GridLAB-D Technical Support Document: Residential Equivalent Thermal Parameter Model September 2025 LE Hinkle CD Goodman RG Pratt Prepared for the U.S. Department of Energy Under contract DE-AC05-76RL01830 This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor Battelle Memorial Institute, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or Battelle Memorial Institute. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. PACIFIC NORTHWEST NATIONAL LABORATORY operated by BATTELLE for the UNITED STATES DEPARTMENT OF ENERGY under Contract DE-AC05-76RL01830 Printed in the United States of America Available to DOE and DOE contractors from the Office of Scientific and Technical Information, P.O. Box 62, Oak Ridge, TN 37831-0062 www.osti.gov ph: (865) 576-8401 fox: (865) 576-5728 email: reports@osti.gov Available to the public from the National Technical Information Service 5301 Shawnee Rd., Alexandria, VA 22312 ph: (800) 553-NTIS (6847) or (703) 605-6000 email: info@ntis.gov Online ordering: http://www.ntis.gov Choose an item. GridLAB-D Technical Support Document: Residential Equivalent Thermal Parameter Model September 2025 LE Hinkle CD Goodman RG Pratt Prepared for the U.S. Department of Energy Under Contract DE-AC05-76RL01830 Pacific Northwest National Laboratory Richland, Washington 99352 Contents 1.0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Available Documentation Resources and Limitations . . . . . . . . . . . . . . . 1 1.2 Scope of Technical Support Document . . . . . . . . . . . . . . . . . . . . . . . 2 2.0 ETP Closed Form Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.1 Derivation of the second-order differential equation . . . . . . . . . . . . . . . . 4 2.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.4 State-space form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.0 GridLAB-D House Parameter Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2 Envelope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.3 Thermal Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.4 HVAC Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.4.1 Sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4.2 Internal Gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.0 HVAC Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.1 Thermostat Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.2 Outdoor air temperature adjustments to capacity and COP . . . . . . . . . . . . 18 4.3 Window opening model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 5.0 Implementation in GridLAB-D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 5.1 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 5.2 Model Update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 5.3 Event-driven timestep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 6.0 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Appendix A Guidance on user-defined parameters calculation . . . . . . . . . . . . . . . . A.1 Contents iv Figures 1 ETP representation of the typical residences in GridLAB-D (2020) . . . . . . . . . . . . 1 2 Equivalent Thermal Parameters Circuit Modeled by House-E in GridLAB-D (2020) . . 4 3 Dependency map of building geometry parameters in GridLAB-D residential model . . 10 4 Dependency map of building envelope parameters in GridLAB-D residential model . . 12 5 Dependency map of building thermal mass parameters in GridLAB-D residential model 14 6 Dependency map of building HVAC design parameters in GridLAB-D residential model 16 A.1 Dependency map of building geometry parameters in GridLAB-D residential model with detailed calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1 A.2 Dependency map of building thermal mass parameters in GridLAB-D residential model with detailed calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2 A.3 Dependency map of building thermal mass parameters in GridLAB-D residential model with detailed calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.3 Tables 1 Geometry parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2 Constant defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3 Thermal integrity level 5 defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4 Thermal mass parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 5 HVAC design parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 6 HVAC equipment parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 7 HVAC System Modes and Activation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 8 Window opening model parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . 21 9 Fraction of heat gain delivered to mass defaults . . . . . . . . . . . . . . . . . . . . . . 22 10 Determining Tevent in order to calculate next timestemp . . . . . . . . . . . . . . . . . . 25 A.1 Detailed geometry parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2 A.2 Detailed thermal mass parameter defaults . . . . . . . . . . . . . . . . . . . . . . . . . A.3 v 1.0 Introduction GridLAB-D is a power distribution systems simulation and analysis tool developed by Pacific Northwest National Laboratory (PNNL) (Chassin et al. (2008)) and has been used in various large scale simulation studies such as (Hansen et al. (2019)), (Reeve et al. (2022)), and (Mukherjee et al. (2023)). Residential thermal behavior is modeled using a thermal heat flow circuit, referred to as the equivalent thermal parameter (ETP) model in GridLAB-D’s online documentation (GridLAB-D (2020)). It is a single zone model that lends itself well to residential buildings. As outlined in Goodman et al. (2022), the thermal dynamics of most homes are well represented by the ETP model. Although GridLAB-D and its underlying residential load model have been used in a variety of power systems analyses, comprehensive documentation of the derivation of the residential house model has not been formally published. 