1、 EditorialModeling high-performance buildingsWe are extremely pleased to present this special issue of HVAC the desiccant wheel dehumidifier and the energy recoveryventilator (ERV) with high-efficiency direct-expansion (DX) dehumidifier were the two options resultingin the smallest increase of sourc
2、e energy consumption for relative humidity set-points between 50%55%.The article by Ruan and Horton presents a model for predicting the long-term performance of groundheat exchangers. Heat transfer inside a borehole was modeled as a one-dimensional (vertical) quasi-steadystate with a time step of 2
3、h. The equivalent circuit method was used to determine a constant heat flux atthe borehole wall, and a two-dimensional (vertical and radial) transient model was developed to analyzeheat transfer in the ground outside the borehole with corrections due to interference of adjacent boreholes.These coupl
4、ed equations were solved using the bisection method. The model was then used for a casestudy: an office building located in Chicago with a ground-coupled heat pump. The results showed thatthe build-up of heat will increase ground temperature at the middle of adjacent boreholes by about 3C(5.4F) afte
5、r a 20-year operation, which will degrade the yearly performance of the ground-coupled heatpump system by about 1.4%.Platt, Ward, and Wall present the Australian view of optimal supervisory HVAC control, with ex-perimental results and analysis from demand response experiments in two large office bui
6、ldings. Simpleset-pointresetsandpre-coolingpriortoset-pointincreasehaveimplicationsintheredistributionofcoolingbetween zones and energy demandwhen temperature is set back to the original set-points. The authorsprove that participation of large multi-zone HVAC systems in such exercises is not straigh
7、tforward andthat a global approach may yield undesirable consequences in a particular zone of a building. Therefore,they implemented an adaptive, intelligent HVAC control system that plans its behavior for the day ahead,re-planning in response to changing environments or network requests. The contro
8、ller uses a cost functiontoevaluatedifferentoperatingstrategies,takingintoaccountenergyconsumption,energycost,andthermalcomfort. The measured impact on the reduction of peak loads and energy demand, as well as on humancomfort, is presented and compared with modeling results. This article shows the n
9、ecessity to developHVAC management systems that can be dynamic and intelligent, respond to changing events, and considera variety of external factors, such as occupancy, human comfort, electricity price, and weather forecast.Multi-family residential buildings represent a significant portion of the b
10、uilding stock in larger cities.Wang, Zhang, Jiang, and Liu conducted a modeling case study about the performance of three typesof HVAC systems in mid-sized apartment buildings for different climatic conditions, system operationschemes, and applicable building codes. A direct-expansion (DX) split sys
11、tem, a split air-source heatpump (ASHP) system, and a closed-loop water-source heat pump (WSHP) system with a boiler and anevaporative fluid cooler as the central heating and cooling source were considered in this analysis. TheDownloaded by T accepted February 22, 2011Jose A. Candanedo, is PhD candi
12、date. Andreas K. Athienitis, PhD, PEng, Member ASHRAE, is Professor, Scientific Director ofthe Canadian Solar Buildings Research Network.following key features for the case of a cold sunnyclimate:a114First and foremost, integrated designthe con-ception of the house sub-systems (lighting,HVAC, applia
13、nces) forms a coherent plan inwhichthereiscomplementarityoffunctions.Thisincludes the building integration of the renewableenergy systems, which play a role as componentsof the building envelope (walls or roofs).a114Passive solar techniquesthis key componentof the overall design strategy includes hi
14、gh-performance windows oriented toward the235HVAC if the consumption of the house exceeds thegenerated power (e.g., at nights or under cloudyconditions), energy is purchased from the utilitygrid.a114Motorized shading devices on windows, elec-trochromic windows or similar technologies thatenable part
15、ial control of the solar heat gains en-tering the space.a114Energy storage devices, such as thermal energystorage (TES) or batteries.a114A centralized supervisory control system that en-ables the implementation of advanced controlstrategies for energy management.Appropriate control strategies are es
16、sential forthe successful operation of high-performance build-ings (Torcellini et al. 2004). For over two decades,optimal control has been investigated as a tool forthe management of passive and active TES capac-ity in buildings (Braun 1990, 2003; Morris et al.1994; Henze et al. 1997, 2004). Optimal
17、 controlalgorithms use estimates of future loads to select asequenceofcontroloperationstooptimizeanobjec-tive function (typically, energy, peak load, or cost).These investigations have addressed mostly optimalcontrol of cooling capacity storage (ice or chilledwater) in commercial buildings (Henze 19
