A Passive Design Strategy for a Horizontal Ground Source Heat Pump Pipe Operation Optimization with a Non-homogeneous Soil Profile

Ao

AGa

b

c

a

ARRA

KHTNOGEC

1

oiaswcacsr

U

(mg

0h

Energy and Buildings 61 (2013) 39–50

Contents lists available at SciVerse ScienceDirect

Energy and Buildings

j ourna l ho me p age: www.elsev ier .com/ locate /enbui ld

passive design strategy for a horizontal ground source heat pump pipeperation optimization with a non-homogeneous soil profile

mir Rezaei-Bazkiaeia,∗, Ehsan Dehghan-Niri c, Ebrahim M. Kolahdouzb, A. Scott Webera,ary F. Dargushb

Department of Civil, Structural & Environmental Engineering, University at Buffalo, Buffalo, NY 14260, United StatesDepartment of Mechanical & Aerospace Engineering, University at Buffalo, Buffalo, NY 14260, United StatesSmart Structures Research Lab, Department of Civil, Structural & Environmental Engineering, University at Buffalo, NY 14260, United States

r t i c l e i n f o

rticle history:eceived 30 August 2012eceived in revised form 4 January 2013ccepted 29 January 2013

eywords:orizontal ground source heat pumpire Derived Aggregateon-homogeneous soil

a b s t r a c t

The effectiveness of a non-homogeneous soil profile for horizontal ground source heat pumps (GSHPs),defined as natural backfill with an intermediate layer of material having different thermal character-istics, is investigated. Steps toward development of a comprehensive model to consider the effects ofthe non-homogeneous layer are described. The developed model is utilized successfully in conjunctionwith a genetic algorithm (GA) search method to obtain the optimized operational parameters for a GSHPin three different climate conditions. A properly sized and engineered non-homogeneous soil profiledemonstrated the potential to increase the energy extraction/dissipation rates from/to the ground to asignificant level. The potential benefit of a recycled product, Tire Derived Aggregate (TDA), as an insu-

ptimizationenetic algorithmnergy efficiencyontrol

lating non-homogeneous layer is assessed. TDA is demonstrated to be more effective in the cold climate(Buffalo) by increasing the energy extraction rates from the ground approximately 15% annually. TDA’seffectiveness is less pronounced in a relatively moderate climate (Dallas) by increasing the energy extrac-tion rates from the ground about 4% annually. For the cooling only scenario (Miami), a high conductiveintermediate layer of saturated sand exhibited greater potential to increase the energy dissipation to theground.

Published by Elsevier B.V.

. Introduction

Building energy consumption comprises a considerable portionf every nation’s energy budget. Within the energy consumingtems in buildings, heating, ventilation and air conditioning (HVAC)ccounts for a majority of energy demand. The reputation of groundource heat pumps (GSHPs) in harnessing the Earth’s clean energyith high performance coefficients combined with the technologi-

al advances in the HVAC industry has introduced these systemss one of the promising technologies to reduce building energy

∗ Corresponding author at: 204 Jarvis Hall, University at Buffalo, Buffalo, NY 14260,nited States. Tel.: +1 716 380 5255; fax: +1 716 645 2549.

E-mail addresses: [email protected], amir rezaee [email protected]. Rezaei-Bazkiaei), [email protected] (E. Dehghan-Niri),

[email protected] (E.M. Kolahdouz), [email protected] (A.S. Weber),[email protected] (G.F. Dargush).

378-7788/$ – see front matter. Published by Elsevier B.V.ttp://dx.doi.org/10.1016/j.enbuild.2013.01.040

network and the circulating pump. The heat pump unit itself con-sists of sub-units which handle the thermodynamic relationshipsbetween the working fluid and the load side (indoor air or DomesticHot Water (DHW) network); compressor, condenser, evaporator,the expansion device and blower fans or circulating DHW pumps.The working fluid flows through the ground pipes and exchangesheat with the surrounding soil medium where it gains/loses heat inheating/cooling modes, respectively. The fluid at the outlet of theground pipe enters the heat pump where the thermodynamic cycleof the heat exchange between the working fluid and the refrigerantis responsible for heat delivery or dissipation in heating and coolingmodes, respectively. A successful design of GSHP involves carefulselection and sizing of both the ground side and the indoor unit sothat heating/cooling loads are met year-round with the least wasteheat/electricity inventory. This selection process usually involvesemployment of simplifying engineering assumptions. For example,one of the most well-known ground pipe sizing semi-empirical for-

4 rgy an

Tpf

vwEpsclcuastvitmr

ptnaptotbocyucatiaopw

bcettsccawtwg[ti

dwiEt

0 A. Rezaei-Bazkiaei et al. / Ene

his assumption, while not completely representative of groundipe operation, is assumed to be within reasonable range of erroror engineering practice.

A heating/cooling load budget analysis for a prospect buildingia accepted methods, such as the bin method or, more popular,ell accepted dynamic thermal load simulation software such as

nergy+and eQuest, is the GSHP design start point. The calculatedeak load data is subsequently used to size and design the GSHPystem. Even with the dynamic software tools, the design for theapacity of the heat pump equipment is usually based on the peakoad (worst case scenario) conditions that is oversized in manyases. Moreover, most ground source heat pump design methodssually consider a constant source/sink strength (ground temper-ture) based on the maximum and minimum observed data oremi-experimental formulas [1]. These assumptions do not capturehe dynamic variation of the boundary conditions and the time-ariable nature of the system functionality. The historical approachn design methods yield operational values for heating and ventila-ion systems that stray from the least waste energy path. This has

ade the dynamic, real-time analysis of GSHPs the focus of severalecent research studies [2–4].

An essential part of the GSHP design, after selection of theipe length and size, is the proper selection of the working fluidemperature range as it is pumped back into the ground. The Inter-ational Ground Source Heat Pump Association (IGSHPA) suggestsn approach that relates the inlet water temperature to the heatump to maximum, minimum and average ambient air tempera-ures. It is the responsibility of the designer to select the appropriateptimal entering water temperature to the heat pump that guaran-ees efficient performance. Given the complexities of an accurateuilding load evaluation including users consumption habits, levelf building insulation, changes in occupancy, and sudden climatichanges that are far different from the long term statistical designear data, etc., and its causal relationship with the GSHP systemnits, the optimization of the overall performance of a GSHP is ahallenge. Compounding the problem is the fact that ground char-cteristics vary in different regions to make it more challengingo track the dynamic building load demand and keep the heat-ng/conditioning process as optimized as possible. Because thebsence of design experience can cause a wide deviation from theptimized cost and operation path, the optimization algorithm pro-osed in this study aims at providing the designers of the GSHPsith beneficial information regarding the GSHP design.

Strategies to control the capacity of GSHP systems to match theuilding energy demand emerged as a response to the large energyonsumption rates in buildings [5]. The capacity control practicesxplore options to satisfy building load requirements and minimizehe waste energy in the heating/conditioning process. Capacity con-rol usually involves either the control on the components of a GSHPystem (e.g. the compressor, condenser, ground pipes, etc.) or ahange in design configurations for different seasons or advancedontrol algorithms [6]. The correlations between different units of

GSHP are quite complex and changes in characteristics of one unitith time affect the overall performance of the system. Introduc-

ion of the building properties and its dynamic energy requirementsill add to the complexity of the near-optimal energy delivery

oal [7–9], especially when integrated building energy solutions10,11] or non-conventional HVAC designs [12,13] are involved, orhe system performance goals require a close examination of thenteractions between the building and the HVAC systems [14].

Researchers have attempted to model these complexities byynamically tracking the interactions between these units via

n combination with thermodynamic data base models such asES [3] or by self developed programs [17,18]. The general struc-ure of such models consists of first-tier models (e.g. ground pipe

d Buildings 61 (2013) 39–50

and heat pump), which each have sub-models of second-tier (e.g.compressor, evaporator, etc.) with higher details. The functionalrelationship between the sub-models is constructed via operationalparameters, such as water/brine and refrigerant flow rates [19,20].The level of complexity of these sub-models significantly dependson the purpose of the modeling effort [6].