1.1 Available Documentation Resources and Limitations Below is a list of published residential building model documentation for GridLAB-D, including a summary and description of limitations. • GridLAB-D ShoutWiki (GridLAB-D (2020)) The primary online documentation for GridLAB-D is located on ShoutWiki. The ”Residential module’s user’s guide” describes the ETP heat balance equations, initializing the room air temperature and the mass temperature, user-defined parameters and default values, design loads and HVAC sizing, thermostat control, and guidance on how to configure the house model. Figure 1 shows the relationship between the major variables that make up the ETP model. The derivation of the ETP model is also included on the Wiki with references to a handwritten derivation and accompanying Excel spreadsheet located on the GitHub reposi- tory. Although the Wiki includes important details on the ETP model and its usage, there are some consistency issues, missing units, inclusion of equations not currently implemented in the GridLAB-D code base, and compiling issues of the LaTeX equations. Details on modeling end-use loads are also included but some sections are incomplete. Figure 1. ETP representation of the typical residences in GridLAB-D (2020) Introduction 1 • GridLAB-D Technical Support Document: Residential End-Use Module Version 1.0 (Taylor et al. (2008)) This documentation describes the modeling assumptions for the end use loads that can be implicitly defined when configuring the residential house model. Other than appliances and lights, this document also covers modeling assumptions for internal gains (plug loads) and heating and cooling loads. Many sections of the ShoutWiki match this technical support document. Some items in the document are not currently implemented in the GridLAB-D code which can be confusing for users. It also outlines a state space description of the ETP model which can help users understand the system of equations but does not describe how the solution is currently implemented in GridLAB-D. The state space representation does not include equations related to HVAC electric power consumption; it is focused only on the system of equations for the thermal dynamics of the home. This document only provides power consumption modeling details for the dishwasher, range, microwave, and internal gains. Details on the thermal models for refrigerators and water heaters are included. • Notes on the GridLAB-D Household Equivalent Thermal Parameter Model (Tesfatsion and Battula (2020)) Iowa State University researchers identified a need for comprehensive documentation of the GridLAB-D residential building model. Some examples of gaps identified by the Iowa State researchers include ambiguity on units of measure and unclear differentiation of structural elements versus data driven parameters. The house ETP model is presented as a complete analytic state space control representation including power consumption (not just thermal dynamics). An alternative solution method than the current GridLAB-D implementation is presented using simple forward finite-difference approximation. Tables of user-defined pa- rameters, derived parameters, variables, and default values are also provided. 1.2 Scope of Technical Support Document This document is to serve as a supplemental reference to the Wiki (GridLAB-D (2020)) and GitHub repository (https://github.com/gridlab-d/gridlab-d). Key documentation gaps that are addressed are listed below. • Complete derivation of the closed-form implicit solution currently implemented in the GridLAB- D codebase. • Step-by-step implementation of the ETP in the GridLAB-D “house_e.cpp” code with reference to the closed-form solution derivation. Implementation steps for the HVAC system operations are included as they relate to the thermal dynamics of the house model. • Comprehensive definitions of the house parameters, including dependencies/relationships be- tween parameters. GridLAB-D has several options for setting parameter values and it is easy for the user to unintentionally override a parameter value due to the complex relationships (user-defined parameters versus calculated parameters). • Tables that outline default values for user-defined parameters. • Excel parameter spreadsheet that describes every GridLAB-D house parameter and the vari- able name in the code. Introduction 2 • Calculation steps for estimating the mass heat capacity and the conductance between the indoor air and the mass if the user would like to modify from the defaults. This document does not cover electrical power consumption modeling methods for HVAC or other end-use loads. The purpose of this technical support document is to serve as a comprehensive resource for the thermal dynamics modeling and house definition. Additional documentation updates are forthcoming for the GridLAB-D software suite. This document reflects the model as implemented in GridLAB-D version 5.3.0; this model has not been substantially changed for many versions and years and we expect both the structure of the model and the default parameter values of said model to be unchanged for the foreseeable future. Introduction 3 2.0 ETP Closed Form Solution The ETP model in GridLAB-D simulates the behavior of a single zone. It is classified as a 2R2C resistance-capacitance (RC) model. It consists of two capacitors that account for the indoor air capacity, Ca , and mass capacity, Cm , and two resistors that represent the conductance between the outdoor air and indoor