Demand Response with Smart Homes and Electric Scooters An Experimental Study on User Acceptance

Transcription

1 Demand Response wit Smart Homes and Electric Scooters An Experimental Study on User Acceptance Alexandra-Gwyn Paetz, Tomas Kascub, Patrick Jocem, olf Fictner Karlsrue Institute of Tecnology (KIT), Cair of Energy Economics ABSTRACT Smart tecnologies and electric veicles are supposed to efficiently use renewable resources by sifting loads and storing surplus electricity. Altoug several field tests wit smart meters, dynamic pricing or electric veicles are conducted, ardly any consumer as experienced a combination of tese components. So far, neiter consumer perceptions nor te effectiveness of tese demand response options are apparent. Test-residents were selected to move into KIT s smart ome for several weeks. It is equipped wit smart appliances and two e-scooters. An energy management system scedules EMS) te operation time of appliances according to external price signals. Everyting can be monitored on touc-screen panels provided in every room and on mobile devices. In tis experimental setting demand response was tested in four pases: First, testresidents were provided wit feedback. Ten, different electricity-tariffs were introduced. Finally, te test-residents were able to fully use te EMS to scedule operation times of te smart appliances automatically. Electric scooters were provided for free use during te wole period. In general, te test-residents sowed ig interest in detailed information on teir consumption. However, feedback alone ad neiter load-sifting nor conserving effects. Tese actions were realized wen dynamic pricing was introduced. Te willingness to sift consumption was limited to a few appliances (e.g., diswaser) and depended on monetary savings. Carging te electric scooters was only sifted, if tere was anoter veicle available for emergencies. Te automatic load-management ensured more convenience in sifting demand, especially in combination wit dynamic electricity prices and load-limits. Introduction Te aim of reducing greenouse gas emissions by 80% in Germany by 2050 wit respect to 1990 causes major canges in te energy sector: electricity is increasingly generated from renewable resources, tat are partly volatile, ardly controllable and to some extent generated decentralized. Tis callenge migt rise wit an increasing electrification of te transportation sector, as electric veicles (EV) will raise peak loads presumably in te evening ours. Different measures tat tackle tese callenges are under discussion and include demand response. Te aim is to matc electricity demand wit (renewable) supply by incentivizing ouseolds to respond accordingly. Tis implies some adaptions to everyday ouseold beavior, e.g., carging te EV at a different time of te day. However, promoting te sustainable use of electricity is particularly difficult, because it differs from oter goods: It is invisible, untoucable, and only consumed indirectly via related activities (Hargreaves, Nye & Burgess 2010). Providing ouseolds wit feedback, offering a price incentive and equipping tem wit automated solutions migt elp to sift loads. Some of tese measures ave been addressed in European and German Policy (directive 2006/32/EC, 21, 40 German Energy Act): utilities ave to offer some kind of electricity tariff tat motivates residential consumers to conserve electricity and/or sift teir demand. Furtermore, new buildings ave to feature smart meters given teir feasibility.

