This project aims to develop hybrid control and machine learning approaches for advanced control in next generation thermal systems, such as low-temperature heating/high-temperature cooling. Such frameworks will help establish standard guidelines for their implementation, and speed up adoption and evaluation of machine learning methods, in such systems. The goal is to improve the intelligence and reduce investment costs, and facilitate fast large-scale implementation, customization, and integration with storage, renewables, end-users, and utilities in future thermal energy networks.
The next generation of heating and cooling systems, such as low-temperature heating/high-temperature cooling, come with higher requirements for system control and load prediction. The challenges are even greater when integrating such systems with storage and renewables. For example, widespread use of heat pumps and storage, and how they are used to manage peak load and operating strategies, will have a direct impact on the role of utilities, such as district heating companies. Modern methods from control and machine learning show potential to solve these challenges, but there are still no established standards for how to implement these, or how to simulate and evaluate them from a system integration perspective.
This project aims to develop novel approaches technically to the design and evaluation of advanced controls for such heating and cooling systems and contribute to standard guidelines for their implementation. Another aim is to develop a hybrid control framework that can reduce costs and facilitate fast large-scale implementation by fundamentally improving the used machine learning methods. Together with industry partners, we will better understand the overall performance and feasibility of these solutions. The results and final report for the project are expected in Dec. 2024.