The operation of a community energy storage system (CESS) is challenging due to the volatility of photovoltaic distributed generation, electricity consumption, and energy prices. Selecting the optimal CESS setpoints during the day is a sequential decision problem under uncertainty, which can be solved using dynamic learning methods. This paper proposes a reinforcement learning (RL) technique based on temporal difference learning with eligibility traces (ET). It aims to minimize the day-ahead energy costs while maintaining the technical limits at the grid coupling point. The performance of the RL is compared against an oracle based on a deterministic mixed-integer second-order constraint program (MISOCP). The use of ET boosts the RL agent learning rate for the CESS operation problem. The ET effectively assigns credit to the action sequences that bring the CESS to a high state of charge before the peak prices, reducing the training time. The case study shows that the proposed method learns to operate the CESS effectively and ten times faster than common RL algorithms applied to energy systems such as Tabular Q-learning and Fitted-Q. Also, the RL agent operates the CESS 94% near the optimal, reducing the energy costs for the end-user up to 12%.@en

The energy transition poses a challenge for the electricity distribution network design as new energy technologies cause increasing and uncertain network loads. Traditional static load models cannot cope with the stochastic nature of this new technology adoption. Furthermore, traditional nonlinear power methods have difficulty evaluating very large networks with millions of cables, because they are computationally expensive. This paper proposes a method which uses copulas for modeling the uncertainty of technology adoption and load profiles, and combines it with a fast linear load flow model. The copulas are able to accurately model the stochastic behavior of solar irradiance, load measurements, and mobility data, converting them into electricity load profiles. The linear load flow model has better scalability and stability compared to traditional load flow models. The models are applied to a case study which uses a real-world dataset consisting of a realistic technology adoption scenario and a low-voltage network with millions of cables, which considers both voltage and current problems. Results show that risk profiles can be generated for all cables in the network, resulting in a valuable map for the district network operator as to where to focus their efforts.@en

The rapid increase in the number of PV installations in current low voltage (LV) distribution networks brings many technical operational challenges. This claims for the deployment of control strategies to deal with these concerns, especially those related to overvoltage issues. Based on this, this paper presents a comprehensive assessment of the performance of PV inverters operating with droop control for overvoltage mitigation using a stochastic methodology based on a Monte Carlo approach. The uncertainty related to the PV generation and the users’ consumption behavior is fully considered through advanced statistical modeling techniques. Voltage magnitude and loading indexes are used as key metrics to assess the technical performance of the distribution network, simulated using OpenDSS, under two droop-based control strategies: Active Power Control (APC) and coordinated Reactive and Active Power Control (RPC-APC). The effects of curtailed energy on the PV users’ revenue is also analyzed. A case of study based on real smart meter data from The Netherlands is used. According to the obtained results, both control strategies are effective to mitigate voltage violations. Nevertheless, for the case of 100% PV penetration, the droop-based coordinated RPC-APC allowed an 18% more of exported energy than the droop-based APC control strategy.@en