air, Ua , and the conductance between the mass and indoor air, Hm . With known outdoor air temperature To , heat gain delivered to the indoor air Qa , and heat gain delivered to the mass Qm , we can solve for the temperature of the indoor air Ta and mass Tm through a second-order differential equation derived from this model. Further, assuming constant To , Qa , and Qm over each timestep, we can find a closed-form solution, thus reducing the computational requirements. The following subsections detail the derivation of the closed-form solution implemented in GridLAB-D. Figure 2. Equivalent Thermal Parameters Circuit Modeled by House-E in GridLAB-D (2020) 2.1 Derivation of the second-order differential equation First, the heat balance on Ta is given as: dTa 0 = Qa − Ua (Ta − To ) − Hm (Ta − Tm ) − Ca (1) dt Rearranging Equation 1 in the form for solving a differential equation: dTa 0 = Qa − (Ua + Hm )Ta + Ua To + Hm Tm − Ca (2) dt dTa Ua + Hm Hm Ua Qa + Ta − Tm − To − =0 (3) dt Ca Ca Ca Ca Solving Equation 3 for Tm gives: Ca dTa Ua + Hm Ua Qa Tm = + Ta − To − (4) Hm dt Hm Hm Hm And differentiating Equation 4 with respect to time: dTm Ca d2 Ta Ua + Hm dTa = + (5) dt Hm dt2 Hm dt ETP Closed Form Solution 4 Next, the heat balance on Tm is given as: dTm 0 = Qm − Hm (Tm − Ta ) − Cm (6) dt Solving Equation 6 for Ta yields: dTm 0 = Q m − H m T m + Hm T a − C m (7) dt Cm dTm Qm Ta = + Tm − (8) Hm dt Hm Then, differentiating Equation 8 with respect to time: dTa Cm d2 Tm dTm = + (9) dt Hm dt2 dt Rearranging Equation 7 yields: dTm 0 = Cm + Hm T m − Q m − Hm T a (10) dt Substituting Equation 4 and 5 into Equation 10 gives: Cm Ca d2 Ta Cm (Ua + Hm ) dTa dTa 0= + + Ca + (Ua + Hm )Ta − Ua To − Qa − Qm − Hm Ta (11) Hm dt2 Hm dt dt Rearranging Equation 11:   C m C a d2 Ta Ua dTa + C m + C m + C a + U a Ta = Q m + Q a + U a To (12) Hm dt2 Hm dt Equation 12 is of the form: d2 Ta dTa a 2 +b + cTa = d (13) dt dt a second-order linear differential equation where Cm Ca a≡ (14) Hm Ua b ≡ Cm + Cm + Ca (15) Hm c ≡ Ua (16) d ≡ Q m + Q a + U a To (17) which assumes that d is constant over each timestep. ETP Closed Form Solution 5 2.2 Solution First, we solve the homogeneous equation to find the complementary solution (Ta )c . Let Ta = ert : dTa = rert dt d2 Ta = r2 ert dt2 ar2 ert + brert + cert = d (18) The homogeneous solution to Equation 18 is given by: ar2 + br + c = 0 √ −b + b2 − 4ac r1 = (19) 2a √ −b − b2 − 4ac r2 = (20) 2a We can then show that b2 − 4ac > 0 which indicates that r1 and r2 are real numbers. The complementary solution (Ta )c is formed by multiplying each solution by an arbitrary constant, A1 and A2 (Ta )c = A1 er1 t + A2 er2 t (21) and the general solution is given by the sum of the complementary solution and particular solution (Ta )p , which is assumed to be a constant Ta = (Ta )c + (Ta )p = A1 er1 t + A2 er2 t + (Ta )p (22) Then, taking the first and second derivative with respect to time: dTa = A1 r 1 e r 1 t + A2 r 2 e r 2 t (23) dt d2 Ta = A1 r12 er1 t + A2 r22 er2 t (24) dt2 Substituting Equations 22, 23, and 24 into Equation 13: aA1 r12 er1 t + aA2 r22 er2 t + bA1 r1 er1 t + bA2 r2 er2 t + cA1 er1 t + cA2 er2 t + c(Ta )p = d (25) Rearranging Equation 25: A1 er1 t (ar12 + br1 + c) + A2 er2 t (ar22 + br2 + c) + c(Ta )p = d which reveals that the particular solution is: d (Ta )p = (26) c So the general solution is given by: d T a = A1 e r 1 t + A2 e r 2 t + (27) c Qm + Qa T a = A1 e r 1 t + A2 e r 2 t + + To (28) Ua ETP Closed Form Solution 6 2.3 Boundary conditions With known boundary conditions Tao and dT ao dto at t = 0, we can solve for A1 and A2 : d T a o = A1 + A2 + (29) c dTao = A1 r1 + A2 r2 (30) dt Multiplying Equation 29 by r2 gives: d r2 Tao = r2 A1 + r2 A2 + r2 (31) c Then we subtract Equation 30 from 31 to eliminate A2 : dTao d r2 T a o − = (r2 − r1 )A1 + r2 (32) dt c r2 Tao − dTdtao − r2 dc A1 = (33) r2 − r1 From Equation 33 and Equation 29, A2 is dT d r2 Tao − dtao − r2 dc A2 = T a o − − (34) c r 2 − r1       r2 d r2 r2 dTao A2 = T a o 1 + + −1 − (35) r2 − r 1 c r 2 − r1 r 2 − r1 dt Substituting Equation 27 and 23 for Ta and dT dt in Equation 4 gives: a   Ca U a + Hm Q m + Q a + U a To Ua Qa Tm = r1 t r2 t (A1 r1 e + A2 r2 e ) + r1 t A1 e + A2 e + r2 t − To − Hm Hm Ua Hm Hm (36) Rearranging and simplifying Equation 36:     Ca U a + Hm r1 t Ca U a + Hm U a + Hm Qa Tm = A1 r1 + e + A2 r2 + e r2 t + Qm + + To (37) Hm Hm Hm Hm Hm U a Ua Let Qm Qm Qa T m = A1 A3 e r 1 t + A2 A4 e r 2 t + + + + To (38) Hm Ua Ua where Ca U a + Hm A3 ≡r1 + (39) Hm Hm Ca U a + Hm A4 ≡ r2 + (40) Hm Hm Equation 37 gives Tm for the next time step, i.e., at t = to + ∆to . Note that the solution is discontinuous at any time ts when Qm , Qa , or To change. At time t+ s , the instant after Qm , Qa , or To changes, then Equation 4 can be solved for dT a | dt ts +: dTa Hm U a + Hm Ua Qa + Tm t s − ts = T at s + Tot s + (41) dt t+ s C a C a C a C a ETP Closed Form Solution 7 2.4 State-space form Although not implemented in GridLAB-D, a state-space form of the ETP model can be derived by rearranging Equations 3 and 10: ẋ = Ax + Bu (42) y = Cx + Du (43) " # " # T˙a Ta ẋ = , x= , u=1 (44) ˙ Tm Tm " # −Ua +Hm Hm Ca Cm A= (45) Hm −Hm Cm Cm " # U a To Qa Ca + Ca B= Qm (46) Cm " # " # 1 0 0 C= , D= (47) 0 1 0 ETP Closed Form Solution 8 3.0 GridLAB-D House Parameter Definitions The ETP model in Section 2.0 is implemented in the GridLAB-D house object. The ETP model parameters Ca , Cm , Ua and Hm are determined based on the house geometry, thermal mass characteristics, envelope, and HVAC design. Each category consists of user-defined parameters and calculated parameters, i.e., those involved in intermediate calculations of the ETP model parameters. This section classifies the parameters and maps their relationships to support users while setting up the house object(s). It also includes default values and references as appropriate. It is important to note calculated parameters and ETP parameters can be defined directly by the user but this is not commonly done. Instead, modelers define values for the more common house parameter values, allowing easier configuration of the models. 