1、CALIBRATION OF AN ENERGYPLUS SIMULATION MODEL BY THE STEM-PSTAR METHOD Antonio Carrillo, Fernando Dominguez and Jose M. Cejudo Energetics Research Group, University of Malaga, ETSII, Industrial Engineering School, Plaza de El Ejido s/n, 29013, Malaga, Spain ABSTRACT The PSTAR (Primary and Secondary
2、Terms Analysis and Renormalization) method was developed by the National Renewable Energy Laboratory (NREL) to determine the key thermal parameters of a building from short-term outdoor test results. This paper shows an application of the PSTAR method as a quantitative guidance to calibrate a detail
3、ed thermal model of a dwelling. An existing dwelling sited in southern Spain is used as a case study. The entire process comprises: A) Starting from audit data, a detailed EnergyPlus model of the building is created. B) Some experimental tests are carried out: blower door test, thermographic inspect
4、ion, determination of thermal resistance of some envelope components using heat flux meters and the STEM (Short Term Energy Monitoring) test procedure. C) The PSTAR method is used to obtain quantitative information about the model ability to reproduce three primary thermal parameters of the actual b
5、uilding: the building heat loss coefficient, the charge/discharge of the building mass coupled with variations of the indoor temperature and the solar gains. D) The EnergyPlus model is reasonably calibrated using all information available. As a result, a calibrated model is obtained, whose performan
6、ce shows good agreement with measured data, yet some caveats still remain in the calibration process. Finally, some modifications to the STEM test are suggested in order to obtain a better experimental data set. INTRODUCTION The basic dilemma in building model calibration is that a realistic thermal
7、 model of a building tends to be complex and has by far more parameters than can be estimated from performance data. Thus, calibrating a model by varying its input parameters can be rather problematic. The STEM-PSTAR method is an attempt to solve this problem. The method has been used generally on d
8、etached houses, see Balcomb et al.(1993) and also has been experimentally validated in Judkoff et al. (2000). PSTAR stands for Primary and Secondary Terms Analysis and Renormalization, and was developed by the National Renewable Energy Laboratory (NREL), see Subbarao (1988). Heat flow into the house
9、 air is mathematically separated into terms relating to the effect causing the heat flow. The sum of the terms should be equal to zero at each hour if energy is to be balanced. The terms are listed in table 1 with the sign convention that a heat flow that heats up the indoor air is positive. Each of
10、 these macro-terms is computed as convenient, usually by simulation of a detailed building model, which is sometimes called “micro” level simulation. This arrangement of the heat flows is directed to a subsequent calibration of the most relevant terms in the heat balance. These are called “primary”
11、terms, and usually are: the building steady heat loss LQ , the charge/discharge of the building mass coupled with variations of the indoor temperature storageinQ,and the solar gains sunQ , see table 1. When measured data is used, the heat balance is not fulfilled and three renormalization factors ar
12、e introduced. These are PL,Pin,storageand Psun, and are simply scale factors for the primary heat flows. They are used to minimize the Root Square Mean Error (RSME) of the heat balance. Measured data is obtained via the STEM procedure. STEM stands for Short Term Energy Monitoring, and is described i
13、n some detail in Subbarao (1988) and Balcomb et al. (1993), and more recently in Judkoff et al. (2000). It is an experimental protocol with the aim to enhance the primary heat flows over the remaining ones in order to allow a robust identification of the renormalization factors. It consists of three
14、 periods, an initial period of co-heating, during which inside air temperatures are maintained at a uniform and constant value, using several portable electrical heaters. The objective is to enhance the steady state heat loss to the environment LQ over the other heat flows. Then, there is a cool-dow
15、n period when the heaters are shut-off and the inside air temperature changes. The objective is to enhance the internal mass charge/discharge heat flow storageinQ,. Finally there is a period when the indoor air temperature is allowed to freely float, and can be used, altogether with the previous per
16、iods, to calibrate the solar gains term sunQ . The result should be a calibrated macro-level building model, i.e. a model that minimizes the heat balance RMSE when using measured data. The problem at this point is that Eleventh International IBPSA Conference Glasgow, Scotland July 27-30, 2009 - 2043
17、 -the PSTAR building model is not a physical model any more, but a grey-box model, so it lacks versatility. Alternatively, this paper proposes to preserve a physical micro level model of the building and to use the PSTAR method to evaluate whether the model captures the primary thermal characteristi
18、cs of the actual building. The renormalization factors will provide quantitative information about how far is the model from the actual building. Certainly, the calibration of a micro level model implies varying its input parameters, which is problematic and even arbitrary because it depends on the
19、judgement of the user, who will need to make some assumptions. Some extra information from additional experimental tests may contribute to support the abovementioned assumptions. In this case the tests were a blower door test, a thermographic inspection and the determination of the thermal resistanc
