- Email: [email protected]
Coupled EnergyPlus and computational fluid dynamics simulation for natural ventilation
Building and Environment 68 (2013) 100e113
Contents lists available at SciVerse ScienceDirect
Building and Environment journal homepage: www.elsevier.com/locate/buildenv
Coupled EnergyPlus and computational fluid dynamics simulation for natural ventilation Rui Zhang a, *, Khee Poh Lam b, Shi-chune Yao b, Yongjie Zhang b a b
IBM, T.J. Watson Research Center, United States Carnegie Mellon University, United States
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 December 2012 Received in revised form 27 March 2013 Accepted 2 April 2013
Energy modeling approaches have continued to advance to cater for emerging new design concepts toward “greener” solutions that optimize energy consumption in buildings while maintaining thermal comfort as well as healthy environment. Increasing attention is given to passive and mix-mode systems in building. Computational Fluid Dynamics (CFD) model has been widely adopted as effective tool for natural ventilation simulations. However, CFD become unstable for conjugate heat transfer model, which is the transient heat transfer between solid and fluid. Solid and fluid has different respond times to thermal energy. Typically walls respond in hours, and air responds in seconds, causing the system to become stiff. In addressing the issue, a coupled lumped heat transfer model (EnergyPlus) and CFD model (Fluent) was implemented, and 8 days of simulation was conducted. The airflow rates of openings and heat transfer coefficients from the airflow network module in EnergyPlus and those from CFD model were compared. Results show that airflow network model generally predict smaller airflow rates for the openings. Airflow network model generates better results for openings on the south and east facade and internal openings, where there is no immediate adjacent building. Among the heat transfer coefficients calculation methods in EnergyPlus, the TARP algorithm generated closest HTC values to the coupled CFD results. The overall heat transfer coefficients of all the enclosure surfaces are calculated and it is found that the overall thermal resistances generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. Ó 2013 Elsevier Ltd. All rights reserved.
Keywords: CFD Energy simulation EnergyPlus Natural ventilation Air change rates Heat transfer coefficient
1. Introduction Energy modeling approaches have continued to advance to cater for emerging new design concepts toward “greener” solutions that optimize energy consumption in buildings while maintaining thermal comfort as well as healthy environment. Instead of the conventional approaches that rely solely on mechanical system to provide the desired thermal conditions, increasing attention is given to passive and mix-mode systems in building. Notwithstanding the advances in system modeling, there are still limitations in the capability of representing the “real” environmental behavior with different building spatial configurations. For example, most current energy models adopt a “nodal” approach to simulate the heat transfer process in the building, with simplified heat resistors and capacitors network. The underlying assumption for nodal model is the uniformity of room temperature, which may * Corresponding author. IBM, T.J. Watson Research Center, Business Analytic and Mathmetical Science, 1101 Kitchawan Rd., Room 32-244, Yorktown Heighs, NY 10598, United States. Tel.: þ1 4126517384. E-mail addresses: [email protected], [email protected] (R. Zhang). 0360-1323/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.buildenv.2013.04.002
apply to most of the building zones with moderate room size. However, limitations also come with such nodal approach: Priori and empirical knowledge of various coefficients are needed for model input, such as wind pressure coefficient, heat transfer coefficient, loss factors, friction factors, etc. Effects of thermal and air flow patterns introduced by building spatial configurations are difficult to be represented into the model. More detailed zonal model, such as COMIS [1], has been developed for ventilation design in complex buildings. However, research [2] shows that the results of zonal model are not satisfiable compared with even coarse-grid CFD models under isothermal conditions. Finite Volume Method (FVM) such as the Computational Fluid Dynamics (CFD) on the contrary will perform detailed computation on the heat transfer and air flow simulation, which could supplement the nodal model for the building energy simulation. Detailed temperature profile and air flow field are calculated with first
R. Zhang et al. / Building and Environment 68 (2013) 100e113
principle based NaviereStokes set of equations and turbulence models. Coefficients, such as the heat transfer coefficients, will be the results of the simulation, which is defined by a set of boundary conditions. However, limitation also comes with CFD. Firstly, the computation resources needed for CFD simulation are much more expensive than the nodal model. Secondly, the coupling between fluid and solid material is difficult and computational more expensive, thus the thermal storage capacity of building component is difficult to be correctly modeled. Therefore, CFD has been limited to detailed studies of given space under a number of specific boundary conditions. However, with the rapid development of computation power, the computation time of CFD is decreasing, it is highly necessary to develop an integrated thermal simulation that combines the advantages of both the nodal and FVM model, providing higher accuracy within acceptable computation time. The nodal model calculation of the heat transfer process in the building is based on the fundamental heat balance principles. Each zone is assumed with uniform state variables of temperature, pressure and density. Empirical coefficients are used to calculate the state variable transfer fluxes between each node. One example of how such nodal system is implemented is shown as following:
Cz
NX Nsl NX surfaces zones X dTz _ i Cp ðTzi -Tz Þ m hi Ai ðTsi -Tz Þ þ Q_ i þ ¼ dt i¼1 i¼1 i¼1
_ inf Cp ðTN -Tz Þ þ Q_ sys þm
(1)
where: CzdTz/dt e room air energy change rate, PNsl _ Q e sum of the convective heat transferred through ini¼1 i ternal heat sources or sinks, PNsurfaces hi Ai ðTsi -Tz Þ e sum of the convective heat transferred i¼1 through building envelope, PNzones _ i Cp ðTzi -Tz Þ e energy from neighbor zones air mixing, m i¼1 _ inf Cp ðTN Tz Þ e total energy of infiltration air, m Q_ sys e system output, In equation Eq. 1, the heat transfer coefficient needs to be determined for a given set of boundary conditions that define the thermal and fluid fields in the space. Various empirical correlations have been developed to calculate the heat transfer coefficient, which are dependent on the ambient air flow characteristics. However, the lack of knowledge of the air flow, will lead to considerable errors in the heat transfer prediction of the nodal model for buildings, where the heat transfer process is complex by nature. Research [3] shows that differences in heat transfer coefficient between CFD and empirical equation can range from 1.42 W/ m2 K to 111.41 W/m2 K. Mora et al. [2] investigated the effectiveness of zonal model in describing the airflows in large indoor space. A coarse-grid CFD simulation was also carried out. Velocity and pressure computation from different formulations of zonal models and coarse-grid ke3 CFD model were compared to measurements in a 2D isothermal room. The results showed that zonal model computation was not satisfactory compared with coarse-grid CFD model (grid of 10 10). Negrao [4] studied the heating process in a room with a thermostat and a heating radiator both in a nodal model and a CFD-nodal model. The ESP-r program, which is a nodal model for building thermal computation, was used to compute the required energy generation by the radiator, based on the calculated room air temperature and thermostat setting. Significant difference in the resulting heat generation requirement and the room air temperature had been observed between the two models. Wang and Wong [5] studied the performance of naturally ventilated residential buildings in Singapore. A zonal model for building
101
thermal performance and airflow network was used, and a manual coupled approach between the zonal model and CFD was also developed. Field measurement was carried out in a real building in Singapore. Compared with measurements, the discrepancy of indoor air velocity and dry bulb temperature predicted by zonal model was larger than that predicted by the coupled model. Nagai and Kurabuchi [6] constructed a nodal model for a high-rise apartment building with central void through the whole height of the building in Japan. The simulation results of the air temperature in the void space and the surface temperatures of the corridors were compared with measured data. Results generated by the nodal were generally lower than the measured data. 2. Previous works Manual coupling of the CFD model and nodal model has been done for many geometrically complicated buildings. Nagai and Kurabuchi [6] used CFD to decide the coefficients in the nodal model for a high-apartment building with central void space throughout the height of the building in Japan. Manz and Frank [7] used a one-way-static coupling to study the thermal performance of double facade buildings. Wong et al. [8] also coupled CFD and nodal model to study the performance of double facade building in a tropical climate in Singapore. Research work on automated coupling of the nodal model and CFD model at run time is dated from 1990s. Negrao [9] implemented a full iterative coupling approach of the nodal modal and CFD model. Selected space was simulated with CFD, and ESP-r was used as the nodal model. The coupled variables were: surface temperature of all the walls and windows, and the pressures at the internal openings that connect the space with the rest of the building. A full iterative strategy has been used, where coupled variables will be exchanged at each iterative step until a convergence criterion was reached at each time step. It was found that low under-relaxation (0.1) was necessary in order for the CFD computation to converge for the coupled system. The number of iteration of the CFD computation was increased compared with the uncoupled CFD computation. Based on the work by Negrao [9,4], Beausoleil-Morrison [10], Beausoleil-Morrison et al. [11], Beausoleil-Morrison and Clarke [12] continued with the investigation of the coupling between nodal model (ESP-r) and CFD model. A conflation controller was implemented to configure the CFD model at each time step. At the start of each time step, the zero-equation turbulence model was employed, and the resulting viscosity field was used to initialize the ke3 turbulence model. Grashof Number (the ratio of buoyancy and viscous forces) and Reynolds Number (the ratio of inertial and viscous forces) were evaluated. The existence of the buoyancy term in zmomentum equation was determined. The conflation controller would then choose the appropriate turbulence model and near wall model. An empirical validation of the coupled model by BeausoleilMorrison [10] is conducted by Bartak et al. [13]. Two CFD models with coarse-grid and fine-grid were constructed and compared with experimental results. It was found that the coarse-grid and fine-grid CFD produced comparable results. Variables validated were air temperature, comfort parameters, and local mean age of air under isothermal and non-isothermal conditions. Good agreements were found between the predicted and measured data. Djunaedy et al. [14], Djunaedy et al. [15], Chen et al. [16] and Zhai et al. [17] discussed both the advantages and disadvantages of internal coupling of the CFD and nodal model. It was concluded that external coupling is more favorable in terms of no stiffness issue, less expensive, nodal and CFD model themselves are independent and can be optimized individually. Zhai and Chen [3,18e22] also investigate the coupling strategy extensively. They implemented a coupling strategy to exchange heat
102
R. Zhang et al. / Building and Environment 68 (2013) 100e113
transfer coefficient and surface boundary conditions between nodal model and CFD model through dynamic or static coupling. Hereby, static and dynamic is the strategy used for data exchange instead of thermal process. A sensitivity study was also carried out to investigate the effectiveness of the coupling method in improving the building energy simulations. Their results showed that for rooms with moderated size, without sensible temperature stratification, the coupling approach shows marginal effect. However for rooms with large temperature stratification, the discrepancy between coupled method and nodal model is significant (42%). Wang [23], Wang and Chen [24] proved theoretically that the coupled model has a solution and it is unique. The investigation used Scarborough criterion to evaluate convergence performances and analyzed the stabilities of three coupling strategies, which are pressure coupling, pressure and flow rate mixed coupling, and flow rate coupling. It was found that the pressure coupling performs best. Wang and Wong [5] developed a text-based interface for automated coupling program to extract and exchange information between EPS-r (nodal model) and Fluent (CFD model). External coupling was used and results showed that coupled program yields better results compared with single nodal or CFD model. Wang and Wong [25] compared velocity inlet coupled and pressure inlet coupled strategy, and found that pressure inlet coupled strategy gave better results. In this paper, we will present a novel approach and platform for the coupling between nodal model, i.e., EnergyPlus and CFD, i.e., Fluent, via open source platform of BCVTB (Building Controls Virtual Test Bed [26]). The coupling platform is configurable, and flexible. One demonstration example upon the coupling platform is described in detail. Two key parameters from the nodal model and coupled CFD model will be compared. One is the heat transfer coefficients computed from the coupled CFD model and various heat transfer coefficient calculation methods in EnergyPlus. The other is the air flow rate through openings from coupled CFD model and Airflow Network model in EnergyPlus.