2 However, te number of German ouseolds tat are actually equipped wit a smart meter, accompanying display products or ave a dynamic electricity tariff is still very low. Only few smart appliances are being offered on te market and teir tecnical integration into some kind of ouseold management system is under researc. So ardly any consumer as experienced a smart future ouseold environment on a daily basis, and terefore neiter te user acceptance nor te effectiveness of tese solutions is clear. At te same time market penetration rates for EV are still low, only representing 0.02% in te car, 0.08% in te scooter, and 5% in te bicycle market (KBA 2011, ZIV 2011). Tus, tere are only few EVusers and even less tat ave additionally experience wit demand response. In order to evaluate demand response we ave analyzed te everyday electricity demand of test-residents in a smart ome wit electric scooters (e-scooters). Te load-sifting potential of te smart ome solutions was tested in four pases: First extensive feedback on te resident s electricity consumption as well as on te power generation by te potovoltaic system (PV) was provided to te test-residents. Ten different electricity tariffs were tested. Afterwards an automated energy management system (EMS) was introduced tat enabled te smart appliances to react automatically to dynamic prices. Two e-scooters were provided for free use during te wole period. Te test-living pase was accompanied by two in-dept interviews, a pre-post-questionnaire and an online blog. Te paper is organized in five sections: First we sortly summarize te state of researc. Ten, we describe te smart ome setting and explain te researc design, before presenting te results. e close wit a discussion of our results. Previous Researc Te main assumption beind most tecnologies introduced to field trials is based on an information deficit model (cf. Hargreaves, Nye & Burgess 2010): It is assumed tat consumers lack awareness of teir electricity consumption due to its invisibility. If consumers terefore ad enoug information, tey could be able to cange teir demand. Feedback is terefore a popular element wit smart meters. Te findings about energy response, i.e., decreasing energy demand, after introducing feedback systems are mixed: ile Darby (2010) indicates tat energy conservation of 5 25% are possible in er review, marginal to zero effects were reported in oter studies (cf. Pyrko 2011). From a micro-economic point of view, offering a financial incentive leads to optimal beavior of ouseolds, as tey are (financially) better off by consuming less at times of ig electricity prices. Results from some field tests confirm tis teory, as te effectiveness of feedback picks up wit financial incentives leading to lower electricity bills (cf. Cassing & Kiesling 2008). Tis can eiter derive from electricity conservation (lower electricity demand as a wole) or from load-sifting (substituting lower-cost off-peak consumption for igercost peak consumption). Mainly te second effect appears to exist: if peak reductions are acieved, tey are often accompanied by an off-peak increase (valley filling) and conservation effects are terefore minor to non-existent (EEE 2006). A meta-review of VaasaETT (2011) reveals tat not all dynamic pricing models ave te same effectiveness: wile Time-of-Use Tariffs (TOU) lead to an average load-sifting effect of 5%, Critical Peak Pricing (CPP) models proved to be most effective (16% less loads during peak priced ours). Figure 1 provides an overview of ow te different pricing models can be classified.

3 Parameter Exemplary TOU (time-of-use pricing) Figure 1: Classification of electricity pricing models Exemplary CPP (critical-peak pricing) Exemplary RTP (real-time pricing) Exemplary RTP wit load-limit Number of Two One Tree Tree Infinite price levels Time sceme Eac level valid twelve ours wit fix sceme (prices apply at same times every day) Flate price applies Eac level valid at for everyday (fix sceme) one our wit dynamic sceme (structure canges daily) Eac level valid at for one our wit dynamic sceme (structure canges daily) Load-limit / / / 3 k / Critical peak event / 6 times per year / / / Electricity price Ct/k 20 Ct/k; additional penalty (+15 Ct/k) wen CP-event is called ) t/k Rate (c a te R Off-peak 15ct/k Peak TOU Rate 25ct/k ) Flat Rate 20ct/k t/k (c a te R CPP Rate 35ct/k (6 days) Ct/k Ct/k; additional penalty (+ 5 Ct/k) wen passing load-limit ) t/k (c a te R Appro ximate Hourly Prices ) t/k (c a te R Appro ximate Hourly Prices Loadlimit Exemplary extreme RTP Hig dynamics, as consumer prices follow market prices Can adopt