E

VE

a

ARRA

KDPRE

1

cenwPitaetc

tetsDctpm

9

0h

Energy and Buildings 68 (2014) 223–231

Contents lists available at ScienceDirect

Energy and Buildings

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

nergy savings from direct-DC in U.S. residential buildings

agelis Vossos ∗, Karina Garbesi, Hongxia Shennergy Analysis and Environmental Impacts Department, Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, CA, USA

r t i c l e i n f o

rticle history:eceived 2 April 2013eceived in revised form 6 July 2013ccepted 3 September 2013

eywords:irect current (DC)

a b s t r a c t

An increasing number of energy-efficient appliances operate on direct current (DC) internally, offering thepotential to use DC directly from renewable energy systems, thereby avoiding the energy losses inherentin converting power to alternating current (AC) and back. This paper investigates that potential for net-metered residences with on-site photovoltaics (PV) by modeling the net power draw of a ‘direct-DChouse’ compared to that of a typical net-metered house with AC distribution, assuming identical DC-internal loads. The model comparisons were run for 14 cities in the United States, using hourly, simulated

hotovoltaics (PV)esidential buildingsnergy conservation

PV-system output and residential loads. The model tested the effects of climate and battery storage. Asensitivity analysis was conducted to determine how future changes in the efficiencies of power systemcomponents might affect potential energy savings. Based on this work, we estimate that net-metered PVresidences could save 5% of their total electricity load for houses without storage and 14% for houses withstorage. Direct-DC energy savings are sensitive to power system and appliance conversion efficienciesbut are not significantly influenced by climate.

. Introduction

A convergence of factors is driving recent interest in using directurrent (DC) from photovoltaic (PV) systems in its DC form to powerlectricity loads in buildings, rather than converting it first to alter-ating current (AC), as is current practice. The new millennium hasitnessed sustained and rapid growth in the adoption of rooftop

V systems, as concerns about climate change have intensified. PVs a DC power source. Batteries also act as a DC source and arehe dominant energy storage technology used with PV systems. Inddition to these two factors, an increasing fraction of the mostfficient electric appliances operate internally on DC [1,2], makinghe direct use of DC (direct-DC) in a building more effective andompelling.

Devices that operate internally on DC, referred to inhis paper as ‘DC-internal’ appliances, include all consumerlectronics—therefore, essentially all advanced communicationsechnologies—fluorescent lighting with electronic ballasts, solid-tate (such as light-emitting diode or LED) lighting, and brushlessC motors. Advanced brushless DC (permanent magnet) motorsan save 5–15% of the energy used by traditional AC induc-

ion motors and up to 30–50% in variable-speed applications forumping, ventilation, refrigeration, and space cooling [3]. DC-otor-driven heat pump technologies for water and space heating

∗ Corresponding author at: 1 Cyclotron Road, Mail Stop MS90-4000, Berkeley, CA4720, USA. Tel.: +1 510 4952521.

E-mail address: evossos@lbl.gov (V. Vossos).

378-7788/$ – see front matter. Published by Elsevier B.V.ttp://dx.doi.org/10.1016/j.enbuild.2013.09.009

Published by Elsevier B.V.

can also displace conventional resistance heating with savings of50% or more. Thus, ‘DC-internal’ technologies tend to be moreefficient than their AC counterparts and are capable of servicingessentially all building loads [3]. These trends make a strong argu-ment for investigating the potential benefits of directly couplingDC power sources with DC loads.

The direct use of DC has been recommended as a key strat-egy for improved reliability and increased energy savings at thebuilding level [4–6], and it is already being implemented in com-mercial buildings, particularly for lighting applications [7], whileDC-compatible appliances are emerging on the market [8]. How-ever, residential applications have received little attention anddiffer considerably from commercial applications. Most impor-tantly, residential loads have poorer coincidence with PV systemoutput than commercial loads and are less predictable. These issueswould appear to make the residential sector a poorer candidate fordirect-DC than the commercial sector. Acknowledging these bar-riers, this study assesses the relative energy savings of direct-DCpower for residential buildings.

The majority of studies that address DC power systems in thecontext of electricity savings have been analytical, rather thanexperimental, in nature. Savage et al. [9] estimated that electric-ity savings of 25% can be achieved in the U.S. residential sector byreplacing appliance AC-to-DC converters with a more efficient cen-tralized rectifier and using DC distribution to power DC-internal

loads. Hammerstrom [10] compared the power conversions forvarious residential appliance categories under AC and DC powerdistribution and found that a residential building coupled with aDC power source will use 3% less electricity with DC distribution,

224 V. Vossos et al. / Energy and Buil

Nomenclature

AC alternating currentBDCPM brushless permanent magnet motorDC direct currentLED light emitting diodeMPPT maximum power point trackerNEMS national energy modeling systemPV photovoltaic

cLilUttcmtaasdcn

pmdbwetttb[td

STctist(o

2

2

mtrtcd

consumption devices and to distribute DC power throughout thehouse with fewer losses. Low voltage is used for low-power loads,like consumer electronics and lighting, to facilitate safer and easier

