How to construct a stochastic power plant - A -Trial Pantelis Koutroumpis, Zeynep Gurguc, Ralf Martin⇤, Mirabelle Muûls†, Tamaryn Napp and Ian Staffel 13th May 2014 Preliminary 1 Introduction The threat of climate change combined with a will to decrease imported energy dependency has resulted in a policy-driven increase in renewable power production. This trend is bound to increase in the coming decades. Figure 1 illustrates this by plotting the half-hourly (log) changes in power demand together with the half hourly changes in wind power for the UK during the first 3 days of 2009. Wind power often grows or shrinks by more than 20% whereas power demand never exceeds a change of 10%. A central challenge for the provision of larger shares of renewable power is their intermittency. Compared to fossil fuel alternatives, renewable power sources are typically more variable and less easy to predict. However, the power market is no stranger to variability and uncertainty. Power demand varies heavily throughout the day, the days of the week and times of the year. What’s more, there can be sudden demand peaks or troughs depending on arbitrary factors such as how entertaining a TV program is or the outcome of a football match. Equally, the power market equilibrium can be heavily affected by technical faults in power plants. At present this variability is by and large managed from the supply side; i.e. power plants are switched on and off to keep a balanced equilibrium of demand and supply. However, many have suggested that it could be vastly more cost effective to address some of this variability through better management of the demand side. Economists in particular have long suggested that the key to demand side management is a more variable pricing of electricity at the retail level (Faruqui and Sergici, 2010, Faruqui and Malko 1983). However, dynamic pricing can only work if power consumers spend cognitive resources on the state of the power market. It’s effect also needs to be persistent through time, yet the literature’s findings on the impact of financial rewards on energy consumption are mixed. Abrahamse et al. (2005) find these to be short lived, while others show they can be large and persistent over time (Dolan and Metcalfe, 2013). Besides, feedback and information to consumers in addition to financial incentives are effective measures to reduce energy consumption (Abrahamse et al. 2005; Allcott Imperial College Business School, South Kensington Campus, London SW7 2AZ, United Kingdom, Grantham Institute on Climate Change, and Centre for Economic Performance (CEP), London School of Economics (LSE). Email: [email protected] † Grantham Institute for Climate Change and Imperial College Business School, South Kensington Campus, London SW7 2AZ, United Kingdom, and CEP. Email: [email protected] ⇤ 1 -.4 -.2 0 .2 .4 Figure 1: Wind power versus total power demand (First 3 days of 2009) 0 50 100 150 half hours Dlntotal Dlnwind and Rogers 2012; Dolan and Metcalfe 2013), whereas social interaction and norms have been proposed as other power demand shifters, though their effect also decreases through time (Hori et al. 2013, Dolan and Metcalfe 2013). This raises the question as to whether ‘social’ demand side management, which is currently being researched by utility companies and others (McMichael 2013, Kellet 2007,Lockwood and Platt 2009) would be persistent enough to face the huge flexibility requirements of an increasingly renewable power provision. Alternatively, there is now much discussion of smart grids and smart devices that interact with smart grids, which would allow to achieve the needed demand side response automatically, without relying on the customer’s behaviour. At present grids are not terribly smart yet and most electrical devices are not smart enough to talk to smart grids. As an alternative we examine in this study a device that is not terribly smart but that potentially allows smart people to manage power demand at low cost and with current grid technology: the Power Balance Plug (POWBP). The POWBP is a simple power plug that can be fitted to any existing power plug. It has a wifi chip that allows it to connect to now common WiFi routers (73.3% of UK households have a WiFi network1 ), allowing to remote control the plug. We have trialled this setup in the context of a university student hall. We thereby contribute to the literature in two dimensions. First, , we assess the effectiveness of the POWBP as a short-term alternative to smart devices. Second, in subsequent waves of experiments, we hope to add to existing literature on the behavioral aspects of low-carbon technologies usage (Jaffe and Stavins 1994) and energy consumption responses to pricing. In this note we discuss our initial findings. 