any price between 10 Ct/k and 35 Ct/k ) t/k (c a te R Appro ximate Hourly Prices Noon Noon Noon Noon Noon Based on Fox-Penner 2009 Residential consumers, owever, don t seem to be very open to beavioral canges in daily routines and require tat demand response options sould not be linked to reductions in comfort (cf. Paetz, Dütscke & Fictner 2012). Consumers terefore indicate ig acceptance of smart appliances and perceive ome automation systems positively. Tis also seems to be true for electric mobility. During a German field test wit battery electric veicles (BEV) dynamic pricing ad been introduced, but ad little to no effects (Paetz, Jocem & Fictner 2012). However, all car users reported teir willingness to drive emission-free and tus carge green electricity. Te idea of aving a smart carging station tat carged at low-price times automatically was tus igly accepted. To sum up, te current literature suggests tat demand response options ave te potential to support (some) reductions of peak-loads wit iger effects wen combining dynamic pricing wit feedback. Regarding user acceptance, te literature indicates tat consumers sow positive reactions wen confronted wit automated solutions. As literature findings do not allow formulating concrete ypoteses, tis study takes an experimental approac looking for first-and user response to a smart ome & mobility environment: How are smart tecnologies used in daily life? ic elements prove to be effective? at drives teir use? Is demand response applicable for electric mobility? Researc Design Smart Home Setting

4 In order to answer tese researc questions outlined above an experimental approac in a smart ome laboratory was employed. Te KIT smart ome 1 represents a building for two residents, wic consists of a 60 sqm two-bedroom apartment and a 20 sqm equipment room. A PV-system is installed on te roof and two e-scooters are integrated over a carging station. 2 Te kitcen is equipped wit smart appliances (wasing macine, dis waser, tumble dryer) tat are connected to a central communication gateway tat provides data on te status of eac appliance and is able to receive control signals. All appliances (smart and conventional) are monitored and integrated to te EMS. Based on te electricity price information te EMS is able to scedule te smart appliances automatically witin certain limits. Tese limits reflect te residents preferences, wic e/se can set by specifying a time frame in wic te appliances may run (degree of freedom). Te EMS ten calculates te most cost-effective running time and communicates it to te user on Energy Management Panels (EMP) tat serve as a connection between te EMS and te user. It displays relevant information and actions of te EMS and is provided on four touc-screen displays and a mobile device (ipod Touc). Two e-scooters are integrated to tis smart ome environment. Tese are Elmoto scooters of type HR2 wit a weel ub motor and a maximum speed of 45 km/. Recarging te litium-ion battery (1.1 k) it takes up to six ours at 350. E-scooters can be driven on all German roads, except for igways, and need an insurance indicator. Experimental Set-Up Tree test-living pases ave been conducted in tis setting (cf. Figure 2). Tis paper focuses on T3, as te results of T1 and T2 ave been publised in Paetz et al Te experimental set-up is structured into four modules wit te objective to analyze different effects isolated from eac oter keeping different variables witin a module consistent. Furtermore, we did not want to overstrain te residents. Figure 2: Experimental Set-Up of test-living pases in te smart ome Module Test T1 (4 weeks) T2 (8 weeks) T3 (5 weeks) 1 Feedback 2 Dynamic Pricing / 3 Automated EMS / 4 Electric Mobility / / Te objective in module one was to analyze te effects of feedback on electricity demand. Te feedback was provided by te EMP. Tus te residents were able to ceck te load of te ouseold and of eac appliance, as well as te PV-power in real-time. Te electricity price was also displayed, but was kept static during te first module. In module two several dynamic pricing models were introduced (cf. Figure 3). All tariffs were calculated wit an average price of Ct/k. ile we tested a variety of TOU and RTP tariffs in T2 experimenting wit time scemes, price levels, and spreads, we used just one 1 Te term smart ome is used for linking different ouseold devices to a network. Te term can terefore include aspects of ambient assisted living, entertainment, and security. Here we focus on energy management. 2 Termic devices (µchp, cooling ceiling) and a mobile storage (BEV) were not yet implemented in tis study.

