Optimization approaches for exploiting the load flexibility of electric heating devices in smart grids

Abstract

Energy systems all over the world are undergoing a fundamental transition to tackle climate change and other environmental challenges. The share of electricity generated by renewable energy sources has been steadily increasing. In order to cope with the intermittent nature of renewable energy sources, like photovoltaic systems and wind turbines, the electrical demand has to be adjusted to their power generation. To this end, flexible electrical loads are necessary. Moreover, optimization approaches and advanced information and communication technology can help to transform the traditional electricity grid into a smart grid. To shift the electricity consumption in time, electric heating devices, such as heat pumps or electric water heaters, provide significant flexibility. In order to exploit this flexibility, optimization approaches for controlling flexible devices are essential. Most studies in the literature use centralized optimization or uncoordinated decentralized optimization. Centralized optimization has crucial drawbacks regarding computational complexity, privacy, and robustness, but uncoordinated decentralized optimization leads to suboptimal results. In this thesis, coordinated decentralized and hybrid optimization approaches with low computational requirements are developed for exploiting the flexibility of electric heating devices. An essential feature of all developed methods is that they preserve the privacy of the residents. This cumulative thesis comprises four papers that introduce different types of optimization approaches. In Paper A, rule-based heuristic control algorithms for modulating electric heating devices are developed that minimize the heating costs of a residential area. Moreover, control algorithms for minimizing surplus energy that otherwise could be curtailed are introduced. They increase the self-consumption rate of locally generated electricity from photovoltaics. The heuristic control algorithms use a privacy-preserving control and communication architecture that combines centralized and decentralized control approaches. Compared to a conventional control strategy, the results of simulations show cost reductions of between 4.1% and 13.3% and reductions of between 38.3% and 52.6% regarding the surplus energy. Paper B introduces two novel coordinating decentralized optimization approaches for scheduling-based optimization. A comparison with different decentralized optimization approaches from the literature shows that the developed methods, on average, lead to 10% less surplus energy. Further, an optimization procedure is defined that generates a diverse solution pool for the problem of maximizing the self-consumption rate of locally generated renewable energy. This solution pool is needed for the coordination mechanisms of several decentralized optimization approaches. Combining the decentralized optimization approaches with the defined procedure to generate diverse solution pools, on average, leads to 100 kWh (16.5%) less surplus energy per day for a simulated residential area with 90 buildings. In Paper C, another decentralized optimization approach that aims to minimize surplus energy and reduce the peak load in a local grid is developed. Moreover, two methods that distribute a central wind power profile to the different buildings of a residential area are introduced. Compared to the approaches from the literature, the novel decentralized optimization approach leads to improvements of between 0.8% and 13.3% regarding the surplus energy and the peak load. Paper D introduces uncertainty handling control algorithms for modulating electricheating devices. The algorithms can help centralized and decentralized scheduling-based optimization approaches to react to erroneous predictions of demand and generation. The analysis shows that the developed methods avoid violations of the residents' comfort limits and increase the self-consumption rate of electricity generated by photovoltaic systems. All introduced optimization approaches yield a good trade-off between runtime and the quality of the results. Further, they respect the privacy of residents, lead to better utilization of renewable energy, and stabilize the grid. Hence, the developed optimization approaches can help future energy systems to cope with the high share of intermittent renewable energy sources.