A Case Study in Actual Building Performance and Energy Modeling with Real Weather Data

As part of this study, an issue has been identified with regards to there being a performance gap with energy efficient buildings. This has been validated through literature review in the areas of occupancy behavior, modeling accuracy and reviewing energy consumption of energy efficient buildings. In order to analyze the error generated between predicted and actual energy performance, a case study approach has been adopted. The Ron Joyce Centre is a LEED Gold Certified building that is part of the McMaster University campus in Burlington. Actual energy performance data has been collected along with detailed drawings to analyze its predicted energy performance using real weather data over a two-year period in eQUEST. The results indicate that eQuest is able to predict electrical consumption within 0.72% of actual on an annual basis. However, natural gas consumption is more erratic and inconsistent based on heating degree days and has fluctuating values with differences ranging between 21% to 4.5% on monthly basis. The overall predicted energy consumption for 2012 is 1096133 kWh and 33227 m. It is not possible to root the cause for this discrepancy with limited data, except to utilize two weather files in generating energy models. The default 30 year average from CTMY for Toronto and another to account for the maximum number of HDD, offering owners a range of natural gas consumption.


Introduction
It is a general knowledge that the current building population consumes a larger portion of the energy produced. In 2010 the commercial building sector consumed about 30% of the total electrical and about 21% of the natural gas produced in Canada (Natural Resoure Canada, 2011). This has been attributed to low efficiencies, poor thermal performance and improper design practices. The introduction of energy efficient buildings is having a steady growth in the Canadian market and aims to make new and renovated buildings more sustainable and efficient. The benefits of green building design help to reduced water consumption; improve indoor air and environment quality due to material selections and low VOC's; and address occupant comfort. Leadership in Energy and Environmental Design (LEED) is one of the most popular optional rating systems that address these issues in seven different areas. One of these is Energy and Atmosphere, which constitutes a big portion of improving a building's energy consumption. However, through literature review including research, reports and studies, evidence shows that these buildings are not performing as predicted through energy modeling simulation.
In order for a building to be certified under LEED, the design team has to generate a base line model in accordance with ASHRAE standard 90.1 or Model National Energy Code for Buildings (MNECB). In doing so, the projected savings for the proposed design may be established (U.S. Green Building Council, 2010). Despite efforts to accurately predict energy consumption pre-construction, there is a significant gap between predicted and actual energy consumption. This has brought about the term known as performance gap with efforts being made to reduce this discrepancy (Carbon Trust, 2011). To give a perspective on the LEED market in Canada, there are over 4000 certified buildings and more still awaiting approval. This has amounted into a steady growth since 2001 of about 50% each year (Canada Green Building Council, 2013). Refer to Appendix A Figure A-1 for further information. The performance gap is not an issue present in all buildings but is an area of concern. As part of this report, the actual versus predicted energy consumption of the Ron Joyce Centre (RJC) will be evaluated. In order to accurately represent the building, eQUEST modeling tool will be utilized in conjunction with actual weather data collected from weather station Hamilton A between January 2011 and April 2013.

