Due to the volatility of renewable energy generation, the probability of intermittent electricity deficiency or excess rises. A stable power grid requires a balanced demand and supply of power. Especially heat pumps in connection with thermal storages that can consume electric power when demand is low and allow for a deferred usage of the stored energy by buffering, bear significant potential for intelligent demand-side management.
In this thesis a genetic algorithm is implemented that optimises the schedule of a heat pump over a specified horizon such that operation costs are minimised while thermal demands are satisfied at any given moment. It is assumed that the demand for flexible device operation resulting from a highly fluctuating supply side is expressed by dynamic electricity prices. High prices prompt a reduction of electricity use while low prices ask for more consumption of electricity. In particular, this work considers the trade-off between the flexible employment of heat pumps and a loss in efficiency due to overheating in the considered heat storages by modelling a variable coefficient of performance. With regards to the application of the genetic algorithm to the heat pump scheduling problem, an optimal set of operators and parameters is sought besides developing a suitable genotype representation and fitness evaluation function. Moreover, in order to enhance performance heuristic approaches are incorporated into the algorithm using problem-specific knowledge. The final parametrisation was evaluated regarding its robustness towards varying price profiles and uncertainty in thermal energy demand.
The operating costs after optimisation were compared to the operation costs of a heat-driven heat pump control. The results of this work reveal that the extent of savings depends on the spreading of the price signal which therefore constitutes a major option to further incentivise demand-side management.