PersonAI - User-Centered AI-based energy services built on personal preference models

Starting point / motivation

The EU wants to increase energy efficiency by 32.5% by 2030 and achieve a 32% share of renewable energies. Currently, the building stock in the EU is energy intensive and mostly inefficient; it is responsible for 40% of the EU's final energy consumption and 36% of its CO2 emissions. In this context, buildings must evolve from their permanent static and inefficient profile to intelligent dynamic actors, while at the same time serving the needs of users.

The past has shown that the successful implementation of energy policy depends heavily on social factors such as social acceptance, tolerance and motivation to participate. Due to the rapid further development of available technologies, AI-supported energy services such as Model Predictive Control (MPC), Demand Side Management, Forecasting, etc. are gaining importance and practical relevance in the building sector. At the same time, compliance with a health-promoting interior quality (temperature, humidity, air quality, etc.) must be guaranteed.

Contents and goals

Even before the COVID-19 pandemic, a majority of the population spent around 90% of their time indoors. Thus, the interior quality in buildings has a significant impact on health and well-being of the users. Nevertheless, compliance with the corresponding comfort criteria is rarely successful.

Depending on the building type, two types of comfort evaluation models are currently used (standardized in ISO 7730:2005 or EN 16798-1:2019): 

1. heat balance models (e.g. predicted mean vote PMV index) or 
2. adaptive models. Both are highly simplified and statistical methods, were determined under laboratory conditions and are intended to predict the average comfort rating of a large group of people.

Several studies have shown that these models do not adequately reflect the complexity of the interactions in the human-environment relationship and also do not take individual circumstances (age, gender, health, clothing, etc.) into account. For the reasons mentioned before, personal comfort models became an innovative and new research field.

Methods

Personal comfort models are created using detailed data such as,s subjective feedback surveys (personal preferences, clothing factor, level of activity, draft, etc.), physiological measurement data (skin temperature, heart rate, steps, etc.), GPS location or environmental conditions (temperatures, humidity).

The models are then trained separately for each participant and can be aggregated again to predict the thermal comfort of a group (e.g. on a floor, in a thermal zone) under the given environmental conditions.

Expected results

Different studies and simulations estimate energy savings between 21.81% and 44.36% through AI-based Energy Services and comfort improvements between 21.67% and 85.77% through Personal Comfort Models. The combination of these two approaches with the development of AI-based personal comfort models forms the core of the project "PersonAI".

For this purpose, the implementation of a broad-based long-term study with 40-50 people and detailed data collection (surveys in combination with skin temperature, heart rate, etc.) of a cross-section of the population that is as representative as possible is planned. The resulting personal comfort models are then aggregated and demonstrated in a proof of concept.

Project management

Technische Universität Graz

Project or cooperation partners

• DILT Analytics GmbH
• Forschung Burgenland GmbH
• Universität Graz - Institut für Öffentliches Recht und Politikwissenschaft

Gerald Schweiger
Inffeldgasse 16b/II
A-8010 Graz
Tel: +43 (316) 873 - 5747
E-Mail: gerald.schweiger@tugraz.at
Web: https://www.tugraz.at/institute/ist/home/