DEVELOPING WINTER RESIDENTIAL DEMAND RESPONSE STRATEGIES FOR ELECTRIC SPACE HEATING

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. DEVELOPING WINTER RESIDENTIAL DEMAND RESPONSE STRATEGIES FOR ELECTRIC SPACE HEATING Leduc Marie-Andrée 1, Daoud Amed 1, and Le Bel Célyn 1 1 Laboratoire des tecnologies de l énergie d Hydro-Québec, Sawinigan, Qc, Canada ABSTRACT Te objective of tis paper is to present space eating strategies leading to a load reduction during critical periods on te grid but also taking into account te occupant termal comfort in order to make tem acceptable. Using TRNSYS, tree types of strategies were investigated: lowering te termostat setpoint, storing eat in te building termal mass during off-peak ours and lowering te setpoint during critical periods, and limiting te available power to baseboards. Results sow tat te strategies can save up to 7 per residence wic corresponds to 9% of te reference eating load of te average residence. INTRODUCTION In te province of Quebec, Canada, electricity is provided by te government owned utility, Hydro- Quebec, wic produces mainly from renewables (98% of its production). Te region s electric grid peaks in winter; around 38 MW as of 211. All-electric ouseolds are largely responsible for tis peak demand. More tan two tird of ouseolds use electricity as teir exclusive source of energy and are terefore defined as all-electric (AE) ouseolds. For approximately 87% of tese AE ouseolds, space eating is done using electric baseboards, located in eac eated room of te ouse, and controlled by teir own line-voltage termostat. During peak ours, te space eating load of AE ouseolds can account for up to 8% of teir total load. Te sape of te region s demand profile is strongly coincident wit te demand profile from tese customers. For te commercial and institutional sector (CI), electricity is te main source of energy for space eating for 6% of te utility s customers, of wic 75% ave baseboards and room termostats similar to te equipment used in te residential sector. For te winter s igest peak our, te combined space eating load, from te residential and CI sectors, represents 4% of te total electric utility peak load. It is expected tat te installation of electric space eating will increase over te coming years due to teir low initial and operating cost, ence reducing te gap between te available resources and te demand during peak ours. Demand response involving residential AE ouseolds could possibly relieve te electric grid and sould terefore be investigated. Many studies involving te residential sector ave been done over te years. California s vast Statewide Pricing Pilot (SPP) (CRA, 25) and Oregon s Olympic Peninsula (PNNL, 27) are probably te most well-known. From te detailed analysis of te SPP data (Herter et al., 27), it was found tat automation ad a great effect on load reduction, greater tan price signal alone. Since electric space eating is te most important load during peak ours, it terefore seems logical to focus tis study on possible automated control strategies for decentralised eating equipments. It as been decided tat tis study sould not involve te use of termal storage devices or alternative fuel sources in order to reduce te complexity and cost of te implementation of tese strategies. Tis would make tem applicable to all AE ouseolds since only teir line-voltage termostats would ave to be replaced by communicating termostats. Tese strategies must lead to a peak eating load reduction but must also take into account te occupant termal comfort in order to make tem acceptable. Te feasibility of suc strategies as to be demonstrated for ligt termal mass woodframed buildings typical for ouses in Quebec. In te past few years, many studies ave been done to evaluate te effect of controlling cooling equipments on te building's total peak load. Most of te work found in literature as been focussing on commercial and institutional buildings. Studies ave been done on te reduction of te cooling load using nigt cooling and on te effect of termal mass and termal storage devices (Braun et al, 23; Yang et al, 28; LBNL, 26; Yin et al, 21). Oters ave focused on optimal control of setpoints and storage (Lee et al., 28; Henze et al, 25). Katipamula (Katipamula et al, 26) and Reddy (Reddy et al, 1991) ave studied control strategies for HVAC equipment in residential buildings. Very few studies are available on winter peak reduction obtained by controlling te space eating