1、4685 (RP-1093) Overview of Existing Literature on Diversity Factors and Schedules for Energy and Cooling Load Calculations Bass Abushakra, Ph.D. Member ASHRAE Member ASHRAE Member ASHRAE Jeff S. Haberl, Ph.D., P.E. David E. Claridge, Ph.D., P.E. ABSTRACT This paper provides an overview of methods re
2、ported in the literature for generating typical load shapes required for the simulation of energy use and peak cooling loads in commercial buildings. It also includes a survey of available databases of monitored commercial end-use electricity data and a review of classzjication schemes of the commer
3、cial build- ing stock. This study was conducted as part of ASHRAE P- 1093, whose objective was to identih and utilize the most appropriate methods for developing diversity factors on rele- vant monitored sets of lighting and equipment data. The paper also provides a discussion offour methods in the
4、literature that were adapted and combined in RP-1093 to develop a library of diversity factors and schedules for use in building energy and cooling load simulation programs. INTRODUCTION In most o%ce buildings, internal heat gains from people, plug loads, and lighting are significant contributions t
5、o the cooling load and, hence, are very important for energy calcu- lations. Energy simulation programs typically require the input of representative hourly values of the electrical consumption of lights and equipment. The heat gains from these sources deviate from their peak values due to people en
6、tering and leaving the building, who switch lights and equipment on and off. Some equipment, such as computers and copiers, automatically switches to ?standby? or ?energy saving? modes when unused for a period of time during the day. This variability in operation of office equipment and lighting is
7、typically treated by entering the peak consumption of lighting and plug loads and then using different sets of 24- hour ?diversity factors? for weekdays, weekends, and any other set(s) of important daily variations. These diversity factors are numbers between zero and one that are used to multiply t
8、he peak consumption that has been entered. Diver- sity factors given in the user?s manuals of individual simula- tion programs or estimates from engineering experience are normally used in simulation. The goal of ASHRAE RP-1093 was to compile a library of schedules and diversity factors based on mea
9、sured electric- ity consumption data for use in energy simulations and peak cooling load calculations in office buildings. This research project derived multiple sets of diversity factors from measured data in 32 office buildings (Abushakra et al. 2002). As part of the research, the methods reported
10、 in the literature for generating load profiles were carefully reviewed, and the methodology suitable for use with one year of hourly profiles was developed and implemented in a spreadsheet. The project also developed a procedure for estimating diversity profiles for loads due to occupants that is b
11、ased on the diversity profiles for lights and plug loads (Abushakra and Claridge 2001). This paper presents a review of the literature on methods for generating load shapes required for simulation of energy use and peak cooling loads and the databases containing end- use energy consumption data for
12、commercial buildings. The literature on diversity factors and load shapes was examined, with particular attention to the descriptions of methods used, references to existing databases of monitored whole-building energy use and end-use data in commercial buildings, and different ways of classieing th
13、e commercial building stock. Literature related to identification and testing of appropriate daytyping methods on monitored data sets of lighting and equipment consumption is also discussed. Table 1 presents a chronological list of the available literature describing data- bases of monitored energy
14、use, load shape determination methods, and commercial building classification schemes. Only one-third of the literature reviewed described the meth- Bass Abushakra is an assistant professor in the Architectural Engineering and Building Construction Department, Milwaukee School of Engi- neering, Milw
15、aukee, Wisc. Jeff S. Haberl is a professor in the Department of Architecture and David E. Claridge is a professor in the Depart- ment of Mechanical Engineering, Texas A Norford et al. (1988); Parti et al. (1988); ASHRAE (1 989); Haberl and Komor (1 989); Verdi (1 989); Akbari et al. (1990); Eto et a
16、l. (1990); Finleon (1990); Gillman et al. (1 990); Haberl and Komor (1 990); Pratt et al. (1990); Schon and Rodgers (1990); Stoops and Pratt (1990); ASHRAE (1 99 i); Katipamula and Haberl (1 991); Bronson et al. (1 992); Mazzucchi (1992); Rohmund et al. (1992); Abbas (1993); Hadley (1993); Akbari et
17、 al. (1994); Alereza and Faramarzi (1994); Halverson et al. (1994); Hamzawi and Messenger (1 994); Jacobs et al. (1 994); Margossian (1 994); Norford et al. (1 994); Szydlowski and Chvala (1 994); Thamilseran and Haberl (1 994); Wilkins and McGaffin (1 994); Bou-Saada and Haberl (1 995); CEED (1 995
