Slide 1

Slide 1 text

Seeing the light The effects of LED light bulb installation on electricity demand in UK households: results of a large n randomised control trial Tom Rushby, Ben Anderson (@dataknut), Patrick James, AbuBakr Bahaj

Slide 2

Slide 2 text

The menu l The problem − Peak electricity demand l The solution − Reducing & shifting demand l Seeing the light − Large n LED light bulb trial l Did it work? l What do we need to do next?

Slide 3

Slide 3 text

Despite this... l The peak is still − Peaky − Expensive − Dirty Source: Staffell (2018) https://doi.org/10.1016/j.enpol.2016.12.037 UK electricity decarbonisation 2009 - 2015 l Solutions − Reduce it − Shift it UK electricity generation 2009 vs 2015

Slide 4

Slide 4 text

The menu l The problem − Peak electricity demand l The solution − Reducing & shifting demand l Seeing the light − Large n LED light bulb trial l Did it work? l What do we need to do next?

Slide 5

Slide 5 text

Study Design l South East England − Stratified random sample − N ~= 4000 households − Randomly allocated to 4 trial groups − http://www.energy.soton.ac.uk/save-data-sources/ Monitoring Data cloud Analysts Statistical power analysis LED trial is one of these • W every 10s • Wh every 15 min

Slide 6

Slide 6 text

Implementation Data: Winter 2017-2018

Slide 7

Slide 7 text

Where were they put?

Slide 8

Slide 8 text

Theoretical ‘saving’ l Maximum theoretical total installed saving = 155 kW across group l Actual total installed = 124 kW Mean: 176W per household

Slide 9

Slide 9 text

What happened? • Mean Wh • 16:00 – 20:00 only

Slide 10

Slide 10 text

What happened? • Consumption as % of control • 16:00 – 20:00 only • • Demonstrates need for difference in difference model

Slide 11

Slide 11 text

What happened? l Winter: w/c 29 Jan 2018 l Differences visible

Slide 12

Slide 12 text

But... • 16:00 – 20:00 • Difference in Difference Model • Huge inter-household variation • Lack of precision

Slide 13

Slide 13 text

Interactions... Household size Retired Children present Difference-in-difference regression interacting attributes with treatment Rented Unemployed

Slide 14

Slide 14 text

Was it worth it? • ‘Biggest’ week (mid-winter): • Median -23W (-33W in peak) per household • Median -3.9 kWh per household per week • ~£0.70 p/w! • Complex payback period calculations • Seasonal effect in load reduction • Install ££ • Network load avoidance • Customer savings l Thank you! l http://www.energy.soton.ac.uk/tag/save/ l @dataknut l [email protected] / [email protected] Work in Progress :-) 176W installed