equipment. Persson (Persson et al, 25) ave explored space eating peak load reduction but teir solutions involved te use of substitution fuel - 1111 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. sources. Lu (Lu et al, 25) as focused on te global effect of controlling termostatically controlled appliances on te electric grid but not on te effect of tis control on occupant termal comfort. Oregon s Portland General Electric Co. ad a pilot project in 24 on direct load control for electric space eating wic sowed tat residential customers could reduce teir peak load by.48 to.73 by reducing teir space eating setpoint by 2-3 C for 2 to 3 ours witout muc complaints (PGE, 24). Te Olympic Peninsula pilot project included te control of resistive eating equipment but only for omes controlled by one or at most two termostats. Te results of tis pilot study sowed tat a significant part of te overall load reduction due to price increase came from a decrease or sift in space eating load (PNNL, 27). SIMULATION TRNSYS 16 was selected to create te building energy model. It is widely used by te building simulation scientific community, particularly in researc. It as been used to evaluate peaking load control by oters (Henze et al, 25, Persson et al, 25, Braun et al, 22). Indeed, tis tool offers great flexibility. It is user-friendly and allows te incorporation of a wide range of systems troug its modular ( TYPE ) structure. Te TYPE56, te model of multizone building, was used to model te ouse described in te next section. TYPE56 is a non-geometric model wit an air node per termal zone. Te flow of eat received by te node is te sum of convective and radiative flows from te adjacent surfaces, gains by infiltration, ventilation, internal eat gains as well as air coupling from oter zones. Tis detailed multizone building model was cosen in order to simulate complex space eating load strategies adapted to te decentralized eating equipment typical of te type of ousing studied. Te time step used for te simulations vary greatly depending of te model. Te coice of 5 min seems reasonable. A smaller time step would proportionally increase te calculation time and would not significantly improve te solution as demonstrated later. Te meteorological file used for te simulation covers two years, from January 1st 23 to December 31t 24. In fact, te year 24 as recorded a peak demand of 36 268 MW on Hydro-Quebec grid on January 15t, for an outside dry bulb peak temperature of -3 C. All strategies are simulated on January 15t only. Te arcitecture of te modeled building is based on te twin experimental ouses built on te site of te Laboratoire des tecnologies de l énergie d Hydro-Québec (LTE). Te building as a total eated area of 234 sq. ft. (9 m 2 ) including a eated garage of 312 sq. ft. (29 m 2 ). Te total fenestration area is 25 sq. ft. (19 m 2 ). In tis model, te orientation of te building facade is facing nort. Table 1 sows te details of te building's structure. Table 1 Building Caracteristics Wall Material Tickness (cm) External Wall - sky view factor =.5 Insulation 127 Seating 12.7 Air vertical - Vinyl External Wall - sky view factor =.5 Insulation 127 Seating 12.7 Air vertical - Brick 12 Roof - sky view factor =.9 (surf37 ) et.85 (surf46 ) Insulation 152 Wool batts Insulation 73 Plywood 12.7 Roofing - Vertical Foundations - sky view factor =.5 Extruded polystyrene 5 Concrete 254 Horizontal Foundations Carpet 25 Extruded polystyrene 25 Concrete 12 Internal Wall (between ouse and garage) Insulation 127 1 From U-value evaluated by TRNSYS. Termal resistance 1 (K m2 / W) Insulation levels were primarily taken from Hydro-Quebec projects (HQ, 1993) (Millette et al., 24) and from Novoclimat tecnical requirements and guidelines (AEE, 21). Occupation scedules were generated using te forcing functions available in TRNSYS in steps of 5 min. Occupation scedules are representative of a family of four persons. A type of activity is assigned to eac area and determines te instant eat gains from te occupants (sensitive and latent), according to te standard ISO 773. Ligting effects were modeled by an internal eat gain equivalent to 5 or 1 W/m 2 (1% radiative). Tese assumptions are due to te coice of compact fluorescent lamps installed in te building. Tis gain is applied according to te corresponding occupational scedule, wic is valid for te winter period wic is te focus of tis study. We consider only te internal eat gain of equipments aving a demand of more tan 1 W (refrigeration/freezing equipments are excluded). 