model and CFD model can be satisfactorily achieved by maintaining each method’s solution algorithm separately. Furthermore, research [29] found that difference in simulation results between internal coupling and external coupling is not significant. However, the benefits of the external coupling are three folds: 1. computationally less expensive; 2. nodal and CFD model can be maintained and updated individually; 3. configurable exchange variables. This study will implement the Progressive-Replacement external coupling strategy. The exchange variables will be exchanged after each model, i.e., the nodal and the CFD model, comes to its converged state at each time step.
3. Coupling method EnergyPlus [27] is chosen as the nodal model, and Fluent software [28] is chosen as the FVM model for the coupled simulation platform. The coupling method and platform are described in detail in the following sections: 3.1. Coupling strategy The coupling approaches between the nodal and FVM model can be characterized into three types: Full internal coupling, where the set of equations for nodal model and CFD model are solved together iteratively. Iterative external coupling, where the set of equations for nodal model and CFD model are solved individually, the variables are exchanged with an iterative procedure until a converged state is achieved. Progressive-Replacement external coupling, where the set of variables are exchanged after each model comes to converged state at each timestep. Research [4] shows that the fully internal coupling and iterative external coupling will require larger number of iterations to reach convergence, e.g., approximately 600 iterations were necessary for a simulation which required approximately 200 iterations for an individual CFD simulation. It is also found that the convergence of the iterative approach is difficult even for simple room simulation. One of the conclusions from the above study is that the nodal
Fig. 1. The flow chart of the FlowPlus program.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
3.2. Coupling variables As previously mentioned, nodal model requires various building specific heat and mass transfer coefficients, which can be obtained from CFD simulation. At the same time, the CFD tool may have stiffness problem in solving transient conjugate heat transfer with conduction and convection between the building envelope components and air in the space. Based on the characteristics of the two models, a collection of coupling variables was selected. The nodal model will provide to the CFD tool all the interior and exterior surface temperatures of the envelope components of the building and outdoor weather conditions. The CFD tool will take temperatures of various building surfaces as settings of the thermal boundary conditions, including both interior surfaces and exterior surfaces. For natural ventilation, wind is the driving force of the flow through the openings in the building. The wind conditions, in terms of speed and direction, provided by the nodal tool from weather data will be used to determine the boundary type of the four boundary surfaces of the simulation domain. For example, if the wind is coming from south, with a speed of 2 m/s, the surface of the south bound will be set as velocity inlet, with incoming wind velocity of 2 m/s. The north bounding surface of the domain will be set as outflow. The east and west boundaries will be set as symmetry assuming there is no shear strain on the surface. The CFD tool will conduct the steady state natural ventilation simulation. A post processing program will be implemented to extract the temperature profile and velocity fields from CFD tool and then to calculate the values for the exchange variables, which include the following: Airflow rates through all the openings. Average air temperatures through the openings. Surface heat transfer coefficients for all the envelop surfaces.
3.3. Coupling platform The coupling platform will be based on Building Controls Virtual Test Bed (BCVTB Wetter et al. [26]), which is a software environment targeted to provide an integration platform for various simulation tools. For example, the BCVTB enables concurrent energy simulation of whole building in EnergyPlus with HVAC system and operating control in cþþ program, or MATLAB/Simulink, while exchanging data between the programs at each time step. 3.3.1. EnergyPlus object in support of the coupling In this study a new EnergyPlus input object called the “ExternalInterface:Airflow” is added for the purpose of coupling between nodal and CFD model for air flow simulations. The entry of the objects are designed as follows:
103
The ExternalInterface:Airflow object is designed to set the airflow rate for each piece of the openings in the building. The input entry of ExternalInterface:Airflow will specify the name of the opening, and zone name that the air is flowing in (positive value) or flowing out from (negative value). The value of heat transfer coefficient will be updated through the ExternalInterface:Schedule object. 3.3.2. FlowPlus program for executing CFD simulation and extracting coupling variables A cþþ program, called FlowPlus, is developed to execute the Fluent software to conduct the CFD simulation and extract the results for the coupling variables. The program will read in the variable configuration file, which specifies the exchange variables that EnergyPlus is sending and also the variables that are needed to be extracted from Fluent simulation results. Then the program will use the values obtained from EnergyPlus to generate a Fluent journal file, which will set the boundary conditions for the CFD simulation. Then Fluent software will be called to execute the CFD simulation according to the boundary conditions that are specified in the journal file. After Fluent finished executing the iteration, the FlowPlus will extract the temperature profile and velocity fields from Fluent and calculate the values for the exchange variables and send the values to EnergyPlus through BCVTB. The flow chart of this program is shown in Fig. 1. 3.3.3. The variable configuration file The variable configuration file is the key component used to specify the exchange variables. The file follows XML file format. An XML schema for the variable configuration file is defined as follows: An example of the variable configuration file is as follows. The attribute name and type together provide the name and type of the variables that EnergyPlus will export. The schedule, actuator and variable correspond to ExternalInterface:Schedule, ExternalInterface:Actuator and ExternalInterface:Variable respectively. The airflowrate and airflowtemp together will be used by the ExternalInterface:Airflow object in EnergyPlus to update the flow rates and flow temperature. The FlowPlus program will report the airflow rates and airflow temperature for the two elements respectively. The sequence of the exchange variable is organized as the same as the sequence of elements in the variable configuration file. Then EnergyPlus and FlowPlus will assemble and disassemble the exchanged data according to the variable configuration file.
104
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 2. The diagram of the coupled simulation implemented in BCVTB.
Fig. 3. The SketchUp model of the Building 661 and the surrounding buildings for the coupling study.
Fig. 4. The zone and surface definition of the Building 661 both in the EnergyPlus model and the Fluent CFD model.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
105
Fig. 5. The flow rates comparison between nodal airflow network model and coupled CFD model for windows on the south (a), north (b) and east (c) facade. Airflow rates comparisons for the clear story (d) and internal doors (e).
106
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 5. (continued).
In the above example variable configuration file, EnergyPlus will export the outdoor dry bulb temperature, the wind speed, and the inside surface temperature of surface int_Blg_661_z1_w. The FlowPlus program will take the exported value from EnergyPlus and report back the airflow rate for opening int_Blg_661_z2_ cc1_win1 and opening int_Blg_661_z2_cc2_win2 and the heat transfer coefficient for surface int_Blg_661_z1_w. 3.4. Coupling process The diagram of the coupling program is shown in Fig. 2. The BCVTB program will invoke the coupled simulation by calling both the EnergyPlus and FlowPlus. First, EnergyPlus will start the initialization process and conduct the simulation for the first time step. After EnergyPlus finishes computation for the first time step, EnergyPlus will send the variable values through BCVTB to FlowPlus and halt till it receives data back from BCVTB for next time step. On receiving the data FlowPlus will start the steady state CFD simulation and wait till a converged stage of the steady state CFD simulation is reached. The FlowPlus program will then extract the values for all the defined exchange variables, i.e., flow rates of openings, average temperature of the flow and surface heat transfer coefficients, and send the data to EnergyPlus through BCVTB. The above process finishes one cycle of data exchange. To continue, on receiving the data back from BCVTB, EnergyPlus will compute the next time step for the next cycle of data exchange and so on. 4. Case study The live retrofit building 661 on the Philadelphia Navy Yard was chosen for the case study of the coupling platform and methodology developed in the previous section. This historic building features a shared open space at the back with gross area of about 1600 m2, and a two story space in the front with gross area of about 800 m2 for each floor. After retrofit, the building will house EEBHub (Energy Efficient Building Hub) personnel, and function as living laboratories for developing tools and methods to transform the building industry’s current fragmented serial method into integrated team efforts. The building model is shown in Fig. 3. 4.1. Simulation scenario The simulation of natural ventilation was chosen as the case study on the coupling between the nodal model and the CFD model. Natural ventilation is considered to be most complicated in terms of airflow simulations, and it is a very important strategy for energy efficient buildings. The accurate simulation of natural ventilation
will provide architects and engineers with more insight during the design process of the energy efficient buildings. The weather condition in Philadelphia is first analyzed. The thermal comfort criterion from ASHRAE Fundamental [31] is used as criterion for natural ventilation deployable conditions. Then the hours where outdoor weather conditions satisfy this criterion are identified. The month of June has the highest number of hours that outdoor conditions are comfortable, and potentially natural ventilation can be deployed. Thus the month of June is chosen as the simulation period for the study of coupling between nodal model and the coupled model. 4.2. EnergyPlus model The geometry set up of building is shown in Fig. 3. The building at the center with pitched roofs and windows is Building 661. Three immediate neighboring buildings are also presented in the figure and modeled in the simulation model as shading surfaces. The zone and opening surface definitions are show in Fig. 4. The AirflowNetwork Model set up is as MultizoneWithoutDistribution. The Window/Door Opening Factor of all the windows and doors are set as 1. Other settings of the AirflowNetwork Model follow the example file of the “AirflowNetwork3Zvent”, which models the natural ventilation for a 3 zone building. 4.3. CFD model The CFD model has exactly the same geometry setup as that of the EnergyPlus Model. The mesh generation process for the selected application building was also carefully studied, and the mesh generation process and results were summarized in a paper [30]. The paper describe in detail the method to generate uniform all-hexahedral and adaptive hexahedral-dominate meshes. The generated meshes own good quality in terms of aspect ratio and squish index. CFD simulation results show that adaptive hexahedral-dominate meshes yield very similar results of air change rate in the space due to natural ventilation, compared to uniform all-hexahedral meshes yet with up to 90% reduction in the number of elements, hence significantly improve the computation efficiency. In the coupled study, the model adaptive hexahedral-dominate mesh with a domain size of and The generated mesh and one set of velocity filed is 1690 m 1345 m 185 m. The Standard ke3 turbulence model and standard wall function [28] is used for the CFD simulation. The boundary condition of the wind and building surface temperature is set dynamically by the FlowPlus program.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
5. Study on airflow rate under natural ventilation
simulations is shown in Fig. 6. The following observations were obtained:
The airflow rates calculation method by the nodal model in EnergyPlus, and the airflow rates calculation method by the coupled CFD model were described in the following sections. 5.1. Airflow rate computation method in nodal model The AirflowNetwork model in EnergyPlus consists of a set of nodes connected by airflow components through linkages. Each zone is a node in the system, and openings will be linkages between the nodes. The pressure and airflow is the relationship between the components (nodes, i.e., the rooms) of the system. The pressure and airflow calculations determine pressure at each node and airflow through each linkage given wind pressure and forced airflows, EnergyPlus [27]. Newton’s method is used to solve for node air pressures. Starting with an initial set of values for the node pressures, an iterative process will be carried out till the mass balance of each node is approaching to zero (less than a convergence value). 5.2. Airflow rate computation method in CFD model The CFD software Fluent deploys turbulence model with NaviereStokes equations to solve the flow field and the temperature profile in the simulation domain. The airflow rate of an opening is calculated as the summation of flow per unit area as shown in equation.