SAM system advisor modelVSD variable speed drive

ompared to AC distribution. Thomas et al. [11] reported that anED lighting system supplied with DC power from PV can reducets levelized annualized cost by 5% on average, as opposed to an LEDighting system supplied with AC power from PV. However, a 2002.K. study [12], found that a residential PV-powered DC distribu-

ion system with net-metering was 3% less efficient compared tohe equivalent AC distribution system. Finally, Sannino et al. [13]ompared the distribution losses of a DC power system in a com-ercial building with different supply voltages ranging from 48 VDC

o 326 VDC, to an AC power system at 230 VAC. According to theirnalysis, the highest tested DC voltage (326 V) was the most suit-ble level, from an economic and technical standpoint. The reportedavings (or losses) of DC distribution in these studies are largelyependent on the varying assumptions about power system effi-iencies, load timing and type, distribution voltage, and whetheret-metering was taken into consideration.

This paper estimates the potential energy savings of direct-DCower systems in net-metered residences in the United States. Net-etered systems are considered explicitly here, both because they

ominate on-site PV generation [14,15] and because savings coulde overestimated if the DC-to-AC power conversions that occurhen excess PV power is delivered to the grid are ignored. How-

ver, because of the increasing capacity of net-metered PV systems,he intermittence of the solar resource may become a barrier toheir future penetration, because of too much power being releasedo the grid during solar peak periods [16]. Therefore, given that localattery storage, if handled properly, could buffer such fluctuations17] and reduce the mismatch between PV generation and load [18],his paper also explores the impact of energy storage systems onirect-DC energy savings.

Because of the large variability in insolation across the Unitedtates, the paper examines energy savings potential in 14 U.S. cities.his paper also includes a detailed load analysis to account for thehanges in the nature of the load needed to facilitate direct-DC ando account for the timing of the load. The latter is essential because,n the absence of energy storage, only loads coincident with PVystem output can benefit from direct-DC. Finally, we investigatehe potential benefits of shifting cooling loads to earlier in the daypre-cooling) to make these loads more synchronous to PV systemutput and, therefore, more able to benefit from direct-DC.

. Direct-DC house modeling

.1. Model inputs

To address the research objectives, we developed a spreadsheetodel of a hypothetical house with a net-metered rooftop PV sys-

em. To test the potential effect of large variations in insolation, we

an the model for an average house in 14 cities distributed acrosshe contiguous United States. These cities, shown in Fig. 1, werehosen because they were the only cities for which consistent resi-ential load data were available in the desired format, as described

dings 68 (2014) 223–231

below. The distribution of the 14 cities is analogous to the solarresource distribution in the United States.

To obtain electricity load data and PV system output for the aver-age house in each of the 14 cities, we used the System Advisor Model(SAM) [20]. The load data are provided in SAM as example averageresidential electricity loads and are climate-simulated for each hourof the year. These loads, therefore, already incorporate any buildingenvelope effects. For the PV output data, we used SAM to generatehourly estimates of PV system output for one year for each of the14 cities.1 It should be noted here that the use of simulated hourlyload profiles and PV output data is likely to overestimate the instan-taneous PV output that can be absorbed by the load [21] and thesystem storage dynamics, thus affecting the final energy savingsestimates.

2.2. Model development

2.2.1. Distinguishing the cooling loadsBecause of the potential importance of load timing and type

on energy savings and the large diurnal and seasonal variability ofcooling loads, we separated cooling loads from non-cooling loads.To do so, we first converted each city’s load data to load data forthe average day of each month and plotted the resulting aver-age diurnal load curves. An example is shown for Sacramento inFig. 2, which also includes the average PV output for June andJanuary, represented by the dotted lines. Based on an examina-tion of the load data, cooling loads are clearly distinguished fromnon-cooling loads. As seen in the graph, six monthly load curveshave clearly distinguishable afternoon-to-evening cooling loads,while the non-cooling load curves of the remaining six months arealmost matching. Accordingly, the cooling load was obtained bysubtracting the non-cooling load from the total load. Note that forcities with a potentially significant heating load during the winterperiod (Seattle, Medford, Helena, Denver, Chicago, Lexington, NewYork), we calculated the baseline load from months with minimumheating or cooling activity (April and October) to avoid includingwinter electric heating load. We used this approach based on themethodology provided by Reichmuth [22].