3.2 3.3 5.4 2.3 1.6 2.9-1112 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. Hence, te following are considered: cooking range, televisions and domestic ot water electric eaters. Clotes dryer was excluded since a large portion of te consumed energy is evacuated outside. Infiltration rates are given in air cange per our (ACH) and are representative of natural ventilation, as required in te model. Infiltration rates are assumed constant and applied to every zone aving an outdoor surface. In tis study we considered an infiltration rate of.213 ACH. Table 2 summarizes all te eat gains in te model. Table 2 Heat gains Simulation Year Infiltrations () Internal gains () Solar gains () 23-6 94 4 5 6 632 24-6 82 4 5 6 751 Equipments (W) Cooking range 45 TV 16 DHW 1 Termal comfort is evaluated according to ISO 773 standard. Tis calculation is included in te TYPE56 and is used to evaluate te comfort of te occupants of te modeled building. Comfort criteria were terefore added in te occupied areas to determine te viability of some load management strategies. Te insulation values of cloting (Englis cloting factor) were cosen based on cloting in winter, period wic is of interest in tis project. Only areas wit a scedule of occupation ave been considered. Te air velocity is estimated at.1 m/s, wic is representative of air flows in residential omes. To bring te effect of termal mass to te building representative of ouses in Quebec, two canges were made. Firstly, te capacitance of air (C p ) was canged to C p = 3 x V zone for all areas except te attic and te garage. Tis rule is consistent wit tat found in te literature (Carron et al., 27). Tis cange was made to obtain sort time constants of te same order of magnitude as tose obtained experimentally (Gunter, 21). Secondly, internal walls were added in te main occupied areas. Tese internal walls are wooden surfaces of 25 cm tick, wit a coefficient of convection for te internal walls of 11 kj/m 2 K. Tis addition allowed te long time constants to be of te same order of magnitude as tose obtained experimentally. Equipment tat could ave an impact on te eat balance of te building as been neglected for reasons of simplification of te model and te calculation time reduction. For example, a number of small appliances aving a low frequency of use were not modelled altoug tey contribute to reduce te eating load by producing internal gains. Local ventilation (kitcen oods and batroom fans) was not modelled because its frequency of use is low. Latent internal gains ave not been quantified for te ouseold appliances included in te model (e.g. te cooking range). Electric baseboard is te only eating tecnology used in te modeled building. Electric baseboard eaters are located in every room of te ouse and sized as tose in LTE's experimental ouses. Te total installed capacity is 17 including te garage. A fraction of te installed capacity may be considered as a gain by radiation tat will apply to all surfaces of te area. Tis as te effect of reducing te portion of eating wic is directly transmitted to te air by convection, as is te case in reality wit te plume of warm air wic warms te surface above te electric baseboard. However, all of te surfaces of te zone enjoy gain wile in reality, tis is te case for only one of te surfaces. Tis would increase te temperature of te surfaces togeter, so to modify te transfer of eat between tese surfaces and te adjacent area. Te effect sould be negligible. Te TYPE56 already integrates a perfect control of eating, namely tat te demand is equal to te eating load at any time. Tis demand can be limited to te installed capacity. Tis type of control is a reasonable approximation for an electronic termostat aving proportional control. Te termostats' setpoint is constant at 21 C except for garage wic is kept at C. VALIDATION In order to validate te model, te annual energy consumption, apparent UA value and time constants for te building ave been compared to eiter experimental data or oter models for similar buildings. Te apparent UA value is defined by Equation 1 and derived experimentally from Equation 2 using nigt time (E solar = ) winter data (to approximate a permanent regime). UA ( Δ T ) = UA ( Δ T ) + m C p ( Δ apparent enveloppe infiltration T ) Eq 1 E = UA ΔT t E E Eq 2 eating apparent ( ) internal solar Table 3 presents tese values for te two years studied. Experimental data was taken from Gunter (Gunter, 21) and te simulation data from Moreau (Moreau et al., 1994). Table 3 Building Caracteristics Building Annual eating () (23) 141 Model (24) 1374 Experiental data Simulation 1383 data 3 UA app (W/C) τ c τ l C c C l () ( /C) 154 43 58 6.6 8.9 158 1 36 2 54 2 6. 