Data-driven flexibility prediction in low voltage power networks

Network digitalization brings new and massive amounts of data, opening opportunities for more secure and efficient operation of power networks, especially in medium and low voltage grids. This paper presents a set of indexes to quantify flexibility in medium and low voltage networks, considering key aspects, such as distance to congestion, phase imbalance and Distributed Energy Resources and their performance. A data-driven approach, based on Random Forest Regression, lets determine short-term flexibility in the network by predicting a set of indexes 15 min and one hour ahead. In addition to this, the operation scheme being experienced in every distribution network element is identified by comparing the succession of predicted indexes over a period of several hours with a set of characteristic curves previously analysed, providing additional enriched information. The proposed approach is validated by using real data from Smart-city Malaga Living Lab, which evidences that flexibility in medium and low voltage networks is not always proportional to demand, evolves differently throughout the time and is severely influenced by DER penetration rates.