20、e of some envelope components using heat flux meters. Next section describes how the EnergyPlus simulation program can be used to obtain the PSTAR macro terms, then a case study that illustrates the whole procedure is presented. ENERGYPLUS TO OBTAIN THE PSTAR MACRO TERMS EnergyPlus is a general purp
21、ose building simulator, and a useful tool for computing the PSTAR heat flow macro terms. Each term requires an specific simulation scenario, in order to isolate the effect studied. The last column of table 1 summarizes the settings required. Most of these simulation scenarios use special environment
22、al conditions that can be obtained using modified weather files. Measured temperatures and heat flows from adjacent spaces can be included as inputs to the simulation using schedules linked to external files. An underlying assumption in the PSTAR method is the linearity of the model, so that the sup
23、erposition principle could be applied and the heat balance fulfilled. Long wave radiation exchange and non constant convection coefficients make a standard EnergyPlus simulation model non linear. A first approach may be to force the model to be linear by assuming some fixed values for the convective
24、 film coefficients and eliminating the long wave radiation exchange by setting a very reduced value of thermal emissivity for the materials. In this paper, the PSTAR is used just as a tool to help in the calibration of the detailed micro level model, so it is not so crucial to keep linearity as far
25、as the renormalization factors keep on providing useful information. Thus, the standard EnergyPlus models accounting for long wave radiation and non constant convection film coefficients are keeped. Another problem is that there are some surfaces where the heat flux is measured, e.g. the wall next t
26、o the health center. The measured flux is composed of various terms: steady-state conduction, charge and discharge effects due to changes in inside and outside temperatures, and also some heat flow that can be related to solar radiation, passing through windows and being distributed on the interior
27、surfaces. It must be avoided to account twice for this last effect, once into the measured heat flux, and twice into the term sunQ . Simply eliminating these surfaces from the simulation model will not work, because the entering solar radiation will be distributed on the remaining interior surfaces.
28、 A way to solve this problem is to use a fictitious surface, with the same solar absorptance than the actual one, but highly conductive and with a very reduced (ideally zero) convection coefficient. On the outer side of the fictitious surfaces, the same temperature of the zone is imposed. Therefore,
29、 solar radiation on those surfaces will be absorbed, but will never be rejected into the room air. CASE STUDY It is a house sited in the village of Montecorto, Spain. lat.: 3649 N, long.: 518E, elevation 500 m. Figure 1 South oriented faade. There are ten houses in a row, the studied one is sited at
30、 the east boundary of the row, quoted as number 1 in the figure 2. There are two main orientations exposed to the exterior environment, the south faade, (see figures 1 and 2) which is turned 14 to east, and the north faade. Part of the east faade is exposed to the exterior environment,and the other
31、part is an internal partition to an adjacent building, which is a health center that can also cast some shadows on the south faade of the house n1 during the first hours of the day, see figure 4. The west wall separates houses number 1 and 2. The roof is flat and there is a slightly ventilated crawl
32、 space beneath the lower storey of the house. There are two storeys: downstairs there is a living room, a kitchen and a small laundry room. Upstairs there are three bedrooms, and a bathroom, see figure 3. - 2044 -Table 1 PSTAR method macro terms and calculation method of each term using EneryPlus si
33、mulations. TERM DESCRIPTION METHOD LQSteady state gain to house air from outside air, calculated by multiplying the building loss coefficiente (BLC) times the outside-inside temperature difference. Simulation with fixed inside and outside temperatures, e.g. 20-0 C. No solar radiation, sky temperatur
34、e matches the outside dry temperature. Other boundary conditions apart from the exterior environment are set equal to the inside air temperature. Of course, the BLC calculation can be done in a “traditional” manner but, with this simulation, one can be sure of the inbuilt BLC of the simulation model
35、. storageinQ,Heat flow positive if the thermal mass of the building is releasing heat into the house air, which occurs when inside air is cooling down. Simulation with fixed outside temperature, e.g 20C, no solar radiation, sky temperature is equal to the outside air temperature. Other boundary cond
36、itions apart from the exterior environment are set equal to the outside temp. An ideal HVAC system forces the inside temperature to match the measured temperature from the actual building during the STEM monitoring period. The sensible heating/cooling rate of this ideal HVAC system is the summation
37、of LQ +storageinQ,. The latter can be obtained by subtraction. sunQHeat flow to the house air due to solar gains. This includes solar gains through the windows, heat stored into building internal mass and heat flow through the external walls due to solar radiation absorbed on the exterior surfaces.