5 type of RTP in T3. In week tree we added a load-limit to it, meaning tat a penalty applied for demand above tis limit. During te last two weeks (module tree) te automation functions of te EMS ad been activated and te full use of te smart appliances was possible. Bot e-scooters (module four) were available for free use during te wole period. Te test-residents did not pay any rent or expenses on energy. In order to offer tem an incentive a bonus-malus system was designed. For eac k consumed in te lowest price zone, te test-residents received one bonus point vice versa a malus point for eac k in te igest price zone as well as above te load-limit. Eac bonus point correlated to 0.5 and te value of te bonus balance was disbursed at te end of te test-living pase. Figure 3: Dynamic pricing models tested in T3 Tariff Standard TOU RTP RTP RTP RTP wit load-limit structure Time sceme Fix Dynamic Load-limit / / / / / 6 k 4 k varying: k Price levels Prices [Ct/k] Metodology and Sample Te experimental design does not allow gaining representative results, tus selecting test-residents is an important step. e recruited a young sample (-31 years), tat was selected wit a screening questionnaire tat aimed at finding tose wo were generally interested in te topic, but ad little knowledge and no experience wit te tecnologies under researc. Final selection took place after a ouse tour. ile T1 was carried out wit two students, two average adults (=non-students) lived in te smart ome during T2 and a mixed group (one student, one non-student) was recruited for T3. Besides te screening questionnaire and a sort standardized pre-post-questionnaire (on attitudes), we conducted two in-dept interviews a metod to explore opinions allowing close interaction wit te researcer and te tecnology. During te test-living pases te residents were able to write about teir experiences in an online blog. it te combination of tese survey types we gained a ric amount of data in te participants own words. Tis qualitative data was analyzed togeter wit quantitative beavioral data from te real-time metering. Results Te average electricity demand during te test-living pase was 68 k/week. Across all weeks te test-residents managed to consume around 50% of teir electricity demand at low-price (7 Ct/k) times and te oter 50% more or less equally split up at mid-price ( Ct/k) and ig-price (37 Ct/k) times (cf. Figure 4). Te functionalities of te automated EMS (in weeks four and five) did not increase te demand during low-price times. For two reasons: Firstly, te test-residents were able to adjust teir energy use well manually to dynamic pricing. Secondly, tey sifted te same appliances during te manual weeks as tose controlled by te EMS.

6 Figure 4: Electricity demand split into price levels 80 k60 in d n a40 e m d ly 20 e k 0 % % 28% 25% 30% 27% 21% 29% 100% 48% 51% 51% 46% Ct/k Ct/k 07 Ct/k eek By comparing te average daily load-curve accumulated over te five weeks wit te German norm load profile (H0), a quite similar load structure is visible during te day (cf. Figure 5). Bot peaks (around noon and in te evening) occur later indicating a load-sift. By adding te price curve over tat period, a reverse effect between electricity price and demand for te test-living profile can be observed, leading to an electricity bill reduction of 6.5%. Figure 5: Accumulated demand and pricing in comparison to normal load profile k20 in d15 n a 10 e m d our PV demand norm load profile H0 tariff c t/k in r if ta In te following we present te results of eac module and illustrate tem wit interview statements, blog posts as well as teir beavioral data 3. Module 1: Feedback Especially during te first days te test-residents made extensive use of te toucscreen panels. Tey ad fun analyzing te consumption and confirmed tat real-time feedback leads to a more sensitive use of appliances. Te test-residents were fascinated by te real-time information on te PV-power generation. As T3 took place during te winter period, tey were a bit disappointed tat tey were not able to consume more of te selfgenerated power (cf. Figure 5). Te toucpads are easy and fun to andle. I m personally very interested in te PV power. Luckily, we are connected to te grid, because oterwise we would sit a lot in te dark. 3 All statements ave been translated into Englis wile retaining te gist of te original German.