Although numerous studies have aimed at controlling the heatpump side of a GSHP system, a much smaller number have focusedon design methods for the ground side characteristics. Groundrelated work is usually associated with additional capital invest-ment that has retarded exploration of design options in the groundcompared to the heat pump unit inside the building. Previousresearch results suggest that utilization of a non-homogeneoussystem above the pipe burial depth might be worth more atten-tion [21]. Based on the results from a comprehensive horizontalground pipe model calibrated to a set of field data from amild climate, the non-homogeneous soil profile demonstrated thepotential to increase the energy extraction rates from the groundby approximately 17%, in a peak heating month, compared to thehomogeneous soil profile [21].

This research is focused on the analysis of the GSHP systemperformance optimization via control on the source/sink (ground)side. The following simple semi-empirical equations for the enter-ing water temperature to the pump in heating and cooling seasonswere adopted from the International Ground Dource Heat PumpAssociation (IGSHPA) design manual [1] to make the link betweenthe ground pipe and the heat pump side of the GSHP system:

Tf,ih= Tf,imin

+Tf,imean − Tf,imin

Tairmean − Tairmin(Tairmax − Tair) (1)

Tf,ic = Tf,imean + Tf,imax − Tf,imean

Tairmax − Tairmean(Tair − Tairmean) (2)

where Tf,ihis the entering water temperature to the ground in heat-

ing mode, Tf,ic is the entering water temperature to the ground incooling mode, Tf,imin

is the minimum design entering water tem-perature in heating, Tf,imean is the average design entering watertemperature in heating or cooling, Tf,imax is the maximum designentering water temperature in cooling, Tairmean is the averageannual ambient air temperature, Tairmin is the minimum annualambient air temperature, Tairmax is the maximum annual ambi-ent air temperature, and Tair is the air temperature at the time ofsimulation.

The model employed for characteristic analysis of ground pipescomprises a comprehensive surface energy balance model, whichis capable of solving for the temperature distribution of the entiresoil profile to obtain the outlet water temperature from the groundpipes. A detailed description of the surface energy balance equa-tions, parameters and the solution methods has been presented inmaterial and methods section.

The developed model was utilized to investigate the potentialbenefits of a non-homogeneous soil profile on the ground sideperformance. Tire Derived Aggregate (TDA) has been proposed asan intermediate layer in the non-homogeneous system [21]. TDAmainly consists of chopped pieces of used tires in a variety of nom-inal sizes ranging from 1 to above 5 in. [22]. The idea to employTDA, also referred to as tire chips, tire shreds, and tire mulch, incivil engineering applications was first initiated by Humphrey [23].New applications for TDA, mostly in civil engineering, have beenproposed based on its unique physical characteristics. TDA’s rel-atively low density compared to conventional backfill makes it aviable alternative fill material, where a lighter fill material is desired

A. Rezaei-Bazkiaei et al. / Energy an

Table 1TDA thermal and physical properties in literature.

Material [reference] Thermalconductivity(W m−1 ◦C−1)

Specificheat(J kg−1 ◦C−1)

Density(kg m−3)

TDA (2-in. nominal) [22] 0.149–0.164 NA 513TDA (1-in. nominal) [25] NA NA 1060–1100TDA (2–12-in.) [26] 0.29 1.15 720

N

spA5tim

trltc

2

2

osg

Q

wsilr(

w

Q

wdTntKa

g

Q

Q

D

D

TDA (2-in. nominal) [27] 0.564 507 641

A: Not available.

tudies in the literature have reported on the thermal and physicalroperties of TDA [22,26,27] and these are presented in Table 1.

thermal conductivity value of 0.29 W m−1 ◦C−1, specific heat of00 J kg−1 ◦C−1 and density of 720 kg m−3 were chosen as represen-ative properties of TDA material for the purpose of the analysesn this paper. In this work, TDA was compared to other potential

aterials for their applicability as the non-homogeneous layer.The overall goal of this research was to determine a GSHP design

hat optimizes energy savings by utilizing an evolutionary algo-ithm to evaluate different materials for use as the intermediateayer. The optimization was carried out for the same pipe charac-eristics, but different climatic conditions to evaluate the effects oflimatic conditions on the optimized parameters.

. Materials and methods

.1. Surface energies

The surface boundary condition takes into account the effectsf energy balance due to a variety of mechanisms responsible forurface–ambient heat interaction. The total energy balance on theround surface (Qt, W) can be written in rate form as [28,29]:

t = Qc + Qe + Qh + Qle + Qli + Qsi + Qp (3)

here Qc is the conduction heat flux through snow layer or groundurface (W), Qe is the turbulent exchange of latent heat (W), Qhs the turbulent exchange of sensible heat (W), Qle is the emittedongwave radiation heat flux (W), Qli is the incoming long waveadiation (W), Qsi is the solar radiation reaching the surface of earthW), and Qp is the heat flux due to precipitation (W).

Heat conduction through the snow and ground layers can beritten as [29]:

c = −(Ts0 − Tb)

(zs

Ks+ zg

Kg

)−1

(4)

here Ts0 is the snow surface or ground surface temperatureepending on whether the surface is covered with snow or not,b is the ground temperature at the bottom of topsoil layer (thick-ess of first discretization element in y direction, �y), zs and zg arehe thicknesses of snow and the top layer of soil (�y), respectively.s and Kg (W m−1 K−1) are the thermal conductivities of the snownd soil layers respectively.

Turbulent exchange of sensible and latent heat, Qh and Qe areiven as [28,29]:

h = �aCp,aDh�(Ta − Ts0) (5)

e = �aLsDe�(

0.622ea − es0

Pa

)(6)

The exchange coefficients for sensible and latent heat, Dh and

e = Dh = �2Us

(ln z/z0)2(7)

d Buildings 61 (2013) 39–50 41

� = 11 + 10Ri

(8)

where �a is the air density equal to 1.275 kg m−3, Cp,a is the specificheat of air assumed to be 1000 J kg−1 K−1, Ta is the air temperature(K), es0 is the surface vapor pressure (Pa), ea is the atmosphericvapor pressure (Pa), Pa is the atmospheric pressure (Pa), Ls is thelatent heat of sublimation of snow (J kg−1), � is the Von Karman’sconstant, Us is the wind speed (m s−1) at reference height z (m), andzo is the roughness length set to 0.5 m.

The Richardson number, Ri, is defined as:

Ri = gz(Ta − Ts0)

TaU2z

(9)

where g is the gravitational acceleration (m s−2), and Uz is the windspeed at the elevation z relative to reference height z0. The emittedlongwave radiation, Qle, was given by:

Qle = �s�(Ts0)4 (10)

where �s is the soil surface emissivity assumed to be equal to0.98, and � is the Stefan Boltzman constant equal to 5.6704 × 10−8

(W m−2 K−4).The incoming longwave radiation was given by the empirical

description [30].

Qli = 1.08(1 − e−(0.01ea)Ta/2016)�(Ta)4 (11)

log10 ea = 11.40 − 2353Tdp

(12)

where Tdp is the daily dew-point temperature (K). The incident solarradiation reaching earth’s surface can be described as follows:

Qsi = (1 − Albedo)[Sm + SaRe(exp(iwt + �1))] (13)

where Sm is the mean annual solar radiation (W m−2), Sa is theamplitude of surface solar radiation (W m−2), w is the angularvelocity (rad), and �1 is the phase angle (rad). Heat flux due toprecipitation can be written as [31]:

Qp = IRCp,w(Ta − Ts0) (14)

where IR is the rain intensity (kg m−2 s), and Cp,w is the specific heatof water assumed to be equal to 4186 J kg−1 K−1.

As some of the energy equations are a function of the surfacetemperature of the solution domain that need to be numericallysolved, utilization of an iterative method was necessary. The surfaceenergy equations, which are obtained from the daily meteorologicaldata, were solved for the unknown surface temperature values, Ts0,using the Newton–Raphson method. This iterative scheme solvesfor temperature values on the surface using Eq. (15):

Tn+1s0 = Tn

s0 − f (Ts0)f ′(Ts0)

(15)

where f(Ts0) takes the following form:

f (Ts0) = Qc(Ts0) + Qe(Ts0) + Qh(Ts0) + Qle(Ts0) + Qli(Ts0) + Qsi(Ts0)

+ Qp(Ts0) = 0 (16)

The iteration continues until the temperature differencebetween two consecutive iterations becomes less than 0.1 ◦C.

2.2. Ground pipe modeling

The ground pipe configuration employed in this research con-sists of a horizontal pipe with a total length of 61 m (L) and 1.9 cmdiameter, 3 m distance between inlet and outlet, buried at the depthof 2 m, with the working fluid flow rate of 0.19 kg s−1, as depicted in

42 A. Rezaei-Bazkiaei et al. / Energy and Buildings 61 (2013) 39–50

FrhumFmaTFw[taatbw

tc

Fig. 1. Configuration of the ground pipe and soil layer (not scaled).

ig. 1. These design characteristics belong to an experimental testoom in Buffalo, NY, with an approximate heating load of 1.5 kW,eated and conditioned with the specified horizontal GSHP, that isnder investigation by these authors. Assuming there is no ther-al interaction between pipes, the solution domain (detailed in

ig. 2) was considered to be from the center line of the pipe to theid-span of the distance between pipes (1.5 m), in the x-direction,

nd from ground surface to the farfield (5 m) in the y-direction.he ground pipe is buried at the mesh level jpipe, as indicated inig. 2. Soil thermal conductivity (Ks) and thermal diffusivity (˛s)ere selected based on the soil and rock classification guideline

32] to be equal to 1.67 W m−1 K−1 and 66 × 10−8 m2 s−1, respec-ively. For the sake of comparison, this ground pipe physical settingnd the original soil properties were assumed to be identical forll the regions simulated in this paper. A comprehensive study ofhe impact of geographical variation of the soil properties has noteen the focus of this study and is the subject of authors’ future

To account for the three-dimensional behavior of the pipe andhe surrounding soil, the effect of the working fluid flow rate wasonsidered along the pipe direction. The third dimension of the

Fig. 2. Schematic of the solution domain mesh showing the re

Fig. 3. Schematic of slices of 3D domain in pipe direction (not scaled).

problem was modeled by splitting the physical domain in the pipedirection into a series of cross-sections (slices) of the soil profilefor each time step, including the nodal temperature of the fluid atthe pipe’s location. Fig. 3 shows how the slices are spaced in thepipe direction to cover the temperature distribution of the entire3D domain. At each time step, the nodal temperatures of each crosssection were obtained and subsequently updated for the next slicealong the pipe’s length to achieve a temperature distribution ofsoil and fluid at the end of the pipe (L). The same process wasrepeated for the next time steps until the end of the simulationtime.

3. Numerical algorithm

The three-dimensional temperature distribution in the soil

lative size of the mesh grid and pipe diameter (scaled).

rgy and Buildings 61 (2013) 39–50 43

ai

whp(hpect

T

w(t(h

d

wo

T

wat(t

w

w

ı

dev

d

A. Rezaei-Bazkiaei et al. / Ene

two-dimensional geometry by inclusion of the fluid temperaturen the domain [33]. The governing thermal diffusion equation:

1˛

∂T(x, y, t)∂t

= ∂2T(x, y, t)∂x2

+ ∂2T(x, y, t)∂y2

(17)

as solved for the entire domain in two-dimensions for the pureeat conduction first, and the solution was then updated along theipe’s length to obtain the solution in three dimensions. In Eq.17), is the thermal diffusivity of the medium through whicheat travels. The output water temperature equation along theipe direction, calculated based on the analytical solution for thenergy balance between surrounding soil medium and pipe [34],onstructs the link between the fluid and the soil temperature inhe model. Thus,

f,out = Ts − (Ts − Tf,i) · exp

(− KsL

mCp,f

)(18)

here Tf,out is the fluid temperature exiting a pipe of length Lm), Ts is the surrounding soil temperature, Tf,i is the initial wateremperature entering the pipe, Ks is the soil thermal conductivityW m−1 ◦C−1), m is the mass flow rate (kg s−1) and Cp,f is the specificeat of the working fluid (J kg−1 ◦C−1).

Boundary and initial conditions for the solution domain areescribed as follows:

Ti = T(x, y), t = 0

∂T

∂x= 0, x = xmax

∂T

∂x= 0, x = 0

Qt (W/m2), y = 0

T(y, t) = T(y, t)Kusuda, y = ymax

hile the initial temperature distribution of the soil profile (Ti) wasbtained from the Kusuda model [35]:

(y, t)Kusuda = Tavg + Tampe−y√

/˛sP cos

(2

t

P− y

√

˛sP

)(19)

here Tavg is the average annual surface temperature, Tamp is themplitude of fluctuation of the annual surface temperature, y ishe depth from ground surface (m), ˛s is soil thermal diffusivitym2 s−1), P is the duration of a year in seconds, and t is the time ofhe year in seconds.

The second partial derivative of temperature in x-direction wasritten with a central difference scheme:

∂2T

∂x2|n(i,j) =

Tni−1,j

− 2Tni,j

+ Tni+1,j

(�x)2+ O(�x)2 ≈

Tni−1,j

− 2Tni,j

+ Tni+1,j

(�x)2

= ı2x Tn

(�x)2(20)

here

2x T = Ti−1,j − 2Ti,j + Ti+1,j (21)

Using the same discretization for the y-direction and a forwardifference to represent the time derivative, the governing heatquation therefore can be expressed in terms of nodal temperaturealues in explicit form as follows:

Tn+1 − Tn[

ı2Tn ı2Tn]

�t= ˛ x

(�x)2+ y

(�y)2(22)

Next, a scheme was selected to sequence through time. Asepicted in Fig. 4, the solution domain was discretized in nx + 1

Fig. 4. Two dimensional grid of the solution domain [21].

nodes in x and ny + 1 nodes in the y-direction, of which inner domainwas used to solve for the temperature distribution of each crosssection of the physical domain. The nodes on the boundary wereseparated to force boundary conditions in x and y direction.

A fully explicit finite difference solution scheme was employedto solve for temperature distribution of the solution domain. Themodel was constructed in MATLAB. Time steps of 1800 s and spacediscretization of �x = �y = 0.1 m were chosen, based on a stabilityanalysis of the model undertaken in a previous work [21].

Fully explicit finite difference formulation of the heat conduc-tion equation takes the form:

Tn+1i,j

− Tni,j = r(Tn

i−1,j − 2Tni,j + Tn

i+1,j) + r(Tni,j−1 − 2Tn

i,j + Tni,j+1) (23)

where �x=�y

rx = �t

(�x)2, ry = �t

(�y)2, rx = ry = r (24)

After rearrangement, the explicit equation will take the form:

Tn+1i,j

= rTni−1,j + rTn

i+1,j + rTni,j−1 + rTn

i,j+1 + (1 − 4r)Tni,j (25)

Once the homogeneous case formulation was performed, thenext step was adjustments to the difference equations to take intoaccount the effects of the internal boundary conditions on the topand bottom of the intermediate layer. To meet the temperature andflux conditions on the internal boundaries, the energy balance onthe interfacial nodes of intermediate layer and soil was obtained bywriting the nodal fluxes and the energy storage term for the controlvolume around each node. Fig. 5 depicts the energy balance for anode on the top and bottom of the intermediate layer interface, thecontrol volume around the node, and the heat fluxes involved. Theresulting energy balance is written:

q1 + q3 − q2 − q4 = (�Cp)avg · �x · �y∂Ti,j

∂t, (26)

where

(�Cp)avg = (�Cp)s + (�Cp)int.

2, Kavg = Ks + Kint.

2

44 A. Rezaei-Bazkiaei et al. / Energy and Buildings 61 (2013) 39–50

d bot

a

q

q

q

q

bd

r

wttdda

mt

variables continues for a number of times (generations) to achievethe set of variables that yield the optimal results. The specificationsof the GA implementation for this work are listed in Table 2. More

Table 2Genetic algorithm terminology and definitions.

Population size Number of individuals that are evaluated in eachgeneration. At each iteration, the genetic algorithm runsthe core finite difference ground pipe model with theselected variables from the current population to producea new population of possible input variables (500, per eachgeneration in this paper)

Generation Each successive population of possible set of inputvariables is called a new generation (50, in this paper)

Crossover GA’s operation used to vary the programming of achromosome or chromosomes from one generation ofinput variables to the next

Mutation GA’s operation used to maintain genetic diversity from onegeneration of a population of input variables to the next.This operator is needed to avoid the search algorithm gettrapped in a local optimum design

Elitism GA’s mechanism that ensures that the highly fittedindividuals of the population of input variables are passedon to the next generation without being altered by GA’soperators. Using elitism ensures that the best inputvariables of the population can never be altered from onegeneration to the next. This operator increases the rate ofconvergence to the optimal point

Fig. 5. The intermediate layer top an

nd the resulting values of heat fluxes are described as follows:

1 = Kavg ·Tn

i−1,j− Tn

i,j

�x· �y (27)

2 = Kavg ·Tn

i+1,j− Tn

i,j

�x· �y (28)

3 = Ks ·Tn

i,j−1 − Tni,j

�y· �x (29)

4 = Kint. ·Tn

i,j+1 − Tni,j

�y· �x (30)

Substituting the flux equations (27)–(30) into the energyalance equation and rearranging the equations results in the intro-uction of the following new parameters in the difference equation:

AVG = Kavg

(�Cp)avg· �t

(�x)2, rT = Kint.

(�Cp)avg· �t

(�x)2,

rS = Ks

(�Cp)avg· �t

(�x)2(31)

here Kavg is the average of the soil and the intermediate layerhermal conductivity (W m−1 ◦C−1), Kint. is the intermediate layerhermal conductivity (W m−1 ◦C−1), and Ks is the soil thermal con-uctivity (W m−1 ◦C−1). The rAVG, rT and rS are the parametersefined to express the finite difference formula for the energy bal-nce on top and bottom of the intermediate layer.

The difference formulas obtained previously for the explicitethod were revised to account for the heat flux through the TDA

op and bottom layers resulting in the relations:

TDA top layer:

Tn+1i,j

= rAVGTni−1,j + rAVGTn

i+1,j + rSTni,j−1 + rT Tn

i,j+1

+ (1 − 2rAVG − rS − rT )Tni,j (32)

TDA bottom layer:

Tn+1i,j

= rAVGTni−1,j + rAVGTn

i+1,j + rT Tni,j−1 + rSTn

i,j+1

+ (1 − 2rAVG − rS − rT )Tni,j (33)

tom interfaces energy balance [21].

4. Optimization algorithm

A genetic algorithm (GA) optimization scheme was used toobtain the operational parameters of a horizontal GSHP that maxi-mizes energy extraction/dissipation rates from/to the ground. Themain advantage of GA over traditional optimization algorithms forthis work is that it does not require and does not depend on gradientinformation of the objective function; instead it uses a popula-tion of design points and randomly utilizes information from eachgeneration to the subsequent one to search for the parametersthat optimize the objective function. The optimization variablesdefine the characteristics (genes) of each design point in the ini-tial pool (population) of the design points. The information abouteach design point is compared to other ones in the pool to trans-fer information to the next set of design points (next generation).The reproduction of new set of design alternatives with updated

Migration The migration algorithm partitions a population of selectedvariables by the algorithm into a set of sub-populationsand shares information between these sub-populations

A. Rezaei-Bazkiaei et al. / Energy and Buildings 61 (2013) 39–50 45

Table 3Working fluid properties selected from the IGSHPA guideline [1].

Index number Fluid Thermal conductivity(W m−1 ◦C−1)

Specific heat(J kg−1 ◦C−1)

Density(kg m−3)

Dynamic viscosity(kg m−1 s−1) × 10−3

1 Water 0.6 4183 998.3 12 6% propylene-glycol and water 0.476 4140 1010 1.53 13% propylene-glycol and water 0.432 4100 1010 1.9

ib

pdooilccphmwol[

smttotteatmysa

tcmtatcshiG

TI

4 18% propylene-glycol and water 0.408

5 24% propylene-glycol and water 0.389

nformation about GA optimization and illustrative examples cane found in [36,37].

Designers often use past experience to select the “best” heatump design that satisfies the energy needs of the building duringifferent seasons. Given that the common design practice is basedn peak load design, GSHP designs are inclined to be far from theptimized scenario to meet the real-time heating loads in a build-ng. Several researchers in the building energy and heat transferiterature have come to the conclusion that optimization schemesan benefit the design process where the effects of the climaticonditions are meant to be closely considered in an efficient designrocess [38–40]. In their comprehensive work on vertical [38] andorizontal [39] ground source heat pumps, Sanaye and Niroomandodeled the thermodynamic cycle of the heat pump in conjunctionith the thermal resistance pipe model to obtain the optimized

perational parameters. Sayyaadi and Amlashi performed a simi-ar study via exergy analysis of a vertical ground source heat pump40].

In the present work, a genetic algorithm (GA) optimizationcheme was designed to obtain the optimum values of the inter-ediate layer configuration (thickness and position), working fluid

ype, and inlet fluid temperature ranges. The results of the simula-ion provide insights on the benefits returned from the introductionf an intermediate layer on the performance of a GSHP throughouthe year. The model outputs help elucidate the optimal proper-ies and configuration of the intermediate layer and the optimalntering water temperatures to the ground throughout the year tochieve the maximum energy extraction/dissipation rates from/tohe ground. The ultimate motivation behind the analysis is to deter-

ine whether the operating parameters for different months of theear are in a range that confirms the benefits of a capacity controltrategy or a set of selected operational parameters can be used forll months without deviating from the optimized monthly values.

Seven input variables (decision variables) were chosen to feedhe core finite difference model in each run of the GA. These inputsomprise working fluid properties, minimum, mean and maxi-um entering water temperature values, the intermediate layer

hickness, position and thermal properties. To let the evolution-ry algorithm search in a broader spectrum of potential designs,he algorithm was allowed to choose between a range of practi-al working fluid properties (Table 3) and also a range of common

ng fluid properties and common range of soil properties for theSHP application were based on the information published in the

able 4ntermediate layer thermal properties selected from the manual for the soil and rock clas

Index number Intermediate material

1 TDA

2 Sand

3 Clay

4 Loam

5 Saturated silt or clay

6 Saturated sand

4060 1020 34020 1020 6.3

guidelines [1,32], respectively. TDA thermal properties were listedas one of the intermediate layer choices that GA can choose from ineach run. The genetic search algorithm selects from the list of pro-vided input parameters, based on the GA options defined in Table 2,then uses these input variables to run the core finite differencemodel that results in calculation of the outlet fluid temperaturesfrom the ground pipe. The outlet fluid temperature values weresubsequently used to calculate the energy extraction/dissipationrates based on the relation:

Eground = m · Cp,f · (Tavgout − Tavg

in) (34)

where the time averaged values of outlet (Tavgout ) and inlet (Tavg

in)

working fluid temperatures for the simulation period (monthly orseasonal) were used to calculate the energy extraction/dissipationrates in heating/cooling modes. m is the working fluid mass flowrate (kg s−1).

The heat pump work rate was calculated as follows:

Epump = m · �P

� · �pump(35)

where �P is the pressure drop in the ground pipe, calculated basedon the friction factor and Reynolds number described in [41], and�pump is the pump efficiency (assumed 85% for a constant speedpump in this study).

