What Works to Shift Consumption? Results from Large Scale Demand Response Trial International Conference on Energy and Cities Southampton - 10th July 2019 Tom Rushby* | Ben Anderson | Patrick James | AbuBakr Bahaj Engineering & Environment (Energy & Climate Change)

Introduction Can household demand response be used to mitigate traditional (expensive) low-voltage network reinforcement?

The problem 00:00 04:00 08:00 12:00 16:00 20:00 00:00 Load (kVA) Time of Day 2030 Low 2030 Central 2030 High Transformer Rating Weekday transformer load profile growth scenarios to 2030. Adapted from EA Technology [1] Distribution network operators face challenges in the transition to a low carbon energy system: • Load growth from • Electric vehicles • Electrification of heat • Intermittent distributed generation (e.g. PV) These low carbon technologies put pressure on existing low voltage network assets.

Project aims • Investigate what works to reduce domestic demand during peak hours? • Use findings to help DNOs to better model domestic demand (and demand response) and inform investment decisions • Provide statistically robust evidence

How? A large-scale randomised control trial Geographical location of SAVE project trial participants • Statistical power analysis to estimate required sample size • N ~ 4,000 households (1,000 per group) • Random, stratified sample in ‘Solent’ area • Random allocation to treatment groups • Household surveys to evaluate sample

Trial interventions Q1 Q2 Q4 Q3 Q1 Q2 Q4 Q3 2017 2018 TP1 TP2 TP3 Trial group 2 Subsidised bulbs Installed bulbs + data informed Trial group 3 Data informed & £ Data informed & £ Banded pricing (Opt-in) Trial group 4 Data informed only Data informed only Banded pricing (Opt-out)

Data collection Trial evaluation Household surveys LED installation data Energy Loop data Time-use diaries

LED installations

LED installations 76% uptake Average of 7 bulbs replaced per household 176 Watts/household theoretical max. reduction

• Max. effect 47 Watts per household (peak hours) • Seasonal effect (daylight availability) • Huge inter-household variation … • Wide confidence intervals (90% CI = 8 to -97 W) • Approx. 1 W observed reduction for every 3 W of theoretical savings LED impact evaluation: winter 2017/18

Banded pricing

Banded pricing Peak-hours consumption threshold Paid for every hour with consumption under threshold Weekly text message balance updates

Banded pricing Two treatment groups: Opt-in (top) Opt-out (bottom) 38% of group 98% of group Max. effect = 17W Max. effect = 44W

Conclusions • Banded pricing and LED upgrades offer similar potential load reduction in peak hours demand (at maximum) • LEDs more consistent but seasonal • Banded pricing provided short-term impact, sensitive to participant engagement • Banded pricing may be inconsistent (unreliable)

Thank you for listening. www.energy.soton.ac.uk/tag/SAVE [email protected] @tom_rushby