analyzed
period in minutes.
Result evaluation block (F7) compares results of consumers classification (F6) and state change detection (F3) and provides percentage of accuracy. Moreover it calculates total worktime (Eq. 9) and consumption (Eq. 10) for every lighting and climate control/conditioning system work mode [14].
n t
F8: twt, = 15 -L (9)
i=1 ti nt
F8: tc, = 15 (10)
i =1 ti
2. Technology description
The developed EMS has two levels: first level — data collection, second level — data processing and storage. Its block diagram is shown at Figure 5.
EMS hardware is based on power analysers UMG104 and UMG604 by Janitza electronics GmbH. These devices are used for reading data and transmitting it to the server. Their main characteristics are shown in Table I.
Software of EMS includes 2 software products:
- GridVis — used for remote setting of
power analysers and data processing for transmitting it to storage server [15].
- OpenJEVis — basis of EMS software. It
consists of various open source program units. This software supports plug-in system, which allows to expand its abilities. Moreover, it has wide integration capabilities and compatibility with other program products [16, 17].
EMS analytical functionality is provided by its integration with complex mathematical software such as MATLAB or Octave via OpenJEVis plug-in system [18, 19, 20].
3. Results
Experimental research was carried out in the typical office using 2 sectional lighting and laboratory equipment for climate control / conditioning (CC/C) HIL simulation [4].
Research was provided with hardware which is shown on Figures 1 and 2 by exporting the office energy data to EMS within the period from 20.11.2015 to 20.12.2015.
Lighting and climate control/ conditioning system energy consumption are presented as a graph of the open source energy visualization tool — OpenJEVis at Figure 6. Filtered total energy consumption (F1) is presented by blue colour.
UMG 104 power analyser was used as a data collector, UMG 604 — as a protocol converter (Modbus RTU/Ethernet) and «data concentrator».
Fig 1. UMG 604 power analyzer (data concentrator) installed into climate/conditioning equipment HIL simulator
Fig 2. UMG 104 power analyzer (data collector) installed into lighting control panel
Office workload analysis parameters are shown in Table III. Maximum and average consumption values for the whole period: Pmax = 2049,67W - h, Pav = 352,03W - h.
Smoothed energy data (F2) is shown on Figure 6 by green colour.
Table I. Power analysers characteristics
Parameter UMG 104 UMG 604
4 inputs for measurement + +
current &voltage
RS-485& RS-232 + +
interfaces
Ethernet port - +
Built-in memory 4 Mb 128 Mb
web interface - +
Modbus master - +
On the difference of original and smoothed graphs, state change graph has been obtained. Ax is used to detect consumers consumption modes and transitions between these modes. This is shown at Figure 7.
Current energy consumption mode can be recurrently (or cumulatively) defined. Transitions from higher consumption mode to lower can be identified by positive Ax value, from lower to higher - by negative Ax value. Insignificant Ax values (Ax<100) means consumption changes inside current consumption mode.
State change detection result was used for classification accuracy checking and confirmation.
After time stamps definition for each data point, clusterization and classification have been provided. All energy consumers in the office can be divided into three main
groups, which can consist of some smaller consumers: lighting (70%), lighting (30%), climate control system (conditioning). Altogether there are 23=8 possible combinations of these consumers switching. Hence, it is necessary to define eight classes and clusters. The results of the clusterization are shown at Figure 8. Time intervals are numbered from 1 to 96 per day (frequency of measuring was 15 min). Time intervals for each consumer can be determined on clusterization results.
Coordinates of cluster centroids are used in classification. Classification results are shown in Table II. For example, full lighting usage can be described by modes 7 and 8. All classes can be described by the following formula: ft = ci, t1 < t < 12.
Table II. Classification (F6) of lighting and climate control / conditioning modes
Class Lighting 70% Lighting 30% CC / C ci9 Wh ti t2
1 0 0 0 4.7 0 96
2 0 0 1 246.1 31 82
3 0 1 0 433.6 35 80
4 0 1 1 704.9 33 90
5 1 0 0 1083.1 31 88
6 1 0 1 1235.7 38 91
7 1 1 0 1484.5 39 89
8 1 1 1 1706.1 42 66
Table III. Office workload (F7) within the period from 20.11.2015 to 20.12.2015
Class Number of energy data points Time, min. Part of total time Average consumption, Wh Number of energy data points Time, min. Part of total time Average consumption, Wh
All time (0:00-23:59) Working hours (8:00-20:00)
1 2156 32340 72.45% 4.76 689 10335 46.30% 5.03
2 24 360 0.81% 246.04 23 345 1.55% 250.76
3 25 375 0.84% 433.62 25 375 1.68% 433.62
4 62 930 2.08% 704.97 59 885 3.97% 700.25
5 142 2130 4.77% 1083.09 140 2100 9.41% 1083.99
6 77 1155 2.59% 1235.71 67 1005 4.50% 1238.18
7 473 7095 15.89% 1484.46 468 7020 31.45% 1484.38
8 17 255 0.57% 1706.01 17 255 1.14% 1706.01
Total 2976 44640 100.00% - 1488 22320 100.00% -
All lighting (3-8) 796 11940 26.75% 1299.81 776 11640 52.15% 1302.27
All CC/C (2,4,6,8) 180 2700 6.05% 965.36 166 2490 11.16% 958.09
Fig 3. Workload calculation and states identification DAS algorithm
f=Ifi(t)
Fig 4. Obtaining and visualization of lighting and climate control/conditioning energy consumption S _r Remote
Fig 5. EMS structure based on the open source software
P, W.h
500
Mon 23.11.15
Tue 24.11.15
Wed 25.11.15
Thu 26.11.15
Fri 27.11.15
Sat 23.11.15
Sun 29.11.15
Fig 6. Filtered (Fi) office energy consumption data and smoothed (F2) data for one week
Fig 7. Difference (F3) between filtered and smoothed energy data for one week
Fig 8. Results of data cluster analysis (F5) which shows different lighting and climate control / conditioning energy consumption modes - classes 1..8 (see Table I) and their centers
Conclusion
The following results of this work have been obtained:
- the module structure and functioning principles of building energy monitoring DAS have been described. The lighting and climate control/conditioning system energy data
applied analysis includes primary data processing: filtering and smoothing for state change detection, time stamps definition, clusterization and classification, statistical parameters calculation for result evaluation;
- DAS key feature is applied office energy data analysis that allows to provide each type of lighting and climate control/conditioning hardware work mode recognition. It based on general energy consumption profile with following energy consumption and workload evaluation;
- two-sectional lighting and climate control (conditioning) HIL model energy consumption within the period of one month has been analysed on practice. Total office workload during working time was 54.91%, lightning workload was 52.15%, climate control/conditioning - 11.16%.