7 Tere were some additional applications on te EMP. Te weater forecast and te remote ligt control were appreciated most. Te test-residents even tougt tat tese were crucial for aving some kind of feedback device at teir own omes and ad many ideas for additional applications (e.g., roller sutter control). A weekly bill provided information on te electricity costs, te carging loads and a comparison of consumption wit an average German ouseold. Altoug one test-resident did not like te comparison, te oter test-resident tougt it was useful and was surprised tat teir consumption was not lower tan te German average. After using eac appliances a few times, tey felt tey knew teir demand quite well, so tere would be no new feedback information, unless a new device was introduced. Furtermore, tey saw teir electricity demand as given and ad no intention reducing it. After using te wasing macine twice, I know its consumption. e consume electricity anyway. I m not going to stop wasing my clotes only because I see ow muc is required for doing so. Module 2: Dynamic Pricing Te introduction of dynamic prices canged te use of te EMP and te devices. Te test-residents cecked te price forecast in te mornings and planned te use of te appliances accordingly. Splitting te overall demand into te single devices allows analyzing wic appliances were used at te different price zones and sifted accordingly (cf. Figure 6). Tis was especially te case for appliances wit time-independent use, e.g., te dis-waser. Appliances wit a ig sare in te red price zone sow an inconvenient load-sift over time. Not surprisingly, tose appliances are ligt, cooker, stove, and TV. For teir comfort and entertainment services te willingness to subject tese activities to dynamic pricing was low. e classify tem as appliances wit immediate use pattern and low sifting potential. In contrast appliances classified as permanent or time-independent are suitable for (automated) load-sifting, since tey ave ig electricity demand and low timing constraints. As an example te test-residents said to be able to postpone te dis-waser use by twelve ours witout constraints in teir daily routines. From teir perspective te dis-waser was te most suitable, te coffee-macine and te stove te least suitable devices for subordinating dynamic prices. Furtermore tey wised for a smart freezer tat would automatically plan its cooling periods according to te prices and allow cost savings in a comfortable way. Figure 6: Electricity demand of certain devices split into price zones e-scooters wasing masine dryer dis waser TV stove cooker ligt 7 ct/k ct/k 37 ct/k demand [k]

8 Te time sceme of te RTP also influenced te ability of te test-residents to sift appliance use. If low-price or mid-price zones applied on te brink of a day, it was perceived as too early or too late as to comply wit tem. Tis weekend we were very flexible and tus tried to use te appliances at lowpriced times. Only on Sunday evening te green zone started after 7 p.m. and we were not able to wait wit cooking for our guests until tat time. Let s see ow we will manage during te week! For sure I will not get up before 6 a.m. to brew coffee wit ceap electricity. In week tree we added a load-limit to te RTP tat was visualized on te EMP wit a separate bar. During teir activities te test-residents cecked to see if tey were able to use furter appliances witout surpassing te limit. In weeks tree and four we ad kept te loadlimit fixed at 6 k and 4 k respectively. Tese limits ad ardly any impact, being only violated during four minutes in eac week wit a total demand below 0.1 k. en starting to vary te load-limit down to 2.5 k, te limit became significant. Even toug, te amount violated ad no effect on costs wit a demand of 0.65 k, te violation time jumped up to 87 minutes. Tis low limit was not accepted very well and te appliances used anyway. Until 3.5 k it s fairly manageable to use te appliances time-delayed, but beneat tat it s impossible. Te stove alone violates te limit. As long as tere s no sortage, we can put up wit te penalty. en asked about teir preferences for dynamic pricing models for teir own omes, bot test-residents refused to ave one wit a load-limit, as tey felt it was too constraining. So tey were in favor for te RTP, even after sowing tem a TOU model. RTP would allow using some appliances also during te day and one would not be forced to wait until te quite static TOU price at nigt. Furtermore, one of te test-residents felt tat e would better get is sare of low market prices tan wit yearly pre-calculated TOU prices. CPP models were not accepted at all, as te test-residents felt, tat utilities sould be able to forecast te main influences on costs and tere sould be terefore no need in calling critical events. Furtermore, bot residents felt tat price levels sould be someow subject to a cap, fearing tat if tere would be no wind power generation over te course of several weeks, prices would stay constantly ig. Not only did tey feel tat it would be unfair, but also difficult to save on te electricity bill. In line wit te willingness to save money, iger