3.1 Geometry The default house is roughly the size of the DOE prototypical residential home (Pacific Northwest National Laboratory (n.d.)) with notable geometric differences, including reducing the number of stories from two to one. The user-defined parameters that affect ETP parameter Ca is floor area and ceiling height. These are used to calculate the volumetric capacity of the interior air in Equation 48. Several geometric parameters also affect Ua which requires surface areas in its calculation. The relationships between geometric parameters are shown in Figure 3, and the default values are provided in Table 1. Note that some calculated parameters do not have an output. These parameters are referenced as inputs in later diagrams. GridLAB-D House Parameter Definitions 9 GridLAB-D House Parameter Definitions Figure 3. Dependency map of building geometry parameters in GridLAB-D residential model 10 Table 1. Geometry parameter defaults Parameter Default Units Notation 2 Floor area 2500 ft Af Floor aspect ratio 1.5 - - Number of stories 1 - ns Ceiling height 8 ft - Window-wall-ratio 0.15 - - Exterior floor, fraction of total floor area 1 - - Exterior ceiling, fraction of total ceiling area 1 - ECR Exterior wall, fraction of total wall area 1 - - Number of exterior doors 4 - - 2 Area of 1 door 19.5 ft Aext door The air capacity ETP parameter, Ca is calculated as shown in Equation 48. Ca = 3cp ρv (48) where cp is the specific heat capacity of air at constant pressure, ρ is standard air density, and v is the volume of the house. The default values for standard air density and specific heat capacity are given in Table 2. The multiplier of three is an engineering judgment estimate. 3.2 Envelope The building envelope is specified by the modeler using a number of model parameters such as surface R-values, window characteristics, and infiltration rate. Alternatively, these values can be defined as a group by defining a single parameter called the “thermal integrity level”; this ensures a realistic combination of envelope parameters. For example, it is unlikely that a home with a low exterior wall R-value would have highly insulated windows with a high R-value. The default envelope values are provided in Table 3, and the relationship between envelope parameters is shown in Figure 4. The conductance between the outdoor air and indoor air, Ua , is defined by Equation 49. Table 2. Constant defaults Parameter Default Units Notation 3 Standard air density 0.0735 lb/ft ρ ◦ Specific heat capacity of air at constant pressure 0.2402 BTU/lb- F cp GridLAB-D House Parameter Definitions 11 GridLAB-D House Parameter Definitions Figure 4. Dependency map of building envelope parameters in GridLAB-D residential model 12 Table 3. Thermal integrity level 5 defaults Parameter Default Units Notation ◦ 2 Exterior wall R-value 19 hr- F-ft /BTU Rext wall Exterior ceiling R-value 30 hr-◦ F-ft2 /BTU Rext ceiling ◦ 2 Exterior floor R-value 22 hr- F-ft /BTU Rext f loor Glass type Low-E - - Number of glazing layers 2 - - Window frame type Thermal break - - Glazing treatment type Clear - - Infiltration air exchange rate 0.5 hr-◦ F-ft2 /BTU ACH ◦ 2 Exterior door R-value 5 hr- F-ft /BTU Rext door Aext net wall Aext ceiling Aext f loor Awindow Aext door Ua = + + + + + (ACH ∗ Cp ∗ ρv) (49) Rext wall Rext ceiling Rext f loor Rwindow Rext door where A is the area of the surface, R is the R-value of the surface, ACH is the infiltration air exchange rate, Cp is the specific heat capacity of air at constant pressure, ρ is standard air density, and v is the volume of the house. 3.3 Thermal Mass Calculating the mass heat capacity Cm requires estimating the house thermal mass. In the current GridLAB-D implementation, the total thermal mass per unit floor area is a user-defined parameter. The default value is based on residential wood-frame construction including furnishings. In the Appendix, we provide a detailed calculation for total thermal mass per unit to help guide users if they wish to adjust it. Similarly, the mass-air conductance Hm is calculated using two aggregate parameters, interior/exterior wall surface ratio IW R and interior surface heat transfer coefficient hi . These values can also be adjusted using the detailed calculations in the Appendix. The relationships between thermal mass parameters are depicted in Figure 5, and the default values are given in Table 4. Cm and Hm are defined by Equations 50 and 51, respectively. GridLAB-D House Parameter Definitions 13 Figure 5. Dependency map of building thermal mass parameters in GridLAB-D residential model Cm = Af mf − (Ca − Cp ρv) (50) where Af is the total floor area, mf is total thermal mass per floor area, Ca is the house air heat capacity, Cp is the specific heat capacity of air at constant pressure, ρ is standard air density, and v is the volume of the house. The default value for mf (total thermal mass per floor area, see Table 4) is based on a survey of moving companies to estimate a typical furnishings weight.    