18、95). Pre-dictive control has also been applied to the case ofsolar buildings (Kummert et al. 1996, 2001; Chen2001).Several building simulation software tools, suchas ESP-r (ESRU 2010) and EnergyPlus (EERE2010), achieve accurate representations of build-ings through careful integration of detailed mo
19、d-els of physical phenomena into a single, compre-hensive tool. Such a model provides a reliable rep-resentation of the buildings response to externalimpulses and its HVAC system, which is particu-larly useful for research purposes. Although a full-scale building simulation tool can be used for thet
20、esting and design of advanced control strategies,this application can be quite cumbersome. The needfor complex building and HVAC models has beenidentified as a hurdle for the deployment of con-trol strategies (Wang and Ma 2008). For example,anticipatory control strategies used to select set-point tr
21、ajectories (Coffey et al. 2006) require es-timating the effect of an action (such as turning ona piece of equipment or changing the position ofa valve) based on expected loads. Predictive con-trol calculations imply performing building energysimulations at regular intervals with a moving timehorizon
22、.It has been found that simplified building mod-els can successfully be used for control applications(Athienitis et al. 1990; Kummert et al. 1996; Fraisseet al. 2002; Kampf and Robinson 2007). These sim-plified models have commonly been based on ther-mal network representations with a limited number
23、of thermal resistances and capacitances (Fraisse etal. 2002). In a recent large-scale project on predic-tive control, a simplified model of a room, latervalidated with a more detailed model, was usedin the development of control strategies (Gyalis-tras 2010). A simpler model has several advantages(s
24、uch as ease of implementation, insight into phys-ical phenomena, computational efficiency); how-ever, deciding a priori the right complexity levelfor a given application is a difficult task. The se-lection of the appropriate level of model complex-ity for an advanced control strategy is always acruc
25、ial decision (Kummert et al. 2006; Gyalistras2010). The difficulty lies on deciding which de-tails can be neglected without jeopardizing the va-lidity of the conclusions. Simplified physical mod-els, or gray-box models, derived from either moredetailed building simulation models or from ac-tual meas
26、urements, have been proposed as a wayto facilitate the implementation of control strate-gies (Wang and Ma 2008). This approach has theadvantage of providing a link between design andcontrol. Control strategies can (and should) becreated during the design of the building by ap-plying the same models
27、with adjusted levels ofresolution.Thisarticleinvestigatestheapplicationofpredic-tive control strategies in a solar house. The controlstrategies are applied at two different control levels:(a) for the control of a high-mass radiant floor heat-ing (RFH) system in a room exposed to solar gainsand (b) f
28、or controlling the charging of a TES tankwith a solar-source heat pump.Downloaded by Tthisindicatesthat the simplified model can predict the momentwhen maximum or minimum values are reached.Predictive control of RFH systemModel Predictive Control (MPC)The basic principle of MPC algorithms consistsof
29、 finding a sequence of values for the manipulatedvariable(s) that will minimize the discrepancy be-tween the reference and the output. The followingequation presents a typical objective function to beminimized with an MPC algorithm:QSk=CHsummationdisplayi=lwfrk+i yk+i2, (4)whereQSkisthesumofthesquar
30、esofthedifferencebetween the reference values (rk+i) and the output(yk+i) measured at a time step k for a horizon goingfrom i to CH(control horizon). The variable wfisthe weighting factor. Although MPC algorithms aremore computationally intensive than conventionalcontrol, the availability of online
31、weather forecastsand micro-controllers facilitates the implementa-tion of MPC in solar houses (Chen 2002). Figure6 shows the implementation in MATLAB/Simulinkofthesimplifiedmodel.ThisimplementationallowsFigure 5. Comparison of room air temperature as calculated with EnergyPlus and with the simplifie
32、d transfer function model underfree-floating conditions (Montreal climate) (color figure available online).Downloaded by T Henzeand Liu 2005), may also be applied to the spe-cific problem of advanced solar homes. An objec-tive function J (e.g., energy use, peak load, or cost)is optimized. For the ca
33、se of a BIPV/T-assisted heatpump,ifthetotalenergyconsumedbytheheatpumpis selected as the objective function over a period ofinterest Lambda1, J is given byJ =integraldisplayLambda10PHP(t)dt (5)in which PHP(t) is the power consumption rate ofthe heat pump. The problem consists of selectingthe tank se
