developing a method for the operational control of an ecovat system
Gijs de Goeijen is a PhD student in the research group Computer Architecture Design and Test for Embedded Systems (CAES). His supervisors are prof.dr.ir. G.J.M. Smit and prof.dr. J.L. Hurink from the Faculty of Electrical Engineering, Mathematics & Computer Science (EWI).
To decrease the emission of greenhouse gases, as well as to reduce our dependency on fossil fuels for satisfying our energy needs, we see a trend towards the use of more sustainable energy sources. While these sustainable energy sources, such as solar and wind, accomplish these mentioned goals they also present new challenges. One of these challenges lies in the fact that these energy sources are intermittent and uncontrollable. One of the consequences is that times of energy production do not necessarily coincide with times of demand. When energy is generated by fossil fuel powered energy plants, it is relatively easy to match the supply and demand of energy. However, this matching is much more difficult when relying on sustainable energy sources, such as solar and wind energy, due to their uncontrollable nature.
One of the solutions to deal with the mismatch of energy demand and production is energy storage. With storage, energy generated during times of excess production may be stored for use during times of energy shortage. In this context, energy storage may be used to cover mismatches occurring during a day, but also to cover the mismatch between different seasons. For the mismatches during a day electrical storage (i.e. batteries) can be used. However, batteries are currently too expensive for the large capacities required for seasonal storage. One of the promising options for seasonal storage is thermal energy storage. The Ecovat system is an example of such a seasonal thermal energy storage, which aims to store excess thermal energy during times of the year with high thermal and/or electrical energy production, generally during the summer, for use during times of the year with high thermal energy demand, generally during winter. The Ecovat system is designed to be able to satisfy the heat demand of a neighbourhood of houses throughout the year.
The Ecovat system consists of a large well insulated underground buffer (i.e. a large water tank), combined with a number of devices, namely photovoltaic-thermal panels, heat pumps, and a resistance heater, to charge the buffer. The buffer of the system consist of a number of segments, which although not physically separated, may be charged or discharged individually through heat exchangers integrated inside the buffer walls. The energy to charge the buffer can be obtained from locally available energy or can be bought on the energy market, preferably when the energy price is low.
In this thesis we focus on the operational control of such an Ecovat system. We develop a model to determine which of the available devices in the system should charge which buffer segment at which point in time. Furthermore, the model also determines which buffer segment should be used to satisfy the heat demand from the neighbourhood. As the developed model should serve as the base for handling the operational control of a real Ecovat system, we are not just interested in an arbitrary model that is able to obtain charging/discharging strategies, but in a model that is able to compute these strategies in a short time (at most a few seconds).
Although we aim for a model with short computation times, we first focus on a model that does not take this restriction on the computation time into account. The goal of this first model is to get insight in the structure of a good charging/discharging strategy, i.e. a strategy that has low operational costs while satisfying the heat demand of the neighbourhood throughout the year. Furthermore, this model acts as a benchmark for other models that do satisfy the short computation time constraint. To this end, the first developed model is based on an integer linear programming (ILP) model of the Ecovat system.
Due to the long time scales involved when dealing with seasonal thermal energy storage (a year), as well as the short time interval lengths (15 minute time intervals) required to incorporate energy markets into the model, the developed (ILP) model can not be solved for an entire year at once. Due to this we developed an approach based on solving the (ILP) model in a rolling horizon fashion. Although this approach leads to a substantial reduction of computation time, we observe that solving the model in this way does not sufficiently take important seasonal effects into consideration.
To ensure that such seasonal effects are also taken into consideration by the model, we extend the model with a long-term planning step, which generates additional input for the previously developed model. In this planning step we determine daily energy targets for the buffer, based on historical data and predictions, which have to ensure that the correct seasonal behaviour is obtained. While the rolling horizon model with this extension is able to provide good charging/discharging strategies, we observe that even with these modifications the (ILP) model based approach is computationally still too expensive to be used in a practical situation, as in some cases it requires multiple days to determine a charging/discharging strategy for a year of operation of the Ecovat system.
Subsequently, we use the insights obtained from the (ILP) model based approach to develop a heuristic method to control the Ecovat system. This method is based on a number of rules of thumb, and contrary to the(ILP) model based approach, it does not require predictions for weather data and energy prices for future time intervals. This heuristic method requires much shorter computation times, namely it takes only a few seconds to simulate a complete year of operation of the Ecovat system. Comparing the results obtained with the heuristic method, with the results obtained with the (ILP) model based approach, we find that the heuristic method on average only increases the operational costs by 5.2%.
To get more insight in the practical use of the Ecovat system and the developed approach we performed a case study, where we simulate a neighbourhood of houses including an Ecovat system in a distributed energy management simulation, using the developed heuristic method to control the Ecovat system. We compare the achieved results with a simulation using gas boilers to satisfy the heat demand of the neighbourhood instead. The results of this comparison show that using an Ecovat system to satisfy the heat demand leads to significant benefits in terms of energy self-consumption within the neighbourhood, as well as a decrease in CO2 emissions compared to using gas boilers. Furthermore, the obtained results show that the developed approach is robust against prediction errors, such as e.g. a winter that is colder than predicted.
