Increased sustainability in the food retail sector through Measurement and Verification of energy conservation measures

Økt bærekraft ved Måling og Verifikasjon av energitiltak i dagligvaresektoren

PhD thesis defence

Alexander Severinsen
Norwegian University of Life Sciences
School of Economics and Business

April 13, 2023

Buildings consume about 40% of all produced energy

Significant contributors to greenhouse gas emission

Energy efficient buildings is vital to reduce emissions

Picture from NASA earth observatory Showing city lights during night It looks beautiful from space But is an ample reminder… people are expanding everywhere

Main objectives of thesis

• Uncertainty about energy savings is a barrier that hinders new renovation projects
• Contribute to lower that barrier
• Promote action to create more energy efficient buildings

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

Agenda

• Background
• Example of energy saving documentation
• Present the 5 papers
• Conclusion

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

Background

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

• Retrofitting projects of grocery stores
• Contracts with guaranteed energy savings
• Energy monitoring system for 2 200 retail stores
• My main responsibilities are to document and monitor the energy savings

Motivation 1) Do a good job 2) Transparent methods 3) Personal reason: environmental impact

Buildings in Norway

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

Buildings in Norway

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

Food retail stores has the largest energy intensity among the building categories

Source: Enovas byggstatistikk 2017. Note: The U.S. Energy information Administration (EIA) reports a similar measure for food retail stores with 524 kWh/m2, the highest energy intensity of any building category

Documentation of energy savings

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

?

Example

• Retrofitting of a grocery store in 2020
• Ventilation and cooling system are upgraded. New LED-lights
• You guarantee the customer a 50% reduction of energy use

A year after the retrofitting the results are finally available…

• Great! A 50% reduction 😎
• Customer argue that the result was because of the mild winter in 2021!
• You correct for the temperature differences¹

1. More information about adjusting for degree days: https://www.enova.no/kunnskap/graddagstall/

Corrected for
temperature

Documentation of energy savings

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

You study the store in more detail…

• Someone forgot the lights on during night
• Ventilation on during off-hours
• Error in the cooling system

It is not possible to change the weather ❄️

The other issues might have been fixed sooner

Documentation of energy savings

Objectives
Agenda
Background
Examples
Paper 1 / 5
Conclusion

Solution:
a model to continuously monitor energy savings

Measurement and Verification

Measurement and Verification

Objectives
Agenda
Background
Examples
Paper 1-5
Conclusion

Measurement and Verification

Objectives
Agenda
Background
Examples
Paper 1-5
Conclusion

Measurement and Verification

Objectives
Agenda
Background
Examples
Paper 1-5
Conclusion

• The models I will show are based on a consensus approach…
• International Performance Measurement and Verification Protocol (IPMVP)
• Initiative from Efficiency Valuation Organization (EVO)
• They developed a consensus approach to measuring and verifying efficiency investments

Modeling energy savings

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Paper 1: Weeks versus hours

• Energy savings in 5 retrofitting project
• Broken line versus Van Tao benchmarking model
• Improved approaches to estimate energy savings

Paper 2: “Easy” versus “advanced”

• Energy savings in 11 retrofitting projects
• Time series models on hourly level
• Van Tao benchmarking versus gradient boosting model

Severinsen, Alexander, and Øystein Myrland. 2022. “Statistical learning to estimate energy savings from retrofitting in the Norwegian food retail market”. Renewable and Sustainable Energy Reviews, Volume 167, 112691. https://doi.org/10.1016/j.rser.2022.112691
Severinsen, Alexander, and Rob J. Hyndman. 2019. “Quantification of Energy Savings from Energy Conservation Measures in Buildings Using Machine Learning”. In ECEEE Summer Study Proceedings, 757–66, https://www.eceee.org/library/conference_proceedings/eceee_Summer_Studies/2019/4-monitoring-and-evaluation-for-greater-impact/quantification-of-energy-savings-from-energy-conservation-measures-in-buildings-using-machine-learning/

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

30 768 kWh in week 9

-16 \(^\circ\)C

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

17 596 kWh in week 30

22 \(^\circ\)C

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

A lot of variation in the year you are comparing against
That is, one reason, why a model is useful!

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

library(segmented)
load(data_2018) # Before retrofitting (2018)
fit.glm <- glm(kwh ~ tam, weight = kwh, data = data_2018)
fit.seg <- segmented(fit.glm, seg.Z = ~tam, psi = 6)
ref_data$pred_kwh <- predict.segmented(fit.seg, newdata = data_2018)

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

• Reliable model CV-RMSE < 0.2¹

¹CV-RMSE = Coefficient of Variation Root Mean Square Error. International Performance Measurement and Verification Protocol (IPMVP), Efficiency Valuation Organization (EVO). https://evo-world.org/en/products-services-mainmenu-en/protocols/ipmvp

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Savings during the winter

Energy Temperature curve

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Savings during the winter

Savings during the summer

• Next we move on from 52 to 8760 observations
• Lets look at the data

A year of hourly data

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Temperature (°C) and energy consumption (kWh) for a grocery store (n=8760).