price spreads were preferred to lower price spreads, as long as tere was a cap on te igest price zone. If te price differences were only 2 Cents, it wouldn t matter. Tere would be no motivation at all. I d ten rater turn on immediately te tumble dryer tan waiting and saving 4 Cents. But wen we talk about 20 or 30 Cents, I would really take care about tat. Te cost-saving expectation was around 150/year. Altoug tis expectation is similar to previous researc results (cf. Paetz, Dütscke & Fictner 2012), it is fairly ig. During tis test-living pase te residents saved around 6.5% on electricity costs. Projecting tis to a complete year te savings would be around 63. ile saving costs was te main driver for one test-resident to adopt is demand to electricity prices, ecological issues were central for te oter test-resident. Especially te idea of using renewable energies preferably from te own PV was appealing. Accordingly, se wised for more information about te generation mix of electricity. Module 3: Automated EMS In weeks four and five te automated EMS was activated and enabled te full use of te smart appliances. So far te test-residents ad managed to comply wit dynamic prices very well, because tey discovered and used te time-clockers integrated in some appliances. Tey did not switc rigt away from tis manual strategy to te automated EMS. ile te automated EMS also considered te load-limits wen setting te smart appliances, tis was a

9 matematical task in te manual case. Tus, te options of te automated EMS were used regularly, altoug te test-residents didn t fully trust it and cross-cecked, if it ad set te appliances correctly. I usually set te automated EMS first and ceck if it delivers viable running times for te appliances. Oterwise, I stop it and set tem again manually. Te automated EMS was mainly set for te dis-waser and te wasing macine. Even toug, te tumble dryer ad te smart functionalities, too, it was not used muc, as te test-residents feared to leave te wet clotes in tere for too long, before te program would start. en tinking about furter appliances tat would be suitable for te automated program, one test-resident tougt of EVs. Te automation would be ideal for an electric car. It could for example be carged in between two peaks of te wasing macine. Houseold appliances ave to run troug, but te carging of te car could be interrupted. Te automation makes most sense for cars. Module 4: E-Scooters Te objective of providing te e-scooters was to analyze te acceptance of te e- scooters itself and te effect of demand response on te carging beavior. As Figure 6 sows, te e-scooters did not represent a ig sare of te total electricity demand. Still, tey were mainly carged during low-price times. Similarly to te oter ouseold appliances tere was a ig willingness to sift te carging process, but late times limited to put tis into action. In addition te carging plugs were outside, wic limited tis action even more. Generally I tried to comply wit te prices, but yesterday for example te low-price was only on at mid-nigt and tat was too late. Some kind of clock-timer or even te full automation would be great, because it is annoying to go out again and plug te scooter in. Instead of plugging te e-scooters immediately after arriving to te smart ome, one test-resident tried to wait until a low-price time and if se didn t manage, se took te oter e-scooter. Tis strategy was possible, because se was te main user. Apart from a few exceptions te oter test-resident stuck to using is car. I ave not found te ideal application for te e-scooter. Maybe because I was never used to riding a scooter or a motorcycle? itin te city I prefer to walk and for longer distances I use my car. Most of te times I also carry tings wit me and te e-scooter as only limited capacities for tat. Te main user rode te e-scooter every morning to go to work and enjoyed not arriving as sweaty as wit te bicycle. In te afternoon se used it to get to er sports activities. According to er perception er mobility patterns did not cange apart from te fact tat se was now using te e-scooter instead of er bicycle. However, te feeling of riding te e-scooter was quite different. ile se enjoyed te fast acceleration, se felt limited not being able to ride faster tan 42 km/, especially on state roads. Furtermore se felt tat many pedestrians did not respect er rigt of way. In contrast to tat te car drivers paid a lot of attention and were very interested in tis electric veicle. E-scooters are still a new ting in traffic and people react quite differently. Most are interested and approac me wit questions. Oters don t seem to take me seriously on tat veicle maybe because it s so quiet? In order to make te e-scooter a more attractive veicle, te test-residents suggested many improvements. Some consider te veicle itself, suc as aving flasing indicators (instead of indicating by and), oter ideas consider te battery. Having an additional battery in excange would increase te flexibility also wit regard to dynamic pricing. Sorter carging times would make carging out of ome (e.g., at te gym) more attractive.