Q ¼
n X
vi $Ai
107
(2)
i¼0
where: Q e the volume flow rate of the opening, m3/s vi e the velocity of mesh element i on opening surface, m/s Ai e the area of mesh element i on opening surface, m2.
5.3. Airflow rates comparison The airflow rates of each of the openings computed from the airflow network model and the coupled CFD model are shown in Fig. 5, and air change rates for the 5 zones are shown in Fig. 7. The averaged percentage difference of airflow rate values between values generated with nodal air flow network and coupled
Airflow network generates same airflow rate for openings with same area, on the same facade of the same thermal zone. Airflow network model generally predicts smaller airflow rates for the openings. Airflow network model generates better airflow rates for openings on the south and east facade and internal openings, with difference ranging from 100% to 200% from the coupled model of about 2.5 m3/s to 6 m3/s. The difference between the airflow network model and the coupled CFD simulation for openings on the north facade and clear story openings are generally bigger (600% for a clear story opening from the coupled model of about 2 m3/s, and 600% for an opening on north facade from the coupled model of about 1 m3/s). The reason for the bigger difference on north facade is because that there is a very close adjacent building to the north of the building which makes the airflow pattern near the north facade more complicated, which is hard to capture with airflow network model. The percentage differences of airflow rate for openings on north facade range from 100% to 770% from the coupled model of about 1 m3/s. Due to the differences in the predictions of airflow rates of all the openings, the resulting air change rates of all the zones are quite different. Especially for zone Z2, which has clear story openings, the airflow network generates relatively higher airflow rates compared to the coupled simulation, therefore, the air change rate prediction from airflow network model of zone Z2 is higher than that computed from the coupled model. Air change rates predicted by nodal model for other zones are generally lower than that predicted by the coupled model, as a result of the differences in airflow rate calculation of the openings.
6. Study on heat transfer coefficient under natural ventilation The objective of this section is to study the differences in the heat transfer coefficients computed from the empirical methods compared to that computed from the coupled CFD simulations under natural ventilation conditions for the case study building of Building 661. EnergyPlus provides multiple algorithms on the computation of heat transfer coefficients for both interior surfaces and exterior surfaces. The algorithms were mainly based on empirical equations.
Fig. 6. The averaged percentage difference of airflow rate values during run period of June 1st to June 8th between values generated with nodal airflow network and coupled simulations.
108
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 7. The Air Change Rate (ACH) values for the 5 zones during run period of June 1st to June 7th between values generated with nodal airflow network and coupled simulations.
Additionally, in many cases multiple researchers have developed competing models for the same situations that differ and there is no way to determine if one is better than another [27]. 6.1. Heat transfer coefficient calculation in nodal model Heat transfer coefficient for internal surfaces of the building: The algorithms of heat transfer coefficients calculation for interior surfaces investigated in this study are: e Adaptive Convection Algorithm [27] Beausoleil-Morrison [10] developed a methodology for dynamically managing the selection of hc equations called adaptive convection algorithm. The algorithm is used to select among the available hc equations for the one that is most appropriate under the given thermal condition and air speed near the surface at a given time. As BeausoleilMorrison notes, the adaptive convection algorithm is intended to be expanded and altered to reflect different classification schemes and/or new hc equations. e TARP Algorithm [27] The comprehensive natural convection model, accessed using the keyword “TARP”, correlates the convective heat
transfer coefficient to the surface orientation and the difference between the surface and zone air temperature [32]. e Simple Natural Convection Algorithm [27] The simple convection model uses constant coefficients for different heat transfer configurations, using the same criteria as the detailed model to determine reduced and enhanced convection. The coefficients are also taken directly from Ref. [32]. Walton [32] derived his coefficients from the surface conductances for 3 ¼ 0.90 found in the ASHRAE Handbook (1985) in Table 1 on p. 23.2. The radiative heat transfer component was estimated at 1.02 0.9 ¼ 0.918 BTU/h ft2 F and then subtracted off. Finally the coefficients were converted to SI units to yield the values below. * For a vertical surface: h ¼ 3.076 *For a horizontal surface with reduced convection: h ¼ 0.948 * For a horizontal surface with enhanced convection: h ¼ 4.040 * For a tilted surface with reduced convection: h ¼ 2.281 * For a tilted surface with enhanced convection: h ¼ 3.870 Heat Transfer Coefficient for External Surfaces of The Building: Substantial research has gone into the formulation of models for estimating the exterior convection coefficient. Since the 1930’s
Fig. 8. The HTC values for surfaces in zone_1 from the coupled CFD results (solid lines) and from adaptive algorithm (dashed lines) during simulation period of June 1st to 7th.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
109
Fig. 9. The average HTC values during run period of June 1st to June 8th generated with simple, adaptive and TARP algorithm and results from the coupled CFD simulation.
110
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 9. (continued).
there have been many different methods published for calculating this coefficient, with much disparity between them [33,34]. More recently Palyvos [35] surveyed correlations cataloging some 91 different correlations into four categories based on functional form of the model equation. The algorithm is taken directly from Ref. [32]. Walton derived his algorithm from ASHRAE literature which can now be found for example in the ASHRAE Handbook [31], Table 5 on p. 3.12, which gives equations for natural convection heat transfer coefficients in the turbulent range for large, vertical plates and for large, horizontal plates facing upward when heated (or downward when cooled). A note in the text also gives an approximation for large, horizontal plates facing downward when heated (or upward when cooled) recommending that it should be half of the facing upward value. Walton adds a curve fit as a function of the
cosine of the tilt angle to provide intermediate values between vertical and horizontal. The curve fit values at the extremes match the ASHRAE values very well [27]. 6.2. Heat transfer coefficient calculation in coupled model In the coupled model, the CFD model will perform FVM calculation to get the detailed temperature profiles and velocity fields in the space. The heat transfer coefficients of the surface is proportional to the heat transfer rate between each surface and the air adjacent to it. The heat transfer coefficient is calculated by the ratio between the unit area heat flux and the temperature difference between the solid surface and the air temperature in the space.
(a)
(b)
Fig. 10. The percentage difference of HTC values during run period of June 1st to June 8th between values generated with simple, adaptive, TARP algorithm and from that generated from the coupled CFD simulations.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
111
Fig. 11. The three consisting parts of the thermal resistance of a wall: exterior convective resistance, conductive resistance and interior convective resistance.
(a)
(b)
Fig. 12. (a) The overall thermal resistance as results from simple, adaptive, TARP algorithm and the coupled CFD simulations. (b) The percentage difference of the overall thermal resistance from simple, adaptive and TARP algorithm from that calculated in the coupled CFD simulation.