Energy Audits
In performing the literature review a number of research articles, reports and studies were studied to understand why buildings are not performing as intended and what steps can be taken to create a more robust energy model. The selection of articles specifically focus on commercial buildings pertaining to their performance, factors of influence over energy consumption and ability to simulate their daily operations. An initial review consists of studies pertaining to building energy audits and a review of their performance; this included the PROBE (Post-occupancy Review of Buildings and their Engineering) study, LEED buildings and energy efficient buildings.
The PROBE study was conducted between 1995 and 2002 involving the evaluation of 23 energy efficient buildings. The results of this study demonstrated that, of the 23 buildings studied; energy use was generally higher than anticipated due to sources such as, computer labs and office equipment adding 25% to 80% to the total building energy consumption . Many of the inaccuracies stemmed during the modeling phase due to unrealistic assumptions . Carbon Trust has further emphasized this in a recent study Closing the Gap, in which they outline key inaccuracies that occur during the modeling phase. As shown in Figure 1, the design prediction comprising of regulated energy use, does not take into account additional variations and inefficiencies that occur within commercial buildings. This invariably adds to the energy consumption, hence generating the performance gap (Carbon Trust, 2011). An earlier study on a low-energy building constructed in the early 1980's also suffered from unrealistic assumptions and occupant behavior. The simulation and calibration of a DOE-2.1 C model was required to understand the sources of the discrepancy that caused more than twice the predicted energy consumption within a building (Norford, Socolow, Hsieh, & Spadaro, 1994 Regulated energy use Actual total energy use calibration process the energy consumption was heavily dependent with the way the building was used rather than its construction. The cause of the increased over the baseline energy consumption was attributed to tenant power requirements contributing to a 64% increase; HVAC systems were in operation for a longer time contributing to a 24% increase; reduced equipment performance and greater envelope heat transfer contributing to the final 12% increase. The performance gap in all three studies noted above stem from energy modeling inputs based on the very best energy behavior and rule out tenant lifestyle and operational inefficiencies (Norford, Socolow, Hsieh, & Spadaro, 1994). Despite the industry having an early understanding of the shortfalls of energy prediction and sources of inaccuracies, today's energy efficient buildings are still challenged by this.
Two similar studies conducted evaluated the post-occupancy energy performance of LEED buildings within the United States (Oates & Sullivan, 2012;Turner, 2006). LEED buildings are set out to be sustainable and energy efficient, with a large portion of the points allocated toward energy savings. The results of both papers arrived at a similar conclusion, with majority of the LEED buildings surveyed not performing as originally predicted and the energy savings not being realized (Oates & Sullivan, 2012;Turner, 2006). This could be attributed to the performance gap and inaccuracies mentioned in the section above. In Turner's results, she mentioned that the buildings were performing below baseline standards and no single building's actual energy use intensity was within 20% of the projected savings demonstrated from the proposed design model (Turner, 2006). In reviewing Oates and Sullivan's paper on medium and high-energy intensity buildings, consistent results to that of Turner and Carbon Trust study were produced. Final results demonstrated that of the 19 buildings sampled, "four of the high energy intensity buildings performed 48% worse than the design case and 24% worse than the baseline model". "The 15 buildings of medium energy intensity underperformed the design model by 74% and the baseline model by 14%" (Oates & Sullivan, 2012). It must be noted that as many of the commercial buildings get bigger heat loss/ gain for heating and cooling will become smaller in comparison to other end uses like plug loads, lighting and ventilation.
Amongst the study conducted by Turner and Oates the concerns about energy savings not being achieved is evident; however, certain areas of concern must be drawn to the two studies. The first being, not all of the sampling for the LEED buildings were randomized and some were dependent on voluntary information prepared by the owner. This brings about a possible bias because the owners could have only provided the information under the pretext that they assumed their building was performing as expected, which through the studies suggest otherwise (Oates & Sullivan, 2012). The studies are limited because they examine a very small population of the total energy efficient buildings currently on the market; and only Oates and Sullivan study noted the possibility of reduced energy consumption past the first year of operations. This is critical because it takes about a year for the building maintenance and operations managers to optimize the performance of the building based on the loads due to occupancy, scheduling and equipment familiarity (Oates & Sullivan, 2012).
The Carbon Trust study and Oates and Sullivan research, address key reasons for the performance gap. This is significant because as presented in Figure 1 and as Oates and Sullivan mention, energy simulation relies on accurate assumptions such as, operational hours, temperature set points, plug loads and number of occupants. These variables if changed have varying effects on the actual energy performance of the building, some more than others (Oates & Sullivan, 2012).