Aext ceiling Hm = hi (Agross ext wall − Awindow − Aext door ) + Agross ext wall IW R + ns (51) ECR where hi is interior surface heat transfer coefficient, A is the area of the surface, IW R is the interior/exterior wall ratio, ns is the number of stories, and ECR is the exterior ceiling, fraction of total ceiling area. GridLAB-D House Parameter Definitions 14 Table 4. Thermal mass parameter defaults Parameter Default Units Notation ◦ 2 Total thermal mass per unit floor area 2 BTU/ F-ft mf Interior surface heat transfer coefficient 1.46 BTU/hr-◦ F-ft2 hi Interior/exterior wall surface ratio 1.5 - IW R 3.4 HVAC Design Although the HVAC design parameters do not inform the ETP model parameters, ETP model parameter Ua is required to calculate the design cooling and heating loads. The heating and cooling loads are used to determine the capacity of the system and are discussed further in Section 3.4.1. The relationships between HVAC design parameters are shown in Figure 6, and the default values are provided in Table 5. GridLAB-D House Parameter Definitions 15 GridLAB-D House Parameter Definitions Figure 6. Dependency map of building HVAC design parameters in GridLAB-D residential model 16 Table 5. HVAC design parameter defaults Parameter Default Units Notation ◦ Design indoor temperature, cooling 75 F Tcool ◦ Design indoor temperature, heating 70 F Theat Latent cooling fraction 0.30 - flatent HVAC oversizing factor 0 - - Duct pressure drop 0.50 in-H2 O - Window exterior transmission coefficient 0.60 - W ET Max summer hourly load to mean annual hourly load ratio 1.32 - - 3.4.1 Sizing As mentioned previously, Ua is required to calculate the design heating and cooling loads, which are given as: Qheat = Ua (Theat − Toutheat ) (52) where Theat is the design indoor temperature for heating and Toutheat is the design outdoor temperature for heating. BT U Qcoolsensible = Ua (Toutcool − Tcool ) + 3.413 Qidesign + Qsdesign Awindow SHGC ∗ W ET (53) hr where Toutcool is the design outdoor temperature for cooling, Tcool is the design indoor temperature for cooling, 3.413 is a conversion factor, Qidesign is the design internal gains, Qsdesign is the design solar radiation, Awindow the surface area of the windows, SHGC is the solar heat gain coefficient of the windows, and W ET is the window exterior transmission coefficient. 3.4.2 Internal Gains The design internal gains account for heat gains from non-HVAC equipment. The equation is based on a regression model that was developed using End-Use Load and Consumer Assessment Program (ELCAP) “Other” category data collected in 1990 (Pratt et al. (1993)). It was fit to an equation of the form y = axb where y is the design end usage, a = 324.9, x is the floor area Af , and b = 0.442. Then, the estimated end usage is scaled by the maximum hourly load of the summer average load shape to the mean load of the year ratio, which is also from ELCAP. Finally, the design internal gains is given as: BT U /hr 3413 Qidesign = a Abf 1.32 kW (54) 8760 hr yr where 1.32 is ratio of the maximum hourly load of the summer average load shape to the mean hourly load of the year. The value 3413 BTkW U /hr is a conversion factor. GridLAB-D House Parameter Definitions 17 4.0 HVAC Models The user-defined HVAC model parameter defaults are provided in Table 6. 4.1 Thermostat Control This section covers the heating and cooling system controls in GridLAB-D. 1. Thermostat control types • Full control: The thermostat adjusts Tcool,ON , Tcool,OF F , Theat,ON , and Theat,OF F to create a hysteresis band around the heating and cooling setpoints. The default deadband is 2.0 ◦ F. There are auxiliary heating controls using Taux,ON . The control model checks for overlapping deadbands (Tcool,OF F < Theat,OF F ) and throws an error indicating the setpoint deadband (DB) needs to be reduced or the difference between the heating and cooling setpoints needs to be increased. This measure prevents simultaneous heating and cooling. DB Tcool,ON = Tcool + 2 DB Tcool,OF F = Tcool − 2 DB (55) Theat,ON = Theat − 2 DB Theat,OF F = Theat + 2 Taux,ON = Theat − DBaux • No control: No thermostat control. The system mode controls the HVAC system and control logic for determining system mode (HEAT, COOL, OFF, AUX) is ignored. An external input can be used to change the system mode. • Band control: Tcool,ON , Tcool,OF F , Theat,ON , and Theat,OF F are set by the user and are not dynamically adjusted directly in the GridLAB-D house model around the heating and cooling setpoints. 2. System modes: There are four different system modes with their activation described in Table 7. Taux,LO is the lockout temperature compared against the outside air temperature to enable the auxiliary heat when temperatures drop. Note, in AUX mode the heat pump is shut down. 3. Overrides: There are override controls to ignore the deadbands with heating and cooling control or to force the systems off regardless of normal setpoint triggers. During window opening (described in Section 4.3), the HVAC system is disabled. 4. Cycle time control: There are controls to minimize cycling on and off of the equipment. The logic prevents the transition of system modes before the cycle time has been reached. 4.2 Outdoor air temperature adjustments to capacity and COP The effects of outdoor air temperature on COP and capacity for heat pumps and air conditioning units are captured in the equations below. The 95◦ F and 47◦ F values, used in HVAC Models 18 Table 6. HVAC equipment parameter defaults Parameter Default Units Cool system type electric - Heat system type heat pump - Heating COP 2.50 - Cooling COP 3.50 - ◦ Thermostat setpoint, heating schedule F ◦ Thermostat setpoint, cooling schedule F ◦ Thermostat deadband 2 F Thermostat cycle time, minimum 2 min Auxiliary heat type electric - ◦ Auxiliary heat deadband 2 F ◦ Auxiliary heat outdoor lockout temp 10 F Auxiliary heat time delay 5 min Fan type 1-speed - Fan power, low-speed, fraction of high-speed 0.50 - Table 7. HVAC System Modes and Activation Mode Activation Condition Description HEAT Ta < Theat,ON heating system is “on” COOL Ta > Tcool,ON cooling system is “on” Deadband strategy: Ta < Taux,ON auxiliary heating system is “on” AUX Lockout strategy: To < Taux,LO heating and cooling systems “off” Timer strategy: ON when recovery timer exceeded Ta > Theat,OF F OFF all systems “off” Ta < Tcool,OF F HVAC Models 19 COPcool adj and COPheat adj respectively, represent the rated outdoor air temperature for cooling and heating mode. Those values are not adjustable.   