34、t-points that would supply enough en-ergy for the heating needs while minimizing the en-ergy consumed by the mechanical system. In otherwords, the optimal control problem would then con-sist of finding the set-point sequence x1,x2.xnthat minimizes J over the control horizon Lambda1.Toaccomplish this
35、, it is necessary to predict (a) howmuch heat will be required by the heating systemand (b) how much heat can be collected from theroof. Forecasts of solar radiation, temperature, andwind speed, the main factors affecting the output ofthe BIPV/T system, enable the estimation of the en-ergy available
36、 from the BIPV/T. If the heat deliveryrate of the RFH has been calculated (for example,with an MPC algorithm), this information can beused to plan a sequence of set-points for the predic-tion horizon. A heuristic approach for this type ofsystem was presented in Candanedo and Athienitis(2010).System
37、modelingA customary method to model the performanceof a BIPV/T roof consists of dividing it in sev-eral sections (usually not more than five or six) inthe streamwise direction and then performing anenergy balance for each control volume (Bazilianet al. 2001; Candanedo et al. 2010). The exit tem-pera
38、ture of each control volume is the inlet temper-ature of the next control volume. Figure 10 showsthe modeling approach used for the BIPV/T roof.Each control volume spans the entire width of theroof (wPV). The length of each control volume in thestreamwise direcation is LCV. Energy balance equa-tions
39、 are written for three nodes per control volume:the PV panel, the air in the channel, and the bottomof the channel. The three equations are given by(TPV Text)ho+ (TPV Tbot)hr+ (TPV Tma)hct+ Pelec= G, (6)(Tma TPV)hct+ (Tma Tbot)hcb+ qrem= 0,(7)Downloaded by T however, its influenceon results is small
40、 due to the amount of insu-lation used under the BIPV/T roof. The volumeflowrateundertheroofisassumedtobeconstantand equal to 1320 m3/h (780 CFM). The massflow rate is then calculated by multiplying by thedensity and then used to calculate the averageBIPV/T air temperature:TairavgAB=1Delta1tintegral
41、displaytBtATairparenleftbigG,Text,Tattic,wspeed, mairparenrightbigdt. (31)6. TheHCavailablefromtheheatpumpduringthatperiod is estimated by introducing the averageBIPV/Ttemperature(Equation27)intoEquation14 and then multiplying by Delta1t.HHPAB= Delta1t HC(Tairavg,AB) (32)7. The average of both set-p
42、oints (i.e., tank tem-perature) is used as the return temperature inEquation 16 to calculate the average supply tem-perature. Finally, the electric energy consumedby the heat pump (EEAB) is calculated by intro-ducing the supply temperature and the averageBIPV/T temperature into Equation 15:Tret,AB=T
43、spAB+ TspAB2, (33)Tsup,AB= Tret,AB+HCparenleftbigTairavg,ABparenrightbigmwcpw, (34)EEAB= Delta1t PHPparenleftbigTsup,AB, Tairavg,ABparenrightbig. (35)If the temperature of the BIPV/T air is below15C(5F), it is assumed that it falls outside theoperating range of the heat pump. The power con-sumed PHP
44、is then assigned a very large number (1MW), which effectively discards this solution whenexecuting the algorithm.8. If the heat available from the BIPV/T-assistedheat pump is greater than or equal to the energyrequired by the tank, then the heat pump can beused to charge the tank. The cost for the p
45、eriodof interest is calculated by proportionally adjust-ing the electric energy estimate according to theheating requirements:if HHP,AB Ereq,ABthenCAB= EEABparenleftbiggEreq,ABHHP,ABparenrightbigg.(36)9. Finally, if the heat available from the BIPV/T-assisted heat pump is smaller than the energyrequ
46、ired by the tank, the heat pump cannot beused to charge the tank. The assigned cost isthen infinite, since the set-point transition is im-possible:if HHP,AB Ereq,AB,CAB=.(37)Any trajectory having such a set-point transitionwould then be discarded.A summary of this algorithm is presented in Fig-ure 1
47、2.The dynamic programming algorithm presentedabove,includingthecalculationofthecostfunction,was applied to the calculation of optimal set-pointpaths for the TES tank. The heat delivery rates cal-culated by the MPC algorithm were introduced inthe algorithm, along with the weather data file forMontrea
48、l.Results with dynamic programmingalgorithmFigure 13 shows sample results obtained withthe dynamic programming algorithm for the TEStank, corresponding to the period between January24 (0:00) and January 26 (0:00). Results from theMPC controller (RFH rate) were introduced in thealgorithm, along with
49、the weather data (solar radi-ation on the roof, temperature, and wind speed) forthat period. The initial state is assumed to be 37.5C(99.5F),andthefinalstateis30C(86F).Forcom-parison, an arbitrary sequence is presented. For theoptimal path, the heat pump consumes 11.26 kWhDownloaded by T the arbitrary path consumes 15.59kWh.Figure 14 shows the optimal path calculated bythe algorithm for January 24 at 12:00, i.e., starting12 h later after the optimal path shown in Fi