Accuracy of simulation tools
It is still unclear whether modeling tools can accurately predict total energy consumption of a building within an acceptable error margin. Menezes, Cripps and Buswell seek to understand this question by utilizing post occupancy evaulation as a tool to accurately predict lighting, plug loads (office equipment) and catering loads. In the study, detailed sub-metering monitored occupant behavior every half-hour for each of the four tenants in the seven-story building. The overall accuracy of the predicted model was within 3% of the actual energy consumption (Menezes, Cripps, & Buswell, 2011). The paper is limited since it does not take into account the interdependencies between the HVAC system, heating and cooling loads and energy losses through the envelope. All these variables increase the level of complexity when predicting energy consumption monthly and annually.
In a similar study real-time simulation was performed to analyze whole building performance. In their research Pang, Wetter, Bhattacharya, and Haves monitored a commerical buliding in real time to compare simulated energy consumption using EnergyPlus to that of actual data. Despite real-time inputs discrepencies were observed pertianing to chiller operation strategies (Pang, Wetter, Bhattacharya, & Haves, 2012).
This can be attributed to the source code and equations used within the program (Zhu, Hong, Yan, & Chuang, 2013).

Sensitivity Analysis
Understanding relationships that exist within a building and factors that have a significant impact on energy consumption can help during the modeling process.
Relationships such as a direct correlation between occupancy and energy consumption will entail time should be spent detailing schedules accurately. Elie Azar and Carol Menassa address the impact of occupant behavior on energy performance in a typical office building using nine different occupant behavior parameters (Azar & Menassa, 2012). These nine parameters will be of key focus when undertaking a site walkthrough and modeling of the RJC. Sensitivity analysis results for climate zone 1 with greater than 7000 HDD ( o F-day) will be reviewed, since this best represents Toronto's climate with 4066 HDD ( o C-day) (Natural Resource Canada, 2012) (Azar & Menassa, 2012).
Taking this into consideration is critical because it helps to understand which parameters within the building for Toronto weather have the greatest effect on energy consumption. As mentioned within the study, errors generated within predicted models are due to, "simplistic and idealistic data that does not represent actual building systems and occupancy" (Azar & Menassa, 2012). This is very closely related to the conclusions also drawn in the PROBE study. Combining this with the analysis of nine different occupancy behavior parameters will help reduce errors. A list of the nine parameters is shown in Table 1 below (Azar & Menassa, 2012 (Menezes, Cripps, & Buswell, 2011). In a similar sensitivity analysis study, the top three factors having the greatest influence on an office building were mechanical ventilation rate during daytime winter, lighting control and infiltration rate during nighttime winter (Heiselberg, Brohus, Hesselholt, Rasmussen, Seinre, & Thomas, 2009). Furthermore, as noted from Figure 1, the Carbon Trust study also noted that lighting, scheduling effects were generally excluded which have a significant impact on energy consumption (Carbon Trust, 2011). With synergy being drawn to a number of studies it is evident that the nine parameters do play a critical role in energy consumption and adds relevance and validity to this study performed.
Thus, based on the literature review it is conclusive that the performance gap is an issue in today's LEED energy efficient buildings. The studies have demonstrated that taking into consideration occupant behavior, scheduling, set points and after hour activities can help develop strong design predictions. Implementing the suggestions and practices mentioned will require a detailed site and systems analysis of the RJC.
By adding a full building analysis as compared to the study performed by Menezes et. al this paper will help fill the gap of modeling accuracy using the most recent building data and specification (Menezes, Cripps, & Buswell, 2011). It will also add to the analysis of occupancy behavior where all the nine occupancy behavior parameters are taken into consideration and not analyzed in isolation. By taking actual data this report will analyze the error difference between actual performance and predicted using real weather data, current scheduling patterns, occupancy loads, lighting systems used and equipment; an effort not represented by current research reviewed.

