Design and operational optimization of thermal networks: A robust approach under energy market uncertainty

Thermal networks can satisfy heating and cooling demands at a lower cost
with reduced greenhouse gas emissions. Nowadays, the fifth generation of
thermal networks is being rolled out. They operate at low network temper-
atures to allow the incorporation of decentralized prosumers. This network
temperature reduction also facilitates waste heat recovery and the integra-
tion of renewable energy sources. Optimizing the design and operation
of these networks is challenging, given the uncertainties for future energy
prices, emission factors, investment costs, and energy demands, because
they will significantly impact the cost, performance, and decarbonization
that can be achieved. Uncertainties must be considered via a so-called un-
certainty quantification (UQ) and sensitivity analysis in search of an optimal
robust design and business case. The analysis provides Sobol’s indices that
pinpoint influential uncertain inputs and enhance computational efficiency
by focusing on the most impactful factors during robust design optimization.
The road maps for the transition towards fifth-generation district heating
and cooling (5GDHC) networks are unclear in the context of future uncer-
tainties and were not reported in previous literature. Hence, the following
research questions emerged:
1. How to optimize the operation of existing networks?
2. How to optimize the design of a potential 5GDHC network under
uncertainties?
To answer the first research question, a model-based demand-responsive op-
erational optimization strategy is formulated for the existing networks to
optimize the network operational parameters that will reduce the network
return temperature. This will enhance the heat recovery from flue gases
at the heat-producing station and improve the economic and environmental
performance of the thermal network. The strategy incorporates the thermo-
dynamic model of the substation and building heating system as opposed to
traditional weather-based adjustments. The estimated heat demands, the
energy prices and CO2 penalties are inputs the optimizer uses to find the
optimal solution. The optimal solution includes control trajectories for local
controllers, such as the desired supply temperature and mass flows from the
central heat production and flow rates and temperatures in the substations.
Unlike existing PI-based substation controls, a constraint model predictive
control (MPC) is implemented to track the building-side supply tempera-
ture and return temperature to the heat-producing system. The proposed
methodology is implemented on the VUB thermal network at our campus.
The total operational expenditure for electricity and gas consumption shows
an 18 % reduction with an 8 % reduction in emissions and a 6 % efficiency
improvement compared to the measured weather-based approach. The de-
veloped strategy will aid the network operators in the economic dispatch of
heat generation while ensuring the user’s thermal comfort.
For the second research question, a generalized robust techno-economic
analysis and optimization framework is proposed to address the hindrance
in the smooth transition towards 5GDHC networks. Several 5GDHC net-
work designs for existing business parks in Flanders are evaluated under
uncertain energy markets and fluctuating energy demands of the involved
prosumers. The framework utilizes a systematic approach by first perform-
ing a global sensitivity analysis, which yields the influencing uncertain input
parameters that affect the KPIs of the 5GDHC network. This reduces the
number of uncertain input scenarios and improves the computational effi-
ciency of solving a multi-objective robust design optimization framework
(MO-RDO) in the next step. Additionally, the MO-RDO employs a ma-
chine learning (ML) model developed using the initial step’s KPI responses.
The MO-RDO provides a set of environmentally and economically optimal
network designs for a selection of operational parameters and highlights
their trade-offs. The designs and operational parameters are evaluated on
financial metrics, such as the net present value (NPV), return on invest-
ment, and discounted payback period. In the case study of a small network
at the Telenet site in Flanders, the environmental design has a 10 % higher
investment cost compared to the economically optimal solution. Still, it
reduces total emissions (TE) by 13 % and improves overall savings, which
is crucial for better financial gains.
The developed optimization framework offers investors distinct design op-
tions, enabling them to choose between more economical or environmentally
friendly solutions, each with associated financial risks.