Energies 2017 , 10, x doi: FOR PEER REVIEW www.mdpi.comjournal energies [616843]

Energies 2017 , 10, x; doi: FOR PEER REVIEW www.mdpi.com/journal/ energies
Article 1
Heating performance and part load ratio 2
characteristics of boiler staging in office building 3
Da Young Lee 1, Byeong Mo Seo 1, Yeo Beom Yoon 2, Kwang Ho Lee 3,* 4
1 Department of Architectural Engineering, Hanbat National University , San 16-1, 5
Dukmyung-Dong, Yuseong-Gu, Daejeon 305-719, Korea 6
2 College of Design, North Carolina State University, 50 Pullen Rd., Campus Box 7701, NC 27695 -7701, U.S.A 7
3 Department of Arc hitectural Engineering, Hanbat National University , San 16-1, 8
Dukmyung-Dong, Yuseong-Gu, Daejeon 305-719, Korea 9
* Author to whom correspondence should be addressed; E -Mail: [anonimizat] ; 10
Tel.: + 82-42-821 -1126; Fax: + 82-42-821-1590. 11
Academic Editor: name 12
Received: date; Accepted: date; Published: date 13
Abstract: Commercial buildings in Korea account for about 30% of the total energy and occupy a 14
large portion. Commercial buildings account for significant portions of the total building energy in 15
Korea and thus a variety of research on the boiler operation was carried out thus far. However, most 16
of the researches were carried out on the Boiler itself, i.e., the part load ratio characteristics and the 17
corresponding Gas energy consumption patterns w ere not analyzed in the existing studies. In this 18
study, the part load ratio and the operating characteristics of gas boiler were anal yzed within an 19
office building equipped with the conventional variable air volume system. In addition, the gas 20
consumption among different boiler staging schemes was comparatively analy zed. As a result, 21
significant portions of total operating hours, heating load and energy consumption was found to be 22
in the part load ratio range of 0 through 40% and thus the energy consumptio n was significantly 23
affected by Boiler efficiency at low part load conditions. It is necessarily means that the boiler 24
operation at the part load is an important factor in commercial buildings. In addition, utilizing 25
sequential boiler staging scheme can sa ve gas usage about 7%. For heating energy saving, applying 26
the sequential control boiler with 3:7 proportion staging during the annual period applying is 27
considered to be the optimal control algorithm for maximum efficiency of boilers. 28
Keywords: EnergyPlus; Boiler; Part Load Ratio; Gas consumption; Office building; Boiler staging 29
30
1. Introduction 31
In December 2015, international society has adopted the Paris Agreement built upon the 32
convention on the climate changes as a way of reducing green -house gases. Hereupon, Korea has 33
decided to pursue a goal of reducing emissions of green -house gases by 37% from business- as-usual 34
(BAU) emissions of 850.6 million ton CO2e until 2030 [1]. Korea has been chosen as one of the top 10 35
countries with the highest level of emissions of green- house gases and using fossil fuel as major 36
energy source in major industries and daily lives in general [2]. Therefore, efforts are required in all 37
fields in society to reduce the level of emissions of green -house gases. In addition, the dependence 38
rate on energy imports in Korea was estimated to be 96% in 2013, and this indicated the highest 39
proportion among major countries in the world. The high dependence rate on energy imports means 40
that a significant amount of expenses is being used on importation of fossil fuels. Therefore, there is 41
a need to prepare measures on aforementioned issues [3]. 42
According to the results of Energy Consumption Survey on Buildings conducted in 2015 by 43
Korea Energy Economics Institute, the amount of energy used in the field of architecture among the 44

Energies 2017, 10, x FOR PEER R EVIEW 2 of 25
entire energy consumed currently found to be approximat ely 30%. It is currently inevitable to reduce 45
the energy used in buildings to reduce the amount of aforementioned green -house gases. In addition, 46
the energy consumption in commercial buildings has been in an increasing trend by 3% on annual 47
average from 20 00 to 2014 [4], and it was confirmed that 35% of the total energy consumed in the 48
field of commerce and public sector was found to be used in heating and production of hot water. In 49
regard of this issue, among the major energy source used in heating of bui ldings and production of 50
hot water in Korea, gas lines constituted 37%, and the necessity of reducing the gas consumption of 51
heat source devices has been emphasized to decrease the amount of heating energy in buildings [4]. 52
As for the practical plan of imp lementing them, an effort is being requested to reduce the amount of 53
gas used in buildings and heating energy consumption by making it obligatory to build zero -energy 54
buildings in the field of architecture in Korea and enhancing the development of new technology for 55
reducing green -house gases [1]. 56
In Korea, many studies have been conducted to reduce the amount of heating energy used in 57
buildings. However, previous studies conducted regarding building facility system focused on the 58
enhancement of performance of heat source devices including the condensing boiler, exhaust gas heat 59
collecting method, and installation of feed water pre -heater. In general, there has been an insufficient 60
amount of studies dealing with device capacity and operating characteristics for solving problems 61
including the increase in initial investment expenses, maintenance costs, and space of installation 62
caused by an excessive calculation of heat source devices [5]. As for the recent trend of studies in 63
regard of aforementioned issues, first of all, 64
Murray et al presents degree- days simulation technique coupled with a genetic algorithms 65
optimization procedure to propose optimal retrofit solutions. The results demonstrate the necessity 66
to carry out analysis of a project before retrofit w orks commence to ensure an optimal approach taken 67
in accordance with the project specific criteria [6]. Weissmann et al analyzed the PLR (Part Load 68
Ratio) index for the central heating system to quantify the effects of identified building and 69
demolition re lated characteristics on the heating load variability in residential areas in Germany. As 70
a result, the amount of consumed gas ended up decreasing when using the single plant with large 71
capacity, and a conclusion was drawn that the entire PLR of heat source devices was influenced by 72
the heating load to be processed. However, due to insufficient analysis on the efficiency in the area 73
with low load and the amount of consumed gas when operating single plant with large capacity and 74
individual plants, there has been lack of consideration on the judgment on the efficiency of overall 75
period [7]. Two of the previous studies used one heat source device and focused on the maximum 76
load when calculating the capacity. However, since the criteria of evaluation for efficie ncy on heat 77
source device is focused on the efficiency based on PLR when processing high load, there was a need 78
to assess the efficiency when the load was low. 79
Among the studies that controlled the number of two or more heat source devices for reducing 80
the energy, Giurca et al have focused on the selection of number and size of boilers required for the 81
heating units of the residential complexes. In this case study they proposed the selection of number 82
and size of boilers for a heating unit of a residential complex located in Tirgu -Mures, Romania. The 83
conclusion of this study is that the operational safety of the heating unit increases once the number 84
of boilers increases, together with the decrease of thermal energy consumption and the increase of 85
investmen t costs respectively. Based on their calculations, it is recommended that the heating unit 86
should be equipped with seven boilers, and should be provided with a controller for managing them 87
in cascade [8]. 88
Sin et al. have analyzed operating characteristics of heat source devices based on heating load of 89
buildings and evaluated the appropriateness of capacity of boilers installed in the buildings. In order 90
to achieve the goal of the research, this study has used combustion hours and the amount of 91
consumed ga s of three gas boilers in order to indirectly analyze operating characteristics of boilers 92
and heating load of buildings [9]. As a result, the calculated capacity of boilers found to be more than 93
twice the requested heating load. In conclusion, one boiler with capacity that could accept the 94
maximum instantaneous load for energy saving purpose was used as a standard, and it was 95
suggested to use two boilers in order to more swiftly remove the heating load. However, there was 96

