Francisco Sousa Lobo

Abstract
This paper describes a study developed to analyse the interest in investing in Li-ion batteries to perform price arbitrage in the power system of Portugal. In this context, it was developed a methodology to identify the most suitable hours for charging and discharging the energy, and the new market prices were estimated for these hours. It was concluded that at current investment costs in this storage technology, and current market prices, this investment would not be viable in the lifetime of the batteries despite the recent rise of electricity market prices and also the larger price spread. This spread is now larger given the depression of prices at sunny hours that is getting typical in the Iberian electricity market.

Abstract
Hybrid storage systems that combine high energy density and high power density technologies can enhance the flexibility and stability of microgrids and local energy communities under high renewable energy shares. This work introduces a novel approach integrating rule-based (RB) methods with evolutionary strategies (ES)-based reinforcement learning. Unlike conventional RB methods, this approach involves encoding rules in a domain-specific language and leveraging ES to evolve the symbolic model via data-driven interactions between the control agent and the environment. The results of a case study with Liion and redox flow batteries show that the method effectively extracted rules that minimize the energy exchanged between the community and the grid.