10 Before T3 ended, te test-residents were asked for teir final evaluations of te tecnologies and wic, if any, tey would coose for teir own omes. Bot said tey would like to ave a dynamic electricity price wit some kind of feedback device tat sould not be too expensive. Altoug tey tougt tat te oter possibilities wit te automated EMS and te smart appliances were interesting, it would be too cost-intensive to install tem in already existing omes. Discussion and Conclusion Summary and Discussion of Results In tis study an experimental test-living pase was conducted wit te objective to analyze te acceptance of smart ome tecnologies and e-scooters in daily life. A smart ome laboratory on KIT s campus was used as a residential setting to provide tecnical innovations tat enable load-sifting as well as electricity conserving: two feedback options, dynamic pricing, and an automated energy management system. Tese tecnologies were tested in a modular way and were accompanied by different surveys. In our study direct feedback options in real-time (in-ouse displays) were more effective tan indirect ones (enanced billing) wit regard to increasing awareness for electricity use. In general, te test-residents sowed ig interest for detailed information on teir electricity consumption. However, no electricity conservations were induced by te feedback itself. Toucscreen display wit real-time data on electricity prices became te crucial source of information, wen dynamic pricing was introduced to te experiment. Motivated by te experimental set-up and te bonus-malus system, te test-residents managed to adapt most of teir electricity demand to dynamic prices and load-limits. Dynamic pricing turned out to be overall effective in increasing te effort of sifting at least some loads. One strategy employed was te use of clock timers. In order to respect te loadlimits in a comfortable way te test-residents made use of te automated energy management system tat controlled te use of te smart appliances (dis-waser, wasing macine, tumble dryer). In our study te functionalities of te automated EMS didn t increase te amount of electricity sifted, because it applied to te same appliances tat were sifted manually beforeand. However, te test-residents confirmed tat in te long-run automated options were needed especially in combination wit electric veicles. For teir own omes, te testresidents reported interest in dynamic pricing (wit a strong preference for RTP over TOU or CPP models) and some simple form of feedback device tat wouldn t cost too muc. Accordingly te central motifs beind te load-sifting efforts were cost-saving potentials as well as te use of (self-generated) renewable power. Bearing te participant selection structure in mind (interested in energy-related topics, open-minded for new tecnologies, eterogeneous wit regard to daily life routines) te analysis of te results is particularly interesting. Even in tis interested sample te testresidents ad little knowledge about teir electricity consumption patterns before moving in and reported aving fun exploring it. Tis points out tat interactive solutions tat enable users to observe teir own electricity consumption face te callenge of marketing electricity services. Te involvement wit te feedback function decreased over time during te testliving pase and monetary savings became more important, wic was also found to be one of te main underlying motives in sifting loads. Results sow tat te dis-waser, te wasing macine and te e-scooters were te devices mainly subjected to load-sifting. As tese appliances also represent a ig sare of te overall ouseold electricity demand, tey are most suitable for load-sifting. However te acceptance of load-sifting was limited by various factors. (1) Not all daily activities and routines are easily siftable, e.g., wen

11 working ours collide. (2) If some activities can be sifted, tey are not necessarily subjected to sifting, because tey deliver comforting, entertaining, or mobility services. Tis again points out tat electricity is not demanded as an independent good, but as an enabling service and adoptions are arder to make. (3) Oter activities tat involve te consumer less, like using te tumble dryer, are fairly easy to perform at anoter time of a day. ile te innovativeness of te smart ome setting was appreciated during tis experimental study, te test-residents would probably not be willing to spend te same amount of time to familiarize wit te tecnical equipment in teir own ouseolds. Terefore convenient solutions and a customer-friendly support are recommended in order to allow acceptance of tese tecnologies by a broader consumer base. Furter value-added features (suc as a eating control) can increase teir attractiveness, too. e recommend tat future smart ome solutions satisfy te consumers desire to get easy and simple advice on ow to save costs, provides information on te energy-mix, and secures ig levels of flexibility. Demand response is especially suitable for electric veicles due to te required carging power, energy volumes of teir