The objective function used in the simulation was calculated asreciprocal of the difference between the ground energy extractionrate and the circulating pump energy consumption rate as follows:

Minimize(F)

F = 1

Eground − Epump

(36)

It was assumed that maximum heating/cooling energy ratesfrom the ground pipes do not exceed 1.5 kW in any of the cities,to match it with the energy demand of the small test room underinvestigation by the authors. This assumption was made to makethe comparison between different regions possible by searchingfor the parameters that maximize the energy extraction rates fromthe ground with a similar upper limit in all the modeled climaticconditions. Although a different objective function could havebeen defined to include the impacts of the dynamic building load

sification for ground heat pump design [32].

Thermal conductivity(W m−1 ◦C−1)

Thermal diffusivity(m2 s−1) × 10−8

0.29 580.77 451.11 540.91 491.67 662.5 93

46 A. Rezaei-Bazkiaei et al. / Energy and Buildings 61 (2013) 39–50

0 50 100 150 200 250 300 350−5

0

5

10

15

20

25

30

35

Air

tem

per

atu

re (

Deg

rees

C) Buffalo

Dallas

Miami

0 50 100 150 200 250 300 35060

80

100

120

140

160

180

200

220

240

So

lar

rad

iati

on

(W

/m2 )

DallasBuffaloMiami

iation

oa

ecuee

sdafpaottTsrptg

fow6

5

ttafictcyfaicwc(c

parameters to the cosine models. A schematic of air temperatureand solar radiation variation for these cities is presented in Fig. 6.A summary of the input parameters to the model for these cities ispresented in Table 5. Here, Us is the average annual surface wind

Table 5Weather data for selected cities as input to the model.

City

Buffalo Dallas Miami

Ta-avg (◦C) 9.3 19.7 23.7Ta-amp (◦C) 14.1 12.2 6.8˛air (month) 10.1 10.1 10.3

Tdew-avg (◦C) 4.3 9.4 18.1Tdew-amp (◦C) 12.2 11.6 7.3˛dew (month) 10.2 10 10.4

Days (starting on October 1st)

Fig. 6. Annual air and solar rad

ptimization of the building load demand as well as the cost savingsssociated with the presence of the intermediate soil layer.

A population size of 500 and 50 generations was selected forach run of the model after a few diagnostic runs to guarantee theonvergence to the optimized values. A single-objective GA wassed to search for the optimal value of the objective function inach population of results or, in other words, maximize the netnergy extraction/dissipation rates from/to the ground.

For each set of parameters in the initial population of the pos-ible design points (500 for these runs) chosen by GA, the finiteifference ground pipe model was run to calculate the energy ratesnd the heat pump work rate required to calculate the objectiveunction. The resulting values of the objective function for all theoints in the pool of initial set of variables were sorted, comparednd ranked by the GA options defined in Table 2 based on theirptimal values of objective function. After all the calculations forhe initial population of possible optimal points were performed,he GA utilizes the mutation and crossover functions as listed inable 2 to create the next generation of the near-optimal points. Aimilar calculation process was repeated for each generation of GAuns where the elite (near-optimal) sets of parameters from eachopulation were passed to the next generations of best results, andhis procedure repeated itself till the last generation (total of 50enerations) of optimal results were obtained.

A total number of 32 processors with 48 GB memory were usedor each monthly run of the model, which resulted in a runtimef about 7 h. A total number of 64 processors with 48 GB memoryere used for each annual run, which resulted in a runtime of about

5 h for each year.

. Effects of climate on optimization

The main focus of this research was to study the effects ofhe climatic conditions on the selection of the optimized opera-ional ground side parameters for a horizontal GSHP employing

non-homogeneous soil profile. To evaluate the potential bene-ts of the intermediate layer in different climate conditions, threeities representing different climates were selected for evalua-ion. These were Buffalo, NY, Dallas, TX and Miami, FL. Buffalo’slimate condition requires space heating for a majority of theear (approximately 838 annual cooling-degree-days, extractedrom www.degreedays.net), so Buffalo was assumed to represent

heating dominated city with eight months of heating (start-ng in October). Miami’s case with approximately 4517 annualooling-degree-days was assumed to be a representative of a

ooling-degree-days) was done with the assumption of six monthsNovember–April) of heating to be representative of a mild climateondition. To simplify the introduction of the weather data, the air

Days (starting on October 1st)

variation in the selected cities.

and dew-point temperature, and solar radiation values for thesecities were introduced to the model via estimation of these inputsby the following cosine functions:

Tair = Ta-avg + Ta-amp cos(

2t

P− ˛air

12

)(37)

where Ta-avg is the average annual air temperature, Ta-amp is theamplitude of fluctuation of annual air temperature, and ˛air is thefitted cosine model phase difference for air temperature, calcu-lated based on the start time of the modeling on October first (theassumed heating season start time). In addition, the dew-pointtemperature Tdewpoint used in calculation of the incoming longwaveradiation described in [21], was modeled as:

Tdewpoint = Tdew-avg + Tdew-amp cos(

2t

P− ˛dew

12

)(38)

where Tdew-avg is the average annual dew-point temperature,Tdew-amp is the amplitude of annual dew-point temperature vari-ation, and ˛dew is the fitted cosine model phase difference fordew-point temperature. Finally,

Qsi = Qsi-avg + Qsi-amp cos(

2t

P− ˛Qsi

12

)(39)

where Qsi-avg is the average annual solar radiation reaching the sur-face, Qsi-amp is the amplitude of fluctuation of the solar radiationthroughout the year, and the ˛Qsi

is the fitted cosine model phasedifference for radiation.

Weather data were obtained form the National OperationalHydrologic Remote Sensing Center (NOHRSC) and curve-fitting wasundertaken in the Microsoft Excel environment to obtain the input

Qsi-avg (W m−2) 142 168 182Qsi-amp (W m−2) 75 59 33˛Qsi

(month) 9.2 10.2 10.1

Us (m s−1) 4.2 4.1 4

A. Rezaei-Bazkiaei et al. / Energy and Buildings 61 (2013) 39–50 47

Table 6Monthly genetic algorithm results for Buffalo, Dallas and Miami.

Month Buffalo Dallas Miami

Intermediatematerial

Eground (W) % Intermediatematerial

Eground (W) % Intermediatematerial

Eground (W) %

October Sat. sand 558 6.5 TDA 1500 0.2 Sat. sand 1499 2.2November TDA 492 4.7 Sat. sand 1071 4.9 Sat. sand 1500 0.1December TDA 515 13.2 Sat. sand 1096 1.7 Sat. sand 1359 7.1January TDA 474 18.8 TDA 1054 0.1 Sat. sand 712 7.4February TDA 375 19.4 Sat. sand 1017 4.9 Sat. sand 649 6.2March TDA 252 6.8 Sat. sand 937 6.8 Sat. sand 1145 8.5April Sat. sand 179 9.1 Sat. sand 825 6.7 Sat. sand 1500 0.6May Sat. sand 295 11.7 TDA 1500 0.3 TDA 1500 1.1June TDA 1500 2.5 TDA 1500 1.7 TDA 1500 0.2

v(

6

6n

masafeteTese

miaRdficul

TiobcfaTftyoscJs

July Sat. sand 1500 1.2 TDA

August TDA 1500 1.2 TDA

September TDA 1500 1.3 TDA

elocity (m s−1) and the rest of the parameters are defined in Eqs.37)–(39).

. Results and discussion

.1. Monthly optimization of depth and placement ofon-homogeneous layer for energy saving

The optimization process was performed separately for eachonth to estimate the range of the optimal operating parameters

nd how the optimal parameters vary from month to month. Theelection of the intermediate layer material type by the algorithmssures a non-homogeneous profile that provides the best per-ormance. The optimized monthly parameters for the three citiesvaluated are presented in Table 6. The columns in this table con-ain the optimized choice of the intermediate layer and the energyxtraction/dissipation rates from/to the ground for each month.he energy rates and the percent difference between the calculatednergy rates with the non-homogeneous soil profile and the corre-ponding values for the homogeneous soil profile are presented forach city as well.

