EU carbon emission targets related to climate change has set in motion a process of transition towards an environmentally clean and sustainable power system. A central focus on this process is the transition from fossil fuel based energy sources to clean Renewable Energy Sources (RES). However, the intermittency of RES (e.g. solar and wind) presents a formidable challenge to achieve a stable and reliable supply-demand balance in grid operations. To achieve high levels of RES deployment, increasing the power system flexibility will be central to accommodate large fluctuations in supply and to cope with peak demand. Prospects of electricity storage technologies have emerged as a potential key technology to manage high levels of RES in the power system.
Recent projections on the cost of electricity storage show a high decrease in the next five years (,, ). As the commercial maturity of batteries might become a reality within the next decade, many questions remain on the role of batteries in the power system, where batteries should be located? What capacity will be optimal? What kind of battery services are the most valuable? How do batteries contribute to the large deployment of distributed RES installations? Significant research has been done on estimating sizing and sitting of storage in power systems. Yet, most of this research treat storage capacity as continuous instead of discrete, i.e. allocating storage by percentages of a total allowed capacity, wherever necessary in the grid. Despite these previous studies have provided interesting contributions on the value of storage in the power system, many of them lack the modeling of power flows, technical limits, or voltage considerations.
This thesis focuses on battery flexibility in medium voltage grids. Specifically, how to define cost-effective strategies to deploy batteries in a medium voltage grid? What is the optimal battery location in a distribution grid? And how do the technical limits of the power grid influence the allocation of storage? To address these questions, an optimization model was developed to simulate half-hourly operational decisions for a distribution grid. The model is multi-period and includes: power flows, diverse technical consideration for different battery sizes, high RES penetration levels, time of use electricity prices (half-hour dynamic prices), load data of actual customers and battery costs. To decide on the battery location, the model employs binary variables to determine the investment and sitting of the battery in a distribution grid. That is, the model is a mixed integer linear program with multi-period features which provides an investment analysis for the cost-effective sitting of batteries in a time horizon of 10 years. The model is implemented to the IEEE 33 bus test system. Results show that in general battery location and size strategies are driven by multiple factors, which can be either fixed or dynamic, like thermal limits and power load consumption, respectively. Some relevant findings are: First, flexibility in terms of power arbitrage delivers costs reductions of around 4% when RES production is low, compared to a No-Batteries case. Moreover, when RES production is high, the reductions in total costs can ramp up to 12% below the No-Batteries case. Also, this model decides to al-
locate batteries only if they are economically feasible for a 10-year time horizon. These results indicate a potential revenue up to 2.1 million pounds (GBP) based on the investment in batteries. And they are based on battery cost prices from 2008, together with several optimistic projections for the next decade. Furthermore, depending on battery size, RES penetration, RES generation and technical limits, batteries tend to be located for buses at the entrance of branches with high loads. Likewise, line limits and voltage limits proved to be decisive in the election of buses and the number of batteries placed.
On one hand, results show that optimal allocation strategies depend on grid topology features and technical limits, on the other hand, they also have a high dependency on time-varying and unsteady factors (e.g. power generation and loads). The optimal location strategies tend to change and adapt to the dynamics of the system. Moreover, with the exception of the slack bus, every bus in the system turned out to be an optimal location, at least once. These results indicate that batteries might be useful in every bus of the distribution grid, but only if each battery is operated in coordination and cooperation with one another. These insights support the idea of designing local electricity markets. Based on a reflection of this work, we recommend a market design that retrieves day-ahead and intraday DSO-reports of the battery operations and the flexibility that is available in the distribution grid. And also a subsequent structure of market incentives and penalties that maximizes the value of flexibility while keeping non-optimal operations to a minimum.
In short, this thesis contributes with a novel modeling approach that can shed some lights on the optimal battery allocation problem for distribution grids. Moreover, it provides insights on how location affects the value of storage, how optimal locations are affected by multiple technical factors. And finally, it also provides some reflections on the need to collectively, cooperatively and coordinately operate the storage resources in the grid by considering market-based solutions.