Connected homes and virtual inhabitants for large-scale energy gains

The arrival of algorithms in eco-design makes it possible to analyse, understand, study and forecast user practices better and better and take them into account to optimise energy efficiency. As part of his PhD research project, Eric Vorger, modelled the energy consumption of buildings resulting from the presence and activities of inhabitants. This occupancy model is coupled with the Pléiades Comfie dynamic thermal simulation tool, developed by MINES ParisTech with the support of the lab recherche environnement.

In France, buildings are responsible for around 45% of energy consumption, which is much more than transport (31.3%). It is a sector which represents an important energy saving potential. In particular, current knowledge makes it possible to design and renovate very low-energy buildings on a large scale. The energy efficiency of buildings is a fundamental market trend which is accelerating, largely under the impetus of certifications such as passive or positive energy buildings. The energy efficiency guarantee, the contract that binds the prime contractor to overseeing the energy consumption of a building, can also encourage renovation work if a return on investment is truly guaranteed.

The energy efficiency of the building is based on solutions such as constructing compact housing, favouring solar gain and natural lighting, using insulating materials allowing air-tightness or heat storage, using double flow ventilation systems which recover and distribute heating, alternative energy sources (solar, geothermal, photovoltaic) and automatic consumption regulation devices. However, low-energy buildings often consume more than expected. According to a CEREMA study conducted between 2012 and 2016, 50% of efficient buildings consume a surplus of 10 KWhep per sq.m per year. In 25% of cases, this excess is 35 kWhep per sq.m per year. According to the study, the differences observed depend in part on user behaviour. As buildings become efficient, these deviations become increasingly significant compared to the overall performance of the project. They can easily reach 20 to 30% of the total consumption.

This discrepancy, due to an imprecise estimate of actual uses, was also observed by Eric Vorger, former doctoral student at MINES ParisTech and co-founder of Kocliko, a start-up that offers digital tools to optimise and guarantee the energy efficiency of buildings. “Human behaviour is simplistically modelled in building energy simulation software. However, its impact is considerable and it is the source of significant variations between simulation results and in situ measurements. Short-term predictions (of the order of 24 hours), used to regulate energy consumption in real-time, as well as medium-term predictions (one year), useful for sizing renovation operations, are marred by very significant uncertainties (often greater than 50%) because of this lack of knowledge of uses”.

Prediction of energy consumption — A comparison between the presence rate in an office according to the conventional scenario of the RT2012 regulations (in black), and a scenario developed by Eric Vorger on the basis of statistics (in red). According to the statistical model, the presence rate is lower compared to the conventional assumption.

The arrival of algorithms in eco-design makes it possible to analyse, understand, study and forecast user practices better and better and take them into account to optimise energy efficiency. As part of his PhD project, Eric Vorger, modelled the energy consumption of buildings resulting from the presence and activities of inhabitants. This occupancy model is coupled with the dynamic thermal simulation tool Pléiades+COMFIE, developed by MINES ParisTech with the support of the lab recherche environnement.
The presence of people in a home and their activities are valuable information that makes it possible to predict the opening or closing of windows and blinds, the use of lighting and domestic equipment, the occupation of certain rooms in the home, the number of people in a room and the metabolism of the occupants which influence the ambient temperature.

Eric has developed probabilistic models based on INSEE data (e.g. the “Time Use” surveys) and measurement campaigns. Thanks to these models, the same home can be simulated 10,000 times with, each time, occupants having different socio-demographic characteristics which determine different schedules. The second graph illustrates the average daily scenario of a virtual individual, produced from a series of simulations. Other simulations were carried out successively with the same dwelling and different types of household: a large family, a working couple, a retired person, etc. By taking into account the composition of households, more reliable forecasts of energy consumption are obtained, which allows contractors to commit to realistic energy efficiency targets and achieve them.

Detailed description of household daily activities — We can observe the habits of the virtual occupant modelled by Eric Vorger, and the fact that they tend to be more present at home in the morning than in the afternoon, that an important part of their activities consists of eating, dressing or showering, unlike the cooking, which is a less frequent activity.

This design support tool enables design to be optimised thanks to realistic usage forecasts. There are therefore fewer design errors, and consumption and comfort forecasts are more reliable. Kocliko has also developed solutions for energy management during the operational phase. Based on the observation that connected objects are now technology that is accessible at low cost, the start-up proposes to exploit the “Internet of things” in order to extend the collection of measurements in a housing unit: temperature, humidity percentage, electricity consumption, thermostat usage, etc. The machine learning algorithms developed by Kocliko combine this data to reconstruct actual uses and manage energy consumption. An energy saving of at least 20% can thus be achieved during the operating phase.