Energies 2017, 10, x FOR PEER R EVIEW 3 of 25
not enough information on the boil er capacity in the suggested method, and the operating order has 97
not been assigned in the devices that the number was controlled. In other words, it is expected to be 98
difficult to accurately judge the efficiency based on the amount of consumed gas of devic es when 99
operating two heat source devices of which capacity and operating order were not assigned [9]. 100
Shide and Jia have introduced an intelligent and optimal control strategy of energy saving for 101
heat supplying process under four gas boiler group run c onditions in a residential district. In 102
addition, to keep boilers run under efficient state with rated load, the tracking Time- Discrete- Control 103
(TDC) and the optimization program of boiler group were further developed. As a result, boiler 104
group’s optimization program greatly reduces the gas consumption and ensures the thermal 105
supplying at the same time. According to statistics during several years run, natural gas consumption 106
was reduced 15% than before [10]. Wei presented a micro -CCHP (combined cooling, he ating and 107
power) system driven by biogas, while adopting the hybrid cooling mode. Moreover, a multi – 108
objective optimization model considering off -design performance of the facilities was proposed to 109
maximize the PESR(primary energy savings ratio) and minimi ze the energy costs. As a result, the 110
highest level of efficiency and the most economical energy costs were shown when PESRs were 111
15.03% and 15.06% in winter season, and thus the model was ultimately suggested. However, while 112
the overall energy costs were saved in winter season, the cost -efficiency was degraded from an 113
increase of heating energy due to the use of supplementary boiler. In addition, it is assumed that 114
there has been insufficient analysis regarding the supplementary boilers that operated in th e time 115
between summer and winter and the annual energy consumption efficiency [11]. 116
Among the studies regarding the amount of consumed energy, expense, and partial load factor 117
while controlling the capacity and number of heat source devices, Renato M. Laz zarin insisted that 118
to obtain the contributions granted by an act of promotion of the energy savings, the existing 119
atmospheric traditional boiler must be replaced with a condensing boiler without any further 120
specification. According to this Paper, Two mode ls of quite different modulating ratios and nominal 121
capacity were considered. As a result, two condensing boilers with lower capacity than that of the 122
previous boilers and with modulation factor of 1:9 were suggested as the best alternatives. Suggested 123
boilers have shown the efficiency that was 40% higher than the previous boilers while reducing the 124
primary energy by 15% [12]. In Liu et al’s study, an optimal strategy was proposed to improve the 125
chiller sequencing control for energy efficient and reliable operation in commercial buildings. In 126
addition, more comprehensive part -load performance curves for the chiller were obtained to predict 127
the operating performance of chiller more accurately. According to the load profiles in typical 128
working days, three diff erent chiller sequencing control strategies were compared for different 129
system configurations. As a result, up to approximately 21.2% of energy was saved [13]. The in situ 130
performance of three different chiller sequencing control strategies were compared a nd evaluated in 131
Sun et al’s study. All three control strategies were operating online in parallel in a super high -rise 132
building. The results of the baseline strategy, using both electrical current and cooling load direct 133
measurement, were taken for actual on-site applications. The overall performance of the improved 134
strategy, using fused cooling load measurement, was the best. It had the largest energy saving, up to 135
6.6% compared with the strategy using both electrical current and cooling load direct measur ement 136
[14]. Sun et al presents a strategy for improving the reliability and the energy efficiency of chiller 137
sequencing control based on the total cooling load measurement of centralized multiple centrifugal 138
chiller plants. The improved chiller sequencing control uses the fused measurement of building 139
cooling load to replace the direct measurement, computes online the chiller maximum cooling 140
capacity by a simplified model, and calibrates it according to the quality of the fused measurement. 141
Hereupon, when a pplying the improved measures in replacement of previous controlling plans, 0.8 142
% of energy was reduced when there was no error in the system, while 2.31 % of energy was reduced 143
when there was an error in the system [15]. Chang proposed a method that emplo ys the central 144
monitoring system to record data and to determine the chiller power consumption model, applying 145
ANN (Artificial Neural Network) technique to determine OCS (Optimal Chiller Sequencing) without 146
the requirement to measure the chilled water flow rates. According to the results of the comparison 147
on the amount of consumed electricity in three freezers operating in each representative day, the 148

Energies 2017, 10, x FOR PEER R EVIEW 4 of 25
expenses for consumed electricity were reduced according to the combination of operating order of 149
three freezers among the freezers number 1 to 6 used in the study. This was identified as a measure 150
of saving the expenses of 1.06 million NT $ during eight months compared to the past [16]. Previous 151
studies were turned out to insufficiently consider the operating order, calculation of capacity, and 152
load assigning methods of each device when controlling heat source devices as well as the efficiency 153
regarding the PLR (Part Load Ratio) of each device. 154
Aforementioned studies have focused on the processing of high load when calculating the 155
capacity of heat source devices and it was found that the efficiency of devices with low load was 156
insufficiently analyzed. On the other hand, Yu et al. have analyzed characteristics of part load in 157
residential buildings and prevented the excessive operation of the system in the low part load 158
condition by controlling the high instantaneous load caused by the time delay and dissatisfaction on 159
pre-set temperature when operating floor heating source devices. As a result, approximately 24.7 % 160
of energy was saved depending on the efficiency of part load. However, due to the difference in 161
schedules of using residential and office buildings and mostly applied heating system, additional 162
controlling method is required to apply them to the office b uildings [17]. In addition, Seo and Lee 163
have supplemented aforementioned studies, proceeding the chiller staging control in office 164
buildings. In addition, the amount of consumed energy was analyzed in each interval of part load. 165
When using the sequential d istribution method suggested in this study, 10% of the entire amount of 166
consumed energy was saved [18]. However, the study conducted by Seo and Lee focused on air – 167
conditioning and insufficiently analyzed the total energy consumed in both air -conditioning a nd 168
heating source devices. Thus, it is required to conduct research applying the efficient load 169
distribution method by controlling the number of heat source devices depending on the necessity of 170
reducing heating energy consumed in office buildings. 171
Theref ore, this study is a follow -up research of aforementioned study of Seo and Lee, and it was 172
conducted regarding the control of the number of boilers by partitioning the capacity according to 173
the characteristics of part load and applying various load distrib ution algorithms for efficient 174
consumption of heating energy used in office buildings. In addition, capacity of devices was 175
partitioned in three ways after analyzing the pattern of graphs of annually consumed amount of gas 176
and annually accumulated operatin g hours of boilers in the office. Moreover, the study aimed to 177
develop a method to save energy by applying load distribution algorithm systems using sequential 178
and uniform methods in order to increase the efficiency of heating energy. 179
2. Methods 180
Prior to p roceeding this study, the necessity of research has been proved through previous 181
studies. EnergyPlus v6.0 developed by US DOE (United States Department of Energy) was used for 182
this study, and while using the EnergyPlus, the heat balance method recommended by ASHRAE 183
(American Society of Heating, Refrigerating, and Air -Conditioning Engineers) was used for 184
interpreting the load through interpretation simulation programs of thermal environment and air – 185
conditioning/heating load in buildings. In particular, when interpreting the load, the degree of 186
freedom in internal heating elements and systems and plant setups is high, following the user’s 187
scheduled control, and detailed interpretation is available [19]. In addition, there are many 188
advantages including the avai lability of interpretation in dynamic simulations in abnormal 189
conditions by integrating the advantages of previous DOE -2 and BLAST. Moreover, in particular, the 190
components among modeling including zone, system, and plant modeling have huge advantage of 191
establishing the integrated relationship in terms of flexible connection. Therefore, they were chosen 192
as interpretative tools as they were appropriate for the purpose of this study. The following Fig. 1 193
indicates the study flow chart. There are five cases used in this study that were classified depending 194
on the boiler capacity and controlling method. In addition, optimal boiler controlling method has 195
been suggested through the analysis of representative days and annual data. 196
197