2.2.2. House configurationsTo quantify the potential energy savings of direct-DC, the model

compares power conversion losses in a house with AC distribution,called the AC-house, and a house with DC distribution called the DC-house, as shown in Fig. 3. The DC-house power system configurationeliminates DC–AC–DC conversion losses to DC-internal applianceswhen adequate PV power is available, but incurs AC–DC losses viathe bidirectional inverter when grid backup power, delivered as AC,is used. In the AC-house, which constitutes the base case, all poweris distributed in AC form to appliances that accept AC power. In theDC-house, all power is distributed in DC-form to appliances thataccept DC power, but these appliances are identical in every otherway to their AC counterparts. That is, the AC appliances are assumedto be the DC-internal appliances with an AC–DC power converteron the input.

As discussed earlier, cooling loads are separated from non-cooling loads, while the latter are further broken up to high- (380 V)and low-voltage (24 V) loads. High voltage is used for high-power

1 The inputs used in SAM to generate PV system outputs are 180◦ azimuth, 20◦

PV array tilt angle, and a 0.85 derate factor. Each house’s PV system DC rating was1 kW, but the actual PV system capacity was later scaled to allow zero-net electricityconsumption for the conventional AC-house, as discussed below.

V. Vossos et al. / Energy and Buildings 68 (2014) 223–231 225

Fig. 1. PV solar resource map.Reproduced with permission from the author [19].

Fig. 2. Average monthly diurnal load curves for Sacramento. From top to bottom,t(T

hrspmfd

iobida

2

tadpcr

Table 1Power system full-load conversion efficiencies.

Power system component Efficiency(%)

Component efficiencyin literature

PV Inverter (AC House), includes MPPTa 95 [25]: 90%, [26]: 95%DC-house rectifier (meter → DC)b 93 [27]: 90%, [26]: 95%,

[28]: 90%DC-house inverter (DC → meter)b,c 97 Not available in the

marketCharge controller or MPPTd 98 [29], [30]: 97–99%DC-house DC–DC converter:

380 V–24Vb95 [24]: 90%, [25]: 95%

Battery (one way) 90 Depends on technologyand state of charge

a Typical of today’s new PV-system string inverters.b Represents best models that could be built today, according to industry experts

interviewed.

he larger peaks correspond to July, August, September, June, October, and Maydescending). The superimposed base load curves correspond to November–April.his variance is attributed to the cooling load.

andling and flexibility. The chosen voltages for the DC-houseeflect existing (24 VDC) and pending (380 VDC) EMerge Alliancetandards for DC distribution [24]. It should be noted that, based onrevious work [13], and the fact that high-power loads eventuallyake up for two-thirds of the total house load (versus one-third

or low-power loads), we assume that AC-house versus DC-houseistribution losses are comparable.

For any given city, the PV arrays for the AC- and DC-houses aredentically sized; that is, they are configured to have the same DCutput. But, the capacity of the PV systems differs from city to city,ecause in each locale the systems are sized so that the AC house

s net-zero in annual electricity usage—that is, the total electricityrawn from the grid equals the electricity delivered to the grid onn annual basis.

.2.3. Power system conversion efficienciesBecause DC power systems are only beginning to emerge on

he market and are not yet produced for residential applications,ll power system component efficiencies were based on similar

evices used for other purposes and are representative of high-endroducts. Table 1 presents the power system conversion efficien-ies assumed in the model and corresponding efficiencies found inecent literature. The listed efficiencies reflect the input of industry

c Today’s PV-system inverter minus the MPPT, which has estimated losses of 2%.d Typical of today’s high-end charge controller efficiencies.

experts at the 2011 Green Building Power Forum, including makersof the new generation of DC power supplies for data centers, andby Emerge Alliance members.

2.2.4. Switching to DC-internal loadsTo make a fair comparison of the performances of the AC- and

DC-houses, their loads needed to be identical except for their powerinput characteristics. To obtain residential end-use consumptionat as high a resolution as possible, we ran the Energy Informa-tion Administration’s 2010 release of the National Energy ModelingSystem (NEMS) using the Annual Energy Outlook reference caseassumption. This resulted in an average annual U.S. residential elec-tricity consumption for 2010 for 32 different appliances. We thendetermined whether these appliances could operate on DC powerby considering the internal functions of the appliances. Table 2summarizes the results of this investigation.

With energy efficiency guiding the selection of the hypotheti-cal suite of appliances for both houses, we decided to: replace allnon-DC-compatible equipment with DC-internal models currently

on the market; replace electric resistance heating applicationswith DC-driven heat pump technologies where applicable modelsexist (electric water heaters, electric dryers, electric furnaces); andreplace all incandescent lights with electronic (fluorescent or LED).

226 V. Vossos et al. / Energy and Buildings 68 (2014) 223–231

Fig. 3. AC- and DC-house power system configuration. Only components that generate, convert, and consume power are shown. The AC-house inverter (top) includes MPPT.T inclu

Ta

fcwcaittGs

he DC-house’s bidirectional inverter (bottom) does not include MPPT, because it is

his suite of appliances constitutes the efficient DC-compatible loadssumed for both the AC- and DC-house load modeling.