1 https://www.strategyanalytics.com/default.aspx?mod=pressreleaseviewer&a0=5193 2 0 5 kdensity kWh 10 15 Figure 2: Average Consumption: Trial versus control group average half-hourly electricity consumption 0 .1 .2 x Control group 2 A .3 .4 Trial group -Trial We are conducting our trial at a new student residence of Imperial College, comprising of more than 500 identical self contained studio flats with bathroom and kitchenette. The trial group consists of 12 students who were given POWBPS.2 The trial group was non-randomly selected on a first come first serve basis after an email was sent round inviting students to participate. Figure 2 shows density plots of the average (pre-trial) energy consumption of trial and control group. Both groups have a by and large a common support but the trial group consumes on average slightly more energy (0.09 vs 0.07) kWh per 30 minutes.3 Figure 3 shows how consumption varies over the course of a day. Not surprisingly peak consumption is in the evening. The lowest average consumption is in the early morning hours. This is comparable to the average UK household consumption pattern as shown in Figure 4. Our main experiment are a series of randomly spaced switch off events. POWBPS plugs were given to each student from the trial group together with a 4-socket extension lead and instructions on how to use the plug. The plugs were then switched off at random points in time for 30 minutes. There would be at least a gap of 3 hours between two switch off events. The trial was conducted from January 14 until March 30. Figure 5 illustrates one week of the trial period (February 20 to 26). The blue vertical lines show the switch off events during this period, whereas the red line shows average consumption. 2 We are adapting wifiplugs from www.wifiplug.co.uk for that purpose. All our experiments below are in conducted within 30 minute intervals. This is the time resolution of the UK wholesale electricity spot-market. To ease comparison we report all figures in terms of 30 minute intervals. 3 3 Figure 3: Average (half-hourly) Consumption over hours of a day Linear Prediction .08 .1 .12 Adjusted Predictions of hours#betaid with 95% CIs Data analysis .04 .06 In this section we look first across all households, for the whole year. Then we pick out a few different sub-groups of households: first single pensioners and other single-person households (because single person households raised some unexpected findings in the initial study), and second we examine electricity use in flats and detached houses (because these dwelling types are at opposite ends of the scale and occupancy spectrums). Finally we compare the electricity profiles for all homes monitored during the coldest and the warmest months. The figures do not match 0 presented 1 2 3 in 4 the 5 original 6 7 analysis 8 9 10 13 14 15 16 17 18 19 because 20 21 22 similar figures of 11 the12 Household Electricity Study of 23 hours households where more than 10% of improved data cleaning. The original analysis also excluded the energy use was unknown. In contrast, we have opted to include all households, in order to betaid=1 keep sample sizes as large as possible. betaid=0 This analysis uses raw data, unadjusted for seasonal effects. This means the sample sizes for short periods of the year are rather small, since the analysis includes only the dwellings monitored over the period selected. All households, whole year 800 Heating 700 Water heating Showers 600 Washing/drying Watts Figure 4: The mean daily profile for all homes included in the study across the whole year indicates that on average cold appliances draw very similar power from the grid throughout the day, see chart below. Conversely, audiovisual, lighting, electric cooking and ‘washing/drying’ (which also includes dishwashing) vary quite significantly through the day. In fact the averaging across homes and times of year reduces the variability of these components of the profile, and in reality they fluctuate much more than the graph suggests for specific groups and/or periods. Average Consumption over hourshousehold of a day - UK Household Electricity 500 Cooking 400 Lighting Cold appliances 300 ICT 200 Audiovisual 100 Other Unknown 0 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 24:00 The evening peak is very pronounced, and made up largely of electricity used for cooking in the early evening, transferring to lighting and audiovisual later in the evening. The evening peak is much more pronounced than the morning peak, and accounts for 50% higher peak load when averaged across all homes and periods. Overall, although there is an identifiable morning peak at about 7.30am, lower demand for lighting, space heating and cooking are almost 4 9 Survey 26/02/2014 25/02/2014 24/02/2014 23/02/2014 22/02/2014 21/02/2014 20/02/2014 0 Power Consumption (kWh) .05 .1 .15 Figure 5: Seven days of switch off events Time 3 Evaluation of the experiment We use a flexible linear regression model to evaluate the effect of the switch off events on the trial group. Our most general specification is as follows: kW hit = SW IT CHt ⇥ CON N ECTit + ↵t + ↵hhi + ✏it where i indexes an individual participant, t indexes a specific half hourly period, SW IT CHt is an indicator variable equal to 1 during a switch off event, CON N ECTit is an indicator variable equal to 1 if a particular plug is connected to our server, ↵t is a period specific effect and ↵hhi is a half-hourly period of a day specific effect for every participating individual. 