batteries, and te long parking times. For tese reasons furter application areas for electric veicles ave to be found and marketed, if a iger sare of electric veicles is to be on our roads. Tis is especially important wit regard to carging management options. Limitations As wit all empirical work, tis study is subject to several limitations. Certainly, te generalizability of te findings is limited. Te test-residents recruited for tis experiment are not representative for any kind of population. Moreover, tey signed up voluntarily, so views and experiences of individuals not interested in tis kind of tecnology are not covered ere. en recruiting test-residents we deliberately tried to engage individuals wo are likely to be among te early adopters. ile te KIT s smart ome offers a unique residential setting, it is also a limitation, because different set-ups may also elicit different beaviors. Tis is also true for te bonusmalus system. Te value of a bonus points was iger tan te real value of te electricity price. ile te test-residents earned 33 bonus points ( 16.5), teir real savings would ave been a fraction of tat ( 4.76). As a lot of tecnical equipment in te front- and back-end tat is required, but not yet available on te market, it was impossible to conduct te experiments wit a control group. Terefore, direct comparisons to any oter ouseolds are limited. Even toug te testliving pase was fairly long for being an experiment under laboratory conditions, no conclusions can be made on te long-term effects and beaviors. Conclusions Our study seds more ligt on te acceptance of demand response wit regard to smart omes and e-scooters. It turns out tat wen smart tecnologies are used, load-sifting efforts are observable. Te incentives for tis load-sifting beavior are monetary savings and te use of renewable energy resources. e furter looked at te use of e-scooters in tis environment and can confirm load-sifting efforts for te carging process, too. Automated solutions, tat enable load-sifting in a comfortable way, become even more important in combination wit e-scooters. Load-sifting potentials migt even increase if electric veicles experience a market penetration. Terefore, furter analyses on te acceptance of sifting car-carging and

12 integrating te car battery into te EMS (e.g., in combination wit te PV system) seems wortwile. Besides te electrical components also termal components could be integrated into te EMS, as energy conserving and monetary savings are igly probable. In order to callenge tis study s results furter experiments wit oter target groups are needed. Based on te results of tis study we recommend tat policies take tese beavioral influences on electricity demand of residential consumers into account. Greater transparency on demand and costs are an incentive in te sort run (and smart meters migt terefore be a good starting point), but in te long run additional benefits ave to be perceivable suc as monetary savings and a sustainable environmental impact. Tese incentives tat go beyond feedback will enable te acceptance of demand response, and are terefore decisive on our progression towards a more efficient energy system. References Cassin, D.P. & L. Kiesling. 2008: Decentralized Coordination troug Digital Tecnology, Dynamic Pricing, and Customer-Driven Control: Te Gridise Testbed Demonstration Project. In Te Electricity Journal 21 (8): Darby, S Smart metering: wat potential for ouseolder engagement? In Building Researc and Information 38: [EEE] Energy & Environmental Economics Inc A survey of time-of-use (TOU) pricing and demand-response (DR) programs. San Francisco, Calif. Fox-Penner, P Smart Power, Climate Cange, te Smart Grid, and te Future of Electric Utilities. asington, DC: Island Press, Hargreaves, T., M. Nye & J. Burgess Making energy visible: A qualitative field study of ow ouseolders interact wit feedback from smart energy monitors. Energy Policy 38; [KBA] Kraftfartbundesamt Bestand an Personenkraftwagen nac Kraftstoffarten sowie Bestand an Elektrokrafträdern in Deutscland. Paetz, A.-G., E. Dütscke, &. Fictner Smart Homes as a Means to Sustainable Energy Consumption: A Study of Consumer Perceptions. Journal of Consumer Policy 35(1): Paetz, A.-G., P. Jocem, &. Fictner Demand Side Management mit Elektrofarzeugen Ausgestaltungsmöglickeiten und Nutzerakzeptanz. In Proceedings of te 12 t Symposium Energy Innovation. Graz: TU Graz. Paetz, A.-G., B. Becker,. Fictner, & H. Scmeck Sifting Electricity Demand wit Smart Home Tecnologies An Experimental Study on User Acceptanc. In 30 t USAEE Conference Proceedings. asington: U.S. Association of Energy Economics. Pyrko, J Am I as smart as my smart meter is? Swedis experience of statistics feedback to ouseolds. In Proceedings of te ECEEE 2011 Summer Study, Presqu'île de Giens: European Council for an Energy Efficient Economy. [VaasaETT] Vaasa Energy Tink Tank Te potential of smart meter enabled programs to increase energy and system efficiency a mass pilot comparison. [ZIV] Zweirad-Industrie-Verband Der deutsce Farradmarkt im ersten Halbjar 2011.