1. An investigation of the feedback and feedforward mechanisms required for Crowdsourcing occupant datasets for a UK school stock model
Duncan Grassie
oC? ppm? kWh?

2. Today’s presentation Modelling school buildings
Why? – Within the context of energy demand reduction targets
How? - Stock modelling and preceding methods
Stock modelling case study
Creation and potential auto-generation of datasets
Issues with stock modelling process
Sensitivities run
Crowdsourcing
How does it work? – Data sharing for shared knowledge
Feedback and feedforward mechanisms
Future Work – a framework for creating and testing a Crowdsourcing platform

3. Modelling school buildings
Why? – Within the context of energy demand reduction targets
How? - Stock modelling and preceding methods
Stock modelling case study
Creation and potential auto-generation of datasets
Issues with stock modelling process
Sensitivities run
Crowdsourcing
How does it work? – Data sharing for shared knowledge
Feedback and feedforward mechanisms
Future Work – a framework for creating and testing a Crowdsourcing platform

4. School buildings in context of energy demand
Non-domestic buildings
Any building that’s not residential! ~1.8 million in UK
Carbon emissions for heating, lighting, power account for 18% of total UK
Schools
Comprise around 47,500 premises in UK
6% of electrical, 13% of non-electrical non-domestic consumption
Discrepancy between:
a) Operation measurements and design predictions
b) Similar buildings with different energy consumptions

5. Modelling techniques Top-down Bottom-up
Statistical analyses of headline figures
Using national level datasets
ie Statistical methods, artificial neural networks
Model informs on a national level
Does not provide causal detail
Bottom-up
Aggregating end-use components
Using individual post occupancy studies
ie Building physics
Performance attributed to end-uses
Intensive datasets/gathering

6. Stock models – National scale building simulation
SimStock – auto-generation of energy simulation models
National level Lidar and taxation databases
Requires input on fabric, energy services offered
Ref: Coffey, B. et al. (2015) ‘An epidemiological approach to simulation-based analysis of large building stocks

7. Modelling school buildings
Why? – Within the context of energy demand reduction targets
How? - Stock modelling and preceding methods
Stock modelling case study
Creation and potential auto-generation of datasets
Issues with stock modelling process
Sensitivities run
Crowdsourcing
How does it work? – Data sharing for shared knowledge
Feedback and feedforward mechanisms
Future Work – a framework for creating and testing a Crowdsourcing platform

8. Case study school campuses
Typically a mixture of Victorian, 1960s and post-2000 builds

9. Data Sources for occupant data
Setpoints and equipment specs
Schedules

10. Use of Archetypes for fabric data
1950s map
1960s map
Ref: Digimaps – online historical OS maps

11. Problems with modelling school buildings
Modern archetypes
How many floors?

12. Results and sensitivities

13. Sensitivity to heating controls

14. Modelling school buildings
Why? – Within the context of energy demand reduction targets
How? - Stock modelling and preceding methods
Stock modelling case study
Creation and potential auto-generation of datasets
Issues with stock modelling process
Sensitivities run
Crowdsourcing
How does it work? – Data sharing for shared knowledge
Feedback and feedforward mechanisms
Future Work – a framework for creating and testing a Crowdsourcing platform

15. Crowdsourcing – data gathering on grand scale
The Problem:
Occupant schedules, equipment, setpoints required for effective feed forward
Head teachers, janitors know this data but what motivation for passing it on?
Crowdsourcing
Recruiting volunteers on a mass scale by sharing aggregated knowledge / reward
Monetary example – Crowdfunding – Brewdog
Knowledge example – Extra terrestrials!
In schools?
Compared to peers/standards
Ref: Carbon Trust
Cost of fixing?

16. Feedback and feedforward mechanisms1 2 3
Future research project:
Design feedback mechanism to occupants based on workshops with school governors/head teachers and buildings engineers on motivations/knowledge required
Design structure of aggregated data to be fed forward based on metrics / sub-sectors of stock specified by
Measure if individual school user data gathered from a crowdsourcing platform improves the accuracy of fedforward individual detail on energy performance

17. Future Work

18. Summary The important role of school buildings in UK energy reduction
As a proportion of UK non-domestic buildings
Access to national datasets (DECs) and energy efficiency measures
Stock modelling can provide:
Causal data (bottom-up) on entire sub-sector performance (top-down)
Case study demonstrates:
Stock models requires individualised data on school schedules and setpoints to feedforward underperforming schools to policy makers
Datasets of occupant data over entire population
Require new methods of motivating / educating volunteers (feedback)

19. Acknowledgements
This work was supported by the EPSRC Centre for Doctoral Training in Energy Demand (LoLo), under Grants EP/L01517X/1 and EP/H009612/1.
Thanks to Dr. Sung-min Hong and Steve Evans for providing the geometry for the three case study schools
Any questions?