4 Results Table 1 shows the regression results. The results suggest that for the average connected plug a switch off event leads to a reduction of consumption by 0.0101 kWh. Put differently: the average user has been connecting a 0.0202kWh (= 0.0101 ⇥ 2) device during the average switch off period. That’s about the combined wattage of an iPad and an iPhone charger. Compared to the overall average consumption it is typically more than 10% (compare with Figure 3). In column 3 we investigate how this impact varies across different times of the day we look at 3 periods: Night (from 1 to 8 o’clock), Day (from 8 to 18 o’clock), Evening (from 18 to 1 o’clock). This suggest that during the night and day the impact is uniformly 0.01 kWh, whereas in the evening it is only 0 .007kWh. In columns 4 and 5 we also look at the level of electricity consumption. This suggests, that trial group residents consume more energy (as we have seen before), that energy consumption goes up over time, that trial group residents consume more during periods when their plugs were 5 Table 1: Seven days of switch off events Dependant Variable Switch off event X Connected Change in Power Consumption -0.0101*** -0.00974*** (0.00297) (0.00297) Connected 0.00107 0.000996 0.000996 (0.00121) (0.00107) (0.00107) Trial group . -0.000150 -0.000150 . (0.00106) (0.00106) Trial Period X Trial Group 0.000122 0.000119 0.000119 (0.00121) (0.00120) (0.00120) Trial Period -0.0000859 (0.000170) Switch X Connected X Night -0.0110** (0.00538) Switch X Connected X Day -0.0108** (0.00443) Switch X Connected X Evening -0.00703 (0.00534) Period fixed effects No Yes Yes Room X Half hour of day fixed effectsYes No No N 1835445 1835445 1835445 Power Consumption -0.00503* -0.00150 (0.00272) (0.00311) 0.00606*** 0.0133*** (0.00111) (0.00112) 0.0336*** (0.00111) -0.0308*** -0.0326*** (0.00110) (0.00126) 0.00844*** (0.000156) No Yes 1835445 Yes No 1835445 connected and that compared to the pre-trial period trial participants’ consumption increased less than consumption of the control. This latter effect could be a result of the incentive scheme we offered to participants which rewarded both, load balancing services as well as reductions in overall consumption. 5 Conclusion The initial results are promising. We can identify a statistically significant and economically meaningful effect of the plugs on energy consumption. In particular this effect appears considerably stronger than the effects identified by studies looking into real time pricing.4 In future work we plan to conduct similar experiments with a larger sample of consumers as well as different types. References Abrahamse, W., Steg, L., Vlek, C., and Rothengatter, T. (2005). A review of intervention studies aimed at household energy conservation. Journal of Environmental Psychology, 25(3):273–291. Allcott, H. and Rogers, T. (2012). The short-run and long-run effects of behavioral interventions: Experimental evidence from energy conservation. Technical Report 18492, National Bureau of Economic Research. Dolan, P. and Metcalfe, R. (2013). Neighbors, knowledge, and nuggets : two natural field experiments on the role of incentives on energy conservation. 4 e.g. Allcott (2009) finds effects that are at best equivalent to a 5% reduction of a households consumption. 6 Faruqui, A. and Malko, J. R. (1983). The residential demand for electricity by time-of-use: A survey of twelve experiments with peak load pricing. Energy, 8(10):781–795. Faruqui, A. and Sergici, S. (2010). Household response to dynamic pricing of electricity: a survey of 15 experiments. Journal of Regulatory Economics, 38(2):193–225. Hori, S., Kondo, K., Nogata, D., and Ben, H. (2013). The determinants of household energy-saving behavior: Survey and comparison in five major asian cities. Energy Policy, 52(C):354–362. Jaffe, A. B. and Stavins, R. N. (1994). The energy-efficiency gap. what does it mean? Energy Policy, 22(10):804–810. Kellet, J. (2007). Community-based energy policy: A practical approach to carbon reduction. Journal of Environmental Planning and Management, 50(3):381–396. Lockwood, M. and Platt, R. (2009). Green streets: Final report to British Gas. IPPR London. McMichael, D. (2013). The value of social networks in the diffusion of energy-efficiency innovations in uk households. Energy Policy, 53(0):159–168. 7
© Copyright 2024