- energy consumers were classified according to clusterization data to eight classes. Recognized average energy consumption during working time for lighting was 1302.27 W-h, for climate control/conditioning - 958.09 W-h.
References
[1] Faizrakhmanov R.A., Frank T., Kychkin A.V., Fedorov A.B. Sustainable energy consumption control using the My-JEVis energy management data system // Russian Electrical Engineering. 82 (11), pp. 607-611 DOI: 10.3103/S1068371211110022.
[2] Khoroshev N.I., Kazantsev V.P. Management support of electroengineering equipment servicing based on the actual technical condition // Automation and Remote Control. 2015. Vol. 76. No 6. pp. 1058-1069.
[3] Kychkin A.V. Intelligent information and diagnostic system for examining blood vessels // Journal of Computer and Systems Sciences International. 2013. L 52. № 3. Pp. 439-448.
[4] Petrochenkov A.B., Frank T., Romodin A.V., Kychkin A.V. Hardware-in-the-loop simulation of an active-adaptive power grid // Russian Electrical Engineering. 2013. L 84. № 11. Pp. 652-658.
[5] Claridge, D. E. (1998). "A Perspective on Methods for Analysis of Measured Energy Data from Commercial Buildings." Journal of Solar Energy Engineering 120(3): 150155.
[6] Jacobsen, F.R. (1985). Energy signature and energy monitoring in building energy management systems. Proceeding of
CLIMA 2000 World Congress, vol. 3: Energy Management, p. 25-31.
[7] L. Prerez-Lombard, J. Ortiz, and C. Pout, "A review on buildingsenergy consumption information," Energy and Buildings, vol.40, no. 3, pp. 394-398, 2008.
[8] Egorov A. V., Kuprijanova N.I. Data clustering methods features // Izvestija JuFU. Technical sciences. - 2011. - №11 -p. 174-178.
[9] J. E. Seem, "Using intelligent data analysis to detect abnormal energy consumption in buildings," Energy and Buildings, vol. 39, no. 1, pp. 52-58, 2007.
[10] D. Liu, Q. Chen, K. Mori, and Y. Kida, "A method for detecting abnormal electricity energy consumption in buildings," Journal of Computational Information Systems, vol. 6, no. 14, pp. 4887-4895, 2010.
[11] Firth, S., K. Lomas, et al. (2008). "Identifying trends in the use of domestic appliances from household electricity consumption measurements." Energy and Buildings 40: 926-936.
[12] J. E. Seem, "Pattern recognition algorithmfor determining days of the week with similar energy consumption profiles," Energy and Buildings, vol. 37, no. 2, pp. 127-139, 2005.
[13] L. C. Braga, A. R. Braga, and C. M. P. Braga, "On the characterization and monitoring of building energy demand using statistical process controlmethodologies," Energy and Buildings, vol. 65, pp. 205-219, 2013.
[14] L. Zhao, J. Zhang, and R. Liang, "Data acquisition and transmission system for building energy consumption monitoring," Abstract and Applied Analysis, vol. 2013, Article ID 613043, 8 pages, 2013.
[15] Janitza - Home. - URL: http://www.janitza.com/ (accessed: 7.04.2016).
[16] Envidatec GmbH in Russia [Envidatec GmbH web-page]. - URL: www.envidatec-ost.ru (accessed: 7.04.2016).
[17] OpenJEVis - The Open Data Monitoring and Storage Solution. - URL: http://www.OpenJEVis.org (accessed: 7.04.2016).
[18] Kychkin A.V. Synthesizing a System for Remote Energy Monitoring in Manufacturing // Metallurgist. 2016. Vol. 59. Iss. 9-10. DOI: 10.1007/s11015-016-0170-5
[19] Lyakhomskii A.V., Perfil'eva E.N., Kychkin A. V., Genrikh N.A software-hardware system of remote monitoring and analysis of the energy data // Russian Electrical
Engineering 86 (6), pp. 314-319. DOI: 10.3103/S1068371215060103 [20] Kychkin A. V., Musikhina K.G., M.G. Razepina. Research the effectiveness of creation and implementation of energy management system in an industrial plant. PNRPU Bulletin. Electrotechnics.
Kychkin Aleksey Vladimirovich,
Associate Professor, Microprocessor Means of Automation Department Electrical Engineering Faculty, Perm National Research Polytechnic University
E-mail: aleksey .kychkin@gmail.com
Mikriukov Georgii Pavlovich,
Master student, Microprocessor Means of Automation Department
Electrical Engineering Faculty
Perm National Research Polytechnic University
E-mail: mikriukov.georgii@yandex.ru
Informational Technologies. Control Systems. 2014. № 1 (9). P. 66-79.
About authors:
