1、 Matthew Tokarik is a MASc student, and Russell Richman is a professor in the Building Science Program, Department of Architectural Science, Ryerson University, Toronto, Ontario, Canada. A Multi-Objective Optimization Analysis of Passive Energy Conservation Measures in a Toronto House Matthew Tokari
2、k Russell Richman Student Member ASHRAE Member ASHRAE ABSTRACT Advancements in whole building energy modeling have coincided with the demand for improved building energy performance and have become a useful tool in determining optimal configurations of energy saving measures on the path to net zero
3、building. This study presents a multi-objective optimization analysis in which passive energy conservations measures of a high performance house in Toronto are evaluated for life cycle cost and performance. The main objective of the study was to identify economically efficient design solutions that
4、may be used to inform future efficient housing design and housing performance standards. An optimization environment was created using the JEPlus software suite where a case study house acted as the reference building. The simulation model of the case study house was calibrated using a data-driven p
5、rocedure with reference to utility bill data, air-temperature sensors, and short term spot measurements. Acceptable CV(RSME) and NMBE tolerances for monthly natural gas and electricity consumption as well as zone air temperature were reached in accordance with ASHRAE Guideline 14 calibration require
6、ments. The optimization varied passive energy efficiency parameters in search of configurations yielding optimal building performance and life cycle cost. The optimization results showed that energy savings of 33% relative to building code minimum were justified at the point of minimal life cycle co
7、st via passive energy saving measures alone before considering active systems. These results suggest that improved thermal envelopes are economically advantageous with good building practice. However, they suggest that the current Passive House standard does not coincide with the economic minimum fo
8、r the local economic and environmental climate. INTRODUCTION Over the last two decades, the residential subsector has consumed approximately 17% of Canadas total secondary energy use and created 14% of Canadas greenhouse gas (GHG) emissions on average (Natural Resources Canadas Office of Energy Effi
9、ciency 2014). To reduce such emissions as the Canadian population grows, the Government of Canada has adopted a residential building goal of net zero energy (NZE), where houses produce as much energy on-site, often with photovoltaics (PV), as they consume annually (Natural Resources Canada 2014). Ty
10、pically, NZE houses require reduction of space conditioning load through building envelope improvements, direct solar exposure through good orientation, and energy demand reduction through selection of highly efficient appliances and space conditioning equipment before utilization of renewable energ
11、y sources can be considered viable. The passive measures also create benefits such as improved thermal comfort through increased interior temperatures of exterior wall and fenestration surfaces, improved structural durability through moisture damage mitigation, and improved resilience against outage
12、s through reduced space conditioning demand and increased heat retention ability. Despite the added benefits, cost continues to be a significant deterrent for majority market adoption of conservationist building practice. In a survey of Canadian residential building contractors, the cost of high per
13、formance building materials and the cost of adoption, including training and certification, were the most significant barriers for energy efficient construction (The Construction Sector Council 2011). This common perception mainly considers initial cost and does not look at other measures of value o
14、r long term capital value. It is generally understood that the greater initial capital costs are required to achieve improved energy efficiency through the inclusion of energy conservation measures (ECM) but reduced energy bills can return this investment over the long-term. The actual financial rel
15、ationship between initial capital cost and long-term savings for energy efficient homes is less obvious because of the variability in market construction rates, diminishing returns on ECM upgrades, mortgage rates, fuel price escalation, weather patterns, housing maintenance, and occupancy behavior.
16、Therefore, while home builders, consumers, and policy makers are aware of the benefits of improved housing efficiency, they are unsure where the effort no longer becomes financially viable. This study will present a virtual optimization environment, which couples a building simulation engine with ba
17、tch handler and optimization algorithm, and use it to investigate the extent to which passive ECMs remain financially viable for housing in the cold climate of Toronto, Canada. The study uses multi-objective optimization to assess parameters for life-cycle cost and building performance, where the de
18、sign intent is to first reduce building loads as much as possible before considering active systems or energy production. Relevant Studies With advances in computer science, designers and researchers are now well-equipped with building simulation programs that can predict the effects of design varia
19、bles on building energy consumption. When such simulation engines are coupled with mathematical optimization algorithms, combinations of chosen variables can be iteratively evaluated relative to each other, to minimize one or more function. The process searches the solution space and objectively det
20、ermines an optimal or near-optimal combination of variables in a relatively short time. Typically, reference buildings in optimization studies are created hypothetically using historic weather data, national building code standards, and representative averages, or use national research houses made t
21、o represent typical dwellings (Nguyen et al. 2014). However, few studies use actual occupied buildings, where the model is calibrated to real-world data (Bucking et al. 2014). Recently, in a method similar to that outlined in the Energy Performance of Buildings Directive (The European Parliament and
22、 the Council of the European Union 2010), the Passive House Institute US (PHIUS) modeled a representative detached house and optimized passive energy saving parameters for 111 climate zones, for life cycle cost and annual energy savings concurrently using BEopt, to determine a cost-competitive perfo
23、rmance level which led to the climate specific PHIUS+ 2015 standard (Wright et al. 2014). In the current study, the approach is similar to that of the aforementioned studies, but utilizes an existing house, with accurate climate data and measured performance levels, as a reference building to create
24、 a calibrated building energy model and improve result reliability. METHODOLOGY A feasibility assessment of passive energy efficiency measures was performed by coupling the EnergyPlus simulation engine with the jEPlus optimization suite (Zhang 2012), which utilizes a genetic algorithm to minimize th
25、e specified objective functions. Specifically, jEPlus+EA uses the Non-dominated Sorting Genetic Algorithm II (NSGA-II), a sub-classification of an evolutionary algorithm to optimize the parameters. Simulation batches were performed on a local 32-core machine at Ryerson University, in Toronto, Canada
26、, which performed approximately 7 simulations per minute. Reference Building The reference building for the optimization study was an existing 3-story single family detached house in Toronto, ON. It was built in the early 1900s with double-wythe structural masonry, a rectangular footprint with the n
27、arrow dimension facing north and south, and flanking houses separated by approximately 1.5 m (3.3 ft) (Figure 1). It underwent an extensive retrofit in 2010 with the objective of creating an exemplar for sustainable housing renovation. The retrofit strategy followed the Passive House design methodol
28、ogy, but did not reach certification levels. The envelope was treated with closed cell spray foam on the interior of the existing brick, and quad-glazed windows with fiberglass frames. Heat is produced by a highly efficient natural gas boiler for space conditioning via in floor radiant tubing, and d
29、omestic water use. Space cooling is delivered from one wall-mounted mini-split air source heat pump, and balanced mechanical ventilation is driven by an ERV. The homeowner provided utility consumption, electricity and natural gas spot measurements, and hourly indoor air-temperature data, making the
30、house a good candidate for research. The EnergyPlus building model was calibrated using a data-driven method, where model input is based on a data reliability hierarchy (Raftery et al. 2011a, 2011b). Acceptable CV(RMSE) and NMBE tolerances were reached for monthly natural gas and electricity consump
31、tion and hourly indoor air temperature for each zone according to the limits set by ASHRAE Guideline 14 (ASHRAE 2002). A summary of each calibrated data set is shown in Table 1. The design variables with their associated performances and costs are shown below (Table 2, Table 3, and Table 4) with the
32、 as-built components bolded. It should be noted that air-tightness was not included as a design variable, as the incremental cost per performance increase could not be ascertained. Costs were collected from as-built receipts and invoices where available, or supplier quotes. All other unlisted variab
33、les remained constant during the simulation procedure to retain the fundamental behavior of the calibrated simulation model. Estimated costs were adjusted to represent the Canadian renovation year, 2010, when the average Canadian Dollar (CAD) exchanged for 0.971 US Dollars (USD) (Bank of Canada 2015