112
R. Zhang et al. / Building and Environment 68 (2013) 100e113
6.3. Heat transfer coefficient comparison between nodal model and coupled model The heat transfer coefficients (HTC) calculated using the simple, adaptive and TARP methods in EnergyPlus are compared with that calculated from the coupled CFD model. Fig. 8 shows the HTCs calculated by the adaptive algorithm (dashed lines) and calculated by the coupled simulation (solid lines) for surfaces in zone_1 during simulation period of June 1st to 7th. As is shown, the difference between the adaptive algorithm and the coupled simulation is not negligible. The adaptive algorithm generate sometimes higher and sometimes lower HTCs compared to the coupled model. There is no general consistency in terms of the difference between the two methods. To provide better overview and understanding of the differences between the three algorithms in EnergyPlus and coupled simulation for all the surfaces, the average values of HTC for the surfaces during the simulation period from June 1st to June 8th are calculated and shown in Fig. 9. The percentage difference of HTC between simple, adaptive and TARP algorithm compared with that from the coupled CFD results are summarized in Fig. 10. The following observations were found: The simple algorithm generates consistently higher HTC values for the exterior surfaces. The averaged percentage difference of HTC values for the exterior surfaces between simple algorithm and the coupled CFD simulation is 188% from the coupled model of 25.05 W/m2 K. The adaptive algorithm also generates higher HTC values for the exterior surfaces comparing with that of the coupled CFD results. However, the difference is smaller than that from the simple algorithm. The averaged percentage difference of HTC values for the exterior surfaces between adaptive algorithm and coupled CFD results is 65% from the coupled model of 25.05 W/m2 K. The TARP algorithm generates closest HTC values to the coupled CFD results. The averaged percentage difference of HTC values for the exterior surfaces between TARP algorithm and coupled CFD results is 7% from the coupled model of 25.05 W/m2 K. All three algorithms in EnergyPlus generate similar HTCs for interior surfaces. The percentage difference of HTC values from algorithms in EnergyPlus and couple CFD simulations for the interior surfaces are between 10% and 80% from the coupled model of 1.31 W/m2 K. The overall thermal resistance will determine the heat transferred through the envelope and then influence the heating and cooling load calculations. Therefore, it is important to investigate the overall thermal resistance as results of different convective heat transfer coefficient calculation algorithms. The overall thermal resistance of wall consists of three parts: convective resistance to the exterior of the envelope, the conductive resistance of the envelope and the convective resistance to the interior of the envelope. The previous study shows that the empirical algorithms in EnergyPlus generate closer HTC values comparing to those from the coupled CFD simulations for interior surfaces, and produce much higher difference for exterior surfaces. However, the contribution of exterior convective resistance in overall thermal resistance is relatively small, thus the overall thermal resistance difference will be dominated by the interior convective HTC calculation. Fig. 11 shows the exterior convective resistance, wall conductive resistance and interior convective resistance for one of the walls. As is shown, the exterior convective resistance is about 2%, while the interior convective resistance is much larger at 20%. Therefore, the
difference of thermal resistance calculation between the three empirical algorithms will be negligible, since the differences in HTC calculation of the three algorithms are negligible. The overall heat transfer coefficients of all the enclosure surfaces are calculated and summarized in Fig. 12. As is shown, the overall thermal resistance generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. The average differences of overall thermal resistance from the empirical algorithms and the coupled CFD simulations range between 28% w 37% from the coupled model of 4.27 m2 K/w. 7. Conclusion A coupled simulation platform between nodal model and Finite Volume Method (FVM) model for advanced building thermal simulation was implemented and tested with a live retrofit building project at the Philadelphia Navy Yard. Heat Transfer Coefficients (HTC) were compared between the EnergyPlus empirical models and the coupled CFD model. It is found that the simple algorithm generates higher HTC values for the exterior surfaces. The averaged percentage difference of HTC values for the exterior surfaces between simple algorithm and coupled CFD simulation is 188% from the coupled model of 25.05 W/m2 K. The adaptive algorithm also generates higher HTC values for the exterior surfaces comparing with that of the coupled CFD results. However, the difference is smaller than that from the simple algorithm. The averaged percentage difference of HTC values for the exterior surfaces between adaptive algorithm and coupled CFD results is 65%. The TARP algorithm generated closest HTC values to the coupled CFD results. The averaged percentage difference of HTC values for the exterior surfaces between TARP algorithm and the coupled CFD results is 7%. All three algorithms in EnergyPlus generate similar HTCs for interior surfaces. The percentage difference of HTC values for the interior surfaces are between 10% and 80% from the coupled model of 1.31 W/m2 K. The overall heat transfer coefficients of all the enclosure surfaces are calculated and it is found that the overall thermal resistance generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. The average differences of overall thermal resistance between the empirical algorithms and coupled CFD simulations range from 28% w 37% from the coupled model of 4.27 m2 K/w. From the study of airflow rates through openings under natural ventilation, it is found that Airflow network model in EnergyPlus generates same airflow rate for openings with same area, on the same facade of the same thermal zone. Airflow network model generally predict smaller airflow rates for the openings. Airflow network model generates better airflow rates for openings on the south and east facade and internal openings, with difference ranging from 100% to 200% from the coupled model of about 2.5 m3/s to 6 m3/s. Due to a very close adjacent building to the north of the building, the difference between airflow network model and coupled CFD simulation for openings on the north facade are generally bigger, which is around 600% for openings on north facade from the coupled model of about 1 m3/s. The percentage difference of airflow rate for openings on north facade ranges from 100% to 770% from the coupled model of about 1 m3/s. The paper presents a modeling study of heat transfer and airflow study for natural ventilation. The primary objective of the work is to compare the relative accuracy of the AirFlowNetwork model in EnergyPlus and a coupled CFD-EnergyPlus model. Our premise, based on previous research work, is CFD will be more accurate than lumped nodal models. Thereby, this paper summarized findings on the differences in HTC and airflow calculation
R. Zhang et al. / Building and Environment 68 (2013) 100e113
between the AirFlowNetwork model and the coupled CFD model. Significant discrepancy has been identified for airflow simulations. Study on the overall thermal resistance shows the empirical method in nodal is acceptable. The method for the coupled CFD and nodal is flexible, configurable and easy. The simulation time for one run of CFD for the demonstration example of Building 661 is about 27 min, which is still relatively long. The paper presented a comprehensive comparative study between nodal model and coupled CFD model, which will serve as valuable reference for future similar research or practice. References [1] Feustel H, Rayner-Hooson A. COMIS fundamentals. Tech. Rep. LBNL-28560. Lawrence Berkeley National Laboratory; 1990. [2] Mora L, Gadgil A, Wurtz E. Comparing zonal and CFD model pre-dictions of isothermal indoor airflows to experimental data. Indoor Air 2003;13:77e85. [3] Zhai Z, Chen Q, Haves P, Klems JH. On approaches to couple energy simulation and computational fluid dynamics programs. Building and Environment 2002;37(8):857e64. [4] Negrao C. Integration of computational fluid dynamics with building thermal and mass flow simulation. Energy and Buildings 1998;27:155e65. [5] Wang L, Wong NH. Coupled simulations for naturally ventilated residential buildings. Automation in Construction 2008;17:386e98. [6] Nagai T, Kurabuchi T. A two-node heat and mass transfer network model for long-term simulation of void space of high-rise apartment houses equipped with gas-fired boilers. In: 11th international IBPSA conference 2009. p. 434e41. [7] Manz H, Frank T. Thermal simulation of buildings with double-skin facades. Energy and Buildings 2005;37:1114e21. [8] Wong NH, Wang L, Aida NC, Pandey AR, Wei X. Effects of double glazed facade on energy consumption, thermal comfort and condensation for typical office building in Singapore. Energy and Buildings 2005;37:563e72. [9] Negrao C. Conation of computational fluid dynamics and building thermal simulation. Ph.D. thesis. University of Strathclyde; 1995. [10] Beausoleil-Morrison I. The adaptive coupling of heat and air flow modelling within dynamic whole-building simulation. Ph.D. thesis. University of Strathclyde; 2000. [11] Beausoleil-Morrison I, Clarke J, Denev J, Macdonald I, Melikov A, Stankov P. Further developments in the conation of CFD and building simulation 2001. p. 1267e74. [12] Beausoleil-Morrison I, Clarke J. The implications of using the standard ke3 turbulence model to simulate room air flows which are not fully turbulent. In: Proceeding of ROOMVENT 1998. p. 99e106. [13] Bartak M, Beausoleil-Morrison I, Clark J, Denev J, Drkal F, Lain M, et al. Integrating CFD and building simulation. Building and Environment 2002;37:865e71. [14] Djunaedy E, Hensen J, Loomans M. Toward external coupling of building energy and airflow modeling programs. ASHRAE Transactions 2003;109(2): 771e87.
113
[15] Djunaedy E, Hensen J, Loomans M. Selecting an appropriate tool for airflow simulation in buildings. Building Services Engineering Research and Technology 2004;25(3):289e98. [16] Chen Q, Peng X, van Passen A. Prediction of room thermal response by CFD technique with conjugate heat transfer and radiation models. ASHRAE Transactions 1995;101:50e60. [17] Zhai Z, Chen Q, Klems J, Haves P. Strategies for coupling energy simulations and computational fluid dynamics programs. In: Proceedings of 7th international IBPSA conference 2001. p. 59e66. [18] Zhai Z, Chen Q. Solution characters of iterative coupling between energy simulation and CFD programs. Energy and Buildings 2003;35(5):493e505. [19] Zhai Z, Gao Y, Chen Q. Pressure boundary conditions in multi-zone and CFD program coupling. In: 1st IBPSA-USA conference. [20] Zhai Z. Numerical determination and treatment of convective heat transfer coefficient in the coupled building energy and CFD simulation. Building and Environment 2004;36(8):1000e9. [21] Zhai Z. Performance of coupled building energy and CFD simulations. Energy and Buildings 2005;37(4):333e43. [22] Zhai Z. Sensitivity analysis and application guides for integrated building energy and CFD simulation. Energy and Buildings 2006;38(9):1060e8. [23] Wang L. Coupling of multizone and CFD programs for building airflow distribution and contaminant transport simulations. Ph.D. thesis. Purdue University; 2007. [24] Wang LZ, Chen QY. On solution characteristics of coupling of multizone and CFD programs in building air distribution simulation. In: Proceeding of 9th international IBPSA conference 2005. p. 1315e22. [25] Wang L, Wong NH. Coupled simulations for naturally ventilated rooms between building simulation (BS) and computational fluid dynamics (CFD) for better prediction of indoor thermal environment. Building and Environment 2009;44:95e112. [26] Wetter M, Nouidui TS, Haves P, Zhang R, Li Z. Building controls virtual test bed interface 2011. [27] EnergyPlus. EnergyPlus documentation 2011. [28] A. Inc.. Ansys fluent documentation 2011. [29] Djunaedy E, Hensen J, Loomans M. Comparing internal and external run-time coupling of CFD and building energy simulation software. In: Proceeding of 9th international conference on airflow in rooms 2004. p. 393e6. [30] Zhang R, Zhang Y, Lam KP. Conformal adaptive hexahedral-dominant mesh generation for CFD simulation in architectural design applications. In: Proceedings of 2011 winter simulation conference 2011. p. 928e42. [31] ASHRAE. ASHRAE fundamental handbook 2004. [32] Walton G. Thermal analysis research program reference mannual. Tech. Rep. NBSSIR 83e2655. National Bureau of Standards (now National Institute of Standards and Technology); 1983. [33] Cole R, Sturrock N. The convective heat exchange at the external surface of buildings. Building and Environment 1977;12:207. [34] Yazdanian M, Klems J. Measurement of the exterior convective film coefficient for windows in low-rise buildings. ASHRAE Transactions 1994;100(1): 1087. [35] Palyvos J. A survey of wind convection coefficient correlations for building envelope energy systems’ modeling. Applied Thermal Engineering 2008;28(8e9):801e8.
