Session: 16-01-01: Poster Session: NSF-Funded Research (Grad & Undergrad)

Paper Number: 99408

99408 - Automated Modeling Technique for the Energy Efficient Design of District Heating Networks 

District heating networks (DHNs) are a promising tool to reduce energy usage in buildings, integrate co-generation into the energy infrastructure and incorporate renewable energy sources into the energy mix. DHNs take water heated at a centralized plant and distribute it to individual buildings, where heat exchangers extract energy to heat the building. While DHNs are inherently more efficient than traditional heating methods, a smart control strategy that meets the buildings’ heat demand while reducing energy losses from heat transfer and pumping costs can further increase their efficiency. Due to the large-scale nature of DHNs, distributed model predictive control (dMPC) is the preferred method for controlling these systems. However, before a dMPC system can be implemented, the network must be partitioned into smaller subsystems. Effective partitioning requires models of the subsystems to ensure the coupling between them does not affect stability or optimality of the overall controller. Automating the modeling process will make the search for an optimal partitioning scheme possible.

The high number of users, pipe segments, and actuators in the system, along with the fact that every DHN has a different topology, makes modeling these systems a manual and time intensive process. Existing modeling techniques either require the mathematical modeling of every component in the network and their connections or large quantities of data to accurately represent a DHN.  Minor changes in a DHN’s configuration, such as a temporary closure for maintenance, can greatly affect the model’s accuracy and therefore the controller’s performance. The automated modeling technique developed here will efficiently generate these models with limited information and ensure the model can be easily adapted when faced with changing network topology. Ultimately, this automated modeling technique will also serve as the basis for further research into partitioning these systems for dMPC design.

Combining a first principles-based modeling approach and a graph-theory based representation of the DHN allows for the automated creation of an accurate state-space model based only on the system topology and a few network characteristics. This modeling technique relies on the identification of repeated structures (the buildings, heat exchangers and bypass valves) that connect an individual user to the network to create fundamental state-space matrix blocks. The DHN can then be reduced to a directed, rooted tree, with the generation plant designated as the root. The fundamental building blocks established for each user are represented as nodes and the connecting pipes are represented as weighted edges. The degree of each node is identified and used to dictate the mass flow rate splits between network branches, with special cases established for leaf users (terminal buildings) and users with degree 2 (buildings connected in series). Finally, the edge weights of the connecting pipes are used to augment the state-space model, connecting the fundamental blocks based on the directed flow through the graph.

The proposed algorithm that generates a state-space model of a DHN based on the topology of the network is implemented in the optimization of the design of the topology of new DHNs. The goal is to reduce the energy lost during peak, steady-state operation of a DHN. A test map of eight users is taken from a portion of the University of Parma DHN, and using a branch and bound algorithm with heuristic reduction of the search space, the optimal tree is found. The relative energy losses of this configuration are compared against the current topology and the layout resulting from minimizing the installation cost (the existing technique employed during new DHN design).

Presenting Author: Audrey Blizard The Ohio State University

Presenting Author Biography: Audrey Blizard is a current Ph.D. student at The Ohio State University, studying Mechanical Engineering. She was awared the NSF GRF in 2021. She completed her undergraduate degree at Penn State University.

Authors:

Audrey Blizard The Ohio State University
Stephanie Stockar The Ohio State University