2 8.9 2 1 building 1993, 2 all buildings average, 3 RSI exterior wall = 3.52, ELA = 319 cm 2-1113 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. To evaluate te time constants of te building, two power cuts ave been simulated for eac year (release test). Te evolution of te average temperature of te building during te release test is sown to follow an exponential decrease and can be represented by a two time constants lumped capacitance model. Te sort time constant is evaluated by taking te data for te first tree ours of te release test and te long time constant wit te remaining data, Equation 3 is used to calculate te termal capacity C. Tint, t Text UAapparent t ln = t = Eq 3 T int, t Text C τ Te sensitivity of te model to certain parameters as been investigated; tese include te internal convection coefficient, time step, termal mass and infiltration rate. Effects on bot te temperature and demand profiles ave been studied for te peaking day of 24 as tis is te period of interest ere. Te internal convection coefficients used in te model can ave a significant impact on te load profile, terefore on te annual energy consumption. Beausoleil-Morrisson (Beausoleil- Morisson, 22) as found tat te use of an adaptive algoritm able to differentiate te surface type and conditions and use te appropriate correlation to calculate te internal convection coefficient can produce a significant difference on te annual energy consumption. Tis difference is greater for poorly insulated buildings. For tis model, two options ave been tested. Te first is te reference case wic uses constant internal convection coefficients for all internal surfaces (walls and windows). Te second uses TRNSYS ability to calculate te internal convection coefficient for eac surface at eac time step, terefore taking into account te effect of te temperature difference between te air and te surface. Te coefficient is calculated using te correlation expressed by Equation 4, were A = 5,76 and n =,3 for vertical surfaces, A = 2,11 and n =,31 for orizontal surfaces if T surf T air, A = 1,87 and n =,25 for orizontal surfaces if T surf < T air ; tese are te default values used by TRNSYS. ( T ) n c = A T Eq 4 surf air Figure 1 sows tat, using te default values, te calculation of te internal convection coefficients produces a difference of up to about,3 in te load profile for te period of interest (January 15t) wic is small but not necessarily negligible. Hence, in future work, tis aspect sould be furter investigated. Effect on te temperature profile is not significant. 15 1 5 reference calculated 98 99 91 911 912 913 Figure 1. Effect of internal convection coefficient on te load profile Te effect of reducing te simulation time step was investigated. Figure 2 sows tat tere is no net advantage in reducing te time step since bot cases produce similar load profiles. Te temperature profiles are identical. 15 1 5 1min 5min 98 99 91 911 912 913 Figure 2 Effect of time step on te load profile As discussed earlier, bot zonal capacitance and internal walls were included in te model in order to account for te termal mass of te building. Te effect of using one or te oter of tese metods was evaluated. Figure 3 sows tat te sole use of termal walls fails to produce a realistic temperature profile for te first our of te release test, as te air temperature drop is too drastic. Te effect of te time step is also very visible, as sown by te ripples in te slope. Deg C 22 2 reference 16 Cair = 1.2xVzone, wit internal mass 14 Cair = 3 x Vzone, witout intermal mass 12 98 99 91 911 912 913 15 reference Cair = 1.2 x Vzone, wit internal mass 1 Cair = 3 x Vzone, witout internal mass 5 98 99 91 911 912 913 Figure 3 Effect of termal mass - 1114 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. Increase in zonal capacitance produces a far more realistic temperature profile but, according to available experimental data, te long time constant (after tree ours of te release test) is sligtly less tan expected. Terefore, te use of bot metods to add termal mass seems to yield te best results but tis would ave to be validated experimentally. 