Updated: Nov 5, 2018
Last week saw the release of version 1.0.3 of SIFT, bringing new analysis features and several smaller improvements here and there. We’re particularly excited about the new Scripting Sandbox where users build their own custom analysis methods within the SIFT ecosystem. Within the Sandbox, the user has free range to wield the power of Python in any way they see fit, while benefiting from having all the data managed for them.
Data management and preparation is not seldom a major time sink in the analysis of large data sets, and with this new module we offer significant time savings to analysts and engineers without sacrificing the flexibility that a scripting environment offers.
Peter van Heck, a data scientist, and Craig Stout, an engineer, both working with large data sets from operational offshore wind farms at the ORE Catapult have been providing pre-release testing support. We caught up with them to find out how they have been using the new Sandbox so far and what’s next on the menu:
SIFT has provided ORE Catapult with a fantastic toolkit for the day-to-day analysis of operational data. With this new intuitive scripting interface, we can apply our in-house scripting capabilities directly in combination with the powerful analysis options within SIFT.
We have initially been using the new sandbox environment to test the in-house logic conditions we use to classify operational states of turbines, thus allowing us to fine-tune complex rule-bases to determine the running condition of turbines in specified intervals. We plan to build on this by implementing machine learning algorithms from within SIFT, with the aim of anomaly detection and the identification of outliers in complex datasets.
To demonstrate the power of the Sandbox we’ve spent a couple of hours building a random forest regressor using Scikit-learn to predict (and detect changes) in the temperature of a gearbox bearing, and another one to automatically classify a range of power curve performance issues. Keep your eyes out for a forthcoming article where we’ll demonstrate the amazingly short process and code required to achieve this.
Sift is a product of Lindahl Ltd. and is available to licensing by your organisation. If you’d like to find out more or get your hands on a free trial, then don’t hesitate to get in touch!