Review of simulation tools
There are currently well over 200 building energy simulation tools on the market for engineers, architects and owners to utilize to evaluate their design. Each program has its own capabilities and limitations for different climate zones. For the purpose of this paper eQuest has been selected. The reason for this is because RJC is a LEED Gold certified building and eQuest is one of the eight approved modeling programs set out by Canada Green Building Council (CaGBC). During the LEED certification process Enermodal Engineering utilized this program for their proposed savings the building would achieve. However, despite this it is important to understand the limitation of this program to help access any discrepancy between actual and predicted energy performance. To understand the limitations of eQuest, comparison will be made to EE4, Loads" seeks to "analyze algorithms, modeling assumptions and simplifications that lead to large discrepancies" in the modeling softwares (Zhu, Hong, Yan, & Chuang, 2013). Therefore, it is not only important to understand the features that are lacking within each program as outline by Crawley et al. (2008) but also how each program inputs "default values, or algorithms for load simulations" (Zhu, Hong, Yan, & Chuang, 2013). Through simple testing and verification such as ASHRAE Standard 140-2007, DOE-2.1E and EnergyPlus "had the fundamental capabilities and appropriate modeling assumptions for load simulation" (Zhu, Hong, Yan, & Chuang, 2013). In further testing, DOE-2.1E was found to be limited in accurately calculating the heat balance between multiple zones because of its simplification for long-wave radiation exchange and multiple zone solutions. Thus, caution should be given when modeling radiant cooling or heating applications and different operating conditions in adjacent spaces. The effect of this can be observed in higher cooling loads and lower heating loads over an annual basis (Zhu, Hong, Yan, & Chuang, 2013). Awareness of these limitations within the source code of the program will help analyze the output data with better understanding and any discrepancies.

Post-Occupancy evaluation
A number of the papers reviewed attribute the evaluation of energy consumption within the building as post-occupancy evaluation (POE). As described by Preiser, post occupancy evaluation seeks to satisfy the needs of the occupants within the building (Preiser, 2013). It evaluates to see if the facilities within the space support the people's needs. This is significantly different from what is described as a technical test such as, energy audits in which the tests are done regardless of its occupants (Preiser, 2013).
This does become confusing because examining the energy performance (energy audit) of a building is dependent on the occupancy load, schedule and the building systems to meet the required set points. Thus, it can be said that an energy audit may be one component of POE because an energy audit may "draw on available knowledge through data collection and instruments in order to predict a building's likely performance over a period of time" (Preiser, 2013).
Therefore, in order to perform an accurate energy audit of the RJC a site assessment will have to take place. This walk through of the site will take note of occupancy load, scheduling, instruments installed within spaces to reduce energy consumption, equipment installed and HVAC system. A more detailed outline will be provided in the methodology.

Limitation in current case study and sources of discrepancy
In comparing the actual energy performance versus modeled energy performance it is critical not to repeat mistakes that cause the performance gap in the first place. Based on the PROBE study and Carbon Trust the major source of discrepancy was a combination of poor assumptions primarily in the area of occupancy behavior due to the increased level of controllability within the space; operational hours; lack of allocating power consumed by small plug loads; and external lighting (CIBSE Energy Performance Group, 2012). All these small differences can amount to 30% of energy consumed in office buildings; followed by lack of monitoring post-occupancy the industry is not able to correct itself to better improve their modeling assumptions and methods (CIBSE Energy Performance Group, 2012). These issues could prevail in this research however; steps will be taken to help reduce such errors, which will be outlined in detail through the methodology.

Methodology
In order to analyze and compare the actual versus predicted energy consumption a case study approach is being adopted. A similar process adopted by the PROBE study with few modifications will guide the process of analyzing the energy consumption of the Ron Joyce Centre.

Actual Data
To create an accurate energy model using eQUEST a complete issue for construction (IFC) package will be obtained from Enermodal Engineering and Sodexo.
The IFC package includes as built AutoCAD drawings outlining architectural plans, lighting schedules and mechanical systems. This information is critical to designing the shell of the building with proper zoning allocations and envelope design. Data pertaining to commissioning and onsite testing will help determine the equipment efficiency and heating and cooling capacity for mechanical systems installed. This is in relation to heat pump CFM, fan coil heaters, exhaust fans and the air-handling unit (AHU). In addition, monthly utility bills and sub-metered data will aid in the calibration and validation process. Regular site visits will be conducted to gather information related to number of computers in use, window blinds closure, café operation hours and class schedules.

Simulation
Gathering all the information from site visits, IFC packages and commissioning reports will be utilized to create an eQUEST model of the building. The initial construction of the building will be created using the Design Development Wizard to input initial zoning allocation, window placement, scheduling and HVAC design. A few deviations in the shell design will be encountered due to simplification (grouping) of the interior spaces. This process will lead to a takeoff and calculations for the building to input lighting density, equipment load and general scheduling depending on zone characteristics. This process will help reduce any assumption entered into the program and reduce calibration and validation time in detailed design mode. There will be a total of four simulation runs being performed including 2011, 2012, 2013 January to April and CTMY-30 year Toronto average.

Weather file
To accurately simulate the energy consumption of the building for the time period of January 2011 to April 2013 real weather data will be retrieved from Environment Canada weather station. Figure 2 below displays the six possible weather stations in proximity to the RJC. Onsite weather data will be collected from the building automation system (BAS) for 3-4 days to help determine the local weather station that best correlates with respect to temperatures and relative humidity the building experienced over the years. Through this process an ".epw" weather file will be created and converted into a ".bin" using eQuest weather conversion software.

Validation and calibration
After creating the eQUEST model using the wizard mode and performing an initial simulation, reference will be made to the actual utility bills for validation and calibration purposes. This process is necessary to make any adjustments to equipment loads, café operations, occupancy loads and scheduling. In order to make the necessary changes adjustments will be made using eQUEST detailed mode where specific input parameters may be adjusted to individual zones and equipment.
The validation technique most likely to take place is empirical test in which calculated results from eQuest will be compared to monitored data from the RJC building (Neymark et al., 2002). This technique offers an approximate truth within experimental accuracy and at any complexity (Neymark et al., 2002). The main disadvantage is the uncertainties, which stem from two main groups, which include external errors and internal errors (Neymark et al., 2002).

External error types
• Differences between the actual microclimate that affects the building versus the weather input used by the program. • Differences between the actual schedules, control strategies, effects of occupant behavior and effects from the building existing in the "real world", versus those assumed by the program user.
• User error in deriving building input files.
• Differences between the actual thermal, optical and other physical properties of the building including its HVAC systems versus those input by the user. Internal error types • Differences between the actual thermal transfer mechanisms taking place in the real building and its HVAC systems versus the simplified model of those physical processes in the simulation (all models are by definition simplifications of reality). • Errors or inaccuracies in the mathematical solution of the models.
The biggest challenge is interpreting the results because "all possible error sources are simultaneously operative" (Neymark et al., 2002). Thus, reviewing and analyzing the output data from eQuest can make adjustments to lighting, café and occupancy schedules, and HVAC system for natural gas consumption. Though this process helps to refine the input parameters into the program it does not eliminate the errors completely.

General
The Ron Joyce Centre located in Burlington, Ontario, is a five story,

Shell and HVAC
The basic design of the building shell comprises of a curtain wall system for maximum daylight penetration. The interior of the building has a central atrium space that allows daylight to enter deep into the core of the structure. The mechanical system is a hybrid system comprised of a water loop heat pump system in conjunction with an outdoor air makeup unit for fresh air ventilation. The system contains an economizer and energy recovery wheel to help save energy. Smaller systems are also incorporated including radiant floor heating on the first floor along the perimeter and fan coil heaters in designated zones.
The big hall and auditorium on main floor have CO 2 sensors to help regulate ventilation within the space. For the purposes of this simulation, the demand ventilation control will not be installed rather operate of a fixed schedule, Monday to Sunday.

Utility Data
Utility bills for gas and hydro from January 2011 to April 2013 have been complied for comparison against simulated energy consumption. The electrical metering is done by Burlington Hydro Inc. In addition, McMaster has an independent sub metering system within the building to measure electrical consumption by end use. The end use categories are lighting, HVAC, elevators, emergency, power panels and café.
However, it must be noted in pervious attempts to record this data, the sub metered readings did not correlate with the utility bills. The sum of sub metered electrical consumption was about up to twice the actual electrical consumption noted by Burlington Hydro Inc. Appendix B Table B In addition, it must be noted from August 2010 to November 2, 2011 the natural gas consumption was estimated by Union Gas. This resulted in no measured natural gas consumption per month, due to an initial and final reading spanning more than a year. This will render cross-referencing between predicted and actual consumption ineffective. By calculating the ratio of heating degree-day to natural gas consumption for 2012 a more accurate estimate may be generated for 2011.

Objective
As part of this case study two questions are going to be answered including: 1. Examine the variation between simulated and real data and be able to explain the cause for the differences. Focus will be on 2012 results, as this year has the most complete data.
2. Compare and analyze the difference between LEED proposed energy consumption and 2012 building energy consumption.
By answering these questions it would be valuable to both the McMaster University and the operation and management team ensuring the building is running efficiently. For future reference McMaster University can have a benchmark to follow in future constructions or in evaluating existing buildings on campus with similar functionality. Though the building is operated by a third party company this data will help improve the building on a yearly basis or just ensure that it's running optimally.

Modeling
Modeling of the RJC building was completed using eQuest v3.64 and the following sections summarizes modeling notes for the design of the building.

Zoning
Each floor of the building was modeled as a separate shell in eQuest to generate the four-story building. Areas of similar use, occupancy schedule and set point temperatures were grouped together to simplify the design of the building. A detailed zoning drawing for each floor is provided in "Appendix C" for review. In creating the zoning for the fifth floor, screened exterior spaces located in the north and south were discounted due to exposure to exterior environment.

Building Shell
The envelope construction of the building consists mainly of a curtain wall system and opaque sections between floors. A detailed material selection and thermal resistant values for different envelope sections are listed in Table 2.

Scheduling
Accurately representing the schedule for the building is a challenge. Students are free to enter and leave during operating hours of the building. Given the class schedule and events that may take place during the weekend a detailed schedule was generated.
Refer to Appendix D for a detailed outline of the schedule for occupancy, lighting, office, and miscellaneous equipment. The occupancy ratio for each month was generated based on maximum capacity for each floor. The capacity for each floor is as follows; first floor 600 persons; second and third 300 persons each; fourth is not applicable (incomplete); and fifth 11 persons. Related schedules such as lighting, HVAC, plug loads, and miscellaneous loads were based on occupancy since there is a direct correlation.

Weather data
Site temperature reading for June 2, 3, 4, 6, 7, 18 and 19 th 2013 were gathered from the BAS. Comparisons were made to local weather stations including Hamilton A,

Burlington Piers (AUT), Oakville Town Ontario, Grimsby Mountain Ontario, Southern
Ontario emergency and Hamilton RBG CS. Station Hamilton A correlated best with site temperatures recorded over the selected days with minor differences in temperature readings. Figure 3 shows the delta temperature between site and Hamilton A station readings. It must be noted the site temperature readings have a consistently high reading between the time of 5pm and 7pm each day. This would be attributed to direct solar radiation on the sensor, generating an inaccurate reading. The average temperature difference as displayed in Figure 3 is 0.32 degrees Celsius.
In generating the weather file for eQuest, station Hamilton A detailed all the required information except global horizontal and direct normal solar radiation. This data was retrieved from the Canadian Weather for Energy Calculations (CWEC) for Toronto weather. Cumulating this information together, weather files were generated for 2011, 2012 and 2013 January to April. Appendix E details the weather file information.

HVAC
The design of the HVAC system in eQuest was modeled based on IFC mechanical engineering drawings. Appendix F details the actual systems within the building including, air schematic and hydraulic schematic. Modeling this information into eQuest is divided into two sections, water-side HVAC and air-side HVAC design.