The house loads used for the modeling were generated asollows: The NEMS load data were separated into cooling and non-ooling loads, and the energy usage for each constituent end-useas adjusted according to the DC-internal savings potential indi-

ated in Table 2. The aggregate percentage savings for the coolingnd non-cooling loads were then calculated and applied to cool-ng and non-cooling loads inferred from the SAM data for each of

he 14 cities. The overall weighted average energy savings relativeo standard residential loads were calculated to be about 33% (seearbesi et al. [3] for details). This had the effect of both scaling andhifting the loads across load categories.

ded separately [23].

2.2.5. AC–DC appliance conversion efficienciesBecause the appliances in both houses were DC-internal, each

AC-house appliance was assumed to have an AC–DC converterappropriate to its power consumption. We based the converterefficiencies on external power supply data from the ENERGY STAR[31] and 80plus [32] databases, both of which include the mostefficient products in the market. Fig. 4 shows the compiled effi-ciencies versus power supply power output from these two datasets. These efficiencies were applied to the NEMS appliance data,

given typical wattages. Average conversion efficiencies were thendetermined by weighting the load fractions. The weighted averageAC/DC appliance converter efficiencies for cooling and non-coolingloads were calculated to be 90% and 87%, respectively.

V. Vossos et al. / Energy and Buildings 68 (2014) 223–231 227

Table 2Residential appliances functions and equivalent DC-internal technologies.

Function within appliance Appliance type Standard technology DC-internal best technology Energy savings comparedto standard technologya

Lighting Incandescent, fluorescent, LED Incandescent Electronic 73%Heating Heater Electric resistance Heat pump operated by

BDCPMb50%

Cooling Motor (compressor, pump, andmotor-driven fan)

Induction motor, single-speedcompressor, pump, and fanwhere applicable

BDCPM operating VSDc 30–50% (VSD)

5–15% (motor onlydepending on size)

Mechanical work Motor Induction motor BDCPM 5–15% (depending on size)Cooking Electric cook top Electric resistance Induction cooker 12%Computing Digital technology Digital technology is already

DCSame 0

2

2

tsseiaul

2

i

TS

g

a Energy savings assuming AC power sourceb BDCPM: brushless DC permanent magnet motorc VSD: variable-speed drive

.3. Modeling scenarios

.3.1. Overview of system configurationsTo compare the energy use of the AC- versus the DC-house and

o test implications of storage and load shifting, we considered theystem configurations presented in Table 3. Note that, for everyystem configuration, the AC-house is identical to the DC-house,xcept for the power system components and the form (AC or DC)n which power is delivered to the loads. Thus, both houses aressumed to have identical electricity storage systems in config-rations where storage is considered (1b and 2b) and the same

oad-shifting mechanisms in configurations 2a and 2b.

.3.2. Configurations with storageTo test the effect of battery storage on the direct-DC energy sav-

ngs, battery storage was included in the model runs for both the

Fig. 4. AC/DC power converter efficiencies of AC-house appliances.

able 3ystem configurations for the modeling scenarios.

Without electricity storage With electricity storage

1a. Average residential loada 1b. Average residential load withstorage

2a. Shifted average residential load 2b. Shifted average residential loadwith storage

a Configuration 1a, average residential load (no energy storage) was presentedraphically in Fig. 3.

AC- and the DC-houses.2 The charge controller, which is assumedto include maximum power point tracking (MPPT) to optimize PVpower output, regulates current to and from the batteries. The stor-age system is charged only by excess PV power, but not by rectifiedgrid power. The AC-house inverter is bidirectional, as is the normfor modern grid-interactive inverters with battery back-up. Fig. 5shows system configuration 1b, Average residential load with stor-age, for both houses.

Battery efficiency was assumed to be 90% one way [33] (81%roundtrip), as shown in Table 1. To identify a reasonable value forthe maximum charging capacity of the battery (in kWh), we ranthe model for one city (Sacramento) and performed a sensitivityanalysis to determine how the amount of excess PV power sent tostorage varied with battery capacity. The results of this analysis arepresented in Fig. 6. For charging capacities of up to about 10 kWh,a linear relationship exists between the charging capacity and thepercentage of excess PV sent to storage. For charging capacitiesgreater than 10 kWh, the relationship becomes one of diminishingreturns. Taking into account the results of this analysis, we assumeda battery capacity of 10 kWh for Sacramento. For every other city,the battery size was scaled to the PV system size, normalized tothe optimal size for Sacramento. In this way, we used an integratedapproach to design each house optimally for each city’s climate. Theminimum battery charge was set to 20% of full capacity, a typicalvalue for deep-cycle batteries.