COPcool  k0 +k1 Tout , if curve type=DEFAULT COPcool adj (curve type) = COPcool , if curve type=FLAT (56)   95◦ F COPcool Tout , if curve type=LINEAR where k0 = −0.01363961 and k1 = 0.01066989.  COPheat  ′ +k T ′2 +k T ′3 ,  k0 +k1 Tout 2 out 3 out if curve type=DEFAULT COPheat adj (curve type) = COPheat , if curve type=FLAT (57)   Tout COPheat max(1, 47 ◦ F ), if curve type=LINEAR where k0 = 2.03914613, k1 = −0.03906753, k2 = 0.00045617, k3 = −0.00000203, Tout ′ is min(Tout , 80). ( Ccool design (k0 + k1 Tout ), if curve type=DEFAULT Ccool adj (curve type) = (58) Ccool design , if curve type=FLAT where k0 = 1.48924533 and k1 = −0.00514995. ( 2 ), Cheat design (k0 + k1 Tout + k2 Tout if curve type=DEFAULT Cheat adj (curve type) = (59) Cheat design , if curve type=FLAT where k0 = 0.34148808, k1 = 0.00894102, and k2 = 0.00010787. 4.3 Window opening model If the window opening model is enabled, and the outdoor air temperature is between the low temperature cutoff Tlow and high temperature cutoff Thigh , there is some probability occupants will open the windows 0 ≤ popen ≤ 1 which is calculated using Equation 60. In order to simulate the occupant behavior, a random value is drawn from a uniform distribution randval ∼ U (0, 1). If randval ≤ popen , then the windows are opened. In this case, the HVAC system is overridden to OFF and Ua is multiplied by 10 (window open scaling factor shown in Figure 4) for all ETP model calculations to simulate an increase in heat transfer. There is no source mentioned for the factor of 10 increase for Ua so it is assumed to be engineering judgment. If the windows are closed, the HVAC system operates normally. Lastly, if the outdoor air temperature changes by more than window temperature delta from the previous state, the model is updated to determine if the windows will remain opened or closed. popen = awin To2 + bwin To + cwin (60) where awin is the window quadratic coefficient, bwin is the window linear coefficient, and cwin is the constant coefficient. Defaults for the window opening model are provided in Table 8. HVAC Models 20 Table 8. Window opening model parameter defaults Parameter Default Units Notation ◦ Low temperature cutoff 60 F Tlow ◦ High temperature cutoff 80 F Thigh Window quadratic coefficient 0 - awin Window linear coefficient 0 - bwin Window constant coefficient 1 - cwin ◦ Window temperature delta 5 F - HVAC Models 21 Table 9. Fraction of heat gain delivered to mass defaults Parameter Default Units Notation HVAC gains delivered to mass 0 - Qh Internal gains delivered to mass 0.5 - Qi Solar gains delivered to mass 0.5 - Qs 5.0 Implementation in GridLAB-D As stated in Section 2.0, solving the ETP model requires known outdoor air temperature, To , heat gain delivered to the indoor air Qa , and heat gain delivered to the mass Qm at each timestep. To is read from the weather file (top of the hour), or linearly interpolated if specified. Qa and Qm are given by the following equations: Qa = (1 − fh )Qh + (1 − fi )Qi + (1 − fs )Qs (61) Qm = (fh )Qh + (fi )Qi + (fs )Qs (62) where fh is the fraction of HVAC heat gains delivered to the mass, Qh is the heat gains from the HVAC system, fi is the fraction of internal gains delivered to the mass, Qi is the internal heat gains, fs is the fraction of solar gains delivered to the mass, and Qs is the solar heat gains. The default fi and fs is set to 0.5 to represent that not all heat gains are delivered directly to the air (Table 9). In order to appropriately model the heat transfer to the air and dampen the response of Ta , half the heat gain is assumed to be delivered to the mass. Tm and Ta are coupled through Hm (see Figure 2). Qi is determined based on probabilistic models built with ELCAP data (Pratt et al. (1993)). For the current end uses and occupants, Qi for each timestep is given by: BT U X N Qi = 3.412 peu fIeu + nocc focc Qocc (63) hr eu=1 where N is the number of end uses, peu is the current real power, fIeu is the fraction of load that is internal to the house, nocc is the number of occupants, focc is the occupancy fraction, and Qocc is the sensible heat per occupant. Qs is calculated using the equivalent solar aperture and incident solar radiation. Incident solar radiation is based on solar radiation from the weather file, where solar radiation is the total over the hour ending at the time of the observation and is not interpolated. Qs is given as: X 8 Qs = Id Awindowd SHGCd (64) d=1 where I is the incident solar radiation, Awindow is the window area, and SHGC is the window solar heat gain coefficient. 5.1 Initialization The model is initialized by calculating all ETP constants: Ca , Cm , Hm , Ua , a, b, c, r1 , r2 , A3 and A4 . The initial Tao is a random value drawn from a uniform distribution between Tlow and Thigh : Tao ∼ U (Tlow , Thigh ) (65) Implementation in GridLAB-D 22 where   Tcool + Tdeadband Tlow = clip , 60, 140 (66) 2   Theat − Tdeadband Thigh = clip , 60, 140 (67) 2 where Tcool is the cooling setpoint, Theat is the heating setpoint, and Tdeadband is the thermostat deadband. We also assume that Tmo = Tao . 