Stochastic optimization of electrified district heating using systems dynamics model
Khurshid, Usama (2026)
Pro gradu -tutkielma
2026
School of Business and Management, Kauppatieteet
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026052553590
https://urn.fi/URN:NBN:fi-fe2026052553590
Tiivistelmä
This study examines the optimization of thermal energy storage operations in district heating when electricity prices act as the key source of uncertainty. A literature examines modelling and optimization approaches currently being employed, as well as prominent control strategies and uncertainty modelling. A particular focus is placed on understanding the need and role of dynamic modelling approaches, specifically System Dynamics modelling, and its potential in handling feedback loops that represent complex district heating operations.
An inherited System Dynamics model developed in Simulink is coupled with an external optimization layer using a Genetic Algorithm in MATLAB to determine offsets to a moving average charging/ discharging logic across three storage configurations (i) heat tank (ii) heat tank and one battery (iii) heat tank and two batteries. Electricity price uncertainty is represented by generating multiple scenarios by fitting an AR(1) GARCH(1,1) model on real electricity prices after preserving deterministic inter-temporal trends. Each scenario contains hourly electricity prices over a one-year horizon, comprising of 8784 timesteps.
The results from the literature review indicate that most district heating optimization approaches rely on closed-form mathematical representations rather than dynamic feedback loops, which are underrepresented. The empirical results indicate that external optimization added to SD models representing DH storage operations have the potential to introduce greater cost savings, however, their potential is limited by the degree of flexibility offered by the storage setup. This is supported by improved cost savings in the tank + 2 batteries storage configuration.
An inherited System Dynamics model developed in Simulink is coupled with an external optimization layer using a Genetic Algorithm in MATLAB to determine offsets to a moving average charging/ discharging logic across three storage configurations (i) heat tank (ii) heat tank and one battery (iii) heat tank and two batteries. Electricity price uncertainty is represented by generating multiple scenarios by fitting an AR(1) GARCH(1,1) model on real electricity prices after preserving deterministic inter-temporal trends. Each scenario contains hourly electricity prices over a one-year horizon, comprising of 8784 timesteps.
The results from the literature review indicate that most district heating optimization approaches rely on closed-form mathematical representations rather than dynamic feedback loops, which are underrepresented. The empirical results indicate that external optimization added to SD models representing DH storage operations have the potential to introduce greater cost savings, however, their potential is limited by the degree of flexibility offered by the storage setup. This is supported by improved cost savings in the tank + 2 batteries storage configuration.