Within the scope of the SUMMER programme, aiming at providing better access to commercial and plant data of redelivery points, Snam needed to find a new approach to manage the issue of Unaccounted-for Gas (UFG). It is the quantity of gas remaining between the gas injected and the gas delivered to end users, net of line-pack variations and the consumption related to the network op. management.

An initial prototyping activity lasting for one year was useful to collect data for the next project phases. We identified six main use cases and adopted a methodological approach to build the data mining model (Crisp-DM). It progressively allowed to identify correlations between UFG & sub-threshold measures. By discussing the outcomes of our analyses with Snam, which we carried out with the help of machine learning activities, we improved the understanding of the phenomena. Hence, we began an industrialisation plan of the models implemented and defined a roadmap to complete the planned cases in 2022. The architecture is based on Microsoft stack: Data Lake Gen2 for data storage, Synapse for data structures, Azure Machine Learning for FE Analytics. UI is designed according to SUMMER guidelines and to market best practices. It allows users to create standard reports independently.