Energies 2017, 10, x FOR PEER R EVIEW 5 of 25
198
Figure 1. Study flow chart 199
3. Simulation condition 200
3.1 Simulation Model 201
This study proceeded the analysis after dividing the representative periods into intermediary 202
seasons (April, May, September, October) when air- conditioning and heating devices are used 203
toge ther and the winter season (November, December, January, February, March) when only heating 204
devices are used. As for weather data used in building model, the ones in the area of Incheon, 205
Republic of Korea, provided by EnergyPlus 6.0 were used, and April 17 th and January 23rd were 206
chosen as the representative days since the dates show the average temperature and load patterns 207
among the representative periods of intermediary seasons and winter seasons for the research. As for 208
the heating schedule on represent ative days, operating hours were chosen as 6 a.m. to 7 p.m. on 209
weekdays in consideration of internal heating factors in office buildings [20]. VAV system was used, 210
and boiler capacity was chosen to be 290,000W, which is the capacity that could handle the m aximum 211
heating load required on the relevant building. As for internal heating and schedule value, 212
9.3m2/person was set up on occupants based on ASHRAE 90.1, 2004, and 9.1W/m2 was set up for the 213
illumination. In addition, 14.4W/m2 was entered for electroni c devices [21]. Fig. 2 indicates the 214
internal heating schedule used in this study, and Table 1 indicates the simulation condition. Fig. 3 is 215
a simulation model that was used in this study, and large -sized office building used in the previous 216
study of Seo a nd Lee was utilized. This model was made according to ASHRAE 90.1, 2004, and is a 217
prototype model provided by EnergyPlus [21]. The size of building includes the width of 48m, length 218
of 73m, and height of 14m, and the window area proportion is 45%. 219
220

Energies 2017, 10, x FOR PEER R EVIEW 6 of 25
221
Figure 2. A Internal Heat Gain Schedule 222
223
Table 1. Simulation conditions 224
Input Elements
Representative day 17th April, 23th January
Terminal Unit Conventional VAV Box with Reheat
Boiler Capacity 290,000W
Heating Set -point(℃) Heating: 21 ℃
225
226
Figure 3. Simulation Model 227 0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Fraction [%]
HourOccupancy Equipment Lighting

Energies 2017, 10, x FOR PEER R EVIEW 7 of 25
3.2 Load Distribution Algorithm 228
Fig. 4 indicates load distribution algorithm. As shown in the figure, EnergyPlus v6.0 suggests 229
three types of load distribution algorithm including optimal, sequential, and uniform [18]. There is a 230
need to enter minimum part load ratio, maximum part load ratio, and optimum part load ratio values 231
in table 2 in priority when establishin g the boiler model in EnergyPlus. Optimal algorithm can 232
determine the priority in the system to distribute the load, and it can also designate the maximum 233
PLR of the operating device. In addition, heat source device operating in priority recognizes the 234
entered optimum part load ratio value as an upper limit, and the remaining load that exceeds them 235
is recognized by the device of the next priority. Sequential algorithm, alike optimal algorithm, firstly 236
distributes the load to the system with top priority. Ho wever, the difference between the two 237
algorithms is that sequential algorithm recognizes the maximum part load ratio as the operating limit. 238
Lastly, as for the uniform algorithm, all the installed heat source devices operate in the case of load 239
regardless of the priority, and the load is identically processed. However, according to the part load 240
ratio input values of boilers applied in the study as shown in Table 2, there is no difference in entered 241
values between optimum part load ratio and maximum part lo ad ratio, which are entered variables 242
of optimal control algorithms as stated above. Thus, optimal control algorithm and sequential control 243
algorithm are identically operated. Therefore, as what is pursued by the analysis of the load 244
distribution control a lgorithm according to the changes made in input values of part load ratio is 245
different from the suggestion of optimal load distribution control algorithm in staging control which 246
is the goal of this study, the analysis on the optimal control algorithm was excluded in the research 247
among the three types of control algorithms [18]. In order to provide more detailed explanation of 248
aforementioned three types of algorithms, following cases of circumstances were used. First of all, 249
the load processed amount accord ing to the load distribution algorithm in the case of 120kW heating 250
load in the buildings that required up to 200kW of heating load was shown in Fig. 5. Assuming that 251
two boilers in the capacity of 100kW were installed in the buildings, first of all, unifo rm algorithm 252
distributes the same load to the device. Therefore, it processes 60kW of load in each boiler at the same 253
time. Then, once sequential algorithm is applied, boiler_1 operates first according to the priority. In 254
addition, once the maximum part lo ad ratio is set up at 1, boiler_1 processes 100kW of load which is 255
the highest level that can be treated, and the remaining load is processed in boiler_2. Optimum 256
algorithm recognizes the optimum part load ratio as the optimal value and also as the limit as 257
explained above. Once the optimum part load ratio in boiler_1 is set up at 0.9, it processes 90kW of 258
load, and the remaining load is processed in boiler_2. 259
260

Energies 2017, 10, x FOR PEER R EVIEW 8 of 25
261
Figure 4. Load distribution algorithm [18] 262
Table 2. Simulation input related to boiler part load ratio 263
Field Input
Minimum Part Load Ratio 0
Maximum Part Load Ratio 1
Optimum Part Load Ratio 1
264

Energies 2017, 10, x FOR PEER R EVIEW 9 of 25
Figure 5 . Load distribution Example 265
3.3 Simulation Cases 266
In this study, annual heating load data in the office buildings were analyzed, and in 267
consideration of the amount of heating energy needed, the boiler staging control was proceeded with 268
a total of two boilers based on the Case_1 boiler with 290kW of capaci ty. First of all, the capacity of 269
one 290kW boiler in the base model was partitioned into 5:5, 3:7, and 7:3. In the previous studies 270
regarding the control of the number of boilers in office buildings, the partition of capacity in 3:7 was 271
found to show the highest level of efficiency among the partitioned capacity of 5:5 and 3:7, etc. [22]. 272
However, capacity partition of 5:5, 3:7, and 7:3 were considered to compare the efficiency as well as 273
the amount of consumed energy from the load distribution algorithm t o proceed the research. 274
Boiler_2 in each case was partitioned in the capacity after excluding the capacity of boiler_1 among 275
the entire capacities. Characteristics of each case are shown in Table 3. 276
277
Table 3. Simulation Cases 278
Field Equipment List Capacity [kW] Operating priority Load
Distribution
1st 2nd
Case_1 Boiler_1 290 Boiler_1 *
Case_2 Boiler_1, Boiler_2 145 145 Boiler_1, Boiler_2 Uniform
Case_3 Boiler_1, Boiler_2 87 203 Boiler_1, Boiler_2 Uniform
Case_4 Boiler_1, Boiler_2 87 203 Boiler_1 Boiler_2 Sequential
Case_5 Boiler_1, Boiler_2 203 87 Boiler_1 Boiler_2 Sequential
279
280
281 0102030405060708090100
Boiler_1 Boiler_2 Boiler_1 Boiler_2 Boiler_1 Boiler_2Heating Load [kW]
.Uniform Sequential Optimal