A year of hourly data

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

Features
Month
Week
Hour
Week x Hour
Temperature
Temperature²
Temperature³
Temperature x Month
Temperature² x Month
Temperature³ x Month
Temperature x Hour
Temperature² x Hour
Temperarure³ x Hour

Hong, Tao. 2010. “Short Term Electric Load Forecasting.” PhD thesis. https://repository.lib.ncsu.edu/handle/1840.16/6457

• Tao Vanilla benchmarking model (Linear regression)
• Benchmark model in the GEFCom2012 load forecasting competition
• Best 25 of 100 teams

Temperature (°C) and energy consumption (kWh) for a grocery store (n=8760).

• We have opening hours
• Weekends
• Shifting seasons
• Temp. difference across day and night..months

Van tao benchmarking model

Objectives
Agenda
Background
Examples
Paper 1-2 / 5
Conclusion

• Train a model on one year of hourly data for 2018
• Temperature and calendar data
• Can the model mimic how the actual data looks like?

• Predict the energy consumption in week 15
• Model seems to perform fairly well
• CV-RMSE = 0.087

• Use the 2018 model to predict the energy consumption in week 15 in 2021
• The model predict the energy consumption as if the building performed similar to 2018…
• …but with the temperature and calendar data from 2021

☀️

🌛

• The energy savings from the retrofitting can be studied in detail
• Day by day and night versus day
• Enables a better understanding of the measures that were taken

• The challenge is that we are depending on the energy analyst!
• We have developed a web framework to set models into production

Monitoring energy savings

Objectives
Agenda
Background
Examples
Paper 3 / 5
Conclusion

• Energy data collected from Elhub on an hourly level
• Temperature data was collected from Met.no through publicly open API
• Building data: Location, name and size through customer API

• ELhub = central repository for meter readings for all buildings in Norway
• Production and consumption
• Owned by Statnett (datterselskap)

Monitoring energy savings

Objectives
Agenda
Background
Examples
Paper 3 / 5
Conclusion

• Reactive programming using R/Shiny
• Fast development & automated energy savings report
• Not dependent on the energy analyst
• Open source and transparent
• As of today 1 241 buildings are monitored continuously

Severinsen, Alexander, and Øystein Myrland. 2022. “ShinyRBase: Near real-time energy saving models using reactive programming”. Applied Energy, Volume 325, 119798. https://doi.org/10.1016/j.apenergy.2022.119798

Getting the methods into production would have never been possible without the genius code from the fingertips of Einar-Johan Hansen and Stian Berger (@Capia) - and years of endless discussions with Petter Arnestad and many other collaborators!

Web application

Objectives
Agenda
Background
Examples
Paper 3 / 5
Conclusion

Step 1:
User chose building

Step 2:
Year to compare against

Step 3:
Date when finished retrofitting

Energy savings = difference between actual and predicted consumption

Model quality can be assessed

Typical retrofitting project

Objectives
Agenda
Background
Examples
Paper 4 / 5
Conclusion

• Study of energy efficiency throughout a renovation project in 34 grocery stores
• Inputs: kWh/m² and demand for heating and cooling
• The most efficienct buildings before and after the renovation project

• Average efficiency before the retrofitting = 74%
• Savings ranged from 56% - no savings
• During implementation the TVB model was used to adjust the installed equipment
• Average efficiency after the retrofitting = 74%

Severinsen, Alexander and Holst, Helen Marita, A 3-Step Framework to Benchmark Potential and Actual Energy Savings in Retrofitting Projects
Available at SSRN: https://ssrn.com/abstract=4181531 or http://dx.doi.org/10.2139/ssrn.4181531

Peak shaving

Objectives
Agenda
Background
Examples
Paper 5 / 5
Conclusion

• The grid owner wants to reduce the peaks to stabilize the grid
• Thus, tariffs for large loads are more expensive
• Objective of paper was to improve the economics of solar panels and batteries with load forecasting
• Building with annual consumption 2900 MWh, 891 kW max peak

Peak shaving

Objectives
Agenda
Background
Examples
Paper 5 / 5
Conclusion

• Battery storage was only economically beneficial when forecasting was deployed
• Most of the savings came from peak shaving, not from increased self-consumption

Fagerström, Jonathan, Kari Aamodt Espegren, Josefine Selj, and Alexander Severinsen. 2019. “Forecasting and Technoeconomic Optimization of PV-Battery Systems for Commercial Buildings.” In ECEEE Summer Study Proceedings, 949–54, https://www.eceee.org/library/conference_proceedings/eceee_Summer_Studies/2019/5-smart-and-sustainable-communities/forecasting-and-technoeconomic-optimization-of-pv-battery-systems-for-commercial-buildings/

The IPCC Sixth Assessment Report was released in 2022

UN Secretary-General António Guterres during the press conference...

• The report exposes how climate change impacts people and the planet
• 234 experts from 64 countries contributed (and agrees!)
• Many ecosystems are at the point of no return

Double-click on the heading to start “voice-over” Academic in early days Politics -> Prime minister of Portugal Later The High Commissioner for Refugees One of the world’s largest humanitarian organizations Since 2017 Guitteres has served as the secretary-general of the United Nations He has probably read thousand of reports! This is what he said during the press conference

Conclusion

Objectives
Agenda
Background
Examples
Paper 1 - 5
Conclusion

• Documented energy savings for grocery stores were 35%
• Savings potential in food retail stores is 700 GWh
• ≈ energy demand for 55 000 households

Credit: OPEN AI DALL-E “Digital art of almost 50 year old male saying thank you to the audience”