Contents lists available at SciVerse ScienceDirect
Building and Environment journal homepage: www.elsevier.com/locate/buildenv
Coupled EnergyPlus and computational fluid dynamics simulation for natural ventilation Rui Zhang a, *, Khee Poh Lam b, Shi-chune Yao b, Yongjie Zhang b a b
IBM, T.J. Watson Research Center, United States Carnegie Mellon University, United States
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 December 2012 Received in revised form 27 March 2013 Accepted 2 April 2013
Energy modeling approaches have continued to advance to cater for emerging new design concepts toward “greener” solutions that optimize energy consumption in buildings while maintaining thermal comfort as well as healthy environment. Increasing attention is given to passive and mix-mode systems in building. Computational Fluid Dynamics (CFD) model has been widely adopted as effective tool for natural ventilation simulations. However, CFD become unstable for conjugate heat transfer model, which is the transient heat transfer between solid and fluid. Solid and fluid has different respond times to thermal energy. Typically walls respond in hours, and air responds in seconds, causing the system to become stiff. In addressing the issue, a coupled lumped heat transfer model (EnergyPlus) and CFD model (Fluent) was implemented, and 8 days of simulation was conducted. The airflow rates of openings and heat transfer coefficients from the airflow network module in EnergyPlus and those from CFD model were compared. Results show that airflow network model generally predict smaller airflow rates for the openings. Airflow network model generates better results for openings on the south and east facade and internal openings, where there is no immediate adjacent building. Among the heat transfer coefficients calculation methods in EnergyPlus, the TARP algorithm generated closest HTC values to the coupled CFD results. The overall heat transfer coefficients of all the enclosure surfaces are calculated and it is found that the overall thermal resistances generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. Ó 2013 Elsevier Ltd. All rights reserved.
Keywords: CFD Energy simulation EnergyPlus Natural ventilation Air change rates Heat transfer coefficient
1. Introduction Energy modeling approaches have continued to advance to cater for emerging new design concepts toward “greener” solutions that optimize energy consumption in buildings while maintaining thermal comfort as well as healthy environment. Instead of the conventional approaches that rely solely on mechanical system to provide the desired thermal conditions, increasing attention is given to passive and mix-mode systems in building. Notwithstanding the advances in system modeling, there are still limitations in the capability of representing the “real” environmental behavior with different building spatial configurations. For example, most current energy models adopt a “nodal” approach to simulate the heat transfer process in the building, with simplified heat resistors and capacitors network. The underlying assumption for nodal model is the uniformity of room temperature, which may * Corresponding author. IBM, T.J. Watson Research Center, Business Analytic and Mathmetical Science, 1101 Kitchawan Rd., Room 32-244, Yorktown Heighs, NY 10598, United States. Tel.: þ1 4126517384. E-mail addresses: [email protected], [email protected] (R. Zhang). 0360-1323/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.buildenv.2013.04.002
apply to most of the building zones with moderate room size. However, limitations also come with such nodal approach: Priori and empirical knowledge of various coefficients are needed for model input, such as wind pressure coefficient, heat transfer coefficient, loss factors, friction factors, etc. Effects of thermal and air flow patterns introduced by building spatial configurations are difficult to be represented into the model. More detailed zonal model, such as COMIS [1], has been developed for ventilation design in complex buildings. However, research [2] shows that the results of zonal model are not satisfiable compared with even coarse-grid CFD models under isothermal conditions. Finite Volume Method (FVM) such as the Computational Fluid Dynamics (CFD) on the contrary will perform detailed computation on the heat transfer and air flow simulation, which could supplement the nodal model for the building energy simulation. Detailed temperature profile and air flow field are calculated with first
R. Zhang et al. / Building and Environment 68 (2013) 100e113
principle based NaviereStokes set of equations and turbulence models. Coefficients, such as the heat transfer coefficients, will be the results of the simulation, which is defined by a set of boundary conditions. However, limitation also comes with CFD. Firstly, the computation resources needed for CFD simulation are much more expensive than the nodal model. Secondly, the coupling between fluid and solid material is difficult and computational more expensive, thus the thermal storage capacity of building component is difficult to be correctly modeled. Therefore, CFD has been limited to detailed studies of given space under a number of specific boundary conditions. However, with the rapid development of computation power, the computation time of CFD is decreasing, it is highly necessary to develop an integrated thermal simulation that combines the advantages of both the nodal and FVM model, providing higher accuracy within acceptable computation time. The nodal model calculation of the heat transfer process in the building is based on the fundamental heat balance principles. Each zone is assumed with uniform state variables of temperature, pressure and density. Empirical coefficients are used to calculate the state variable transfer fluxes between each node. One example of how such nodal system is implemented is shown as following:
Cz
NX Nsl NX surfaces zones X dTz _ i Cp ðTzi -Tz Þ m hi Ai ðTsi -Tz Þ þ Q_ i þ ¼ dt i¼1 i¼1 i¼1
_ inf Cp ðTN -Tz Þ þ Q_ sys þm
(1)
where: CzdTz/dt e room air energy change rate, PNsl _ Q e sum of the convective heat transferred through ini¼1 i ternal heat sources or sinks, PNsurfaces hi Ai ðTsi -Tz Þ e sum of the convective heat transferred i¼1 through building envelope, PNzones _ i Cp ðTzi -Tz Þ e energy from neighbor zones air mixing, m i¼1 _ inf Cp ðTN Tz Þ e total energy of infiltration air, m Q_ sys e system output, In equation Eq. 1, the heat transfer coefficient needs to be determined for a given set of boundary conditions that define the thermal and fluid fields in the space. Various empirical correlations have been developed to calculate the heat transfer coefficient, which are dependent on the ambient air flow characteristics. However, the lack of knowledge of the air flow, will lead to considerable errors in the heat transfer prediction of the nodal model for buildings, where the heat transfer process is complex by nature. Research [3] shows that differences in heat transfer coefficient between CFD and empirical equation can range from 1.42 W/ m2 K to 111.41 W/m2 K. Mora et al. [2] investigated the effectiveness of zonal model in describing the airflows in large indoor space. A coarse-grid CFD simulation was also carried out. Velocity and pressure computation from different formulations of zonal models and coarse-grid ke3 CFD model were compared to measurements in a 2D isothermal room. The results showed that zonal model computation was not satisfactory compared with coarse-grid CFD model (grid of 10 10). Negrao [4] studied the heating process in a room with a thermostat and a heating radiator both in a nodal model and a CFD-nodal model. The ESP-r program, which is a nodal model for building thermal computation, was used to compute the required energy generation by the radiator, based on the calculated room air temperature and thermostat setting. Significant difference in the resulting heat generation requirement and the room air temperature had been observed between the two models. Wang and Wong [5] studied the performance of naturally ventilated residential buildings in Singapore. A zonal model for building
101
thermal performance and airflow network was used, and a manual coupled approach between the zonal model and CFD was also developed. Field measurement was carried out in a real building in Singapore. Compared with measurements, the discrepancy of indoor air velocity and dry bulb temperature predicted by zonal model was larger than that predicted by the coupled model. Nagai and Kurabuchi [6] constructed a nodal model for a high-rise apartment building with central void through the whole height of the building in Japan. The simulation results of the air temperature in the void space and the surface temperatures of the corridors were compared with measured data. Results generated by the nodal were generally lower than the measured data. 2. Previous works Manual coupling of the CFD model and nodal model has been done for many geometrically complicated buildings. Nagai and Kurabuchi [6] used CFD to decide the coefficients in the nodal model for a high-apartment building with central void space throughout the height of the building in Japan. Manz and Frank [7] used a one-way-static coupling to study the thermal performance of double facade buildings. Wong et al. [8] also coupled CFD and nodal model to study the performance of double facade building in a tropical climate in Singapore. Research work on automated coupling of the nodal model and CFD model at run time is dated from 1990s. Negrao [9] implemented a full iterative coupling approach of the nodal modal and CFD model. Selected space was simulated with CFD, and ESP-r was used as the nodal model. The coupled variables were: surface temperature of all the walls and windows, and the pressures at the internal openings that connect the space with the rest of the building. A full iterative strategy has been used, where coupled variables will be exchanged at each iterative step until a convergence criterion was reached at each time step. It was found that low under-relaxation (0.1) was necessary in order for the CFD computation to converge for the coupled system. The number of iteration of the CFD computation was increased compared with the uncoupled CFD computation. Based on the work by Negrao [9,4], Beausoleil-Morrison [10], Beausoleil-Morrison et al. [11], Beausoleil-Morrison and Clarke [12] continued with the investigation of the coupling between nodal model (ESP-r) and CFD model. A conflation controller was implemented to configure the CFD model at each time step. At the start of each time step, the zero-equation turbulence model was employed, and the resulting viscosity field was used to initialize the ke3 turbulence model. Grashof Number (the ratio of buoyancy and viscous forces) and Reynolds Number (the ratio of inertial and viscous forces) were evaluated. The existence of the buoyancy term in zmomentum equation was determined. The conflation controller would then choose the appropriate turbulence model and near wall model. An empirical validation of the coupled model by BeausoleilMorrison [10] is conducted by Bartak et al. [13]. Two CFD models with coarse-grid and fine-grid were constructed and compared with experimental results. It was found that the coarse-grid and fine-grid CFD produced comparable results. Variables validated were air temperature, comfort parameters, and local mean age of air under isothermal and non-isothermal conditions. Good agreements were found between the predicted and measured data. Djunaedy et al. [14], Djunaedy et al. [15], Chen et al. [16] and Zhai et al. [17] discussed both the advantages and disadvantages of internal coupling of the CFD and nodal model. It was concluded that external coupling is more favorable in terms of no stiffness issue, less expensive, nodal and CFD model themselves are independent and can be optimized individually. Zhai and Chen [3,18e22] also investigate the coupling strategy extensively. They implemented a coupling strategy to exchange heat
102
R. Zhang et al. / Building and Environment 68 (2013) 100e113
transfer coefficient and surface boundary conditions between nodal model and CFD model through dynamic or static coupling. Hereby, static and dynamic is the strategy used for data exchange instead of thermal process. A sensitivity study was also carried out to investigate the effectiveness of the coupling method in improving the building energy simulations. Their results showed that for rooms with moderated size, without sensible temperature stratification, the coupling approach shows marginal effect. However for rooms with large temperature stratification, the discrepancy between coupled method and nodal model is significant (42%). Wang [23], Wang and Chen [24] proved theoretically that the coupled model has a solution and it is unique. The investigation used Scarborough criterion to evaluate convergence performances and analyzed the stabilities of three coupling strategies, which are pressure coupling, pressure and flow rate mixed coupling, and flow rate coupling. It was found that the pressure coupling performs best. Wang and Wong [5] developed a text-based interface for automated coupling program to extract and exchange information between EPS-r (nodal model) and Fluent (CFD model). External coupling was used and results showed that coupled program yields better results compared with single nodal or CFD model. Wang and Wong [25] compared velocity inlet coupled and pressure inlet coupled strategy, and found that pressure inlet coupled strategy gave better results. In this paper, we will present a novel approach and platform for the coupling between nodal model, i.e., EnergyPlus and CFD, i.e., Fluent, via open source platform of BCVTB (Building Controls Virtual Test Bed [26]). The coupling platform is configurable, and flexible. One demonstration example upon the coupling platform is described in detail. Two key parameters from the nodal model and coupled CFD model will be compared. One is the heat transfer coefficients computed from the coupled CFD model and various heat transfer coefficient calculation methods in EnergyPlus. The other is the air flow rate through openings from coupled CFD model and Airflow Network model in EnergyPlus.