Energies 2017, 10, x FOR PEER R EVIEW 10 of 25
3.4 Boiler Performance Curve 282
This study used the boiler efficiency cubic performance curve of EnergyPlus in order to produce 283
a better simulation of the quantitative boiler performance [23]. The efficiency of the performance 284
curve tends to change depending on PLR, and PLR shows the load operated in comparison with the 285
entire boiler capacity during the operati on of the boiler. The following formula (1) is the 286
aforementioned boiler efficiency cubic curve. Table 4 shows the coefficient of applied performance 287
formula. A total of four entered coefficients of a, b, c, and d have been calculated through a series of 288
procedures according to the catalogue which was produced based on the actual survey data from the 289
company A in the United States. 290
𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩𝑩 =𝒂𝒂+𝒃𝒃∗𝑷𝑷𝑷𝑷𝑷𝑷 +𝑩𝑩∗𝑷𝑷𝑷𝑷𝑷𝑷𝟐𝟐+𝒅𝒅∗𝑷𝑷𝑷𝑷𝑷𝑷𝟑𝟑 (𝟏𝟏) 291
Where, 292
PLR = Boiler Part -load ratio 293
Table 4. Boiler Cubic curve input dat a 294
Coefficient Entry Input Data
a 0.986083
b 0.939675
c 0.823611
d 0.575395
4. Results Analysis 295
4.1 Variations of boil er load and outdoor temperature 296
In order to identify the load patterns in each time prior to the analysis of the annual amount of 297
consumed energy, representative days were chosen to analyze the heating load. The following Fig. 6 298
shows the heating load of AHU heating coil and temperature of open air in representative days. As 299
mentioned above, representative days were April 17th and January 23rd. Open air temperature in 300
representative day of the intermediary seasons ranged from 6 to 16 ℃, and the one in the winter 301
season ranged from -4.5 to 0.5℃. In ad dition, the load pattern of the representative day in the 302
intermediary seasons ranged from 0 to 88kW, and the load pattern of the representative day in the 303
winter season ranged from 30 to 287kW. As shown in Fig. 2, internal heat generation is reduced by 304
5% in human bodies, 10% in illumination, and 20% in electronic devices after 8 p.m. Indoor 305
temperature is the lowest during night times when system is not operated and the outside 306
temperature is the lowest, while heating load is the highest at 6 a.m. when bo iler starts operating to 307
keep up with the pre- set temperature. 308
309

Energies 2017, 10, x FOR PEER R EVIEW 11 of 25
310
Figure 6. AHU Heating Rate and Outdoor Air Temperature variations 311
4.2 Part Load Ratio variations 312
Fig. 7 indicates PLR patterns in representative days in the intermediary seasons and winter 313
season in Case_1. Case_1 indicates that one boiler is in charge of all the heating loads without 314
controlling the number of boiler, and hence shows the identical pat terns with the heating load in 315
representative days explained in Fig. 5. As explained above, the highest level of PLR is shown 316
between 6 and 7 a.m. when heating system starts operating, and heating load decreases due to an 317
increase of open air temperature a nd internal heating sources. Therefore, PLR decreases as well. Open 318
air temperature at 7 p.m. is similar with the one at 6 p.m. However, indoor heating load increases 319
due to a decrease in internal heating load, making PLR increase as well. PLR patterns are identical in 320
the intermediary seasons and winter season. However, due to the difference in heating load values, 321
PLR in the representative day of the winter season tends to be higher than the one in the 322
representative day of the intermediary seasons. 323
324 024681012141618
050100150200250300
123456789101112131415161718192021222324
Outdoor Bulb [ ℃]Heating Rate [kW]
HourIntermediate Season Heating Rate Heating Season Heating Rate
Intermediate Season Outdoor Bulb Heating Season Outdoor Bulb

Energies 2017, 10, x FOR PEER R EVIEW 12 of 25
325
Figure 7 . Part Load Ratio (Case_1) 326
Fig. 8 indicates PLR in representative day of the intermediary seasons in Case_2 and Case_3, and 327
Fig. 9 shows PLR in representative day of the intermediary seasons in Case_4 and Case_5. Case_2, 3, 328
4, and 5 use two boilers , but there is a difference in terms of boiler capacity and load distribution 329
algorithm. As for Case_2, two boilers with the identical capacity were controlled with uniform 330
algorithm. As explained above, uniform algorithm evenly deals with the heating load as all the 331
systems operate at the same time once indoor heating load occurs. Therefore, as for Case_2 where 332
two boilers with the same capacity were installed, two boilers showed the same PLR. As for Case_3 333
and Case_4, the capacity of the first boiler was 30% of the entire boiler capacity, and the capacity of 334
the second boiler was 70% of the entire boiler capacity. However, there was a difference in terms of 335
load distribution algorithm. As for Case_3, the identical uniform algorithm was used with the one in 336
the Case_2, and the two boilers tended to share the same amount of load. However, due to the 337
difference in capacity, boiler_1 with relatively low capacity tend to show higher PLR than the one of 338
boiler_2. 339
On the other hand, Case_4 uses the sequential alg orithm. As explained above, according to the 340
sequential algorithm, the second boiler deals with the remaining heating load once PLR of the first 341
boiler reaches 100%. However, during the representative day of the intermediary seasons, the first 342
boiler is ab le to deal with all the heating loads. Thus, boiler_2 does not operate and the PLR is shown 343
as 0%. As opposed to Case_3 and _4, Case_5 indicates that the capacity of the first boiler is 70% of the 344
entire boiler capacity, and the capacity of the second boil er is 30% of the entire boiler capacity, and 345
the sequential algorithm is applied in the same manner with the Case_4. Since the capacity of the first 346
boiler is greater in the Case_5 than in the Case_4, PLR of the first boiler in the Case_5 is lower than 347
the one of the first boiler in the Case_4. As for the second boiler, the first boiler handles all the heating 348
loads in the same manner with the Case_4. Therefore, PLR is 0%, and, hence, the boiler does not 349
operate. 350
351 0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Part Load Ratio [%]
HourCase_1 Intermediate Season PLR Case_1 Heating Season PLR

Energies 2017, 10, x FOR PEER R EVIEW 13 of 25
352
Figure 8. Part Load Ratio in Intermediate Period (Case_2, 3) 353
354
355
Figure 9 . Part Load Ratio in Intermediate Period (Case_4, 5) 356
357
Fig. 10 and 11 indicate the patterns of PLR by time during the representative day of winter 358
season. As for Case_2, all the systems operate in the same capacity and time period. Therefore, two 359
boilers indicate the same PLR in the same manner as seen during the representative day of the 360
intermediary seasons. However, as the heating load is higher than the one during the representative 361 0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Part Load Ratio [%]
HourCase_2 Boiler_1 Case_3 Boiler_1
Case_2 Boiler_2 Case_3 Boiler_2
0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Part Load Ratio [%]
HourCase_4 Boiler_1 Case_5 Boiler_1
Case_4 Boiler_2 Case_5 Boiler_2

