TECH

A new modelling approach could make it easier to predict home energy use

Residential energy consumption plays a crucial role in the overall energy landscape, and it accounts for a significant portion of global energy demand. This sector encompasses various energy-intensive activities, including space heating, water heating, lighting, and the operation of electric appliances and devices in millions of households worldwide. The global population continues to grow, which leads to an increase in the number of residential units, and subsequently to higher energy requirements in meeting basic living needs. Additionally, rapid urbanization and improved living standards in many regions have resulted in increased energy consumption per household. The sector’s energy demand is further influenced by factors such as population density, climate conditions, building characteristics, and socio-economic factors. It is therefore crucial to understand and address the complexities of energy consumption to transition to a low-carbon future.

Energy fuels economic growth by powering industries, transportation, and essential services. As societies continuously evolve and technology becomes more pervasive, energy needs continue to increase. Nevertheless, the world is grappling with the adverse consequences of excessive energy use, such as greenhouse gas emissions, climate change, and resource depletion. It is therefore imperative to recognize that energy use is inextricably linked to global challenges, which makes it vital to explore diverse sustainable practices to find those that balance economic growth with environmental responsibility.

Improving energy efficiency and promoting sustainable energy practices have emerged as critical global priorities to reduce energy consumption in the residential sector. The understanding of residential energy consumption patterns is a crucial step to design effective policies, implement targeted interventions, develop sustainable energy systems, and reduce our environmental footprint. Mitigating energy consumption in the residential sector requires a multi-faceted approach. Key strategies include the promotion of energy-efficient technologies, behavioral changes, and policy interventions. Adopting energy-efficient appliances, optimizing heating and cooling systems, implementing better insulation, and utilizing renewable energy sources are all effective measures through which to reduce energy consumption.

Researchers studied many tools and technologies that facilitate the reduction in energy consumption in households. For example, smart meters can, depending on the technology and legal situation, enable real-time energy monitoring, which empowers homeowners to track and improve their consumption. Home energy management systems  allow for the automatic control and adjustment of energy demand and price signals based on electrical equipment. In addition, the development of energy storage and distributed energy resources offer opportunities to efficiently integrate renewable energy sources in residential systems.

Modeling plays an important role in shaping our understanding of energy use patterns and potential pathways toward sustainability. It also provides a systematic framework through which to analyze the complex interplay of the various factors influencing residential energy use. The developed mathematical and computational models that simulate energy use offer a systematic and structured approach in studying this field.

Predicting how much energy individual homes consume is a persistent challenge for researchers and urban planners. Models often group houses into broad categories based on their age or other characteristics, overlooking differences in their size, construction and energy performance.

A new approach developed by Concordia researchers with colleagues at McGill University, Natural Resources Canada and the National Research Council Canada, could make these predictions considerably more accurate. As described in the journal Energy & Buildings, the model offers a better way to assess energy use across residential neighbourhoods and identify opportunities for reducing consumption.

Their method, called Three-Layer Filtering, progressively refines how buildings are represented in urban building energy modeling. These computer models help researchers understand energy consumption across neighbourhoods and cities, and can inform decisions about building efficiency and decarbonization, the researchers say.

The team tested their model on 100 randomly selected detached houses in Montreal that use electric baseboard heating. They found their approach reduced the average error in predicting individual homes' annual electricity consumption from 80.7 per cent when using existing methods to just 4.3 per cent, providing more accurate data for urban planners and utility officials.

Analysis by archetype...The first step narrows the analysis to comparable buildings, avoiding comparisons between fundamentally different types of homes. The second groups houses according to several characteristics, including construction year, floor area, number of storeys, volume and estimated energy performance, rather than relying on age alone. The houses are then grouped into seven distinct clusters based on their similarities, with each cluster representing an archetype.

In the final step, the researchers adjust the model for individual houses based on their corresponding archetypes. Machine-learning tools help identify the building characteristics that most influence energy consumption, while a simplified version of the model learns from detailed energy simulations to quickly estimate energy consumption under different parameters. Together these measures significantly reduce the computing power needed to make these adjustments.

After the initial grouping, the average prediction error fell to 23.3 per cent. Following individual calibration, 83 per cent of the houses had predictions within five per cent of their measured annual electricity consumption.

This approach avoids having to build and run a separate, resource-intensive model for each property. This could make detailed, building-level energy estimates more practical for municipalities working with large housing stocks, rather than limiting them to small samples or broad neighbourhood averages.

The simplified computational approach offers an additional speed advantage. In a test involving one representative house, the researchers' streamlined model performed almost 22 times faster than conventional simulations, while producing comparable results.

A scalable model for a diverse country...The findings suggest that more accurate, building-specific energy estimates could eventually help municipalities identify homes with high energy consumption, prioritize retrofits and plan the transition to more efficient buildings.

“A city like Montreal is made up of many different neighbourhoods, and each one has a unique energy consumption pattern,” says lead author Wei Zhou, a PhD candidate at the Centre for Zero Energy Building Studies (CZEBS) in the Department of Building, Civil and Environmental Engineering.

“With our model, urban planners can do a speedy and accurate analysis of specific areas and determine which retrofit scenario is most appropriate.”

While the study remains a proof of concept, Zhou says it is scalable to other cities, building stocks, heating systems and climates.

A new modeling approach called Three-Layer Filtering developed by researchers at Concordia University, McGill University, Natural Resources Canada, and the National Research Council Canada significantly improves the prediction of home energy use.

How the tree-layer filtering approach works:

• Progressive Refinement: Instead of building heavy, separate computer models for every single house, the method progressively refines how buildings are represented within urban housing stocks.

• Archetype Grouping: Buildings are initially grouped and adjusted based on corresponding architectural archetypes.

• Machine Learning & Simplification: Machine-learning tools identify which specific building traits drive energy use the most, while a streamlined model learns from detailed energy simulations to evaluate different parameters quickly.

Performance and results:

• Accuracy: Tested on 100 detached homes in Montreal using electric baseboard heating, the method cut the average error in predicting annual electricity consumption from 80.7% down to 4.3%.

• Individual Calibration: After individual adjustments, 83% of the homes had predictions within 5% of their actual measured annual electricity consumption.

• Speed: The streamlined computational approach performed nearly 22 times faster than conventional energy simulations.

Researchers at Concordia University, alongside partners at McGill University, Natural Resources Canada, and the National Research Council Canada, have developed a new urban building energy modeling approach called Three-Layer Filtering.

As detailed in the journal Energy & Buildings, the method significantly improves how residential electricity and energy consumption are predicted across neighborhoods.

Key details of the model:

• Three-Layer Filtering: The method progressively refines how individual buildings and entire housing stocks are represented in computer models, avoiding the need to run heavy, separate simulations for every single property.

• Accuracy Improvement: In a test of 100 randomly selected detached houses in Montreal with electric baseboard heating, the approach cut the average error in predicting annual electricity consumption from 80.7% down to 4.3%.

• Calibration Milestones:

• After initial grouping by archetype, average prediction error dropped to 23.3%.

• After individual calibration via machine-learning and simplified simulation tools, 83% of the houses had predictions within 5% of their actual measured annual electricity use.

• Scalability: While tested in Montreal, researchers note the proof-of-concept framework is designed to scale across different cities, building types, heating equipment, and climates to assist urban planners and utility companies with retrofit planning.

mundophone