34、), and the Historical Index conversion from 2015 to 2010 was 0.905 (RSMeans 2015). Table 1: Calibration Summary of Building Energy Model Data set Monthly Hourly Error CV(RMSE) NMBE CV(RMSE) NMBE Tolerance 15% +/- 5% 30% +/- 10% Natural gas 12.0% 0.3% - - Electricity 4.6% -1.7% - - Basement - - 3.2%
35、0.6% Basement S - - 11.6% 5.9% Living - - 5.4% 0.1% Kitchen - - 6.3% 2.9% Bedroom 1* - - - - Bedroom 2 - - 7.0% 1.9% Baths - - 6.2% 0.7% Bedroom 3 - - 5.4% 0.6% Bedroom 4 - - 6.5% -1.7% Third Bath - - 6.0% -1.9% * No air temperature sensors located in Bedroom 1 Figure 1: Graphical representation of
36、simulated reference building Table 2: Glazing Package Design Variables (5 Options) Description Assembly U-Value, W/m2K (Btu/hrft2F) COG U-Value, W/m2K (Btu/hrft2F) Frame U-Value, W/m2K (Btu/hrft2F) SHGC Cost, 2010 CAD/m2 (2010 USD/ft2) Glazing Package 1, double, low-e, vinyl frame 1.726 (0.304) 1.49
37、6 (0.263) 2.271 (0.400) 0.52 483.66 (43.62) Glazing Package 2, triple, low-e, fiberglass frame 1.087 (0.191) 0.875 (0.154) 1.703 (0.300) 0.41 810.50 (73.10) Glazing Package 3, triple, low-e, fiberglass frame 0.930 (0.164) 0.668 (0.118) 1.703 (0.300) 0.34 975.62 (88.00) Glazing Package 4, triple, low
38、-e, wood frame 0.763 (0.134) 0.676 (0.119) 0.795 (0.140) 0.50 1394.77 (125.80) Glazing Package 5, quadruple, low-e, fiberglass frame 0.652 (0.115) 0.385 (0.068) 1.476 (0.260) 0.23 862.83 (77.82) Table 3: ERV Design Variables (3 Options) Description Sensible Effectiveness 100% Latent Effectiveness 10
39、0% Average Power, W (Btu/hr) Cost, 2010 CAD/ea (2010 USD/ea) ERV Unit 1 0.73 0.55 91 (311) 3567 (3269) ERV Unit 2 0.84 0.71 128 (437) 4305 (4180) ERV Unit 3 0.93 0.55 155 (529) 7669 (7446) Table 4: Opaque Assembly Design Variables (9 Wall Options, 7 Slab Options, and 14 Roof Options) Assembly Descri
40、ption Reff, m2K/W (hrft2F/Btu) Cost, 2010 CAD/m2 (2010 USD/ft2) Above Grade and Below Grade Wall (9 Options Each) 64 mm (2.5 in.) cavity CCSF + 38 mm (1.5 in.) CCSF c.i. 3.76 (21.3) 49.72 (4.48) 64 mm (2.5 in.) cavity CCSF + 51 mm (2 in.) CCSF c.i. 4.22 (24.0) 55.93 (5.04) 64 mm cavity (2.5 in.) CCS
41、F + 76 mm (3 in.) CCSF c.i. 5.46 (31.0) 68.36 (6.17) 64 mm cavity (2.5 in.) CCSF + 89 mm (3.5 in.) CCSF c.i. 6.03 (34.2) 74.58 (6.73) 64 mm cavity (2.5 in.) CCSF + 102 (4 in.) mm CCSF c.i. 6.59 (37.4) 80.79 (7.29) 64 mm cavity (2.5 in.) CCSF + 127 mm (5 in.) CCSF c.i. 7.71 (43.8) 93.22 (8.41) 64 mm
42、cavity (2.5 in.) CCSF + 152 mm (6 in.) CCSF c.i. 8.82 (50.1) 105.65 (9.53) 64 mm cavity (2.5 in.) CCSF + 178 mm (7 in.) CCSF c.i. 9.94 (56.4) 118.08 (10.65) 64 mm cavity (2.5 in.) CCSF + 203 mm (8 in.) CCSF c.i. 11.05 (62.7) 130.51 (11.77) Basement Slab (7 Options) None 0.00 (0.0) 0.00 (0.00) 25 mm
43、(1 in.) XPS c.i. 0.88 (5.0) 12.45 (1.12) 51 mm (2 in.) XPS c.i. 1.76 (10.0) 24.90 (2.25) 76 mm (3 in.) XPS c.i. 2.64 (15.0) 37.35 (3.37) 102 mm (4 in.) XPS c.i. 3.52 (20.0) 49.80 (4.49) 127 mm (5 in.) XPS c.i. 4.40 (25.0) 62.25 (5.61) 152 mm (6 in.) XPS c.i. 5.28 (30.0) 74.70 (6.74) Roof (14 Options
44、) 127 mm (5 in.) cavity CCSF 3.77 (21.4) 62.15 (5.61) 152 mm (6 in.) cavity CCSF 4.53 (25.7) 74.58 (6.73) 178 mm (7 in.) cavity CCSF 5.28 (30.0) 87.01 (7.85) 203 mm (8 in.) cavity CCSF 6.04 (34.3) 99.44 (8.97) 229 mm cavity CCSF (9 in.) 6.79 (38.6) 111.87 (10.09) 241 mm cavity CCSF (9.5 in.) 7.17 (4
45、0.7) 118.08 (10.65) 241 mm (9.5 in.) cavity CCSF + 25 mm (1 in.) XPS c.i. 8.31 (48.4) 130.53 (11.77) 241 mm cavity (9.5 in.) CCSF + 51 mm (2 in.) XPS c.i. 9.35 (53.1) 142.98 (12.90) 241 mm (9.5 in.) cavity CCSF + 76 mm (3 in.) XPS c.i. 10.35 (58.8) 155.43 (14.02) 241 mm (9.5 in.) cavity CCSF + 102 m
46、m (4 in.) XPS c.i. 11.32 (64.3) 167.88 (15.14) 241 mm (9.5 in.) cavity CCSF + 127 mm (5 in.) XPS c.i. 12.26 (69.6) 180.33 (16.27) 241 mm (9.5 in.) cavity CCSF + 152 mm (6 in.) XPS c.i. 13.19 (74.9) 192.78 (17.39) 241 mm (9.5 in.) cavity CCSF + 178 mm (7 in.) XPS c.i. 14.12 (80.2) 205.23 (18.51) 241
47、mm (9.5 in.) cavity CCSF + 203 mm (8 in.) XPS c.i. 15.03 (85.3) 217.68 (19.63) Objective Functions Objective functions are selected simulation results which vary depending on the parametric input combinations, and are the values to be minimized by the optimization algorithm. For this case, performan
48、ce and cost objectives were optimized simultaneously, in so-called multi-objective optimization. One performance objective selected was total peak design load. That is, the sum of heating and cooling peak design load, in W/m2 (Btu/hft2), as calculated by EnergyPlus using the heat-balance method (US
49、DOE 2013). This measurement was selected since it best reflects the performance of passive features, which were of greatest interest to the design. A second performance objective, annual energy savings relative to code minimum, was used as an objective function for a second simulation batch, where the intent was to test the theory that the Passive House standard should be set at the point of “the most energy efficient house design that is cost-competitive” (Feist et al. 2005). For this simulation batch, the model heating and cooling setpoints were changed to 20C (68F) and 25C (77F),