Energies 2017, 10, x FOR PEER R EVIEW 14 of 25
day of the intermediary seasons, they show relatively high PLR. Case_3 also indicates similar patterns 362
with the ones seen duri ng the representative day of the intermediary seasons. However, as boiler_1 363
is unable to process all the heating loads after reaching 100% of PLR at 6 a.m., boiler_2 deals with the 364
remaining heating load. Thus, there is a relatively low difference in PLR b etween the boiler_1 and 365
boiler_2 compared to other time periods. While Case_4 and Case_5 use the same load distribution 366
methods, the capacity of the boiler_1 and boiler_2 are different. The overall pattern is very similar, 367
but there is a difference in the pattern of the boiler_2. This is because of the difference in the capacity 368
of the boiler_1. Boiler_1 in Case_4 with the capacity that is 30% of the entire capacity shows 100% of 369
PLR at 6 a.m. to 8 a.m., and 7 p.m., making boiler_2 operate and handle the re maining load. However, 370
as for the Case_5, boiler_1 with larger capacity deals with most of the loads, and boiler_2 operates 371
only at 6 a.m. when the heating load is the highest. Hereupon, it was confirmed that different PLR 372
values were shown due to the diff erence in capacity of boilers and load distribution algorithm. This 373
is expected to directly influence the gas consumption for the heating in buildings. 374
375
376
Figure 10. Part Load Ratio in Heating Period (Case_2, 3) 377
378 0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Part Load Ratio [%]
HourCase_2 Boiler_1 Case_3 Boiler_1
Case_2 Boiler_2 Case_3 Boiler_2

Energies 2017, 10, x FOR PEER R EVIEW 15 of 25
379
Figure 11. Part Load Ratio in Heating Period (Case_4, 5) 380
4.3 Gas Consumption in Representative Days 381
The following has analyzed the gas consumption in representative days based on the previously 382
analyzed PLR. Previously analyzed PLR is the value that was calculated by dividing the heating load 383
in the system by the system capacity. The pattern of gas consumption seen during the representative 384
days is the same as the pattern of aforementioned PLR. Therefore, analysis of gas consumption 385
pattern in each time during representative days has not been i ncluded in this study. Instead, 386
comparative analysis has been conducted on the entire gas consumption used for the entire 387
representative day. Gas consumption is closely related to the PLR and the efficiency of boilers. 388
Matters related to the efficiency of boilers are presented in the analysis of annual gas consumption in 389
the paragraph 4.4, and the relationship between PLR and gas consumption is stated in the paragraph 390
4.3. The following Table 6 and 7 represent the results of the gas consumption of boiler_1 and boiler_2 391
and the entire gas consumption during representative days of the intermediary seasons and winter 392
season. Case_4 indicated the least gas consumption, and the gas consumption was recorded in an 393
ascending order of Case_5, Case_1, Case_2, and Case _3. Case_1 and Case_2 indicate the same amount 394
of gas consumption. In Case_1, only one boiler was used, thus there was no gas consumption in the 395
boiler_2. As explained above, two identical boilers operated at the same time in Case_2. Therefore, 396
there was a difference in values between the intermediary seasons and winter season, but the amount 397
of gas consumption used in each boiler was identical. When comparing the Case_2 with Case_1, the 398
heating load applied to each boiler in Case_2 was only a half of that in Case_1. However, as the 399
system capacity was reduced to half, PLR turned out to be identical in the end. Thus, Case_1 and 400
Case_2 indicated the same PLR and the same amount of consumed gas, and the only difference was 401
in the number of boilers. The higher amount of gas consumption was shown in Case_3 compared to 402
Case_1 where only one boiler was used. In Case_3, when heating load is required in the uniform 403
algorithm, two boilers operate at the same time. As explained above, since the capacity of the boiler_2 404
is greater than the one of the boiler_1 in Case_3, it was found that the PLR of the boiler_2 was 405
significantly lower than the one of the boiler_1 when loads were identically applied. According to 406
the characteristics of boilers, as the PLR gets lower, the efficiency of boilers also gets lower. Through 407
the lower PLR in the boiler_2, it is assumed that more gas has been consumed. However, the 408 0%10%20%30%40%50%60%70%80%90%100%
123456789101112131415161718192021222324Part Load Ratio [%]
HourCase_4 Boiler_1 Case_5 Boiler_1
Case_4 Boiler_2 Case_5 Boiler_2

Energies 2017, 10, x FOR PEER R EVIEW 16 of 25
difference of the amount of consumed gas between Case_3 and Case_1 was very small. The capacity 409
of boilers entered in Case_4 was the same as Case_3, but the load distribution algorithm was different. 410
As for the sequential algorithm entered in Case_4, only the boiler_1 operates for most of the time 411
except for certain hours when there is much heating load as the boiler_2 op erates while PLR of the 412
boiler_1 reaches 100%. Therefore, higher level of PLR is maintained in Case_4 than in Case_3. 413
Hereupon, the amount of consumed gas seems to be less. As for Case_5, the capacity of the boiler_1 414
is greater than the one of the boiler_2 . Therefore, boiler_1 deals with all the heating loads in the 415
intermediary seasons when there is less heating load. Hereupon, the amount of consumed gas in the 416
boiler_2 is 0. In winter season, only one boiler_1 with large capacity operates for most of the time, 417
and the boiler_2 operates only during certain hours when heating load is much required. Therefore, 418
among the cases except for Case_1 when there was only one boiler, the amount of consumed gas in 419
the boiler_1 in Case_5 seemed to be the lowest. 420
Table 5. Daily Gas Consumption data (Intermediate Season) 421
Case Total Gas consumption
Boiler_1 Gas
Boiler_2 Gas
Case _1 366.8 366.8 ─
Case _2 366.8 183.4 183.4
Case _3 367.3 172.4 194.9
Case _4 323.8 323.8 0
Case _5 341.3 341.3 0
422
Table 6. Daily Gas Consumption data (Heating Season) 423
Case Total Gas consumption
Boiler_1 Gas
Boiler_2 Gas
Case _1 1430.7 1430.7 ─
Case _2 1430.7 715.3 715.3
Case _3 1433.1 642.8 790.3
Case _4 1373.9 1009.5 364.4
Case _5 1379.6 1293.8 85.8
4.4 Detailed analysis on annual data 424
The following Fig. 12 indicates the boiler efficiency curve in each PLR section of boiler. As 425
explained above, the following curve is the efficiency curve made according to actual survey data by 426
the company A in the United States. All the boilers used in the relevant study have the identical 427
efficiency curve. As the PLR increases, the efficiency of boilers increases as well. Once the PLR reaches 428
100%, up to 84% of the boiler efficiency is achieved. In this study, the low efficiency section was 429
assumed as PLR of 0% to 40% where the efficiency was less than 70%, and the high efficiency section 430
was assumed from PLR of 40 to 100% where the efficiency was 70% or higher. Therefore, analysis 431
was carried out by dividing the sections into two for precise analysis: PLR of less than 40% and PLR 432
of 40% or higher. 433
434