18、); Dhar (1 995); Bou-Saada et al. (1 996); Emery and Gartland (1 996); Floyd et al. (1 996); Kati- pamula et al. (1996); Nordman et al. (1996); Parker (1996); Noren (1 997), and Noren and Pyrko (1 998a and b), discussing methods used in Europe; EPRI (1999); Keith and Krarti (1 999); and Thamilseran
19、(1 999). Additional papers gave an insight into approaches to derive the diversity factors, categories of office equipment and their typical energy use, comparison of engineering methods with statistical methods of deriving load shapes, and other 166 ASHRAE Transactions: Research Table 2. Typical Lo
20、ad Shape Samples of Office Buildings from Existing Literature Typical Load Shape Samples TLS for equipment in small office buildings TLS of office buildings in the Pacific Northwest Reconciled TLS for large office buildings Reconciled TLS for small office buildings Whole-building TLS for commercial
21、buildings TLS (July weekdays) for large office buildings TLS of whole-building electric, with electric space heating, for an office building TLS for ofice buildings TLS from measured and simulated data for office buildings Measured and simulated TLS for a large institu- tional building TLS for a med
22、ium office building Description Reference Weekday and weekend (W) Disaggregated average January and July weekday and weekend WBE (MW) Disaggregated WBE into various end-uses (kW). Disaggregated WBE into various end-uses (kW). Restaurant, grocery store, and large office (% of WBE and disaggregated en
23、d-use (% of peak). Disaggregated WBE for a cold day and an average January weekday (W/f?) Normalized measured and simulated lighting and equipment loads (0-1) January, April, July, and Summer peak day for gov- ernment office and private office buildings (kW) Weekday and weekend WBE for different aca
24、demic year periods (kW) Normalized (average kW/max kW) based on manual, EMCS, and occupant sensor controls Norford et al. 1988 Baker and Guliasi 1988 Akbari et al. 1989 Akbari et al. 1989 Baker 1990 peak). Baker 1990 Giliman et al. 1990 Stoops and Pratt 1990 Barrar et al. 1992 Bronson et al. 1992 Di
25、amond et al. 1992 TLS for a large federal office building Disaggregated WBE for working day and nonwork dav kW TLS of end-uses (metered and adjusted) in office buildings TLS of end-uses in office buildings TLS of workstations in office buildings TLS for a large institutional building TLS for a large
26、 federal office building Average summer TLS of medium office buildings TLS derived with a weather-daytyping approach Mazzuchi 1992 Disaggregated WBE for summer typical day and winter typical day (W/fi2) Measured and adjusted winter cold day and typical day, and summer hot day and typical day (W/fi2)
27、 Workday and nonworkday normalized demand (0-1) Pre- and post-retrofit everyday, weekday, and week- end (Wf?) Weekday and weekend disaggregated WBE (kW) Three office buildings WBE based on various levels of monitored data availability (kW). Measured and DOE-2 simulated weekday and week- end (kWh/h)
28、Rohmund et al. 1992 Rohmund et al. 1992 Szydlowski and Chvala 1994 Thamilseran and Haberl 1994 Halverson et al. 1994 Alereza and Faramarzi 1994 Bou-saada and Haberl 1995 TLS of a personal computer in an office building Weekdays, weekends, and holidays TLS for office buildings in Sweden denved with a
29、 weather day- typing approach TLS for a large federal office building Pre- and post-retrofit weekday, weekend summer, and weekend winter Bou-saada et al. 1996 Average power (W) with power management Normalized according to different temperature bins Nordman et al. 1996 Noren 1997 TLS for an Energy-S
30、tar-compliant personal com- puter in an office building Normalized (0-1) I Katipamula et al. 1996 TLS of occupancy and lighting in an institutional building Normalized (0-1) I Keith and Krarti 1999 ASHRAE Transactions: Research 167 related topics. The papers included Pratt et al. (1 990), Owashi et
31、al. (1 994), Floyd et al. (1 996), and Komor (1 997). Table 3 presents the unique methods identified in the U.S. and Europe to develop typical load shapes for weather-depen- dent and weather-independent end-uses. Other methods not included in Table 3 used simple averaging techniques or appeared to d
32、uplicate some previously used methods. The identified methods are either deterministic or statistical meth- ods. Deterministic Methods Four deterministic methods were identified. The end-use disaggregation algorithm (EDA) (Akbari et al. 1988) is an engineering method that primarily utilizes the stat
33、istical char- acteristics of the measured hourly whole-building load and its statistical dependence on temperature. In the EDA, the sum of the end-uses is constrained at hourly intervals to be equal to the measured whole-building consumption, providing a reality check that is not always possible wit