It can be concluded from these results that including an inter-ediate non-homogeneous layer of TDA provides more benefits

n certain climate conditions and within each climate its potentialdvantage is more pronounced in some months more than others.esults for Buffalo (Table 6) show that TDA was selected as theominant intermediate layer for eight months of the year of whichve months (November–March) were in the heating season. In theooling season, TDA was not selected only in July where the sat-rated sand tends to exhibit a dominant effect as an intermediate

ayer.The results for Dallas (Table 6) show a different trend where

DA has been selected as the dominant intermediate layer mostlyn warmer months of the year, whereas saturated sand yields theptimal energy extraction rates in the colder months. The reasonehind this observation probably lies in the difference in the phasehange angle of the annual temperature and solar radiation for Buf-alo and Dallas. The maximum and minimum ambient temperaturend solar radiations occur with a time lag for these cities (Fig. 6).herefore, the interplay of energy exchange processes on the sur-ace seems to favor the most conductive intermediate layer versushe least conductive layer or vice versa in certain months of theear. TDA was selected by the algorithm in the hottest monthsf the year in Miami (Table 6) to yield the highest energy dis-

1500 1.7 TDA 1500 0.91500 0.9 TDA 1500 0.61500 1.7 TDA 1500 0.3

optimization algorithm for these two months were marginally con-vergent to select TDA as the best intermediate material, whereasif the optimization would have continued longer and for largernumber of generations, saturated sand would have been selected.

An interesting observation from the monthly results was thatonly the two intermediate materials with lowest and highest ther-mal conductivities were selected as the choice non-homogeneouslayer material through the optimization algorithm. The fact thatTDA (the least conductive) or saturated sand (the most con-ductive) were selected among the provided list of intermediatematerial suggests that the ground pipes can benefit from a non-homogeneous layer above the pipe burial depth throughout theyear, but not necessarily from the one with the highest insulationproperties.

It should be noted that this modeling procedure does not takeinto account other characteristics of the TDA or saturated sand.Characteristics such as the porosity and water holding capacity,which might potentially contribute to considerably different per-formance results than that only based on the heat conduction in thesoil medium. TDA’s porous structure can potentially enhance themoisture migration to the underlying layers of soil, where highermoisture can contribute to higher thermal conductivities of the soilaround the pipes. It is expected that this characteristic of TDA hasa more substantial effect in the summer time, especially in regionswith less rainfall events.

The values in the last column for each city in Table 6 representthe percentage energy extraction/dissipation rate increase com-pared to the homogeneous soil profile in different months. Thereare trends in comparing the results for similar seasons in differentcities. The percentage increases in energy extraction rate for thecoldest months in Buffalo (January–February) are as high as 18–19%as compared to similar time periods for Dallas with highest valuesof 5–6%. This finding can be translated into the potential for higherenergy harvesting potential with a non-homogeneous soil profile inthe heating season in a colder climate (Buffalo). It should be notedagain that the material choice for the intermediate layer in the coldseason in Buffalo was TDA versus the saturated sand for Dallas.

A similar comparison for the cooling season in all cities revealsan interesting observation. The highest increase in the energy dis-sipation rate, for cooling in Miami, happens in the coldest monthof the year with an increase in energy dissipation rates of approx-imately 6–8%. The optimization algorithm tends to choose higherworking fluid temperatures, compared to the warmer months, withthe highest intermediate layer conductivity (saturated sand) tomaximize the heat flux to the ground and subsequently increase

4 rgy and Buildings 61 (2013) 39–50

rsmefmdewhowi

6

mpttbvtketrwtrmtsmiatacm

aflotacwtcto

Table 7Optimal values of working fluid temperature and TDA configuration from annualGA simulations for Buffalo, Dallas and Miami (to be used as annual design values).

Parameter Buffalo Dallas Miami

Tf,imin(◦C) −2.3 −2.7

Tf,imean-heating(◦C) 7 23.5

Thickness, b (m) 0.5 0.5 0.8Position, d (m) 0.3 0.4 0.4

TV

8 A. Rezaei-Bazkiaei et al. / Ene

ates from the ground are achievable with the non-homogeneousoil profile for cooling purposes. It should be noted that the perfor-ance of the homogeneous system is very close to the maximum

xpected cooling capacity (1500 W) in warmest months of the yearor all these cities, which contributes to the marginal improve-

ents with a non-homogeneous soil profile. In other words, if theesigned ground pipe network was supposed to provide highernergy rates, the percentage increase in energy extraction ratesould probably be higher and more pronounced with the non-omogeneous soil profile. It seems reasonable to conclude that usef a non-homogeneous soil profile for cooling-only purposes, in aarm climate, can be more beneficial if a relatively high conductive

ntermediate layer is employed.

.2. Annual optimization

The intention behind performing the optimization for eachonth was to gain an understanding of the key parameters that

rovide optimal energy rates for the ground pipe and their shortererm variation in different regions. The intermediate layer selec-ion by the optimization algorithm was performed to provide aasis for comparing the ground pipe performance each monthersus the results of annual optimization presented in this sec-ion. After performing the analysis for each month and gainingnowledge of the relationship between selected variables, mod-ling was focused on finding the optimal values of the inlet fluidemperature and the configuration of the intermediate layer year-ound. This approach provides the designer with the optimizedorking fluid temperatures and intermediate layer’s configura-

ion which yields maximum annual energy extraction/dissipationates in each season. One of the most important findings from theonthly simulation results was that the two dominant choices for

he intermediate non-homogeneous layer were TDA and saturatedand, the two intermediate material with lowest and highest ther-al conductivity. Although the selection of the antifreeze solution

s mandated by the pipe characteristics, pressure drop calculationsnd other site specific considerations, the working fluid proper-ies was set to 13% propylene-glycol and water mixture for all thennual runs. This working fluid property was the most dominanthoice of the GA search in the monthly optimization to achieveaximum energy extraction/dissipation rates from/to the ground.Given that the TDA material can be implemented practically as

value-added passive design with non-homogeneous soil profileor the ground pipes, the annual optimization simulations wereimited to the use of TDA. After obtaining the optimized valuesf the entering fluid temperatures and TDA layer configuration,he energy extraction/dissipation rates were calculated for heatingnd cooling seasons. Energy extraction/dissipation rates were cal-ulated for a non-homogeneous profile with saturated sand layerith the same optimized operational parameters achieved for TDA

o compare the annual results. This approach provides a base toompare the difference between the energy harvesting rates fromhe ground with the two choice intermediate material in identicalperational conditions.

able 8alues of seasonally averaged energy extraction/dissipation rates in heating/cooling seas

Parameter Buffalo

Heating Cooling

TDA energy (W) 324 1497

Sat. sand energy (W) 274 1375

Homog. energy (W) 280 1391

% TDA 15.7 7.6

% Sat. sand −2.1 −1.2

Tf,imax (◦C) 45.7 52.6 73.0Tf,imean-cooling

(◦C) 31.7 29.6 55.0

Results from the annual simulations are presented inTables 7 and 8. The values of TDA thickness and position and theoperating fluid temperatures in Table 7 were obtained by runningthe optimization for the entire year, so these values should beused as the annual design values that maximize the energy extrac-tion/dissipation rates from/to the ground. The reported energyextraction/dissipation rates in Table 8 are the average values foreach season. The upper and lower limits for the values of the work-ing fluid temperatures were chosen to cover the below freezingpoint temperatures in heating season as well as high temperaturesin cooling season. The introduction of the broad range of temper-ature values to the search algorithm was intentionally made toutilize the capacity of the search algorithm to find the optimal solu-tion in a bigger space of design options. Although some of the highertemperature values might not be in the common practice range forthe ground pipe design, the goal of this research to explore thepotential new design options for the ground pipe has ruled theneed for keeping the upper limits higher than the common workingfluid temperature ranges. The aim of this study has been to explorethe options for the ground pipe design with new initiatives thathave not been considered before, yet require further investigation.Interesting observation from the optimization results for the max-imum and average working fluid temperature is that the search forthe optimal working fluid temperatures has resulted in values thatare considerably higher than the common upper range design val-ues (approximately 35 ◦C). These findings can be a starting pointfor investigating new ground pipe technologies that are capable ofdelivering higher temperatures from the ground to achieve optimalenergy extraction rates. Nonetheless, the practical aspects of imple-mentation of the higher working fluid temperature values requiresfurther investigation.