Energies 2017, 10, x FOR PEER R EVIEW 17 of 25
435
Figure 12. Boiler Efficiency Curve 436
437
Fig. 13 indicates the amount of consumed gas in the low efficiency section where PLR is 0 to 438
40%. According to Fig. 13, Case_4 shows the lowest gas consumption of 26.6MWh in the low 439
efficiency section of boilers, and Case_1 and Case_2 both show the highest gas consumption of 440
88.0MWh. The reason why Case_1 and Case_2 show the identical values w as that, as explained in the 441
paragraph 4.3, one boiler was divided into two boilers and this led the load to be handled in one 442
boiler to reduce to half. However, the capacity of boilers was reduced to the half as well, and thus 443
there was no change in the P LR. In addition, as they had the same boiler efficiency curves, the amount 444
of consumed gas was found to be the same. As for Case_3, boiler_1 that constituted 30% of the entire 445
capacity and boiler_2 that had 70% of the entire capacity operated at the same time. As the two boilers 446
operated at the same time even when low heating load was required and the boiler_2 that constituted 447
70% of the entire capacity maintained low PLR, it seemed that the amount of consumed gas was 448
found to be high in Case_3 in the low e fficiency section. As for Case_5, boiler_2 does not operate until 449
PLR of the boiler_1 that constitutes 70% of the entire capacity reaches 100% according to the 450
characteristics of load distribution algorithm. However, since the intermediary seasons require lower 451
heating load than the winter season, boiler_1 with larger capacity show lower PLR. Therefore, Case_5 452
shows less amount of consumed gas compared to Case_1, _2, and _3. As for Case_4 that had the least 453
amount of consumed gas, boiler_1 constituted 30% o f the entire boiler capacity, thus high level of 454
PLR is maintained in winter season when high heating load is required, and relatively high PLR is 455
maintained compared to other cases due to the low capacity in the intermediate seasons when low 456
heating load is required. This shows that Case_4 has the least amount of consumed gas in the end. 457
Table 7 indicates operating hours of each boiler, and this directly influences the amount of consumed 458
gas shown in Fig. 13. As explained above, Case_1 and Case_2 have the different capacity of boilers 459
but show identical PLR. Therefore, their operating hours are the same as well. As for Case_3, two 460
boilers operate at the same time in the same manner with Case_2. However, as the capacity of boiler_2 461
is greater in Case_3 than in Case_2, PLR is maintained lower than that of the boiler_1, even if the 462
heating loads applied are identical. In addition, when distributing the load, the remaining load is 463
processed by the boiler_2 if there is a load that cannot be handled by the boiler_ 1 with smaller 464
capacity. Therefore, operating hours of the boiler_2 tend to be relatively longer than the ones in 465 0102030405060708090100
0 10 20 30 40 50 60 70 80 90 100Boiler Efficiency [%]
Part Load Ratio [%]

Energies 2017, 10, x FOR PEER R EVIEW 18 of 25
Case_1 and Case_2. On the other hand, as the boiler_1 in Case_3 has small capacity than other boilers, 466
heating load is processed in the relati vely high PLR section even if the same heating load is dealt 467
with. Thus, less operating hours are shown compared to Case_1 and Case_2 in the section of 0 to 40% 468
part load. As for Case_4 that consumed the least amount of gas in the low efficiency section, operating 469
hours were found to be the shortest as well. First of all, according to the characteristics of load 470
distribution algorithm, once the PLR of the boiler_1 reaches 100%, boiler_2 operates. However, since 471
the capacity of the boiler_1 is low, higher PL R is shown in the intermediary seasons when low heating 472
load is required, and even higher PLR is shown in winter season when high heating load is required. 473
Therefore, operating hours of the boiler_1 in the low efficiency section are the shortest among thos e 474
in all cases. Boiler_2 shows less operating hours than other cases. However, they are still longer than 475
the ones in Case_5. This is due to the difference in capacity of the boiler_1. Boiler_1 in Case_5 has 476
larger boiler capacity. Boiler_1 is able to hand le heating load occurring during most of the 477
intermediary seasons and in winter season at day times. Therefore, there are longer operating hours 478
of the boiler_1, but the operating hours of the boiler_2 are the least among the entire cases. Low 479
efficiency s ection is where the boiler shows low efficiency, and as the operating hours are shorter and 480
the amount of consumed gas is smaller in the section, the efficiency of the operation of boilers turns 481
out to be higher. 482
483
484
Figure 13. Gas Consumption in each PLR (0 ≤ PLR < 40) 485
486
487
488
489
490
491
492
493
494 0102030405060708090100
0≤ PLR＜10 10≤ PLR＜20 20≤ PLR＜30 30≤ PLR＜40 TotalGas Consumption [MWh]
Part Load Ratio [%]Case_1 Case_2 Case_3 Case_4 Case_5

Energies 2017, 10, x FOR PEER R EVIEW 19 of 25
Table 7. Cumulative operation hours in each PLR (0 ≤ PLR < 40) 495
Part Load Ratio
[%] Case 1 Case 2 Case 3 Case 4 Case 5
Boiler Boiler Boiler Boiler Boiler
1 1 2 1 2 1 2 1 2
0 ≤ PLR < 10 747 747 747 474 1,035 251 123 550 47
10 ≤ PLR < 20 596 596 596 421 527 223 63 485 17
20 ≤ PLR < 30 251 251 251 366 175 210 40 353 7
30 ≤ PLR < 40 116 116 116 224 61 211 10 174 6
Total 1,710 1,710 1,710 1,485 1,798 895 236 1,562 77
496
Fig. 14 indicates the amount of consumed gas in the high efficiency section where PLR is 40% or 497
higher. According to Fig. 14, Case_1 and Case_2 showed the lowest amount of consumed gas of 498
32.7MWh in the high efficiency section of the boiler, and Case_4 showed the highest amount of 499
consumed gas of 86.3MWh. The reason for such results can be interpreted through Table 8. As for 500
Case_1 and Case_2, boilers operate for 1,710 hours on an annual basis in Table 7, which indicates the 501
low efficiency section. On the other hand, they only operate for 152 hours in the high efficiency 502
section. As the capacity of boilers is designed to handle the maximum heating load, PLRs of boilers 503
are low due to the excessive capacity during most of the period except for several special days. 504
Therefore, it seems that as operating hours in the high efficiency section are too short, the amount of 505
consumed gas is also low. As for Case_3, longer boiler operating hours are shown compared to 506
Case_1 and Case_2 in the high efficiency section. Ev en if the load distribution algorithm is used in 507
the same manner with Case_2, the capacity of the boiler_1 is as low as 30% of the entire capacity. 508
Therefore, high PLR is continuously maintained during the intermediary seasons and in winter 509
season at day times when heating load is low. Therefore, they show higher amount of consumed gas 510
compared to Case_1 and Case_2. Case_4 indicates the highest amount of consumed gas among cases 511
analyzed in this study due to the capacity of boilers and interaction of applie d load distribution 512
algorithm. According to the characteristics of sequential algorithm, boiler_2 operates once the PLR of 513
the boiler assigned in priority reaches 100%. However, Case_4 shows the longest operating hours in 514
the high efficiency section compar ed to other cases due to the low capacity of the boiler_1. Case_5 is 515
the identical load distribution algorithm with Case_4. However, the boiler_1 with the highest priority 516
constitutes 70% of the entire boiler capacity. Thus, there is longer operating hours in the low efficiency 517
section compared to the high efficiency section where PLR is 40% or higher. 518
519

