Communication-Based Distributed Control of Large-Scale District Heating Networks

TL;DR

The paper addresses the control of large-scale District Heating Networks (DHNs) where non-convex, bilinear coupling makes centralized optimization impractical. It develops a tailored non-cooperative distributed MPC with a dedicated information-passing scheme and a feasibility restoration mechanism to drive subsystems toward a global-consistent solution (Nash equilibrium) while minimizing local costs. The approach is validated on an 18-user DHN partitioned into six subsystems, achieving a $14\%$ reduction in network losses and a $\,37\%$ reduction in average return temperature compared to a centralized reference with zero flexibility envelopes. The method demonstrates scalable coordination and efficient use of building flexibility, with future work focusing on convergence enhancement, early stopping, and deployment on longer simulations or real DHNs.

Abstract

This paper presents a non-cooperative distributed model predictive controller for the control of large-scale District Heating Networks. To enable the design of this controller a novel information passing scheme and feasibility restoration method are created, allowing the local controllers to achieve a global consensus while minimizing a local cost function. The effectiveness of this controller is demonstrated on an 18-user District Heating Network decomposed into six subsystems. The results show that the developed control scheme effectively uses flexibility to manage the buildings' heat demands reducing the total losses by 14% and the return temperature by 37%.

Figures (7)

• Figure 1: Flowchart of steps taken in the presented dMPC scheme.
• Figure 2: Graphs of case study network with user edges numbered.
• Figure 3: Communication graph between subsystems.
• Figure 4: The nominal demands of the buildings in the DHN, organized by subgraph.
• Figure 5: Network level results comparing the optimized and nominal cases.