An additional objective in sizing the battery storage was to havethe single battery configuration work reasonably well for all cities,so that the performance intercomparisons were affected by cli-mate alone. Therefore, after sizing the battery, the model was runwith the 10 kWh-storage capacity for all cities and the followingoutcomes were examined:

• The percentage of time the battery spends at minimum and max-imum capacity: ideally not too large a fraction of the time shouldbe spent in either state, but both states should be manifested.That is, a battery that is never maxed out is larger than needed.If the battery never drew down all the way, this would indicatethat the battery capacity was not being effectively accessed.

• The percentage of loads that are not coincident with PV outputbut are serviced by the battery: this should be reasonably high toindicate that the battery is being effective at servicing loads.

2 Because the model compares energy losses between the AC-house and the DC-house, only the storage system efficiency affects the modeling results and not theassumed storage technology.

228 V. Vossos et al. / Energy and Buildings 68 (2014) 223–231

gurat

•

aAaDmtotsa

Fig. 5. House confi

The percentage of excess PV power that would have been sent tothe grid in the absence of storage but is sent to storage instead.This should be reasonably high for storage to be effective atbuffering the grid from large variations in power output.

As shown in Table 4, on average the DC-houses’ batteries spendn estimated 73% of their time in the active state. On average, theC-houses’ batteries spend an estimated 66% of their time in thective state. In addition, for the same size battery in the AC- andC-houses, the DC houses’ batteries tend to spend more time atinimum capacity, favoring smaller, and therefore, lower cost bat-

eries. The modeling also indicates that a relatively large fraction

f excess PV is captured by batteries and, therefore, used to servicehe on-site load rather than being sent to the grid. Thus the batteryizing appears to be both adequate for all of the cities and effectivet redirecting excess PV to the house load.

ions with storage.

2.3.3. Configurations with load shiftingTo test the potential of load shifting to improve direct-DC sav-

ings, we modeled the impact of shifting the residential cooling loadto start two hours earlier in the day throughout the cooling months(May–October). The cooling load was shifted because cooling dom-inates residential electricity use in general, it is possible to shift toearlier hours (as opposed to other residential loads such as lighting,cooking, refrigeration) and because the residential cooling load isskewed toward evening hours, as shown in Fig. 2. The intent herewas to capture the potential of pre-cooling to improve direct-DCenergy savings, not the usage of specially designed thermal stor-age systems. Therefore, load shifting was limited to two hours.

The house configurations with load shifting do not require anyadditional power system components, apart from a home energymanagement system, which is assumed to have a negligent effecton house electricity consumption.

V. Vossos et al. / Energy and Buildings 68 (2014) 223–231 229

Table 4Storage system performance in the AC and DC houses.

# Cities Cooling loadfraction (%)

Time battery is atminimum capacity (%)

Time battery is atmaximum capacity (%)

Loads non-coincident withPV serviced by battery (%)

Excess PV sent tobattery (%)

AC DC AC DC AC DC AC DC

1 Phoenix 66 23 19 11 12 54 62 69 642 Tampa 56 26 18 7 9 64 76 83 763 Houston 48 22 14 6 9 68 79 87 794 Fort Worth 43 19 13 6 10 68 76 87 775 Sacramento 32 32 23 6 9 68 78 87 806 Atlanta 28 22 14 5 8 70 80 90 827 Lexington 19 26 17 5 8 69 80 88 828 Medford 15 34 24 9 10 63 73 80 759 Los Angeles 15 28 15 4 7 72 86 92 88

10 New York 13 28 17 6 9 69 80 89 8311 Denver 11 25 14 5 8 72 84 92 8712 Helena 10 29 20 8 11 64 73 82 7513 Chicago 10 29 18 7 10 66 76 85 7914 Seattle 3 34 25 10 11 54 63 70 66

2

pctsDcccai

3

dieraf

wfso

Fw

Averages 27 18

Standard deviation 5 4

.3.4. Model runsThe model tracks the efficiency losses throughout the houses’

ower conversion systems and in the AC appliances’ AC–DC poweronverters. The model calculates the impact of net-electricity athe electric meter for both houses over a one-year period for eachystem configuration. The reported energy savings are the direct-C savings as a percentage of the total AC-house load for eachity. The model was run for all house configurations and for allities. In addition to these model runs, sensitivity analyses wereonducted to test the effect of the power system converters oper-ting under partial load conditions and possible future technologymprovements.

. Results

The energy savings reported in this section address only theirect-DC energy savings. The overall 33% appliance efficiency sav-

ngs, which were obtained from switching existing appliances tofficient DC-internal appliances, are excluded. Table 5 shows theesults for system configurations 1a and 1b (with and without stor-ge, but no load-shifting). The cities are ranked by their cooling loadractions to reflect the effect of climate.

The model predicts that the direct use of DC will save energy

ith respect to conventional AC distribution and that the savings

or battery-integrated systems are about twice that of non-storageystems. Averaging over all cities, direct-DC saves an estimated 7%f total (AC-house) electricity use without storage (1a) and 13%

ig. 6. Excess PV to storage versus maximum battery charging capacity (consistentith Mulder et al. [34].