model and CFD model can be satisfactorily achieved by maintaining each method’s solution algorithm separately. Furthermore, research [29] found that difference in simulation results between internal coupling and external coupling is not significant. However, the benefits of the external coupling are three folds: 1. computationally less expensive; 2. nodal and CFD model can be maintained and updated individually; 3. configurable exchange variables. This study will implement the Progressive-Replacement external coupling strategy. The exchange variables will be exchanged after each model, i.e., the nodal and the CFD model, comes to its converged state at each time step.
3. Coupling method EnergyPlus [27] is chosen as the nodal model, and Fluent software [28] is chosen as the FVM model for the coupled simulation platform. The coupling method and platform are described in detail in the following sections: 3.1. Coupling strategy The coupling approaches between the nodal and FVM model can be characterized into three types: Full internal coupling, where the set of equations for nodal model and CFD model are solved together iteratively. Iterative external coupling, where the set of equations for nodal model and CFD model are solved individually, the variables are exchanged with an iterative procedure until a converged state is achieved. Progressive-Replacement external coupling, where the set of variables are exchanged after each model comes to converged state at each timestep. Research [4] shows that the fully internal coupling and iterative external coupling will require larger number of iterations to reach convergence, e.g., approximately 600 iterations were necessary for a simulation which required approximately 200 iterations for an individual CFD simulation. It is also found that the convergence of the iterative approach is difficult even for simple room simulation. One of the conclusions from the above study is that the nodal
Fig. 1. The flow chart of the FlowPlus program.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
3.2. Coupling variables As previously mentioned, nodal model requires various building specific heat and mass transfer coefficients, which can be obtained from CFD simulation. At the same time, the CFD tool may have stiffness problem in solving transient conjugate heat transfer with conduction and convection between the building envelope components and air in the space. Based on the characteristics of the two models, a collection of coupling variables was selected. The nodal model will provide to the CFD tool all the interior and exterior surface temperatures of the envelope components of the building and outdoor weather conditions. The CFD tool will take temperatures of various building surfaces as settings of the thermal boundary conditions, including both interior surfaces and exterior surfaces. For natural ventilation, wind is the driving force of the flow through the openings in the building. The wind conditions, in terms of speed and direction, provided by the nodal tool from weather data will be used to determine the boundary type of the four boundary surfaces of the simulation domain. For example, if the wind is coming from south, with a speed of 2 m/s, the surface of the south bound will be set as velocity inlet, with incoming wind velocity of 2 m/s. The north bounding surface of the domain will be set as outflow. The east and west boundaries will be set as symmetry assuming there is no shear strain on the surface. The CFD tool will conduct the steady state natural ventilation simulation. A post processing program will be implemented to extract the temperature profile and velocity fields from CFD tool and then to calculate the values for the exchange variables, which include the following: Airflow rates through all the openings. Average air temperatures through the openings. Surface heat transfer coefficients for all the envelop surfaces.
3.3. Coupling platform The coupling platform will be based on Building Controls Virtual Test Bed (BCVTB Wetter et al. [26]), which is a software environment targeted to provide an integration platform for various simulation tools. For example, the BCVTB enables concurrent energy simulation of whole building in EnergyPlus with HVAC system and operating control in cþþ program, or MATLAB/Simulink, while exchanging data between the programs at each time step. 3.3.1. EnergyPlus object in support of the coupling In this study a new EnergyPlus input object called the “ExternalInterface:Airflow” is added for the purpose of coupling between nodal and CFD model for air flow simulations. The entry of the objects are designed as follows:
103
The ExternalInterface:Airflow object is designed to set the airflow rate for each piece of the openings in the building. The input entry of ExternalInterface:Airflow will specify the name of the opening, and zone name that the air is flowing in (positive value) or flowing out from (negative value). The value of heat transfer coefficient will be updated through the ExternalInterface:Schedule object. 3.3.2. FlowPlus program for executing CFD simulation and extracting coupling variables A cþþ program, called FlowPlus, is developed to execute the Fluent software to conduct the CFD simulation and extract the results for the coupling variables. The program will read in the variable configuration file, which specifies the exchange variables that EnergyPlus is sending and also the variables that are needed to be extracted from Fluent simulation results. Then the program will use the values obtained from EnergyPlus to generate a Fluent journal file, which will set the boundary conditions for the CFD simulation. Then Fluent software will be called to execute the CFD simulation according to the boundary conditions that are specified in the journal file. After Fluent finished executing the iteration, the FlowPlus will extract the temperature profile and velocity fields from Fluent and calculate the values for the exchange variables and send the values to EnergyPlus through BCVTB. The flow chart of this program is shown in Fig. 1. 3.3.3. The variable configuration file The variable configuration file is the key component used to specify the exchange variables. The file follows XML file format. An XML schema for the variable configuration file is defined as follows: An example of the variable configuration file is as follows. The attribute name and type together provide the name and type of the variables that EnergyPlus will export. The schedule, actuator and variable correspond to ExternalInterface:Schedule, ExternalInterface:Actuator and ExternalInterface:Variable respectively. The airflowrate and airflowtemp together will be used by the ExternalInterface:Airflow object in EnergyPlus to update the flow rates and flow temperature. The FlowPlus program will report the airflow rates and airflow temperature for the two elements respectively. The sequence of the exchange variable is organized as the same as the sequence of elements in the variable configuration file. Then EnergyPlus and FlowPlus will assemble and disassemble the exchanged data according to the variable configuration file.
104
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 2. The diagram of the coupled simulation implemented in BCVTB.
Fig. 3. The SketchUp model of the Building 661 and the surrounding buildings for the coupling study.
Fig. 4. The zone and surface definition of the Building 661 both in the EnergyPlus model and the Fluent CFD model.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
105
Fig. 5. The flow rates comparison between nodal airflow network model and coupled CFD model for windows on the south (a), north (b) and east (c) facade. Airflow rates comparisons for the clear story (d) and internal doors (e).
106
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 5. (continued).
In the above example variable configuration file, EnergyPlus will export the outdoor dry bulb temperature, the wind speed, and the inside surface temperature of surface int_Blg_661_z1_w. The FlowPlus program will take the exported value from EnergyPlus and report back the airflow rate for opening int_Blg_661_z2_ cc1_win1 and opening int_Blg_661_z2_cc2_win2 and the heat transfer coefficient for surface int_Blg_661_z1_w. 3.4. Coupling process The diagram of the coupling program is shown in Fig. 2. The BCVTB program will invoke the coupled simulation by calling both the EnergyPlus and FlowPlus. First, EnergyPlus will start the initialization process and conduct the simulation for the first time step. After EnergyPlus finishes computation for the first time step, EnergyPlus will send the variable values through BCVTB to FlowPlus and halt till it receives data back from BCVTB for next time step. On receiving the data FlowPlus will start the steady state CFD simulation and wait till a converged stage of the steady state CFD simulation is reached. The FlowPlus program will then extract the values for all the defined exchange variables, i.e., flow rates of openings, average temperature of the flow and surface heat transfer coefficients, and send the data to EnergyPlus through BCVTB. The above process finishes one cycle of data exchange. To continue, on receiving the data back from BCVTB, EnergyPlus will compute the next time step for the next cycle of data exchange and so on. 4. Case study The live retrofit building 661 on the Philadelphia Navy Yard was chosen for the case study of the coupling platform and methodology developed in the previous section. This historic building features a shared open space at the back with gross area of about 1600 m2, and a two story space in the front with gross area of about 800 m2 for each floor. After retrofit, the building will house EEBHub (Energy Efficient Building Hub) personnel, and function as living laboratories for developing tools and methods to transform the building industry’s current fragmented serial method into integrated team efforts. The building model is shown in Fig. 3. 4.1. Simulation scenario The simulation of natural ventilation was chosen as the case study on the coupling between the nodal model and the CFD model. Natural ventilation is considered to be most complicated in terms of airflow simulations, and it is a very important strategy for energy efficient buildings. The accurate simulation of natural ventilation
will provide architects and engineers with more insight during the design process of the energy efficient buildings. The weather condition in Philadelphia is first analyzed. The thermal comfort criterion from ASHRAE Fundamental [31] is used as criterion for natural ventilation deployable conditions. Then the hours where outdoor weather conditions satisfy this criterion are identified. The month of June has the highest number of hours that outdoor conditions are comfortable, and potentially natural ventilation can be deployed. Thus the month of June is chosen as the simulation period for the study of coupling between nodal model and the coupled model. 4.2. EnergyPlus model The geometry set up of building is shown in Fig. 3. The building at the center with pitched roofs and windows is Building 661. Three immediate neighboring buildings are also presented in the figure and modeled in the simulation model as shading surfaces. The zone and opening surface definitions are show in Fig. 4. The AirflowNetwork Model set up is as MultizoneWithoutDistribution. The Window/Door Opening Factor of all the windows and doors are set as 1. Other settings of the AirflowNetwork Model follow the example file of the “AirflowNetwork3Zvent”, which models the natural ventilation for a 3 zone building. 4.3. CFD model The CFD model has exactly the same geometry setup as that of the EnergyPlus Model. The mesh generation process for the selected application building was also carefully studied, and the mesh generation process and results were summarized in a paper [30]. The paper describe in detail the method to generate uniform all-hexahedral and adaptive hexahedral-dominate meshes. The generated meshes own good quality in terms of aspect ratio and squish index. CFD simulation results show that adaptive hexahedral-dominate meshes yield very similar results of air change rate in the space due to natural ventilation, compared to uniform all-hexahedral meshes yet with up to 90% reduction in the number of elements, hence significantly improve the computation efficiency. In the coupled study, the model adaptive hexahedral-dominate mesh with a domain size of and The generated mesh and one set of velocity filed is 1690 m 1345 m 185 m. The Standard ke3 turbulence model and standard wall function [28] is used for the CFD simulation. The boundary condition of the wind and building surface temperature is set dynamically by the FlowPlus program.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
5. Study on airflow rate under natural ventilation
simulations is shown in Fig. 6. The following observations were obtained:
The airflow rates calculation method by the nodal model in EnergyPlus, and the airflow rates calculation method by the coupled CFD model were described in the following sections. 5.1. Airflow rate computation method in nodal model The AirflowNetwork model in EnergyPlus consists of a set of nodes connected by airflow components through linkages. Each zone is a node in the system, and openings will be linkages between the nodes. The pressure and airflow is the relationship between the components (nodes, i.e., the rooms) of the system. The pressure and airflow calculations determine pressure at each node and airflow through each linkage given wind pressure and forced airflows, EnergyPlus [27]. Newton’s method is used to solve for node air pressures. Starting with an initial set of values for the node pressures, an iterative process will be carried out till the mass balance of each node is approaching to zero (less than a convergence value). 5.2. Airflow rate computation method in CFD model The CFD software Fluent deploys turbulence model with NaviereStokes equations to solve the flow field and the temperature profile in the simulation domain. The airflow rate of an opening is calculated as the summation of flow per unit area as shown in equation.