5.2 Model Update After the initial timestep, Ta and Tm are solved using the ETP model equations. For each timestep: 1. Solve for air and mass temperatures and determine window opening (if used). a. Compute Ta using Equation 28 b. Compute Tm using Equation 38 c. Determine probability of window opening from Equation 60 and compare it with a random value (described in Section 4.3). 2. Run HVAC system based on the updated temperatures a. Update climate parameters, including To and solar radiation. These are either fixed constants or, more commonly, are updated using time-series data such as in a TMY3 weather file that is defined by the modeler. b. Update COPcool adj , COPheat adj , Ccool adj , and Cheat adj using Equations 56, 57, 58, and 59, respectively. Then update cooling demand Qcool and heating demand Qheat using the following equations (where st is system type): Ccool adj 0.001W Qcool = (68) COPcool adj 3.4120W hr  Cheat adj 0.001W   COPheat adj 3.4120W hr , if st =heat pump   Qheat (Cheat adj , COPheat adj , st ) = 0, if st =gas (69)     0.001W Cheat design 3.4120W hr , if st =resistance c. Compute system capacity based on the system type (st ) and system mode (sm ):   BT U Cheat adj + Pf an ff an heat gain 3412 kW hr if sm = HEAT & st = heat pump     BT U if sm = HEAT & st = gas Cheat design + Pf an ff an heat gain 3412 kW hr BT U Qh = Cheat design + Pf an ff an heat gain 3412 kW hr if sm = HEAT & st = resistance     Ccool adj −Coolon 1+flatent RH adj + Pf an ff an heat gain 3412 kW hr if sm = COOL & st = electric BT U    BT U Caux + Pf an ff an heat gain 3412 kW hr if sm = AUX (70) where Caux is the auxiliary heat capacity, flatent RH adj is defined by Equation 71 below, Pf an is the fan power, and ff an heat gain is the fraction of heat gain from the fan power. If Implementation in GridLAB-D 23 fan heat gain is included, the default fraction is 1. The system modes COOL, HEAT , and AU X are defined in Table 7. flatent flatent RH adj = 0.1 + (71) 1 + exp(4 − 10RH) where flatent is the latent cooling fraction and RH is the outdoor relative humidity. d. Update electric power consumption of customer end uses. The modeling of these devices is not covered in this paper. e. Compute Qi using Equation 63 f. Update Tcool,On , Tcool,Of f , Theat,On , Theat,Of f , and Taux,On for current setpoints. g. If window is open, enable off override on HVAC system. h. Determine the system mode using the thermostat logic and updated Ta . 3. Compute the initial conditions for the next timestep a. Compute Qs using Equation 64 b. Compute heat gains Qa and Qm , using Equation 61 and 62, respectively c. Compute d using Equation 17 d. Compute dT dt a using Equation 41. t+ s e. Compute A1 and A2 using Equation 33 and 34, respectively f. Update Tevent using Table 10 and calculate when Tevent will occur. This step is detailed in the next section. 5.3 Event-driven timestep GridLAB-D uses an event-driven timestep approach, advancing the simulation to the time the next HVAC operation is predicted to occur. This results in variable length timesteps. First, the next indoor air temperature threshold Tevent at which the HVAC system should change its state based on the current thermal conditions and thermostat settings is calculated using the logic outlined in Table 10. The various event temperatures (e.g., Tcool,ON ) are outlined in Section 4.1. Note that dT dt is solved for in Equation 41. a After Tevent is determined from the current system mode and air temperature gradient, the time interval required to reach Tevent is calculated by solving Equation 28 for t and adding it to the current timestamp, which determines when the next update will occur. There is a minimum off cycle time to prevent short cycling of the HVAC equipment. If the time interval required to reach Tevent is less than the minimum cycle time, the next timestep is set to the last cycle time plus the minimum cycle time. It is important to note that there may be small discrepancies in the solution when comparing an independent run of GridLAB-D versus one conducted in co-simulation using Hierarchical Engine for Large-scale Infrastructure Co-Simulation (HELICS). The outdoor temperature is assumed constant over the timestep period. When running GridLAB-D with a co-simulation, the simulation may be interrupted prior to reaching the calculated next timestep. At that time, GridLAB-D reads the new To , updates Ta , and recalculates Tevent and therefore when the next timestep will be for the GridLAB-D ETP model. Because the outdoor air temperature may not be constant during the timestep, this could result in a different thermal response than running GridLAB-D as a standalone simulation. Implementation in GridLAB-D 24 Table 10. Determining Tevent in order to calculate next timestemp Current System Mode Condition (if applicable) Tevent dTa HEAT dt > 0 Theat,OF F dt ≤ 0 dTa HEAT Taux,ON AUX deadband strategy selected AUX n/a Theat,OF F COOL n/a Tcool,OF F dTa OFF dt < 0 Theat,ON dTa OFF dt > 0 Tcool,ON dTa OFF dt = 0 Ta Implementation in GridLAB-D 25 6.0 Conclusion This technical support documentation provides a comprehensive review of the ETP model and its implementation in GridLAB-D, including implementation steps for HVAC operation. The intent of this document was to improve the existing residential building model documentation, focusing on the current implementation in GridLAB-D. The accompanying Excel spreadsheet is intended to provide users with a comprehensive library of the modeling parameters and a testbed for the ETP model. The inclusion of the ETP model closed-form solution in the companion Excel spreadsheet has the added benefit of demonstrating the efficiency/simplicity of the modeling method. Updates to the documentation of the GridLAB-D software suite are forthcoming. Conclusion 26 References Chassin, David P, Kevin P Schneider, and C Gerkensmeyer (May 2008). “GridLAB-D: An open- source power systems modeling and simulation environment”. In: IEEEPES Transmission and Distribution Conference and Exposition. IEEE, pp. 1–5. ISBN: 978-1-4244-1903-6. DOI: 10. 