7 9 66 76 84 782 1 6 7 7 7

with storage (1b). The results show only a weak trend betweencooling load fraction and direct-DC savings. The average fractionof the load serviced directly by the PV system is significant, butvirtually the same, for the AC- and DC-house, 37% and 38%, respec-tively. For system configurations that include load shifting (2a and2b), the results show no significant impact on direct-DC savings,from the two-hour load shift, compared to their corresponding non-load shift configurations (1a and 1b). This is because the load shiftincreased the fraction of load serviced directly by the PV systemonly modestly and by about the same amount (by 5%) – to 42% and43% – in both the AC- and the DC-houses, respectively.

3.1. Sensitivity analyses

3.1.1. Technology improvementsAs discussed earlier, direct-DC savings depend inherently on the

relative efficiencies of the power system components and the appli-ance converters. Although we use current high-end efficiencies forthe modeling, it is likely that these technologies will improve inthe future. Therefore, we ran the model for all cities testing effi-ciency improvement scenarios. These scenarios, and their resultingdirect-DC energy savings for system configurations 1a and 1b, arepresented in Table 6.

As expected, if rectifier and DC/DC converter efficienciesimprove, direct-DC energy savings increase. The opposite occurs ifappliance AC–DC conversion efficiencies improve. Given that suchimprovements are likely to proceed together, the relative effects arelikely to cancel each other out, and, therefore, the model estimatesof energy savings will be relatively insensitive to future changes inthe efficiencies of power system components and appliance powersupplies.

3.1.2. The effect of variable loads on energy savingsPower converter efficiencies are lower under low-load condi-

tions. Given that power converters are sized to meet maximumloads, the variability of residential loads should reduce the energysavings potential below what would be achieved at the ratedfull-load efficiencies. To model the magnitude of the impact thatpart-load conditions might have on direct-DC energy savings esti-

mates, we assigned part-load efficiencies for five power systemcomponents, as shown in Table 7. Part-load efficiencies were con-sidered for load levels below 20% of full load, because power systemefficiencies drop sharply below that level [30,35].

230 V. Vossos et al. / Energy and Buildings 68 (2014) 223–231

Table 5Direct-DC savings and load serviced directly by PV (system configurations 1a and 1b, Table 3).

Cities Cooling load (%) Fraction of load serviced directly by PVsystem (%)

Direct-DC savings as percent of totalAC house load (%)

AC-house DC-house No storage Storage

Phoenix 66 41 42 7.6 11.8Tampa 56 44 45 8.0 12.6Houston 48 43 44 7.9 13.0Fort Worth 43 40 41 7.6 13.2Sacramento 32 37 38 7.4 13.1Atlanta 28 38 40 7.5 13.3Lexington 19 37 38 7.4 13.2Medford 15 34 35 7.2 12.2Los Angeles 15 36 37 7.3 13.2New York 13 36 37 7.3 13.2Denver 11 34 35 7.2 13.4Helena 10 35 36 7.2 12.8Chicago 10 35 36 7.2 12.9Seattle 3 32 33 7.0 11.9

All cities (averages) 37 38 7.4 12.8

Table 6Direct-DC savings for improved power system and appliance technologies.

Scenario description Efficiency improvements Non-storage savings (%) Storage savings (%)

Improved power systemconversion efficiencies

House rectifier: 93% → 95%9.3 13.7DC/DC converter: 95% → 97%

Improved appliance AC–DCconversion efficiencies

Cooling loads: 90% → 95%Non-cooling loads: 87 %→ 90%

Table 7Power system components part-load efficiencies.

Power system component Full-loadefficiency (%)

Part-loadefficiency (%)

AC-house inverter, includes MPPT 95 90a

DC-house rectifier (meter→DC) 93 84b

DC-house inverter (DC→meter) 97 92c

Charge controller or MPPT 98 94d

DC–DC converter: 380 V–24 V 95 87b

a Based on the efficiency curve of the grid-connected string inverter SMA SunnyBoy 7000US [35].

b Based on 220 V AC–DC and 400 V DC–DC step-down power supply efficiencycurve [Personal Communication, Tony Lai, Delta Corp.].

c Based on SMA Sunny Boy 7000US efficiency curve, excluding MPPT losses (2%).d Based on the efficiency curve of the MorningStar SunSaver charge controller

with MPPT [30].

Fig. 7. Effects of part-load conditions on direct-DC savings for the average city.

4.0 9.3

We incorporated the part-load efficiencies in the model (config-urations 1a and 1b) for all cities. Fig. 7 shows the direct-DC energysavings for the average city.