Q ¼
n X
vi $Ai
107
(2)
i¼0
where: Q e the volume flow rate of the opening, m3/s vi e the velocity of mesh element i on opening surface, m/s Ai e the area of mesh element i on opening surface, m2.
5.3. Airflow rates comparison The airflow rates of each of the openings computed from the airflow network model and the coupled CFD model are shown in Fig. 5, and air change rates for the 5 zones are shown in Fig. 7. The averaged percentage difference of airflow rate values between values generated with nodal air flow network and coupled
Airflow network generates same airflow rate for openings with same area, on the same facade of the same thermal zone. Airflow network model generally predicts smaller airflow rates for the openings. Airflow network model generates better airflow rates for openings on the south and east facade and internal openings, with difference ranging from 100% to 200% from the coupled model of about 2.5 m3/s to 6 m3/s. The difference between the airflow network model and the coupled CFD simulation for openings on the north facade and clear story openings are generally bigger (600% for a clear story opening from the coupled model of about 2 m3/s, and 600% for an opening on north facade from the coupled model of about 1 m3/s). The reason for the bigger difference on north facade is because that there is a very close adjacent building to the north of the building which makes the airflow pattern near the north facade more complicated, which is hard to capture with airflow network model. The percentage differences of airflow rate for openings on north facade range from 100% to 770% from the coupled model of about 1 m3/s. Due to the differences in the predictions of airflow rates of all the openings, the resulting air change rates of all the zones are quite different. Especially for zone Z2, which has clear story openings, the airflow network generates relatively higher airflow rates compared to the coupled simulation, therefore, the air change rate prediction from airflow network model of zone Z2 is higher than that computed from the coupled model. Air change rates predicted by nodal model for other zones are generally lower than that predicted by the coupled model, as a result of the differences in airflow rate calculation of the openings.
6. Study on heat transfer coefficient under natural ventilation The objective of this section is to study the differences in the heat transfer coefficients computed from the empirical methods compared to that computed from the coupled CFD simulations under natural ventilation conditions for the case study building of Building 661. EnergyPlus provides multiple algorithms on the computation of heat transfer coefficients for both interior surfaces and exterior surfaces. The algorithms were mainly based on empirical equations.
Fig. 6. The averaged percentage difference of airflow rate values during run period of June 1st to June 8th between values generated with nodal airflow network and coupled simulations.
108
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 7. The Air Change Rate (ACH) values for the 5 zones during run period of June 1st to June 7th between values generated with nodal airflow network and coupled simulations.
Additionally, in many cases multiple researchers have developed competing models for the same situations that differ and there is no way to determine if one is better than another [27]. 6.1. Heat transfer coefficient calculation in nodal model Heat transfer coefficient for internal surfaces of the building: The algorithms of heat transfer coefficients calculation for interior surfaces investigated in this study are: e Adaptive Convection Algorithm [27] Beausoleil-Morrison [10] developed a methodology for dynamically managing the selection of hc equations called adaptive convection algorithm. The algorithm is used to select among the available hc equations for the one that is most appropriate under the given thermal condition and air speed near the surface at a given time. As BeausoleilMorrison notes, the adaptive convection algorithm is intended to be expanded and altered to reflect different classification schemes and/or new hc equations. e TARP Algorithm [27] The comprehensive natural convection model, accessed using the keyword “TARP”, correlates the convective heat
transfer coefficient to the surface orientation and the difference between the surface and zone air temperature [32]. e Simple Natural Convection Algorithm [27] The simple convection model uses constant coefficients for different heat transfer configurations, using the same criteria as the detailed model to determine reduced and enhanced convection. The coefficients are also taken directly from Ref. [32]. Walton [32] derived his coefficients from the surface conductances for 3 ¼ 0.90 found in the ASHRAE Handbook (1985) in Table 1 on p. 23.2. The radiative heat transfer component was estimated at 1.02 0.9 ¼ 0.918 BTU/h ft2 F and then subtracted off. Finally the coefficients were converted to SI units to yield the values below. * For a vertical surface: h ¼ 3.076 *For a horizontal surface with reduced convection: h ¼ 0.948 * For a horizontal surface with enhanced convection: h ¼ 4.040 * For a tilted surface with reduced convection: h ¼ 2.281 * For a tilted surface with enhanced convection: h ¼ 3.870 Heat Transfer Coefficient for External Surfaces of The Building: Substantial research has gone into the formulation of models for estimating the exterior convection coefficient. Since the 1930’s
Fig. 8. The HTC values for surfaces in zone_1 from the coupled CFD results (solid lines) and from adaptive algorithm (dashed lines) during simulation period of June 1st to 7th.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
109
Fig. 9. The average HTC values during run period of June 1st to June 8th generated with simple, adaptive and TARP algorithm and results from the coupled CFD simulation.
110
R. Zhang et al. / Building and Environment 68 (2013) 100e113
Fig. 9. (continued).
there have been many different methods published for calculating this coefficient, with much disparity between them [33,34]. More recently Palyvos [35] surveyed correlations cataloging some 91 different correlations into four categories based on functional form of the model equation. The algorithm is taken directly from Ref. [32]. Walton derived his algorithm from ASHRAE literature which can now be found for example in the ASHRAE Handbook [31], Table 5 on p. 3.12, which gives equations for natural convection heat transfer coefficients in the turbulent range for large, vertical plates and for large, horizontal plates facing upward when heated (or downward when cooled). A note in the text also gives an approximation for large, horizontal plates facing downward when heated (or upward when cooled) recommending that it should be half of the facing upward value. Walton adds a curve fit as a function of the
cosine of the tilt angle to provide intermediate values between vertical and horizontal. The curve fit values at the extremes match the ASHRAE values very well [27]. 6.2. Heat transfer coefficient calculation in coupled model In the coupled model, the CFD model will perform FVM calculation to get the detailed temperature profiles and velocity fields in the space. The heat transfer coefficients of the surface is proportional to the heat transfer rate between each surface and the air adjacent to it. The heat transfer coefficient is calculated by the ratio between the unit area heat flux and the temperature difference between the solid surface and the air temperature in the space.
(a)
(b)
Fig. 10. The percentage difference of HTC values during run period of June 1st to June 8th between values generated with simple, adaptive, TARP algorithm and from that generated from the coupled CFD simulations.
R. Zhang et al. / Building and Environment 68 (2013) 100e113
111
Fig. 11. The three consisting parts of the thermal resistance of a wall: exterior convective resistance, conductive resistance and interior convective resistance.
(a)
(b)
Fig. 12. (a) The overall thermal resistance as results from simple, adaptive, TARP algorithm and the coupled CFD simulations. (b) The percentage difference of the overall thermal resistance from simple, adaptive and TARP algorithm from that calculated in the coupled CFD simulation.
