Delgaz Grid Non-Technical Losses Forecast

The distributor needed a solution to fully automate anomaly detection in the grid and to improve the consumption anomaly detection capabilities for the 54 SRMs along the whole distribution network.

We were faced with incomplete non real-time data as well as a great variability in the patters. On top of that, the aging infrastructure had leaks and made it difficult to both identify anomalies as well as verifying the output of our solution.

We delivered our solution for anomaly detection that used 4 main methodologies for detection, including change point detection and clustering and unsupervised learning.

We also implemented a kanban workforce management module, automating alarms and checks for anomalies with high probabilities of occurrence.