15 1 5 reference.14 ACH.32 ACH 98 99 91 911 912 913 Figure 4 Effect of infiltration on load profile Te effect of infiltration rate as been investigated. As anticipated, te effect is important as presented in Figure 4. For te annual energy consumption, an increase of te infiltration rate to.32 ACH produces a 14% increase in energy consumption. A decrease of te infiltration rate to.14 ACH produces a 9% decrease in energy consumption. Te model, using ideal eating control, seems to produce reasonable results at least for te purpose of tis evaluation. Modeling of te basement could be improved by incorporating a finite difference 3- D model. Te infiltration rate could easily take into account te effect of meteorological conditions suc as wind speed, by incorporating Type75 (Serman-Grimsrud model) into te model. One consideration wort noting is te fact tat baseboard eating produces important temperature stratification in te zone wic will in turn affects te eat transfer troug floors and ceilings. Tis as not been taken into account. RESULTS Tree groups of strategies were evaluated. Table 4 presents te detail of tese strategies. Peak ours are defined as 6 to 1 and 16 to 2. Te reference case is for a constant temperature setpoint of 21 C. Figures 5, 6 and 7 sow te effect of te different group of strategies on te space eating load during te two daily peak periods. Te 912 t our corresponds to 6 on te peak day of 24. In order to evaluate te simulated strategies, tree indicators were used. Te evaluation of tose indicators for te tree groups of strategies is presented in Table 5 for te morning (am) and afternoon (pm) peak periods. Te first indicator, called PLR for power load ratio (Eq 5), represents te maximum power load during peak ours in comparison to te reference case. P PLR = P t fin max, strategy t t fin max, reference t Eq 5 It will be considered tat a PLR lower tan.8 is good (indicated by pale green sading in Table 5), between.8 and.99 acceptable. A PLR of 1. gives te indication tat tere is no power reduction during peak ours. Table 4 Load control strategies Strategy 1.Setback strategy Setback only during peak ours, instantaneous reset algoritm 2.Preeat strategy Envelope termal storage during offpeak ours and setback during peak ours, linear reset algoritm 3.Power limitation strategy Limitation of available capacity, instantaneous reset algoritm 1a- Second floor only @ ºC 1b- Basement only @ ºC 1c- First floor only @ ºC 1d- Second floor and basement @ ºC wit a progressive increase of te setpoint after peak ours on te second floor over 1 until te setpoint 21ºC is reaced 1e- Strategy 1a @ 16 C 1f- All floor @ C 1z- All unoccupied zones @ C 2a- Setpoint @ 24 ºC from, setpoint @ ºC during peak ours, setpoint increase @ 24 ºC from 1 until 16, setpoint @ 21ºC during evening 2b- Strategy 2a wit a morning preeat starting at 3 2c- Strategy 2a wit setpoints @ 25 ºC instead of 24 ºC 2d- Setpoint @ 24 ºC in unoccupied zones between -6, setpoint @ ºC during peak ours, setpoint increases during one our until 24 ºC from 1 until 16, setpoint @ ºC during peak ours setpoint @ 21ºC during evening from 2 2e- strategy 2c wit setpoints @ 23 ºC instead of 24 ºC 3a- 25% of te installed power available 3b- 5% of te installed power available 16 Ref 1a 1b 1c 1d 1e 1f 1g 1z 14 12 1 8 6 4 2 998 912 916 911 9114 91 9122 9126 913 Figure 5 Strategy 1 setback only. - 1115 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. 16 Ref 2a 2b 2c 2d 2e 14 12 1 8 6 4 2 998 912 916 911 9114 91 9122 9126 913 16 14 12 1 Figure 6 Strategy 2 preeat and setback. Ref 3a 3b 8 6 4 2 998 912 916 911 9114 91 9122 9126 913 Figure 7 Strategy 3 power limitation. Te second indicator is te predicted percentage dissatisfied (PPD) index, a comfort indicator. Te PPD indices are calculated by TRNSYS according to ISO 773, at every time step. Te PPD presented in Table 5 are for te worst instantaneous conditions observed for te occupied zones (excluding te batroom) during te preeating and critical periods, regardless of ow long tese conditions may ave lasted. Since te minimum air temperature in eac zone is set to C, air temperature never drops below tis setpoint, as sown in Figure 8. Table 5 Indicator values for te different strategies AM PM PLR PPD RR PLR PPD RR 1a.78 11.7 1.8.83 29.2 1.45 1b.77 11.7 1.3.74 29.1 1.46 1c.79 1.6 1.13.81 52.7 1.41 1d.57 11.6 1.42.57 29.3 1.91 1e.57 11.6 1.16.57 29.3 1.56 1f.67 11.7 1.8.64 29.2 1.45 1g.43 1.3 1.81.49 56.4 2.35 1z.41 11.3 1.81.61 29.8 2.5 2a.14 21.5.57.11 32.6 2.12 2b.22 21.5.58.11 32.6 2.17 2c.8 31.7.37.7 24.2 1.52 2d.22 19.9.7.23 35. 