Master's thesis Thomas Schütz


Development of an agent-based solution for energy management

Master's thesis Schütz Copyright: EBC Results of a one-day, centralized scheduling in the grid connected setting.

This work investigates control strategies of microgrids either connected to a macrogrid or operating in an island mode. The control strategies regard the scheduling of the installed heating devices, such as combined heat and power units, heat pumps and boilers. The microgrid is modeled with a multi-agent approach; the scheduling models are mixed integer linear optimization programs. Two different settings are investigated, a state-of-the-art dimensioning (2013) and a larger dimensioning providing more scheduling flexibility and thus allowing for the self-sustaining island mode (2020). In the grid connected mode, the results are benchmarked with reference cases that only use boilers to cover the thermal demand of each house and a case in which the controller runs a heat driven schedule. The developed logics are applied to a microgrid comprising four houses and renewable energy sources in form of solar and wind energy. Full year simulations are conducted to compare the different scenarios.

In the grid connected setting, the centralized and decentralized controls outperform the reference studies monetarily as well as ecologically. The centralized strategy has pay off times ranging from 6 to 9 years, while the decentralized strategy is able to generate monetary advantages after 6 to 10 years. The 2013 dimensioning does not allow for much flexibility, therefore the results of both strategies barely vary. In the 2020 simulations, the central control proves to be favorable in terms of economy and grid interaction, because it minimizes the dependency on the macrogrid and generates electrical surplus to support the macrogrid at most times. The decentralized scheduling performs best, when comparing the primary energy consumption and CO2 emissions in the 2020 setting. On the other hand, the decentralized strategy is less well coordinated and produces more volatile schedules, requiring more additional electricity from the macrogrid, while simultaneously providing larger amounts of electrical surplus.

The results of the island mode setting attest that the centralized control is more suitable than the decentralized control strategy. The centralized strategy pays off after 12 years, whereas the decentralized strategy does not pay off at all, during the life-expectancy of its components. The centralized strategy also surpasses the decentralized strategy ecologically, reducing the CO2 emissions by 11%. Compared to reference case of the grid connected setting, the centralized control strategy is able to cut the CO2 emissions by 31%.