112
R. Zhang et al. / Building and Environment 68 (2013) 100e113
6.3. Heat transfer coefficient comparison between nodal model and coupled model The heat transfer coefficients (HTC) calculated using the simple, adaptive and TARP methods in EnergyPlus are compared with that calculated from the coupled CFD model. Fig. 8 shows the HTCs calculated by the adaptive algorithm (dashed lines) and calculated by the coupled simulation (solid lines) for surfaces in zone_1 during simulation period of June 1st to 7th. As is shown, the difference between the adaptive algorithm and the coupled simulation is not negligible. The adaptive algorithm generate sometimes higher and sometimes lower HTCs compared to the coupled model. There is no general consistency in terms of the difference between the two methods. To provide better overview and understanding of the differences between the three algorithms in EnergyPlus and coupled simulation for all the surfaces, the average values of HTC for the surfaces during the simulation period from June 1st to June 8th are calculated and shown in Fig. 9. The percentage difference of HTC between simple, adaptive and TARP algorithm compared with that from the coupled CFD results are summarized in Fig. 10. The following observations were found: The simple algorithm generates consistently higher HTC values for the exterior surfaces. The averaged percentage difference of HTC values for the exterior surfaces between simple algorithm and the coupled CFD simulation is 188% from the coupled model of 25.05 W/m2 K. The adaptive algorithm also generates higher HTC values for the exterior surfaces comparing with that of the coupled CFD results. However, the difference is smaller than that from the simple algorithm. The averaged percentage difference of HTC values for the exterior surfaces between adaptive algorithm and coupled CFD results is 65% from the coupled model of 25.05 W/m2 K. The TARP algorithm generates closest HTC values to the coupled CFD results. The averaged percentage difference of HTC values for the exterior surfaces between TARP algorithm and coupled CFD results is 7% from the coupled model of 25.05 W/m2 K. All three algorithms in EnergyPlus generate similar HTCs for interior surfaces. The percentage difference of HTC values from algorithms in EnergyPlus and couple CFD simulations for the interior surfaces are between 10% and 80% from the coupled model of 1.31 W/m2 K. The overall thermal resistance will determine the heat transferred through the envelope and then influence the heating and cooling load calculations. Therefore, it is important to investigate the overall thermal resistance as results of different convective heat transfer coefficient calculation algorithms. The overall thermal resistance of wall consists of three parts: convective resistance to the exterior of the envelope, the conductive resistance of the envelope and the convective resistance to the interior of the envelope. The previous study shows that the empirical algorithms in EnergyPlus generate closer HTC values comparing to those from the coupled CFD simulations for interior surfaces, and produce much higher difference for exterior surfaces. However, the contribution of exterior convective resistance in overall thermal resistance is relatively small, thus the overall thermal resistance difference will be dominated by the interior convective HTC calculation. Fig. 11 shows the exterior convective resistance, wall conductive resistance and interior convective resistance for one of the walls. As is shown, the exterior convective resistance is about 2%, while the interior convective resistance is much larger at 20%. Therefore, the
difference of thermal resistance calculation between the three empirical algorithms will be negligible, since the differences in HTC calculation of the three algorithms are negligible. The overall heat transfer coefficients of all the enclosure surfaces are calculated and summarized in Fig. 12. As is shown, the overall thermal resistance generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. The average differences of overall thermal resistance from the empirical algorithms and the coupled CFD simulations range between 28% w 37% from the coupled model of 4.27 m2 K/w. 7. Conclusion A coupled simulation platform between nodal model and Finite Volume Method (FVM) model for advanced building thermal simulation was implemented and tested with a live retrofit building project at the Philadelphia Navy Yard. Heat Transfer Coefficients (HTC) were compared between the EnergyPlus empirical models and the coupled CFD model. It is found that the simple algorithm generates higher HTC values for the exterior surfaces. The averaged percentage difference of HTC values for the exterior surfaces between simple algorithm and coupled CFD simulation is 188% from the coupled model of 25.05 W/m2 K. The adaptive algorithm also generates higher HTC values for the exterior surfaces comparing with that of the coupled CFD results. However, the difference is smaller than that from the simple algorithm. The averaged percentage difference of HTC values for the exterior surfaces between adaptive algorithm and coupled CFD results is 65%. The TARP algorithm generated closest HTC values to the coupled CFD results. The averaged percentage difference of HTC values for the exterior surfaces between TARP algorithm and the coupled CFD results is 7%. All three algorithms in EnergyPlus generate similar HTCs for interior surfaces. The percentage difference of HTC values for the interior surfaces are between 10% and 80% from the coupled model of 1.31 W/m2 K. The overall heat transfer coefficients of all the enclosure surfaces are calculated and it is found that the overall thermal resistance generated from the three convective HTC algorithms are almost the same, yet with observable difference from coupled CFD simulations. The average differences of overall thermal resistance between the empirical algorithms and coupled CFD simulations range from 28% w 37% from the coupled model of 4.27 m2 K/w. From the study of airflow rates through openings under natural ventilation, it is found that Airflow network model in EnergyPlus generates same airflow rate for openings with same area, on the same facade of the same thermal zone. Airflow network model generally predict smaller airflow rates for the openings. Airflow network model generates better airflow rates for openings on the south and east facade and internal openings, with difference ranging from 100% to 200% from the coupled model of about 2.5 m3/s to 6 m3/s. Due to a very close adjacent building to the north of the building, the difference between airflow network model and coupled CFD simulation for openings on the north facade are generally bigger, which is around 600% for openings on north facade from the coupled model of about 1 m3/s. The percentage difference of airflow rate for openings on north facade ranges from 100% to 770% from the coupled model of about 1 m3/s. The paper presents a modeling study of heat transfer and airflow study for natural ventilation. The primary objective of the work is to compare the relative accuracy of the AirFlowNetwork model in EnergyPlus and a coupled CFD-EnergyPlus model. Our premise, based on previous research work, is CFD will be more accurate than lumped nodal models. Thereby, this paper summarized findings on the differences in HTC and airflow calculation
R. Zhang et al. / Building and Environment 68 (2013) 100e113
between the AirFlowNetwork model and the coupled CFD model. Significant discrepancy has been identified for airflow simulations. Study on the overall thermal resistance shows the empirical method in nodal is acceptable. The method for the coupled CFD and nodal is flexible, configurable and easy. The simulation time for one run of CFD for the demonstration example of Building 661 is about 27 min, which is still relatively long. The paper presented a comprehensive comparative study between nodal model and coupled CFD model, which will serve as valuable reference for future similar research or practice. References [1] Feustel H, Rayner-Hooson A. COMIS fundamentals. Tech. Rep. LBNL-28560. Lawrence Berkeley National Laboratory; 1990. [2] Mora L, Gadgil A, Wurtz E. Comparing zonal and CFD model pre-dictions of isothermal indoor airflows to experimental data. Indoor Air 2003;13:77e85. [3] Zhai Z, Chen Q, Haves P, Klems JH. On approaches to couple energy simulation and computational fluid dynamics programs. Building and Environment 2002;37(8):857e64. [4] Negrao C. Integration of computational fluid dynamics with building thermal and mass flow simulation. Energy and Buildings 1998;27:155e65. [5] Wang L, Wong NH. Coupled simulations for naturally ventilated residential buildings. Automation in Construction 2008;17:386e98. [6] Nagai T, Kurabuchi T. A two-node heat and mass transfer network model for long-term simulation of void space of high-rise apartment houses equipped with gas-fired boilers. In: 11th international IBPSA conference 2009. p. 434e41. [7] Manz H, Frank T. Thermal simulation of buildings with double-skin facades. Energy and Buildings 2005;37:1114e21. [8] Wong NH, Wang L, Aida NC, Pandey AR, Wei X. Effects of double glazed facade on energy consumption, thermal comfort and condensation for typical office building in Singapore. Energy and Buildings 2005;37:563e72. [9] Negrao C. Conation of computational fluid dynamics and building thermal simulation. Ph.D. thesis. University of Strathclyde; 1995. [10] Beausoleil-Morrison I. The adaptive coupling of heat and air flow modelling within dynamic whole-building simulation. Ph.D. thesis. University of Strathclyde; 2000. [11] Beausoleil-Morrison I, Clarke J, Denev J, Macdonald I, Melikov A, Stankov P. Further developments in the conation of CFD and building simulation 2001. p. 1267e74. [12] Beausoleil-Morrison I, Clarke J. The implications of using the standard ke3 turbulence model to simulate room air flows which are not fully turbulent. In: Proceeding of ROOMVENT 1998. p. 99e106. [13] Bartak M, Beausoleil-Morrison I, Clark J, Denev J, Drkal F, Lain M, et al. Integrating CFD and building simulation. Building and Environment 2002;37:865e71. [14] Djunaedy E, Hensen J, Loomans M. Toward external coupling of building energy and airflow modeling programs. ASHRAE Transactions 2003;109(2): 771e87.
113
[15] Djunaedy E, Hensen J, Loomans M. Selecting an appropriate tool for airflow simulation in buildings. Building Services Engineering Research and Technology 2004;25(3):289e98. [16] Chen Q, Peng X, van Passen A. Prediction of room thermal response by CFD technique with conjugate heat transfer and radiation models. ASHRAE Transactions 1995;101:50e60. [17] Zhai Z, Chen Q, Klems J, Haves P. Strategies for coupling energy simulations and computational fluid dynamics programs. In: Proceedings of 7th international IBPSA conference 2001. p. 59e66. [18] Zhai Z, Chen Q. Solution characters of iterative coupling between energy simulation and CFD programs. Energy and Buildings 2003;35(5):493e505. [19] Zhai Z, Gao Y, Chen Q. Pressure boundary conditions in multi-zone and CFD program coupling. In: 1st IBPSA-USA conference. [20] Zhai Z. Numerical determination and treatment of convective heat transfer coefficient in the coupled building energy and CFD simulation. Building and Environment 2004;36(8):1000e9. [21] Zhai Z. Performance of coupled building energy and CFD simulations. Energy and Buildings 2005;37(4):333e43. [22] Zhai Z. Sensitivity analysis and application guides for integrated building energy and CFD simulation. Energy and Buildings 2006;38(9):1060e8. [23] Wang L. Coupling of multizone and CFD programs for building airflow distribution and contaminant transport simulations. Ph.D. thesis. Purdue University; 2007. [24] Wang LZ, Chen QY. On solution characteristics of coupling of multizone and CFD programs in building air distribution simulation. In: Proceeding of 9th international IBPSA conference 2005. p. 1315e22. [25] Wang L, Wong NH. Coupled simulations for naturally ventilated rooms between building simulation (BS) and computational fluid dynamics (CFD) for better prediction of indoor thermal environment. Building and Environment 2009;44:95e112. [26] Wetter M, Nouidui TS, Haves P, Zhang R, Li Z. Building controls virtual test bed interface 2011. [27] EnergyPlus. EnergyPlus documentation 2011. [28] A. Inc.. Ansys fluent documentation 2011. [29] Djunaedy E, Hensen J, Loomans M. Comparing internal and external run-time coupling of CFD and building energy simulation software. In: Proceeding of 9th international conference on airflow in rooms 2004. p. 393e6. [30] Zhang R, Zhang Y, Lam KP. Conformal adaptive hexahedral-dominant mesh generation for CFD simulation in architectural design applications. In: Proceedings of 2011 winter simulation conference 2011. p. 928e42. [31] ASHRAE. ASHRAE fundamental handbook 2004. [32] Walton G. Thermal analysis research program reference mannual. Tech. Rep. NBSSIR 83e2655. National Bureau of Standards (now National Institute of Standards and Technology); 1983. [33] Cole R, Sturrock N. The convective heat exchange at the external surface of buildings. Building and Environment 1977;12:207. [34] Yazdanian M, Klems J. Measurement of the exterior convective film coefficient for windows in low-rise buildings. ASHRAE Transactions 1994;100(1): 1087. [35] Palyvos J. A survey of wind convection coefficient correlations for building envelope energy systems’ modeling. Applied Thermal Engineering 2008;28(8e9):801e8.