2.24 2e.23 17.7.59.2 39.4 2.24 3a.52 11.7..62 29.3. 3b.93 11.7 2.6 1. 29.3 2.3 Deg C 24 21 2 occupation period #1 occupation period #2 occupation period #3 Z1 kitcen (S) Z2 dining room (S) Z3 living room (N) Z7 bedroom (S) Z8 bedroom (N) Z9 bedroom (N) 996 91 914 9 9112 9116 912 Figure 8 Temperature profile, strategy 2b. Tis first assessment of comfort is far from perfect and seems to yield questionable results, especially regarding preeating periods were little discomfort is sown even if air temperatures reac ig levels (for winter cloting conditions). Future work is needed to better evaluate comfort levels associated wit tese strategies. Keeping in mind tat tese peak load control strategies would only be used for a few ours every winter, it was considered arbitrarily tat a PPD lower tan 5% was good (indicated by pale green sading in Table 5), between 5 and 75% acceptable and iger tan 75% not acceptable. Finally, te tird indicator, called RR for reset ratio (Eq 6), represents te maximum load for te 1 min interval immediately following te end of te peak period (subscript end) in comparison to te reference case. Tis indicator evaluates te reset strategy. tend+ 2ts Pmax, strategy t RR = Eq 6 tend+ 2ts P max, reference t It will be considered tat an RR lower tan 1. is good (indicated by pale green sading in Table 5), between 1. and 2. is acceptable wile iger tan 2. is not acceptable. One must bear in mind tat te RR ratio is dependent on te reset algoritm tat as been adopted for te particular strategy. Strategies 1 and 3 use instantaneous reset strategies and terefore sow iger RR ratio. Reset strategies could be quite easily modified to reduce te impact on te grid but migt require more complex ardware equipment for field implementation. Energy consumption for eac strategy as also been evaluated, toug it will not be discussed in full details in tis paper. It was found tat energy consumption will sligtly increase for strategies involving preeat prior to setback (1-6 %); tis was evaluated over a 48 ours period starting at on te day wen te strategy was applied (January 15t), after wic load profiles wit and witout te strategy are similar again. In terms of termal comfort, all te strategies give good indicators for te morning peak period wit a better performance for te Setback strategy and te Power limitation strategy (~11% of PPD). For te - 1116 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. afternoon peak period te strategies 1c and 1g are performing poorly compared to te oter strategies wit PPDs of 52.7% and 56.4% respectively. In terms of PLR, all te strategies are interesting for te morning peak period except for te strategy 3b (PLR=.93). In te afternoon period, te performance is also good for te majority of te strategies except for te 1a and 1c wic give acceptable values (.83 and.81) and 3b wic gives a bad performance indicator (1.). Te Preeat strategy presents a net advantage over te oter two strategies on tis aspect. Te largest peak reduction occurs wit strategy 2c and equals 7 during te morning period. In terms of RR, for te morning peak period, only te strategy 3b gives a bad indicator, te Preeat strategy and te strategy 3a give good RR indicators and te Setback strategy yields an acceptable one. For te afternoon peak period, only strategy 3a gives a good RR value, several strategies (1g, 1z, 2a, 2b, 2d, 2e and 3b) result in bad RR indicator, te rest provides acceptable values. Tis empasizes te importance of managing te increase of te setpoint after te critical periods in order to avoid creating a new peak on te grid. CONCLUSION In tis paper, te first step in te development of efficient space eating load control strategies, for electric baseboard eating systems widely present in Quebec, is presented. Some strategies evaluated in tis study are interesting in terms of bot load reduction during peak and reset periods witout affecting te occupant termal comfort. Tis study ence demonstrates te feasibility of possible load control strategies witout using any termal storage equipment for typical ligt wood-framed ousing in tis region. Tese results are valid only for te simulated building and extrapolation to te wole building stock must be done carefully. In order to do so, buildings wit different caracteristics sould be modeled including different insulation and fenestration caracteristics, infiltration rates, orientation, termal