Smart Thermal Grids.

Energy efficiency in the built environment plays a key role in the transition towards a sustainable zero-carbon future. More specifically, renewables and industrial waste heat should be integrated in today's energy distribution systems. This integration is facilitated by so-called thermal grids, i.e. large systems at building or district level that consist of heat (and/or cold) sources and sinks, which are all connected by distribution pipes. The operation of thermal grids has highly complex dynamics because of two reasons. First, -analogue to the electrical grid- the intermittent loads of the sources and sinks should be aligned to ensure thermal and sanitary comfort of the end-users. Second, each type of thermal load (space heating, cooling and domestic hot water) requires a different temperature level. These temperature levels strictly affect both distribution losses and production efficiency. Currently, thermal grids are operated with static and mostly linear rule-based fuzzy logic control structures. Because of the simplicity and compactness of the linguistic approach of these types of controllers, trajectory following problems (such as heating and cooling reference tracking etc.) can successfully be accomplished. Yet, even though these solutions perfectly fit specific industrial applications, they do not offer any contribution to energy saving for complex thermal grids. Thus, the potential of thermal grids cannot be fully exploited by using conventional approaches. Indeed, primitive rule-based perspectives cannot fully optimize the alignment between production and demand, or the temperature set points along the grids. To sum up, they are designed for reliability, not for optimal efficiency. Optimizing controller dynamics of complex systems has been tackled in numerous areas of industrial applications: automotive, avionics, process industry etc. However, all these subsystems have mostly time-invariant dynamics and considerably less uncertainties that have serious effects on the aimed goal. This means that the data-driven algorithm can stop pre-processing input-output data after proposing an optimal solution under strict assumptions and constraints. On the other hand, environments such as thermal grids, which include high non-linearity, high complexity and time-varying parameters, require novel trends towards data-driven control methodologies. Despite their advantages and maturity, data-driven approaches have not adapted and penetrated into thermal grid applications (or HVAC systems in general). The reason is a lack of existing frameworks for implementation of these approaches and insufficient joint forces of HVAC and AI multi-disciplinary expertise. With this project, ID-Lab and EMIB, aim to set up a strong collaboration and obtain a leading role in the research of data-driven methodologies for optimizing energy efficiency in the building sector.

Keywords:DISTRICT HEATING, CONTROL SYSTEMS, ARTIFICIAL INTELLIGENCE (AI)

Disciplines:Energy in buildings and built environments, Sustainable buildings and cities, Modelling and simulation