The percentages in Table 8 refer to the percent increase in theenergy extraction/dissipation rates compared to the homogeneoussoil profile with TDA or saturated sand layers. The obtained opti-mized temperature values and intermediate layer configurationfor TDA were used to run the homogeneous and saturated sandscenarios for comparison.

A comparison between the Buffalo annual energy extrac-tion/dissipation rates shows 15.7% and 7.6% higher rates with TDAlayer versus the homogeneous case, in heating and cooling seasons,

on for Buffalo, Dallas and Miami.

Dallas Miami

Heating Cooling Cooling

635 1431 1298608 1378 1482612 1382 1456

3.8 3.5 −10.9−0.7 −0.3 1.8

rgy an

iw

oisia(dTirt

payescitfrtaAs5esd

nitstbiewia(oifsaorBOfstfTeat

aa

A. Rezaei-Bazkiaei et al. / Ene

n heating and cooling seasons as compared to 0.7% and 0.3% lowerith saturated sand layer in similar seasons (Table 8).

For Buffalo, higher energy extraction/dissipation rate increasever the homogeneous profile was observed with TDA layer in heat-ng than cooling season (15.7% versus 7.6%). Comparing the heatingeason energy extraction rates over the homogeneous soil profilen Buffalo to Dallas shows that relatively higher efficiencies werechievable in Buffalo (15.6% versus 3.8%). The results for MiamiTable 8), on the other hand, suggest a significantly better energyissipation rates to the ground with saturated sand as compared toDA layer (1.8% versus −10.9%) for cooling purposes. However, its worth noting that the percentage increase in energy dissipationate to the ground (1.8%) with saturated sand is considerably lowerhan the monthly values obtained for Miami (Table 6).

The optimized energy extraction rates with the monthly designarameters obtained from the genetic algorithm deviate consider-bly from the values obtained from the simulations for the entireear. It is harder for the search algorithm to find operational param-ters for the ground pipes that decrease the difference betweeneasonal energy rates and highest achievable monthly values. Thisauses the annual energy rates to be lower than the ones attainablef the monthly control on the operating parameters (working fluidemperatures and intermediate layer configuration) was practicallyeasible. For example, the maximum attainable energy extractionate in heating season for Buffalo is 324 W (Table 8) as comparedo maximum obtainable monthly (October) rate of 558 W (Table 6),

reduction of approximately 40% compared to the monthly value. similar comparison for Dallas shows that the average heatingeason energy extraction rate of 635 W (Table 8) is approximately7% lower than the maximum monthly (October) attainable energyxtraction rate of 1500 W (Table 6). The cooling season energy dis-ipation rates to the ground in all the cities have considerably lowereviance from the corresponding monthly optimized values.

It is worth raising a discussion regarding the physical phe-omenon that leads to the selection of TDA versus saturated sand

n the search algorithm. Conceptually, there is a need for balancinghe amount of heat penetrating the soil surface, mainly from theolar gains and surface heat conduction, and the amount of heathat needs to be extracted/dissipated depending on the season. Thisalance point, of course, has been altered by the introduction of the

ntermediate layer in this study which can be translated into morengineering control on this natural process occurring in the soil. Theay this energy extraction/dissipation demand becomes satisfied

s a function of both the physical soil profile as well as the ambientir temperature and solar radiation variation throughout the yearFig. 6). The combination of these parameters promotes utilizationf an intermediate layer, which best serves the heat transfer regimen the soil profile. For cities with climatic conditions similar to Buf-alo, this balance seems to shift toward the need to maintain thetored heat in the bulk of soil at the pipe level, given the lowermount of heat entering the soil profile from the surface becausef less incident solar energy and lower air temperature. This is theeason TDA has been selected by the algorithm for cold months inuffalo. A similar comparison in the cooling season for Miami, fromctober to April, shows that a higher conductive intermediate layer

avors the heat dissipation process from the pipe to the surroundingoil by letting more heat escape the immediate bulk of soil adjacento the pipe. A closer look at the cooling season optimization results,rom May to September, for Dallas and Miami (Table 6), shows thatDA has been selected as the choice intermediate layer. This can bexplained as an attempt to maintain the ground cooler by havingn overlaying insulation layer so that there is a higher heat flux to

Because the intermediate layer configuration can not bedjusted once the system is installed, it is clear that the intermedi-te layer dimensions that optimize the annual energy extraction

d Buildings 61 (2013) 39–50 49

rates are needed for each of these climates as summarized inTable 7. More flexible control on the energy extraction rates in eachmonth can subsequently be obtained via control on the heat pumpcharacteristics. Variable refrigerant flow or multi-stage heat pumpscan be designed in conjunction with the desirable intermediatelayer properties and configuration to assure the least deviance fromthe actual monthly building energy demand. Moreover, a detailedanalysis of the impact of the geographical variation of soil prop-erties is recommended to be considered in the future studies. Thenumerical results from the presented analyses in this paper aresubject to verification with the expected field data from an exper-imental GSHP facility under investigation by the authors. Authorsexpect that with the results from their field experiment the respon-sible mechanisms associated with the implementation of the TDAmaterial can be revealed with more evidence, so that alternativematerials can be explored, subsequently.

7. Conclusions

A numerical model was developed for the ground pipe of a hor-izontal GSHP with a non-homogeneous soil layer. The model wascoupled with genetic algorithm to search for operational param-eters that maximize energy extraction/dissipation rates from/tothe ground. The search algorithm was given a range of workingfluid properties, intermediate layer thermal properties, a range ofoperating fluid temperatures, and the intermediate layer config-uration to search for the optimized energy extraction/dissipationrates from/to the ground. The optimization was performed for threecities representing a cold (Buffalo), moderate (Dallas) and warm(Mmiami) climate to evaluate the impact of climate on the opti-mization. Despite different performance achievements with eithera low conductance (TDA) or a high conductance (saturated sand)intermediate layer, a non-homogeneous soil profile demonstratedthe potential for increasing the energy extraction/dissipation ratesfrom/to the ground. A shift in perspective toward more controlstrategies for GSHPs with control on the ground pipe side of the sys-tem is suggested based on the model results. Further investigationof other attributes of a non-homogeneous system which potentiallycan enhance the GSHP’s performance is still required. A summaryof findings are listed below:

• A non-homogeneous soil profile exhibited a great potential forenhancing a horizontal GSHP’s pipe performance by increasingthe energy extraction/dissipation rates from/to the ground.

• TDA demonstrated higher benefits in colder climates by increas-ing the energy extraction rates from the ground in the heatingseason.

• TDA demonstrated a marginal enhancement during coolingcycles due to insignificant difference between achievable energyextraction rates with the non-homogeneous and homogeneouscases.

• Saturated sand demonstrated potential for increasing the energydissipation rates to the ground in warm climate.

• The optimized seasonal energy extraction rates from the groundexhibited significant difference (an upper range of 40–60% less)from the highest achievable monthly values in the heating season.

• Minimal difference was observed between the cooling sea-son optimized energy dissipation rates and the correspondingmonthly values.

Acknowledgments

Partial funding for this research was provided by EmpireState Development’s Environmental Services Unit throughthe New York State Tire Derived Aggregate Program at the

5 rgy an

UwnWc

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

0 A. Rezaei-Bazkiaei et al. / Ene

niversity at Buffalo’s Center for Integrated Waste Management:ww.tdanys.buffalo.edu/UB. Authors would like to thank Dr. Ken-eth Fishman and Mr. Louis Zicari from the Center for Integratedaste Management for their constant support and continuing

ontribution to this project.

eferences

[1] IGSHPA, Ground Source Heat Pump Residential and Light Commercial Designand Installation Guide, International Ground Source Heat Pump Association,Oklahama State University, Division of Engineering Technology, Oklahoma, OK,2009.

