The demand for electricity has exploded. The Dutch power grid is under strain and network operators are unable to keep up with developments. This has become a problem for both the business sector and members of the general public. How can this problem be solved? Network administrator Alliander sought help from the team of data scientists that is headed by Roel Bouman of Radboud University. Without energy, the Netherlands would come to a standstill. Companies like Alliander develop and manage the power grid. Households and businesses receive electricity and gas via cables and pipelines. Alliander manages these types of services for over three million customers. “Energy transition, digitalisation, residential construction and economic growth are putting an increasing strain on the electricity grid. In certain locations and at particular times the strain is immense,” explains Jacco Heres, who is a data scientist at Alliander. “We’re working hard to expand the power grid, but in more and more locations, the demand for electricity is growing faster than we can supply it,” Jacco Heres continues. “And we won’t be able to offer industry, offices and supermarkets extra capacity in those locations until the grid has been expanded. In addition to expansion, we’re working on ways in which we can use the power grid more intelligently and more efficiently so that we can prevent congestion.” Project STORM In order to prevent such congestion, greater insight into all of the current that flows through the grid is required. STORM, which is a collaborative project that is being managed by Roel Bouman, is helping to achieve this; Alliander and the Data Science department at Radboud University are both taking part in this project. Roel Bouman is responsible for the project management and technical overview within the project team. He holds a degree in Chemistry and Computer Science, and specialises in Data Science and Machine Learning. “In order to give the project a boost, we organised a hackathon on campus in November 2021,” says Roel Bouman. “The aim of the event was that the participants would give us ideas for using algorithms to make fully automated predictions about when switch events would occur in the data.” Automatic filtering “Alliander already uses measurements to try and predict when and where most activity will take place. But measurement data invariably contain errors and irregularities. In addition to this, there are ‘redundant’, or alternative, routes that are used when, for example, a cable breaks somewhere in the grid. This consequently increases the activity on other routes. If you want to measure the actual load and make predictions for optimal use, you’ll have to filter out the incorrect data. And this should preferably be done automatically,” explains Roel Bouman, “because if it’s done by people, it takes a great deal of time and you’re also dependent on their expertise and availability.” Mathematical model “Using machine learning means always being current and that you have up-to-date data at any given time.” Roel Bouman’s team developed a mathematical model that can use algorithms to analyse this data. “It’s like solving a puzzle, sort of like doing a super sudoku. It involves doing and thinking in equal measure. We’ve made a demonstrator in which the data analysis is fully digital and automatic. It works well and Jacco and his team are now implementing it at Alliander.” So, is this the solution? “Yes, it is,” says Roel Bouman. “Despite the fact that updates will always be needed, because the truth is that every system requires maintenance.” This article appeared earlier (in Dutch) on the website of TechGelderland. Photo via Piqsels.