Water-side HVAC
Three boilers were modeled in total for the system. Boiler one provides hot water for baseboard heaters, unit heaters and fan coil heaters located on the first, fourth and fifth floor. Boiler two ties in with the water loop heat pump for the first, second and third floor. Boiler three supplies hot water to the outdoor air makeup unit (OAMU) to precondition the outdoor air for the whole building. The OAMU is assigned to a dummy

Temperature difference between site and Hamilton A station
zone within the building. Section 7.5.2 will go into more detail with regards to this. It must be noted that the load from boiler three is taken into consideration for total electricity consumption in the building and does not have a value of zero. Four electric hot water tanks located throughout the building supply domestic hot water to washrooms, kitchens and showers.

Air-side HVAC
Heat pumps, unit ventilators, fan coil heaters and/or baseboard heaters supply each of the zones. Fresh air for the building is supplied by the OAMU, which has an actual output of 7931 L/s (16,805 CFM). The OAMU is assigned to a dummy zone within the building since eQuest cannot model multiple HVAC systems for a given zone.
The dummy zone has no internal loads such as people, lighting, equipment, exterior surfaces and envelope. The energy consumed by the OAMU supplying the dummy zone will be included in the total electrical and natural gas consumption.
With a total area of 7024 m 2 (75,623 ft 2 ) covering the first to third floor, the average actual ventilation rate is 1.13 L/s.m 2 (0.22 CFM/ft 2 ). With a maximum of 1200 persons the average actual ventilation rate per person is 6.6 L/s.person (14 CFM/person). This compares to AHSHRAE 62.1 ventilation rates ranging from 0.6-0.9 L/s.m 2 (0.12-0.18 CFM/ft 2 ) and 3.8-5 L/s.person (7.5-10 CFM/person). Thus, the building is well within the requirements of AHRAE 62.1. With the future competition of the fourth floor the total area will be an additional 2063m 2 and 300 persons, totaling 9087 m 2 with 1500 total person capacity. The ventilation rates will be 0.88 L/s.m 2 (0.17 CFM/ft 2 ) and 5.33 L/s.person (11.2 CFM/person), which is well within the requirements of ASHRAE 62.1.
For each zone the total heating and cooling capacity was calculated along with their respective coefficient of performance (COP). Table 3 & 4 provides detailed specifications for heat pumps and heating systems on each floor. Test results of actual air flow output for each heat pump was taken into consideration when assigning overall airflow into each zone. In addition, exhaust fans (EF) 1.1, 1.2, 2, 4, and 6 were modeled into the building; with an exclusion of EF 3, which is used under emergency situations only. Table 5 below details each EF and their corresponding airflow.

HDD (° C-day) 2012
HDD (°C-day) Natural gas consumption (m^3) x10 Altering this reference temperature to increase R squared required a number of iteration to arrive at a final T ref = 13 °C and R 2 = 0.92448. Figure 7 illustrates the relationship between HDD and natural gas consumption for a reference temperature of 13 °C.

Results and Analysis
Based on the modeled information, the output results from eQuest for 2011 to 2013 and CTMY Toronto weather average are presented in the following section. The generated output is divided into two energy consumption sources including electricity (kWh) and natural gas (m 3 ).

2011 Results
Figure 9 offers a break down of the electrical consumption predicted by eQuest and actual based on utility bills per month. The total electrical consumption for 2011 predicted by eQuest is 1097115 kWh with actual totaling 1031486 kWh, which is only a 6.38% difference. On a monthly basis, May experienced the largest difference between predicted and actual, 94938 kWh versus 62636 kWh respectively.
The monthly electrical consumption is subdivided by end use in Figure 10.

2012 Results
For 2012 the total predicted electrical consumption is 1096133 kWh compared to the actual 1088331 kWh, which is less than a 1% difference, as shown in Figure 12.
The largest variation is in April with a 17.79% difference. The end use electrical consumption demonstrates a similar pattern to 2011 with no variations, as per Figure   13.