Energies 2017, 10, x FOR PEER R EVIEW 20 of 25
520
Figure 14. Gas Consumption in each PLR (40 ≤ PLR ≤ 100) 521
522
Table 8. Cumulative operation hours in each PLR (40 ≤ PLR ≤ 100) 523
Part Load Ratio
[%] Case 1 Case 2 Case 3 Case 4 Case 5
Boiler Boiler Boiler Boiler Boiler
1 1 2 1 2 1 2 1 2
40 ≤ PLR < 50 78 78 78 109 11 195 10 105 5
50 ≤ PLR < 60 16 16 16 84 19 171 17 73 3
60 ≤ PLR < 70 20 20 20 64 10 127 10 43 3
70 ≤ PLR < 80 16 16 16 37 12 97 12 19 5
80 ≤ PLR < 90 12 12 12 22 5 64 5 12 0
90 ≤ PLR ≤ 100 10 10 10 61 7 313 8 48 4
Total 152 152 152 377 64 967 62 300 20
524
Table 9 and 10 indicate the graph that shows the annual operating hours and the amount of 525
consumed gas in every PLR section of each case during the operating period. Operating hours are 526
indicated as (h), and the amount of consumed gas is shown as MWh. In t he section where PLR is 0 to 527
40% and where there is low efficiency according to the characteristics of performance curve, Case_1 528
has consumed approximately 73% of gas compared to the annual amount of consumed gas. In the 529
same section, Case_2 has consumed a pproximately 73% of gas, followed by approximately 69% in 530
Case_3, approximately 24% in Case_4, and approximately 58% in Case_5. According to the results of 531
the comparison of the amount of consumed gas in every PLR section, Case_4 showed the least 532
amount of consumed gas. This was because the capacity of the boiler_1 was low, but it is also assumed 533
that the role of load distribution algorithm was heavily influential. As mentioned above, according 534
to the characteristics of sequential algorithm, boiler_2 operat es once the PLR of the boiler_1 assigned 535
in priority reaches 100%. However, due to the low capacity of the boiler_1, it is assumed that there 536 0102030405060708090100
40≤ PLR＜5050≤ PLR＜6060≤ PLR＜7070≤ PLR＜8080≤ PLR＜9090≤ PLR ≤ 100 TotalGas Consumption [MWh]
Part Load Ratio [%] Case_1 Case_2 Case_3 Case_4 Case_5

Energies 2017, 10, x FOR PEER R EVIEW 21 of 25
was longer operating hours in the high efficiency section compared to other cases. This is the number 537
of hours tha t are six times higher than the ones in Case_1 and Case_2 and also three times higher 538
than the ones in Case_5, except for Case_4. Therefore, it shows the highest amount of consumed gas 539
in the high efficiency section compared to other cases. The amount of g as consumed in Case_4 in the 540
high efficiency section is 86.3MWh which is close to 88.0MWh consumed in Case_1 and Case_2 in the 541
low efficiency section. However, there was a huge difference on heat load that was processed due to 542
the difference in boiler capa city. This indicates that Case_4 shows the least amount of consumed gas, 543
as load is more efficiently processed even if the amount of load processed is ident ical to the one in 544
other cases. 545
546
Table 9. Annual PLR date (Case_1, 2, 3) 547
Part Load Ratio
[%] Case_1 Case_2 Case_3
Boiler_1 Boiler_1 Boiler_2 Boiler_1 Boiler_2
(h) MWh (h) MWh (h) MWh (h) MWh (h) MWh
0 ≤ PLR < 10 747 16.4 747 8.2 747 8.2 474 3.1 1,035 16.3
10 ≤ PLR < 20 596 34.1 596 17.05 596 17.05 421 7.4 527 20.4
20 ≤ PLR < 30 251 22.8 251 11.4 251 11.4 366 10.3 175 11.3
30 ≤ PLR < 40 116 14.7 116 7.35 116 7.35 7.35 8.7 61 5.4
40 ≤ PLR < 50 78 12.8 78 6.4 78 6.4 109 5.4 11 1.3
50 ≤ PLR < 60 16 3.2 16 1.6 16 1.6 84 5.1 19 2.7
60 ≤ PLR < 70 20 4.9 20 2.45 20 2.45 64 4.7 10 1.7
70 ≤ PLR < 80 16 4.4 16 2.2 16 2.2 37 3.1 12 2.4
80 ≤ PLR < 90 12 3.8 12 1.9 12 1.9 22 2 5 1.1
90 ≤ PLR ≤ 100 10 3.6 10 1.8 10 1.8 61 6.7 7 1.8
Total 1,862 120.7 1,862 60.35 1,862 60.35 1,645 56.5 1,862 64.4
548
Table 10. Annual PLR date (Case_4, 5) 549
Part Load Ratio
[%] Case_4 Case_5
Boiler_1 Boiler_2 Boiler_1 Boiler_2
(h) MWh (h) MWh (h) MWh (h) MWh
0 ≤ PLR < 10 251 1.4 123 1.3 550 8.3 47 0.2
10 ≤ PLR < 20 223 3.9 63 2.6 485 19.7 17 0.2
20 ≤ PLR < 30 210 5.8 40 2.6 353 22.5 7 0.2
30 ≤ PLR < 40 211 8.1 10 0.9 174 15.1 6 0.2
40 ≤ PLR < 50 195 9.5 10 1.2 105 11.8 5 0.3
50 ≤ PLR < 60 171 10.3 17 2.2 73 10.3 3 0.2
60 ≤ PLR < 70 127 8.9 10 1.7 43 7.1 3 0.2
70 ≤ PLR < 80 97 7.9 12 2.3 19 3.5 5 0.4
80 ≤ PLR < 90 64 5.9 5 1.1 12 2.6 0 0
90 ≤ PLR ≤ 100 313 33.7 8 1.6 48 11.8 4 0.4
Total 1,862 95.4 298 17.5 1,862 112.7 97 2.3
550
551

Energies 2017, 10, x FOR PEER R EVIEW 22 of 25
4.5 Monthly Gas Consumption analysis 552
The following Fig. 15 and Table 11 indicate the monthly amount of consumed gas in each case. 553
Monthly amount of consumed gas in June, July, and August in summer when heating was not used 554
were excluded. As explained above, Case_4 showed the least amount of c onsumed gas of 112.9MWh 555
in both intermediary seasons and the winter season, and Case_5 showed high amount of consumed 556
gas of 115.0MWh, followed by 120.7MWh in Case_1 and Case_2 and 120.9MWh in Case_3 in the 557
ascending order. In the monthly breakdown, there was no change in the aforementioned order. In 558
this study, it is confirmed that while the capacity of each boiler is one of the very important elements, 559
load distribution algorithm has higher influence on the amount of consumed gas than the capacity of 560
boilers. By comparing Case_3 and Case_4 which use the identical boiler capacity, it is found that 561
8.0MWh of gas can be saved on an annual basis only by having the difference of load distribution 562
algorithm, and approximately 7% of heating energy can be saved in Case_4 where the amount of 563
consumed gas is relatively low. When comparing Case_4 and Case_5 which use the same load 564
distribution algorithm but have different boiler capacities, it is known that Case_4 uses less amount 565
of gas by 2.1MWh a year compared to C ase_5. This is the reduced amount of gas by approximately 566
2%, which was less influential than the load distribution algorithm. By comparing Case_1 and Case_4 567
which use one boiler, it is found that it is possible to save 7.8MWh of the gas consumption a year , 568
corresponding to 7% of heating energy. 569
In this study, it was shown that sequential algorithm that progressively operated boilers in the 570
order designated by users was more efficient in terms of energy over uniform algorithm that evenly 571
distributed heati ng load to all boilers. If not using the same boiler capacities, it was confirmed that 572
placing boilers with less capacity in priority was more efficient. 573
574
575
Figure 15. Monthly Gas Consumption 576
577
578
579
580
581 051015202530
1 2 3 4 5 9 10 11 12Gas Consumption [MWh]
MonthCase_1 Case_2 Case_3 Case_4 Case_5

