In Denmark, around 50 percent of electricity comes from renewable sources, mostly wind power. Our mandate is to increase that to 100 percent by 2030. This creates some challenges for Energinet, Denmark’s electric transmission systems operator, because renewable energy always fluctuates. We have to manage the grid carefully to maintain the security of supply.

We’ve developed tools to manage the amount of renewable energy we have today, but as it increases, we’ll need new and better tools. Otherwise, we’ll likely have to make costly infrastructure investments or face brownouts and blackouts.Our current control room tools are good at modeling the grid to simulate error conditions, but the simulations and live data feeds generate big data that remains untapped. That led us to wonder whether an analytical tool could discover insights to improve grid management.

Big data and AI: Advancing grid management decision making

To test the concept, we collaborated with IBM Services on a pilot project. The result was a real technological leap for Energinet—a multicloud solution that gleans operational predictions from big data using AI.

Accessing the system from a web interface, operators get help answering questions like, “What would happen if we took equipment out of service at this time?” or “Based on past experience, which assets are at risk of failing?” It’s a huge step forward in decision support.

An important use case is helping operators evaluate planned maintenance. If the maintenance team wants to take down a line or transformer, operators need to assess the risks. The system’s predictions are likely to be more accurate than their intuitions. Other uses include assessing grid operations, understanding system bottlenecks and suggesting cost-effective investments.

Interested in learning more about solutions for energy and utilities?

From design to proof of concept in three months

Energinet personnel had the idea for the solution, but participating in design thinking sessions helped us understand what is possible and how to do it. Then, with an agile approach we developed the proof of concept in just three months. That’s very fast and cost effective compared to traditional systems development for the control room.

Key to the analytical power is preparing the big data for AI. Systems running on the Microsoft Azure cloud first create simulation and real-time datasets. IBM Cloud Pak for Data on Azure allows users to query the system and AI generates the analysis.

Of course, the usefulness depends on operators trusting the AI. The pilot addressed this by offering explanations for its predictions. We tested the capability by simulating outages with known causes and remedies. Experienced operators easily recognized what to do and why, and then compared their thinking to the AI analysis. The fact that they generally agreed increased trust in the system.

A positive step for a green future

In conceiving the solution, we aimed to help operators understand the risks of removing equipment from the grid. The project proved that possibility and more.

In the future, we plan to advance the concept to where we can look ahead, perhaps over the next 24 hours, to suggest actions that prevent a cascade of problems that might come later. Such AI capabilities can help assure a secure and cost-effective renewable energy supply.

Watch Einar Ritterbusch discuss about moving to renewable energy in a cost-effective way: