Root-cause analysis of power quality disturbances utilizing Machine Learning

Purpose and goal

The project aimed to develop a Minimum Viable Product (MVP) based on Eneryield´s machine learning algorithms together with electricity utilities, and to develop an IP strategy for Eneryield. An MVP for automatic reporting of root causes of disturbances, based on proprietary methods of machine learning, has been successfully developed and tested in this project. An IP strategy to strengthen the control position for Eneryield´s innovations at market introduction has been established.

Expected results and effects

Results show that machine learning can be implemented on the electricity grid as well as important conditions and limitations for this. The project has led to deeper insights into disturbances on the electricity grid, e.g. how different disturbances relate to each other. The project has created collaborations to bring the innovation to market in a short amount of time. Additionally, the project provided guidelines for increased control of intellectual property. Expected effects are shorter disturbance analysis, which will lead to increased reliability and minimize energy losses.

Planned approach and implementation

At the time of application, the hypothesis was that work packages would run sequentially. However, it was shown early that a more agile approach would be required. This is because data collection posed a greater challenge than expected. Utilities have different routines and processes for collecting and storing data, which required adapted and continuous processes. At the same time, better knowledge about the value chain has been obtained and more actors have been involved in the project. Which has contributed to further areas of application for the technology have been discovered.