Energies 2017, 10, x FOR PEER R EVIEW 23 of 25
Table 11. Monthly Gas Consumption [MWh] 582
583
5. Conclusion 584
This study has modeled the large-sized office buildings with EnergyPlus while applying the division of the 585
number of boilers and load distribution algorithm to precisely analyze the annual operating performance of 586
boilers according to the characteristics of part load. Hereupon, research has been conducted to achieve the goal 587
of suggesting the optimal capacity dividing ratio and heat load distribution algorithm in large-sized office 588
buildings. Conclusion has been drawn in this study as follows. 589
590
 According to the results of the analysis of annual operating performance based on characteristics of part 591
load of boilers in each case, Case_1 as a base model was found to use approximately 73% of the annual 592
gas consumption in the section with low PLR of 0 to 40%. As for Case_2 and Case_3 where two boilers 593
operated at the same time uniformly distributing the load, Case_2 and Case_3 were found to use 594
approximately 73% and 69% of annual energy consumption, respectively, in the same PLR section. In 595
addition, a s for Case_4 and Case_5 where two boilers sequentially operated according to their own 596
priority, Case_4 and Case_5 were found to use approximately 24% and approximately 58% of annual gas 597
consumption, respectively, in the same PLR section. 598
 According to the results of analysis of gas energy consumption in the intermediary seasons and winter 599
season, Case_4 was found to use the least amount of gas both in the intermediary seasons and winter 600
season according to the interactions between PLR from load distribution method and boiler efficiency. 601
 According to the results of comparison on annual gas consumption, it was found that the annual gas 602
consumptions in Case_2 and Case_3 that used uniform algorithm were estimated to be identical or higher 603
than the one in Case_1 as a base model due to the characteristics of the load distribution method. 604
Therefore, it is assumed to be difficult to expect reasonable energy consumption in case of boiler staging 605
control. In comparison with Case_1, Case_4 saved the hi ghest amount of energy. To be specific, Case_4 606
saved approximately 14% and also approximately 6% of energy, respectively, in the intermediary seasons 607
and winter season. In addition, Case_4 saved approximately 7% of total annual gas consumption in 608
compariso n to Case_1. 609 Month Case _1 Case_2 Case _3 Case _4 Case _5
1 28.3 28.3 28.3 27.1 27.3
2 20.7 20.7 20.7 19.6 19.8
3 17.2 17.2 17.2 15.9 16.3
4 7.8 7.8 7.8 6.9 7.2
5 2.3 2.3 2.3 1.9 2.1
9 0.4 0.4 0.4 0.3 0.3
10 4.8 4.8 4.8 4.1 4.4
11 15.5 15.5 15.5 14.4 14.7
12 23.8 23.8 23.9 22.7 22.9
Intermediate period 15.2 15.2 15.2 13.1 14.0
Heating period 105.5 105.5 105.7 99.8 101.0
Total 120.7 120.7 120.9 112.9 115.0

Energies 2017, 10, x FOR PEER R EVIEW 24 of 25
 In other words, Case_4 applied with sequential algorithm was found to use the least amount of energy 610
on an annual basis, and, hence, sequential algorithm is assumed to be the most efficient algorithm. 611
 In conclusion, in this study applying the sequential control algorithm by dividing the capacity of boilers 612
in the ratio of 3:7 is assumed to be the optimal algorithm for staging control for the maximum efficiency 613
of boilers. 614
References 615
1. Kim, T.H.; Jung, Y.S.; Jung, H.g. A trend analysis of greenho uses gas reduction technique deve 616
lopment for building sector in Korea. Architectural Institute of Korea, 2016 617
2. Korea Energy Economics Institute. World Energy Market Insight, 2016 618
3. Lee, S.J. Post 2020 types of greenhouse gas reduction contribution. Korea Ene rgy Economics In 619
stitute, 2016 620
4. Choi, D.Y. Building energy consumption standing sampling survey. Korea Energy Economics I 621
nstitute, 2015 622
5. Seo, J.H.; Energy saving in boiler. Korean Journal of Korean Association of Air Conditioning R 623
efrigerating and Sanitary E ngineers, 2009 624
6. Murray, S.Y.; Walsh, B.P.; Kelliher, D.; O’Sullivan, D.T.J. Multi -variable optimization of thermal 625
energy efficiency retrofitting of buildings using static modelling and genetic algorithms ─ A case s 626
tudy. Building and Environment, 2014; pp.9 8-107 627
7. Weissmann, C., Hong, T., Graubner, C.A., Analysis of heating load diversity in German reside 628
ntial districts and implications for the application in district heating systems., Energy and Buildings 629
., 2017; pp.302 -313 630
8. Giurca, I.; Badea, G.; Așchilean, I.; Naghiu, G.S.; Megyesi, E. Selecting the number and size of 631
boilers used within the heating units of the residential complexes. Energy Procedia, 2017; pp.134-14 632
1 633
9. Sin, Y.G.; Jang, Y.S.; Baeck, Y.J.; Kim, Y.I. In -situ diagnosis of boiler performance inclu ding me 634
asurement of dynamic heating load. Korean Journal of Air -Conditioning and Refrigeration Engineeri 635
ng, 2002; pp.22-27 636
10. Shide, W.; Jia, L. Intelligent and optimal control of energy saving of gas boiler group. IEEE, 2 637
010; Volume.3, pp.50-53 638
11. Wei, D.; Chen , A.; Sun, B.; Zhang, C. Multi -objective optimal operation and energy coupling a 639
nalysis of combined cooling and heating system. Energy, 2016; pp.296 -307 640
12. Lazzarin, R.M. The importance of the modulation ratio in the boilers installed in refurbished b 641
uildings ., Energy and Buildings, 2014; pp.43 -50 642
13. Liu, Z.; Tan, H.; Luo, D.; Yu, G.; Li, J.; Li, Z. Optimal chiller sequencing control in an office b 643
uilding considering the variation of chiller maximum cooling capacity. Energy and Buildings, 2017; 644
pp.430 -442 645
14. Sun, Y.J.; Wang, S.; Xiao, F. In situ performance comparison and evaluation of three chiller se 646
quencing control strategies in a super high -rise building. Energy and Buildings, 2013; pp.333 -343 647
15. Sun, Y.J.; Wang, S.; Huang, G. Chiller sequencing control with enhan ced robustness for energy 648
efficient operation. Energy and Buildings, 2009; pp.1246 -1255 649
16. Chang, Y.C. Sequencing of chillers by estimating chiller power consumption using artificial ne 650
ural networks. Building and Environment, 2007; pp.180 -188 651
17. Yu, B.h.; Seo, B .M.; Moon, J.E.; Lee, K.H. Analysis of part load ratio characteristics and gas en 652

Energies 2017, 10, x FOR PEER R EVIEW 25 of 25
ergy consumption of a hot water boiler in a residential building under Korean climatic conditions. 653
Korean Journal of Air -Conditioning and Refrigeration Engineering, 2015; pp.4 55-462 654
18. Seo, B.M.; Lee, K.H. Detailed analysis on part load ratio characteristics and cooling energysavi 655
ng of chiller staging in an office building. Energy and Buildings, 2016; pp.309-322 656
19. Kong, S.H. EnergyPlus program. Korean Journal of Air -Conditioning and Refrigeration Enginee 657
ring, 2000; pp.56-59 658
20. Lee, D.Y.; Seo, B.M.; Kwon, H.J.; Lee, K.H. Heating performance and partial load ratio charact 659
eristics of boiler staging in office building. Korean Journal of Air -Conditioning and Refrigeration En 660
gineering, 2017 661
21. American Society of Heating, Refrigerating and Air -conditioning Engineers. ASHRAE Standard 662
90.1. Energy Standard for Buildings Except Low -Rise Residential Buildings, 2004 663
22. Sohn, J.Y.; Kim, S.H.; Ahn, B.W.; Shin, Y.T. Multiple units control of boiler and refr igerator H 664
VAC system. Korean Journal of Air -Conditioning and Refrigeration Engineering, 1992 665
23. The U. S. Department of Energy. EnergyPlus Input Output Reference. The encyclopedic referen 666
ce to EnergyPlus input and output, 2010 667