[2] O. Zogou, A. Stamatelos, Optimization of thermal performance of a buildingwith ground source heat pump system, Energy Conversion and Management48 (2007) 2853–2863.

[3] H. Madani, J. Claesson, P. Lundqvist, Capacity control in ground source heatpump systems. Part II: Comparative analysis between on/off controlled andvariable capacity systems, International Journal of Refrigeration 34 (2011)1934–1942.

[4] A. Rezaei-Bazkiaei, E. Dehghan-Niri, A.S. Weber, G.F. Dargush, A new controlstrategy to improve the ground source heat pump’s efficiency with a recycledproduct, in: Proceedings of the ASME 2012 International Mechanical Engineer-ing Congress & Exposition IMECE2012, November 9–15, Houston, Texas, USA,2012.

[5] T. Qureshi, Variable-speed capacity control in refrigeration systems, AppliedThermal Engineering 16 (1996) 103–113.

[6] H. Madani, J. Claesson, P. Lundqvist, Capacity control in ground source heatpump systems. Part I: Modeling and simulation, International Journal of Refrig-eration 34 (2011) 1338–1347.

[7] F.F. Al-ajmi, V.I. Hanby, Simulation of energy consumption for Kuwaiti domesticbuildings, Energy and Buildings 40 (2008) 1101–1109.

[8] Y. Gao, J.J. Roux, L.H. Zhao, Y. Jiang, Dynamical building simulation: a loworder model for thermal bridges losses, Energy and Buildings 40 (2008)2236–2243.

[9] L. Wang, J. Gwilliam, P. Jones, Case study of zero energy house design in UK,Energy and Buildings 41 (2009) 1215–1222.

10] Y. Hang, M. Qu, S. Ukkusuri, Optimizing the design of a solar cooling systemusing central composite design techniques, Energy and Buildings 43 (2011)988–994.

11] C. Xi, L. Lin, Y. Hongxing, Long term operation of a solar assisted ground cou-pled heat pump system for space heating and domestic hot water, Energy andBuildings 43 (2011) 1835–1844.

12] A. Khandelwal, P. Talukdar, S. Jain, Energy savings in a building using regener-ative evaporative cooling, Energy and Buildings 43 (2011) 581–591.

13] P.J. Martinez, A. Garcia, J.M. Pinazo, Performance analysis of an air conditioningsystem driven by natural gas, Energy and Buildings 35 (2003) 669–674.

14] Z. Lin, C.K. Lee, K.F. Fong, T.T. Chow, Comparison of annual energy performanceswith different ventilation methods for temperature and humidity control,Energy and Buildings 43 (2011) 3599–3608.

15] O. Siddiqui, A. Fung, H. Tse, D. Zhang, Modelling of the net zero energytown house in toronto using trnsys, and an analysis of the impact usingthermal mass, in: 2008 Proceedings of the 2nd International Conference onEnergy Sustainability, ES 2008, vol. 2, Jacksonville, FL, United states, 2008,pp. 297–304.

16] E. Kjellsson, G. Hellström, B. Perers, Optimization of systems with the combi-nation of ground-source heat pump and solar collectors in dwellings, Energy35 (2010) 2667–2673.

17] R.G. Kapadia, S. Jain, R.S. Agarwal, Transient characteristics of split air-

18] Y. Nam, R. Ooka, S. Hwang, Development of a numerical model to predict heatexchange rates for a ground-source heat pump system, Energy and Buildings40 (2008) 2133–2140.

[

[

d Buildings 61 (2013) 39–50

19] Y. Zhou, J. Wu, R. Wang, S. Shiochi, Energy simulation in the variable refrigerantflow air-conditioning system under cooling conditions, Energy and Buildings39 (2007) 212–220.

20] X. Liu, T. Hong, Comparison of energy efficiency between variable refrigerantflow systems and ground source heat pump systems, Energy and Buildings2010 (2010).

21] A. Rezaei, B.E.M. Kolahdouz, G.F. Dargush, A.S. Weber, Ground source heat pumppipe performance with Tire Derived Aggregate, International Journal of Heatand Mass Transfer 55 (2012) 2844–2853.

22] J. Shao, J. Zarling, Thermal conductivity of recycled tire rubber to be used asinsulating fill beneath roadways, Technical Report, Institute of Northern Engi-neering, University of Alaska Fairbanks, Fairbanks, 1995.

23] B. Lawrence, D. Humphrey, L.-H. Chen, Field trial of tire shreds as insulationfor paved roads, in: Cold Regions Engineering: Putting Research into Practice,ASCE, Hanover, NJ, 1999, pp. 428–439.

24] D. Humphrey, Tire shreds as lightweight fill for retaining walls – results offull scale field trials, in: International Workshop on Lightweight Geomaterials,2002, pp. 261–268.

25] H. Moo-young, K. Sellasie, D. Zeroka, G. Sabnis, Physical and chemical proper-ties of recycled tire shreds for use in construction, Journal of EnvironmentalEngineering (2003) 921–929.

26] D. Humphrey, A. Fiske, R. Eaton, Backcalculation of thermal conductivity of tirechips from instrumented test section, in: Transportation Research Board 81stAnnual Meeting, 2002.

27] H.L. Wappett, J.G. Zornberg, Full scale monitoring for assessment of exother-mal reactions in waste tires final report, Technical Report 27, RecycledMaterials Resource Center, Project No. 27, University of Texas, Austin,2006.

28] F. Ling, T. Zhang, A numerical model for surface energy balance and thermalregime of the active layer and permafrost containing unfrozen water, ColdRegions Science and Technology 38 (2004) 1–15.

29] G.E. Liston, D.K. Hall, An energy-balance model of lake-ice evolution, Journal ofGlaciology 41 (1995) 373–382.

30] R. Fleagle, J. Businger, An Introduction to Atmospheric Physics, vol. 25,Cambridge University Press, New York, 1980.

31] H. Demir, A. Koyun, G. Temir, Heat transfer of horizontal parallel pipe groundheat exchanger and experimental verification, Applied Thermal Engineering 29(2009) 224–233.

32] J. Bose, Soil and rock classification for the design of ground-coupled heat pumpsystems-field manual, Electric Power Research Institute Special Report, EPRICU-6600, 1989.

33] H. Demir, A. Koyun, S.O. Atayilmaz, Determination of optimum design param-eters of horizontal parallel pipe and vertical U-tube ground heat exchangers,in: ASME Conference Proceedings, 2009, pp. 621–627.

34] A. Koyun, H. Demir, Z. Torun, Experimental study of heat transfer of buriedfinned pipe for ground source heat pump applications, International Commu-nications in Heat and Mass Transfer 36 (2009) 739–743.

35] T. Kusuda, P.R. Achenbach, Earth temperature and thermal diffusivity atselected stations in the United States, Technical Report, National Bureau ofStandards, Washington, DC, 1965.

36] D. Goldberg, Genetic Algorithms in Search; Optimization and Machine Learn-ing, Addison-Wesley, Massachusetts, 1989.

37] M. Mitchell, An Introduction to Genetic Algorithms, MIT Press, Cambridge, MA,1996.

38] S. Sanaye, B. Niroomand, Thermal-economic modeling and optimization ofvertical ground-coupled heat pump, Energy Conversion and Management 50(2009) 1136–1147.

39] S. Sanaye, B. Niroomand, Horizontal ground coupled heat pump: thermal-economic modeling and optimization, Energy Conversion and Management 51(2010) 2600–2612.

40] H. Sayyaadi, E. Amlashi, Multi-objective optimization of a vertical groundsource heat pump using evolutionary algorithm, Energy Conversion and Man-agement 50 (2009) 2035–2046.

41] S. Kakac , H. Liu, Heat Exchangers: Selection, Rating and Thermal Design, seconded., CRC Press, Boca Raton, FL, 2002.