38、Simulation with fixed outside temperature, e.g. 20C, solar radiation as measured, sky temperature is set equal to the outside air temperature. Other boundary conditions apart from the exterior environment are set equal to the outside temp. An ideal HVAC system forces the inside temperature to match
39、the outside temperature. The cooling load is sunQ . auxQMeasured heat flow from the electric heaters. It is measured in field during the STEM monitoring period. ilQinfHeat flow due to infiltration air. Can be estimated using the Sherman-Grismrud model, based on the measured leakage area, the inside-
40、outside temperature difference and the wind speed. storageoutQ,Heat flow to the house air due to changes in outside temperature. Positive when outside temperature is cooling down. Simulation with outside temperature as measured, no solar radiation, sky temperature is set equal to the outside air tem
41、perature. Other boundary conditions apart from the exterior environment are set equal to the inside temp. An ideal HVAC system forces the inside temperature to be constant e.g. 20 C. The sensible heating/cooling rate of this ideal HVAC system in each hour is the summation of LQ +storageoutQ,. The la
42、tter can be obtained by substration. skyQHeat flow to the room due to the depression in sky temperature below outside air temperature. Normally negative Simulation with outside temperature fixed, e.g. 20C, no solar radiation, sky temperature matches the depression value below outside air temperature
43、.An ideal HVAC system forces the inside temperature to match the outside air temperature. The sensible heating rate of this ideal HVAC system in each hour is -skyQ . The sky temperature is estimated based on the measured outside temperature, clearness index and relative humidity, using expressions f
44、rom Martin and Berdahl (1984) and Kasten and Czeplak (1979). tempadjQ,Heat flow due to conduction from an adjacent space, where the temperature can be measured. In this case, the crawlspace. Simulation with outside temperature fixed, e.g. 20C, no solar radiation, sky temperature equal to the outside
45、 temperature. Temperature of the adjacent space as measured. Inside temperature constant and equal to the outside air temperature. The sensible heating/cooling of the HVAC system is crawlQ . This term is composed of two parts, a steady-state conduction part, and a transitory part due to changes in t
46、he adjacent space temperature. These parts can be calculated if necessary in the same way as storageoutQ,fluxadjQ,Heat flow due to conduction from an adjacent space, not accesible, so the temperature cannot be measured. A heat flux meter is used to measure the actual heat flux on the inside surface.
47、 Measured via one or more heat flux meters. - 2045 -Figure 2 Plan of the studied house (number 1) orientation, and boundaries. Figure 3 Interior distribution of the house. Table 2 Summary of opaque constructions. Data extracted from building plans. Element Description Solar Absorp. U (no film) W/m2-
48、K weightKg/m2External Walls Cavity wall, PUR 30 mm 0.3 0.68 298 Flat Roof Concrete, XPS 30 mm 0.6 0.80 423 Floor Concrete No isolated 0.3 3.50 326 Internal Floor Concrete 0.3 - 346 Internal partitions Brick wall 0.3 3.33 133 Partitions to adjacent buildings Brick wall,no isolation 0.3 3.33 133 The h
49、ouse was built in year 2006, and the experimental tests were performed during December 2007. The house had never been occupied, thus it had no furniture at all. Table 2 shows a summary of opaque constructions. Windows have clear simple 6 mm glazing with a 12 mm width, al-no break frame. The outside reveal depth is about 25 cm. The test was performed with the windows blinds fully raised (opened). There is a wooden exterior door in the north faade. Table 3 shows data about walls with windows and/or doors and table 4 summarizes the wall areas. Table 3 Data about walls with windo