Partial-load effects reduce estimates of direct-DC energy sav-ings from 7.4% to 5.0% for the non-storage case, but increase themfrom 12.8% to 13.5% for the storage case. The decrease in savings forthe non-storage configuration (1a) is because of the low part-loadefficiency of the DC-house rectifier. On the other hand, the increasein savings for the configuration with storage (1b) is because of thehigher AC- versus DC-house losses incurred between the batteriesand the loads, due to the presence of the inverter in the AC-house.Given the reduction in the already modest estimates of DC energysavings without storage, and the fact that actual loads are signif-icantly more variable than the modeled average loads, savings inthe field may be lower than estimated by the model. On the otherhand, given the uncertainties in the input values in Table 7, based,as they are, on a snap-shot of an emerging market, these resultsmight not persist in the long term.

4. Conclusions

This paper finds that direct-DC could yield significant energysavings in U.S. houses with net-metered PV systems, if the entireload is constituted of DC-powered appliances, especially if thosesystems incorporate battery storage of sufficient capacity to sig-nificantly buffer the grid from PV system output fluctuations.Accounting for variable loads, for the average city direct-DC is esti-mated to save about 5% of total house electricity consumption forthe non-storage case and about 14% for the storage case. Addi-tional energy savings of approximately 33% would be obtained bycompleting the current transition toward the use of DC-internaltechnologies to supply residential electricity demands. While thistransition is occurring even in the absence of direct-DC power

systems, largely because of the efficiency advantages, this benefitdemonstrates that the appliance modifications needed to accom-modate direct-DC are consistent with overall energy efficiencygoals and trends.

d Buil

tetidtcdtt

wslDtthcitiwHnccc

A

tDac

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

V. Vossos et al. / Energy an

The modest savings for the non-storage case are a reflection ofhe fact that residential loads, which peak in the late afternoon andvening, do not facilitate the direct use of PV power. The largerhe mid-day and summertime loads, the greater the energy sav-ngs potential will be. On the other hand, on-site storage favorsirect-DC, because excess power from the PV system can be sent tohe batteries for storage directly in its DC form without DC–AC–DConversion losses. Given that high PV penetration rates desired forecarbonization of the electricity supply also require on-site elec-ricity storage, direct-DC could facilitate by significantly reducinghe effective PV system load.

In conclusion, while the modest energy savings for a houseithout electricity storage might not provide the impetus for tran-

itioning away from an entrenched AC energy infrastructure, thearger savings for a house with storage might. More likely, if direct-C takes off in the residential sector it will be as a spin-off of

he commercial sector, for which products are already enteringhe market, mainly because commercial buildings tend to haveigher day-time loads than residential buildings do, so their loadsoincide better with DC output from PV. Starting with feasible DCnfrastructure investments, one can then move an application at aime, implementing those that are most advantageous (e.g., light-ng, sensors, projectors, fans). In the residential sector, one might

ant to start with high power hard wired systems like solar assistedVAC. We also note that, for technologies that are already DC inter-al, there is no inherent reason why a DC power system shouldost more than an AC system. In fact, DC appliances should be lessostly compared to AC appliances because they eliminate the AC/DConverter.

cknowledgements

The authors thank the following people for their significant con-ributions to this project: Robert Van Buskirk for initiating theirect-DC Power Systems project and providing vision and encour-gement along the way, Eric Fry and Tony Lai for input on poweronversion technology, and Mary James for editing.

eferences

[1] K. George, DC Power Production, Delivery and Utilization, Electric PowerResearch Institute, 2006 (White Paper).

[2] Galvin Electricity Initiative, The Galvin Path to Perfect Power – A TechnicalAssessment, Galvin Electricity Initiative, Palo Alto, CA, 2007.

[3] K. Garbesi, V. Vossos, H. Shen, Catalog of DC Appliances and Power Systems,Lawrence Berkeley National Lab, Berkeley, CA, 2011.

[4] M. McGranaghan, et al., Renewable Systems Interconnection Study: AdvancedGrid Planning and Operations, Sandia National Laboratories, 2008.

[5] Zero Energy Commercial Buildings Consortium (CBC), Barriers and IndustryRecommendations for Commercial Buildings, Next Generation Technologies,2011.

[6] M. Sechilariu, B. Wang, F. Locment, Building-integrated microgrid: advancedlocal energy management for forthcoming smart power grid communication,Energy and Buildings 59 (2013) 236–243.

[7] Nextek Power Systems, Direct-Coupling Demo. Available from:

http://www.nextekpower.com/technology/direct-coupling-demo (cited3 January 2013).

[8] EMerge-Alliance, Registered Products. Available from: http://www.emergealliance.org/Products/RegisteredProducts.aspx (cited 10 October2011).

[

dings 68 (2014) 223–231 231

[9] P. Savage, R.R. Nordhaus, S.P. Jamieson, DC Microgrids: Benefits and Barriers, inFrom Silos to Systems: Issues in Clean Energy and Climate Change, REIL, Editor,Yale Publications, 2010.

10] D.J. Hammerstrom, AC Versus DC Distribution Systems. Did We Get it Right?Power Engineering Society General Meeting, IEEE, 2007.

