Description

Increasing numbers of household prosumers create big amounts of load and generation data available for analysis. ESO being a proactive distribution network operator started utilizing this data for creating and enabling additional services for clients. Modern machine learning algorithms and data accessibility enables development of innovative smart services for clients.

Solution

During summer period 2021 ESO tested novel idea on data analysis. During this period our organization already had huge amount of hourly electrical energy consumption data from household prosumers. Also it wasn't unheard of failures of PV systems installed at consumer premises. The idea came whether it's possible to distinguish failed PV system data pattern from the others. Few different machine learning algorithms were employed to do the task. After some optimization steps some of them started to show promising results. Prompt decision was taken to test it for one summertime season on real prosumers.

Status and Progress

The pilot project was successful, during the period summer of 2021 in total 60 clients were informed about the potential malfunction of their PV systems. Post-pilot survey confirmed successful result showing more than 90% true-negative prediction accuracy. Today new service is developed and already implemented in ESO every day activity.