Cooperative energy management and cross-domain optimization for electro-thermal devices at city-district and city-level
Stoyanova, Ivelina Evtimova; Monti, Antonello (Thesis advisor); Müller, Dirk (Thesis advisor)
1. Auflage. - Aachen : E.ON Energy Research Center, RWTH Aachen University (2021)
Book, Dissertation / PhD Thesis
In: E.ON Energy Research Center : ACS, Automation of complex power systems 88
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme
Dissertation, RWTH Aachen University, 2020
In the last few decades, factors such as climate change and catastrophic events contributed to the ongoing shift of collective mindset from possession to sharing and from excessive exploitation to efficiency and sustainability. Opposed to the conventional design and operation of urban infrastructures with high security margins to compensate possible variations, in recent years the novel challenges arising from population growth, resource scarcity and pollution require a holistic view and novel approaches. Here, optimization is a central tool to improve efficiency and sustainability, as the combined optimization of multiple domains enables the identification of flexibilities and the exploitation of emerging crossdomain synergies. In this context, cities need to develop to more efficient, environmentally and technologically advanced and socially inclusive living spaces based on digital technologies, or so-called Smart Cities. A central factor for the transition to a Smart City is the efficient collaboration of components and subsystems in order to optimize processes, exploit synergies and maximize the efficient deployment of resources. Furthermore, to achieve the ambitious plans of the EU for the decrease of CO2 emissions, the increase of Distributed Energy Resources (DER), integration of Renewable Energy Sources (RES) and customer involvement are essential. Their fluctuating character and the dispersed location require a transformation of the grid to a resource-driven system, in which customer-side flexibilities play a major role. In order to exploit distributed flexibilities in an optimal way and integrate RES fluctuations, reliable, scalable and fair control and energy management mechanisms have to be developed accordingly. This dissertation presents comprehensive cross-domain strategies on different scales, starting from urban level with Smart Cities through considerably large areas with limited data availability down to the city-district level. First, at urban level, a concept for cross-domain optimization of heterogeneous domains is introduced, which enables the definition of interfaces among energy and non energy domains. For this purpose, a modular modeling and simulation approach was developed, which facilitates the integration of heterogeneous domains. To demonstrate the concept, the conventional single-domain optimization is compared to the hierarchical optimization of gas and electricity for a scenario with ten residential buildings. Furthermore, the results of the cross-domain optimization for the minimization of gas import and CO2-emissions are presented, along with the defined interface components among domains. The results show that the epsilon-constraint method appears to be better suitable for heterogeneous domains, as the weighted-sum method is highly sensitive to weight setting for heterogeneous objectives and, therefore, less feasible for the cross-domain case. The computational performance is very good in terms of computation time, however, an aggregation of buildings was necessary to avoid numerical infeasibilities for around 1300 buildings. Considering the original purpose of the method to solve high-level problems of smaller dimensions, this can be considered an acceptable limitation. Then, an adapted method based on statistical process control is presented, which applies statistical markers to classify load variations as common, originating from the stochastic nature of user behavior, or as special-cause variations which require compensation actions, with minimal data requirements. The applied methods, Shewhart Chart (SC) and Exponentially Weighted Moving Average (EWMA), perform best with a window between five and fifteen minutes, with EWMA performing slightly better than SC. Furthermore, the effect of limited anonymized additional data on the performance of the method was investigated and quantified, and found to be significant. Finally, at city-district level, a distributed cooperative energy management concept enables the tracking of a preset load curve and the maximization of the integration of renewable energy sources based on model predictive control(MPC). Here, several optimization strategies are discussed, which are applied to compensate short-term schedule and forecast deviations during runtime. The developed distributed MPC (dMPC) strategy is compared to a continuous rescheduling during runtime. The analysis of the methods showed that dMPC enables the tracking of a negotiated curve, however, the method requires certain forecast accuracy, which is not always guaranteed, especially in grids with high share of RES. The rescheduling strategy is more flexible and maximizes RES utilization, as the reference is updated continuously, but requires significant computation and communication resources. Then, a combined method is defined, which benefits from the advantages of the dMPC and the continuous rescheduling, as it applies a dynamic RMSE-based threshold to enable adaptive decision making according to the adequacy of day-ahead schedules. The method offers a good trade-off between efficient compensation and low computation cost of dMPC compensation with the adaptivity to changing conditions and the improved RES integration of rescheduling.