11] B.A. Thomas, I.L. Azevedo, G. Morgan, Edison revisited should we use DC circuitsfor lighting in commercial buildings? Energy Policy 45 (2012) 399–411.

12] DTI, The Use of Direct Current Output from PV Systems in Buildings,2002.

13] A. Sannino, G. Postiglione, M.H.J. Bollen, Feasibility of a DC network for commer-cial facilities. Industry applications, IEEE Transactions on Industry Applications39 (5) (2003) 1499–1507.

14] U.S. Energy Information Administration (EIA), Domestic Shipments of Photo-voltaic Cells and Modules by Market Sector, End Use, and Type, Available from:http://www.eia.gov/cneaf/solar.renewables/page/solarphotv/table3 7.html(cited 3 January 2013), 2010.

15] NCSC and IREC, Database of State Incentives for Renewables and Effi-ciency: Net Metering Policies Summary Map, 2011, Available from:http://www.dsireusa.org/solar/summarymaps/ (cited 3 January 2013).

16] P. Denholm, et al., The Role of Energy Storage with Renewable Electricity Gen-eration, National Renewable Energy Laboratory, Golden, CO, 2010.

17] P. Denholm, R.M. Margolis, Evaluating the limits of solar photovoltaics (PV) inelectric power systems utilizing energy storage and other enabling technolo-gies, Energy Policy 35 (9) (2007) 4424–4433.

18] H. Lund, A. Marszal, P. Heiselberg, Zero energy buildings and mismatch com-pensation factors, Energy and Buildings 43 (July (7)) (2011) 1646–1654.

19] B. Roberts, Photovoltaic Solar Resource of the United States, 2008, Avail-able from: http://www.nrel.gov/gis/images/map pv national lo-res.jpg [cited15 August 2012).

20] SAM, System Advisor Model. Available from: https://sam.nrel.gov/ (cited 31August 2012).

21] H. Lund, Excess electricity diagrams and the integration of renewable energy,International Journal of Sustainable Energy 23 (4) (2003) 149–156.

22] H. Reichmuth, A Method for Deriving an Empirical Hourly Base Load Shape fromUtility Hourly Total Load Records. ACEEE Summer Study on Energy Efficiencyin Buildings, ACEEE, Pacific Grove, CA, 2008.

23] F.C. Lee, et al., Proposal for a Mini Consortium on Sustainable Buildings andNanogrids, Center for Power Electronic Systems, Virgina Tech., Blacksburg, VA,2010.

24] EMerge Alliance, Standard, 2012, Available from:http://emergealliance.org/Standard/Overview.aspx (cited 15 August 2012).

25] P. Paajanen, T. Kaipia, J. Partanen, DC supply of low-voltage electricity appli-ances in residential buildings, in: CIRED 2009. 20th International Conferenceon Electricity Distribution, Prague, 2009.

26] S.A. Zabalawi, G. Mandic, A. Nasiri, Utilizing energy storage with PV for resi-dential and commercial use, in: Industrial Electronics, 2008. IECON 2008. 34thAnnual Conference of IEEE, 2008.

27] H. Pang, E. Lo, B. Pong, DC Electrical Distribution Systems in Buildings, in: 2ndInternational Conference on Power Electronics Systems and Applications, 2006.

28] M.R. Starke, L.M. Tolbert, B. Ozpineci, AC vs. DC distribution: A loss comparison,in: T&D, IEEE/PES Transmission and Distribution Conference and Exposition,2008.

29] D. Brearly, Distributed PV System Optimization: Microinverters, DC-to-DC andTwo-Stage Inverters, SolarPro, 2010, pp. 32–58, Home Power Inc.

30] MorningStar Corporation, SunSaver MPPT. Available from:http://www.morningstarcorp.com/en/sunsavermppt (cited 16 August2012).

31] Energy Star Energy Star EPS Specifications (dataset used to deter-mine Final Draft Version 2.0, Specification Levels). Available from:http://www.energystar.gov/index.cfm?c=archives.power supplies (cited15 May 2012).

32] ECOS, Power supplies efficiencies, 2010, Available from:http://www.80plus.org (cited 10 July 2010).

33] J.W. Stevens, G.P. Corey, A study of lead-acid battery efficiency near top-of-charge and the impact on PV system design, in: Photovoltaic SpecialistsConference, IEEE, Washington, DC, 1996.

34] G. Mulder, F.D. Ridder, D. Six, Electricity storage for grid-connected household

dwellings with PV panels, Solar Energy 84 (7) (2010) 1284–1293.

35] SMA, SUNNY BOY 5000-US/6000-US/7000-US/8000-US, 2010, Available from:http://www.sma-america.com/en US/products/grid-tied-inverters/sunny-boy/sunny-boy-5000-us-6000-us-7000-us-8000-us.html (cited 16 August2012).