1109/TDC.2008.4517260. URL: http://ieeexplore.ieee.org/document/4517260/. Goodman, Carolyn D., Laura E. Hinkle, Trevor D. Hardy, and Hayden M. Reeve (Nov. 2022). Building Model Calibration: Validation of GridLAB-D Thermal Dynamics Modeling. Tech. rep. Pacific Northwest National Lab. (PNNL), Richland, WA (United States). DOI: 10.2172/1897180. URL: https://www.osti.gov/biblio/1897180. GridLAB-D (2020). GridLAB-D Main Documentation. URL: http://gridlab-d.shoutwiki.com/ wiki/Main_Page. Hansen, J., T. Hardy, and L. Marinovici (Feb. 2019). “Transactive Energy: Stabilizing Oscillations in Integrated Wholesale-Retail Energy Markets”. In: 2019 IEEE Power Energy Society Innovative Smart Grid Technologies Conference (ISGT), pp. 1–5. DOI: 10.1109/ISGT.2019.8791658. URL: https://ieeexplore.ieee.org/document/8791658. Mukherjee, Monish, Manisha Maharjan, and Trevor Hardy (July 2023). “Transactive Emergency Power Allocation”. In: 2023 IEEE Power and Energy Society General Meeting (PESGM), pp. 1– 5. DOI: 10.1109/PESGM52003.2023.10252732. Pacific Northwest National Laboratory, U.S. Department of Energy (n.d.). Prototype Building Mod- els. Accessed: 2025-08-05. URL: https://www.energycodes.gov/prototype- building- models. Pratt, R.G., C.C. Conner, B.A. Cooke, and E.E. Richman (1993). “Metered end-use consump- tion and load shapes from the ELCAP residential sample of existing homes in the Pacific Northwest”. In: Energy and Buildings 19.3. A Special Issue devoted to the End-Use Load and Consumer Assessment Program (ELCAP), pp. 179–193. ISSN: 0378-7788. DOI: https: //doi.org/10.1016/0378- 7788(93)90026- Q. URL: https://www.sciencedirect.com/ science/article/pii/037877889390026Q. Reeve, Hayden M., Steven E. Widergren, Robert G. Pratt, Bishnu Bhattarai, Sarmad Hanif, Sadie R. Bender, Trevor D. Hardy, and Mitchell A. Pelton (Jan. 2022). Distribution System Operator with Transactive (DSO+T) Study Volume 1: Main Report. Tech. rep. Pacific Northwest National Laboratory. DOI: 10.2172/1842485. URL: https://www.osti.gov/biblio/1842485. Taylor, Zachary T, Krishnan Gowri, and Srinivas Katipamula (July 2008). GridLAB-D Technical Support Document: Residential End-Use Module Version 1.0. Tech. rep. Pacific Northwest National Lab. (PNNL), Richland, WA (United States). DOI: 10.2172/939875. URL: https: //www.osti.gov/biblio/939875. Tesfatsion, Leigh and Swathi Battula (July 2020). Notes on the GridLAB-D Household Equiva- lent Thermal Parameter Model. ISU General Staff Papers 202007150700001059. Iowa State University, Department of Economics. URL: https : / / ideas . repec . org / p / isu / genstf / 202007150700001059.html. REFERENCES 27 Appendix A – Guidance on user-defined parameters calculation This section provides detailed calculations that can be used to inform user-defined parameters. Specifically, it is intended to help users define the interior/exterior wall ratio IW R, interior surface heat transfer coefficient hi , and total thermal mass per floor area mf . Figure A.1 shows additional user-defined geometric parameters required to calculate the interior wall area, which is used in the calculation of mf and IW R. Defaults are provided in Table A.1. Figure A.1. Dependency map of building geometry parameters in GridLAB-D residential model with detailed calculations Guidance on user-defined parameters calculation A.1 Table A.1. Detailed geometry parameter defaults Parameter Default Units Room area 250 ft2 Fraction of open interior partitions 0.2 - Number of partitions along long axis 2 - Number of partitions along short axis 3 - Furthermore, there is an additional set of user-defined parameters required to calculate mf regarding the furnishings and appliances. The defaults are provided in Table A.2 and their relationships described in Figure A.2. Figure A.3 shows an updated version of Figure 5 with mf , IW R, and hi as calculated parameters. Figure A.2. Dependency map of building thermal mass parameters in GridLAB-D residential model with detailed calculations Guidance on user-defined parameters calculation A.2 Table A.2. Detailed thermal mass parameter defaults Parameter Default Units Furnishings weight per floor area 3.4 lb/ft2 Number of major appliances 5 - Appliance weight 200 lb Interior horizontal surface heat transfer coefficient 0.49 BTU/hr-◦ F-ft2 Interior vertical surface heat transfer coefficient 0.49 BTU/hr-◦ F-ft2 Figure A.3. Dependency map of building thermal mass parameters in GridLAB-D residential model with detailed calculations     Happliances Hf urnishings (Aint wall + Aceiling ) + hi + hi IW R = (A.1) Agross ext wall where Happliances is the thermal conductance of the major appliances and Hf urnishings is the thermal conductance of the furnishings. Hint surf aces hi = (A.2) (Aint wall + Aceiling ) where Hint surf aces is the thermal conductance of the interior surfaces. Guidance on user-defined parameters calculation A.3 Cint surf aces + Cf urnishings + Capplicances mf = (A.3) Af where Cint surf aces is the thermal mass capacity of the interior surfaces, Cf urnishings is the thermal mass capacity of the furnishings, and Cappliances is the thermal mass capacity of the major appliances. Guidance on user-defined parameters calculation A.4 Pacific Northwest National Laboratory 902 Battelle Boulevard P.O. Box 999 Richland, WA 99352 1-888-375-PNNL (7675) www.pnnl.gov