Leveraging AI to transform incident analysis for grid operators

The energy industry is undergoing a significant transformation, with the integration of renewable energy sources, electrification of transport, and increasing digitalization of the grid. Innovation is essential for transmission system operators (TSOs) to maintain a reliable and efficient grid. In 2021, Elia launched a Moonshot program, aiming to tackle some of the biggest challenges in the energy transition. One area where the moonshot programme is focusing and where innovation and technology can play a critical role is improving the identification of incident causes in overhead lines. In this context, Elia and N-SIDE have co-developed a tool that leverages AI to help in the identification of incidents.

Traditionally, TSOs have relied on sending patrols to the field to investigate incidents, which can be time-consuming, costly and have safety concerns. The use of machine learning algorithms enables near-real-time incident cause recognition. With the benefit of initial information on the cause of incidents, teams in the field can better prioritize their efforts on incidents with highest severity and impact. This allows TSOs to make risk-informed and faster decisions, which results in lower downtime, reduced costs, and increased safety .

Collected data from multiple sources

Incidents in overhead lines, such as lightning strikes, strong winds, animals, objects, and others, require the collection of data from multiple sources. Our solution makes use of historical weather conditions, asset properties, fault recordings, and incidents characteristics. This diverse and comprehensive dataset is processed and prepared for use in our machine learning algorithm, optimized to classify incidents based on their cause.

Architecture of solution. Multiple streams of data are collected, processed and used as an input to the ML model

Leveraging AI to provide decision support to operators

The success of the tool can be attributed to several factors:

• Elia's experts provided their knowledge and expertise as an input when developing the model.

• State-of-the-art machine learning techniques were implemented to handle small data.

• Signal processing techniques were applied to voltage and current signals.

• Emphasis was placed on ensuring the model's transparency and explainability to promote user trust and understanding.

These factors have led to the development of a powerful and effective tool that can greatly improve the efficiency and reliability of TSO operations. 

To assess performance,the tool was tested on unseen events, achieving an accuracy of 80% in the classification of incidents among five different causes.

The future of system operations

Our tool demonstrates the significant potential of using data-driven approaches to enhance TSO operations. As the project enters its third phase of development, it is expected to achieve even greater accuracy and performance. This will be achieved by increasing the amount of historical data used to train the model. In addition, Elia’s expert knowledge will be translated into additional features.

The future of system operations will likely involve increased collaboration between human experts and AI-powered tools, enabling TSOs to harness the full potential of their data.

In addition to incident cause recognition, the tool can be implemented in other areas of transmission system management, such as predictive maintenance, asset management, and grid optimization. As the energy industry continues to evolve, innovative solutions like ours will play an essential role in ensuring the resilience and sustainability of electrical networks.

We invite interested parties to reach out to us for further information and collaboration opportunities, as we continue to refine and expand our AI-based solutions to better serve the needs of the energy industry.