With the growing availability of (near) real-time measurement data from the electricity grid, it will become possible for energy parties to take maintenance activities to another level. Grid Ripple is a data science application for network operators. It identifies critical nodes in the electricity network early on and prevents unplanned outages through preventive maintenance and proactive load distribution. Similar to the effect of a drop of water in a still pool, Grid Ripple realizes the absorption of locally deviating behavior through automatic up and down switching of capacity per node at the grid level. Critical nodes can be spared until maintenance is carried out, by proactively utilizing overcapacity and redundancy.
In 2018, Grid Operator Stedin launched a 'Starthubs' challenge for detecting disruptions in the electricity grid. In September 2018, Stedin named West IT Solutions as the winner of the challenge. After that we promptly started with the pilotfase. Stedin and West worked closely together in this pilot to realize a proof of concept for the Grid Ripple smart grid application. After a successful Demo Day on December 20, 2018, we started with the next fase to further develop the solution. During the pilot, West and Stedin focussed their efforts on the development and validation of the Grid Ripple detection module.
First of all the relevant data sources needed to be identified, linked and cleaned up. After that, we experimented with large data sets to understand the meaning and hidden relationships of the data sets. Finally, an algorithmic model was trained; first on a part of the data set, and later on it was scaled it up to the rest of the grid.
We chose a technical approach in which transparency of the results was the most important result. In doing so people are able to easier understand how the model makes decisions about the different signals involved in a malfunction. The project generated the required internal support due to the positive results of the pilot. These first successes will be the building blocks of future development and implementation.
Ranko Stojakovic of Stedin explains: â€œThe 'Starthubs' challenge is intended to see whether we can achieve an automatic fault detection with our current data set. This concerns SCADA data, data about malfunctions and data from our activities. But apart from collecting this data, interpreting it is perhaps even more important. And after choosing West, colleagues from the Operations Center put quite a lot of time into it. Bart Bikker, a colleague from the Operations Center mentioned: "We have validated (tested) and interpreted the data together with West. We are currently working on a DecisionTreeClassifier, among other things. With this we hope to automatically generate a decision tree based on our signals with which we can detect malfunctions."