mass, location, etc. An experimental validation of te simulated strategies could be done using te twin ouses test benc already built at te LTE site. Te building simulation could also be improved by te coupling of an airflow network model in order to evaluate te effect of te air transfer between termal zones, wit and witout mecanical ventilation, on te strategies performances. NOMENCLATURE ACH : air canges per our C : termal capacity (/ C) E eating : energy consumed by te space eating equipements () E internal : internal gains () ELA : effective leakage area (cm 2 ) E solar : solar gains () c : internal convection coefficient (W/m 2 K) m infiltration : infiltration (kg) P : average power demand over te timestep () PLR : power load ratio PPD : predicted percentage dissatisfied RR : reset ratio ts : time step () t : time (s) T : air temperature ( C) UA apparent : apparent termal losses troug te envelope and including te effect of infiltration (W/ C) UA enveloppe : apparent termal losses troug te envelope (W/ C) τ : time constant () REFERENCES AEE, Fice Aide-mémoire Novoclimat, www.aee.gouv.qc.ca, 21 Beausoleil-Morrisson, I., Te adaptive simulation of convective eat transfer at internal building surfaces, Building and Environment, 37, 22, pp. 791-86 Braun, J.E., Caturvedi, N., An inverse gray-box model for transient building load prediction, HVAC&Researc, vol.8, no.1, January 22. Braun, J.E., Zong, Z., Development and evaluation of a nigt ventilation precooling algoritm, Purdue University, for te California Energy Commission, August 23. Carron, R., Atienitis, A., Verification of a low energy solar ome model to be used wit a GA optimisation tool, CETC Number 27-145, August 27. CRA 1995. Impact evaluation of te California Statewide pricing pilot, prepared for te California Energy Commission by CRA International, Marc 25, USA Gunter, S., Projet panne de courant, pase II: Comportement termique des bâtiments pendant et après une panne de courant, LTE, Hydro-Québec, 21 - confidential Henze, G.P., Kalz, D.E., Liu, S., Experimental analysis of model-based predictive optimal control for active and passive building termal storage inventory, HVAC&R, vol.11, no2, April 25 Herter, K, McAuliffe, P, Rosenfeld, A., An exploratory analysis of California residential customer response to critical peak pricing of electricity, Energy, vol. 32, 27 HQ, Rapport executif final sur le potential tecnico-économique d amélioration de l enveloppe termique des abitations du - 1117 -

12t Conference of International Building Performance Simulation Association, Sydney, 14-16 November. Québec, Hydro-Québec, December 1993 confidential. Katimpamula, S., Lu, N., Evaluation of residential HVAC control strategies for demand response programs, ASHRAE Transactions, 26 LBNL, 26, Demand sifting wit termal mass in large commercial buildings: field tests, simulations and audit, prepared for te California Energy Commission, CEC-5-26-9, January 26 Lee, K.., Braun, J.E., Development of metods for determining demand-limiting setpoint trajectories in building using sort term measurements, Building and Environment, vol.43, 28, pp. 1755-1768. Lu, N., Katimpamula, S., Control strategies of termostatically controlled appliances in a competitive market, IEEE, 25 Millette, J., Lapperrière, A., Desbiens, G., Carrette, D., Manuel tecnique : JAG2., LTE, Hydro-Québec, 24 - confidential Moreau A., Le Bel C., Assessment of te impact of internal gains on te termal loads in te residential sector, Canadian Electricity Association, December 1994 Persson, T., Nordlander, S., Rönnelid, M., Electrical savings by use of wood pellet stoves and solar eating systems in electrically eated single-family ouses, Energy and buildings, 37, 25, pp. 92-929 PGE, Direct load control pilot for electric space eat Pilot evaluation and impact measurement, 24 PNNL, 27. Pacific Nortwest GridWiseTM Testbed Demonstration Projects, Part I. Olympic Peninsula Project, prepared for te U.S. Department of Energy, USA. Reddy, T.A., Norford, L.K., Kempton, W., Saving residential air-conditioner electricity peaks by intelligent use of te building termal mass, energy, vol.16, no7, 1991, pp.11-11 Yang, L., Li, Y., Cooling load reduction by using termal mass and nigt ventilation, Energy and Buildings, 4, 28, pp. 252-258. Yin, R., Xu, P., Piette, M.A., Kiliccote, S., Study on Auto-DR and pre-cooling of commercial buildings wit termal mass in California, Energy and Buildings, vol42, no7, July 21, pp.967-975. - 11 -