CTMY Toronto
The CTMY Toronto weather file is a 30 year average and the most appropriate weather file for energy modelers to use for buildings in Southern and Central Ontario.
The output of this weather file will be compared to 2012 actual energy consumption from the RJC.
In evaluating the electrical consumption there is no significant difference between actual (2012) versus predicted data. In Figure 18 the monthly results from eQuest is compared with 2012 actual utility data. On an annual basis the CTMY Toronto average has a discrepancy of only 0.32%, while the 2012 weather file has a discrepancy of 0.72%. Both weather files accurately predict electrical consumption for the building to within less than one percent; however, CTMY has a smaller error.

Comparison to LEED submission
As illustrated through the literature review, many of the LEED certified buildings are not meeting their expected energy consumption. The sources of generated error have been documented and presented along with means of resolving the performance gap. Since the RJC is a LEED Gold certified building, validating its current energy consumption to the proposed design would serve as a relevant analysis.
Before analyzing and comparing the results between the LEED proposed, 2012 actual and CTMY a preliminary review of the inputs in the program was performed. Key differences to take note of between the initial LEED simulation and current simulation runs mainly include schedules and occupancy loads. In the LEED proposed design the building has an occupancy load of 90%. In comparison, the current simulations have a significantly lower occupancy load at around 50% of maximum capacity, as illustrated in

Modifications
In current simulation runs the building is not operating at full capacity, with an average occupancy load of about 50%. This proved to be challenging when making a comparison with the LEED proposed design, which had an occupancy load of 90%.

Modified simulation, electrical consumption
Modified LEED proposed gains due to the increased occupancy load. The overall annual difference for natural gas due to the changes amounts to 13% annually.
Through the process of increasing the occupancy load to 90% of 1200 persons, the output values correlate very closely with the initial LEED proposed energy consumption. This process makes for a more valid comparison and demonstrates initial estimates hold validity. The predicted natural gas consumption is harder to analyze with certainty due to the lack of accurate data. There are high levels of fluctuation in the results given each year has a varying number of HDD. This has resulted in differences ranging from 33% for 2011 annually to a best result of 4.51% annually in 2012. This inconsistency makes it hard to validate actual building energy consumption to LEED proposed energy consumption and expected savings. As noted by the study conducted by Zhu et al.
eQuest's results for natural gas consumption where not lower than actual. The results fluctuated between simulation runs as noted in In comparison to the LEED proposed design overall the building performing 23% better for electrical consumption but has a 37% increase in natural gas consumption at current occupancy loads. However, by increasing the occupancy load within the building the LEED proposed design holds more validity with electrical consumption increasing and natural gas consumption decreasing.

Recommendations
When evaluating a building's actual energy consumption versus predicted it is hard to find the root cause of the difference because all possible errors within the program are occurring simultaneously. This includes occupancy load, scheduling, lighting schedule, temperature set points and HVAC design.
For the study involving the RJC building the overall electrical consumption was highly accurate based on inputs into the program. However, natural gas consumption is highly erratic and consistency has not been demonstrated with the output data. This once again could be attributed to errors generated with inputs and settings placed by the modeler or an inherent discrepancy within the program.
In future studies involving the RJC it would be of primary importance to recalibrate the sub-metering system. In the process accurate end use electrical consumption within the building can be properly verified and cross-referenced with simulation runs and utility bills. This will help to accurately input values for plug loads, office equipment and miscellaneous equipment, which have a direct effect on internal heat gains. It must be noted that there will be an increase in electrical consumption as the occupancy load increases.
As the modified simulation proposes, with the increased occupancy load the natural gas consumption could reduce. This is due to increased internal heat gains; however, with the fluctuations in natural gas consumption this is not certain. Alternately, by installing a weather station at the building, daily weather data can be collected to create a customized weather file. This will be the most accurate way of simulating the environment and element the RJC is exposed to.