Living Loads is a joint project that develops realistic demand profiles for buildings and districts by combining artificial intelligence with behavioral simulation. The project aims to create web-based software solutions that represent user-dependent energy demands more realistically. This includes electricity, hot water, occupancy and mobility, helping to improve the planning and evaluation of energy systems.

The project addresses a central practical problem: user-dependent energy demands are still often represented by standard load profiles. These profiles are easy to use, but only represent the diversity of real buildings and their occupants to a limited extent. Factors such as the number of occupants, local weather, socio-demographic characteristics and, at district level, the simultaneity of individual loads can have a major impact on actual demand. At the same time, the way buildings are used continues to change, with more diverse and flexible patterns of occupancy and energy use, making reliance on a fixed set of standard profiles increasingly problematic. If these effects are not represented adequately, energy systems may be over- or undersized.

The idea behind Living Loads grew out of practical planning experience, previous research projects and the work on the overarching software project QuaSi, especially during the development of GenSim. These experiences showed the need for a method that combines the realism of detailed behavioural simulation with the simplicity required in planning practice. AI models trained on simulations with the open-source LoadProfileGenerator translate complex usage patterns into a small set of practical input parameters, without requiring a detailed behavioural simulation for every planning task.

Here you can find the newest posts about Living Loads on the QuaSi website.

1.    August 21st 2026 - Launch of new website for Living Loads
2.    January 5th 2026 - Start of new project: SLP_Ade

Living Loads bridges the gap between detailed behavioral simulation and the simple input required in planning practice. At its center is an AI model that uses a small set of key information about buildings and their occupants, including building type, floor area, household size, local weather and socio-demographic characteristics.

From these inputs, the model generates time series for plug loads, domestic hot water and electric vehicle charging, as well as profiles for the presence and activity of occupants. Presence, activity and plug loads can then be used as internal loads in a thermal building simulation to determine heating and cooling demand. In this way, user behavior and building physics are brought together in one consistent demand model.

The AI model is trained on data generated with the open-source simulation tool LoadProfileGenerator (LPG), which simulates people's energy-related activities at household level. LPG has been in use for over a decade and can produce high-quality, realistic load profiles. As part of Living Loads, the underlying simulation data basis of the LPG will be further developed and updated to better reflect current occupant characteristics and changing patterns of residential use. At the same time, detailed behavioral simulations require extensive input data and computational effort. By training the AI model on systematically generated LPG simulations, Living Loads reduces these complex inputs to a small set of key parameters, making realistic demand profiles much easier and faster to generate.

Whole-district profiles are aggregated from individual units so that the stochastic spread and simultaneity of loads are retained. This is particularly important at district level, where coincident loads determine the resulting peak demand and therefore influence the dimensioning of energy systems. Beyond residential use, Living Loads will also extend the LPG to represent small businesses from trade, commerce and services. This enables mixed-use buildings and districts to be represented within the same modeling approach.

Three organisations are working on bringing Living Loads to life. The system analysis group ICE-2 of the Institute of Climate and Energy Systems of the Jülich Forschungszentrum are the developers of LoadProfileGenerator. Their work on the project includes updating LPG and implementing new features, extending the model scope to include small and medium enterprises, as well as applying their expertise with LPG where it is needed.

FACT, part of the TMM Group alongside Digital Building Industries AG, bring experience with creating AI models based on simulations to the project. Their work involves exploring the space of possible AI model architectures to find the best approach and building the training pipeline to create the models. They also work on developing apps based on the models to integrate into a larger ecosystem of related software.

Acting as project coordinator, siz energieplus works on many aspects of the project. This includes developing a model for the charging of electric vehicles, specifying and generating the training data, validating results using measurement data, coordinating the network of associated partners (see below) and developing other open-source software as part of the project.

A variety of organisations are supporting the joint partners working on Living Loads. These associated partners take on various roles ranging from beta-testing the developed software, lending their expertise, validating results to connecting research and education through the project and its results.