Development and application of innovative methods for smart control of district heating networks

The mitigation of the effects of human activities on the climate and environment has become essential to guarantee environmental sustainability and safety. As regards the energy sector, reduction in greenhouse gas emissions and energy efficiency improvement are fundamental targets of the clean energy transition to be pursued over the next decades. Since it accounts for half of final energy use in the European Union, the heating and cooling sector offers significant opportunities for decarbonization. In particular, district heating networks are regarded as highly promising due to their ability to distribute thermal energy in urban areas more efficiently compared to individual heat generation devices, to the possibility to integrate renewable energy sources, and to their flexibility potential. However, the complexity of these systems is increasing and their traditional management approaches, based on the experience of the operators, are not able to fully unlock their benefits. On the other hand, optimal controllers, which are made possible by new digital technologies, may allow this goal to be achieved. Model Predictive Control (MPC) is a smart control strategy which takes advantage of the prediction of the system behavior over a future horizon to optimize its operation. Therefore, it is an adequate solution to cope with the high variability of the external conditions and to perform system optimization. The scope of this thesis is to investigate and develop a complete set of original methods for the application of MPC to district heating networks with different sizes and levels of complexity. Since MPC requires a dynamic model of the system and a computationally efficient optimization algorithm, these two fundamental tools are developed for small-scale and large-scale networks. In particular, the models are control-oriented and physics based and, thus, maintain the representation of the main governing phenomena and physical parameters, such as the heat capacity of the end-users connected to the network. The developed tools are embedded within MPC solutions and their performance is verified in Model-in-the-Loop simulation environments, which enable a reliable comparison of different control strategies without affecting the real system. As for small-scale district heating, the novel optimization algorithm is based on Dynamic Programming and is particularly suitable for a multi-agent hierarchical control architecture. This is tested on a case study located in northern Italy and achieves both minimization of the heat supplied to the end-users and reduction in the production unit operating cost, with reference to a traditional control strategy. In addition, the potential of the system controlled by the MPC in providing flexibility service to the power grid in presence of uncertainty is investigated and verified. Concerning large-scale district heating, the developed optimization algorithm aims to shift the peaks of energy supplied to the various regions of the network by storing heat in their thermal capacity and, at the same time, to reduce the distribution temperature. The algorithm is embedded in an MPC and its application to a city district heating in central Sweden results in up to 16 % peak shaving and up to 20 % reduction in heat losses, with reference to historical data. Overall, the proposed solutions for smart control bring noteworthy advantages to district heating networks in terms of energy and cost saving. Their versatility and independence from the specific problem can aid the extension of MPC to multi-source networks, toward its implementation in real-life cases. This constitutes a promising step in the direction of smart, optimal and efficient energy systems.