34、h pure simulation. The intent of the method is to supply reasonable end-use break- downs when detailed measured information is scarce. This method is a hybrid method that uses monitored data, statistical disaggregation, and prorating based on simulation. The variance allocation method (Schon and Rod
35、gers 1990) is a hybrid engineering/statistical approach to end-use load shape estimation for the commercial sector. The method (1) identifies systematic biases in hourly end-use load esti- mates derived from an engineering model, (2) adjusts the engi- neering model to significantly reduce these bias
36、es for individual building end-use estimates, and (3) uses a variance- weighted approach to reconcile adjusted engineering esti- mates with whole-building metered data. To reconcile the sum of the hourly end-use load estimates with each individual facil- itys hourly research data, the variances obse
37、rved for each regression coefficient are used. The difference between simu- lated and metered totals is prorated based on statistical varia- tion in the simulated end-use loads. The largest and most variant end-uses receive the largest portion of the difference between the engineering simulation and
38、 the metered whole- building load. The Stephan-Demingalgorithm (SRC 1988, ref. Eto et al. 1990) is a statistical adjustment procedure in which elements of an end-use matrix are adjusted when the terminal values (i.e., total hourly loads) are known. When only the hourly whole-building load is known,
39、a weighted distribution of the difference between the measured total and the sum of the simulated end-uses, based on the magnitude of the original simulated estimates, is applied. The heuristic pattern recognition algorithm (Margossian 1994) is used to disaggregate premise-level load profiles. This
40、algorithm uses 5-minute or 15-minute residential premise- level load data as input and also requires connected load esti- mates of the cooling, heating, and water-heating appliances as input. The algorithm first scans the premise-level load profile and identifies all spikes in the profile that are l
41、arge enough with respect to the connected load of the space-conditioning appliance, and categorizes these spikes with various attributes, such as shape, timing, magnitude, and duration. In a second stage-the classification stage-the algorithm decides whether or not to attribute each of the identifie
42、d spikes to the space-conditioning appliance. The resulting spikes compose the end-use load profile for the space conditioning appliance on that day. For example, the load profile of the water-heating appliance is derived from the residuals of the premise-level load profile-after subtracting the spa
43、ce-conditioning appli- ance load profile-using the scanning and classification stages. Statistical Methods Eleven statistical methods were identified, four of which appeared to contain elements most suitable for use in the RP- 1093 project: the mean/standard deviation/regulatory index (Katipamula an
44、d Haberl 199 i), interquatile analysis (Abbas 1993), inverse binning method (Thamilseran and Haberl 1994), and Duncans multiple range test and frequency univariate analysis (Dhar 1995). These methods are described in the next section, while the remaining statistical methods are described in this sec
45、tion. The conditional energy demand (CED) technique (Parti et al. 1988) is a technique where the end-use metered consump- tion information is used only for comparison to the CED esti- mates of end-use load shapes. The CED approach (also called the conditional demand approach CDA) is essentially a co
46、rrelation analysis of the energy use of many separate premises against the energy-using equipment in each of these premises. The analysis seeks to determine the difference in observed load due to the presence of a given energy-using device, all other things being held equal. The difference is taken
47、to be the energy contribution of the device. The tech- nique was first applied to annual and monthly billing data. With the availability of whole-building load shape data, the technique was extended to an hourly time step. The CED carries out the disaggregation of the total load into its end-use com
48、ponents by applying multiple linear regression (MLR) analysis to a data set composed of total load data, survey, and weather information. The model breaks down the hours of the day into four general hourly categories: (1) night, (2) morning, (3) midday, and (4) evening. The statistically adjusted en
49、gineering approach (SAE) (CSI, CA, ADM 1985, ref. Eto et al. 1990) is very similar to the previously described deterministic methods. First, an engi- neering simulation is developed to provide an initial estimate of end-use loads. Next, the initial estimates are regressed against control totals, which are averages of hourly energy use for typical days. The estimated coefficients can then be thought of as adjustment factors that reconcile the initial esti- mates to the control total. Schrock (1997) stated that SAE models that are a subclass of the CED models are used to quickly re
