The use of digital technologies such as cloud computing, big data management and predictive analytics are changing the status quo of the energy market. An increasing number of wind farm operators are starting to implement these techniques, so as to control their costs. But, if the implementation of predictive maintenance is one of your digital business transformation goals, how will you make sure it works?
Early stage failure detection by real-time predictive analytics
When talking about predictive analytics, one of the core features is detection. Beyond that, the real need is data reliability in order to detect failures and relevant anomalies to schedule a maintenance intervention. But, is the required wind turbine data quality good enough?
Let’s start with an overview of the wind energy market grading. This analysis has been carried out by taking a representative sample of data coming from European wind farms.
The following figure represents the wind industry grading mix from a typical European utility, which represents the sample for the analysis of this article.
This grading takes into consideration the evolution of wind industry technology both, on turbine operations and the data variety needed for big data and predictive analytics services.
90% of the wind turbine technology mix comes from assets with less than 2MW capacity per turbine, where the developments at data variety measurements have evolved from 20-80 variables on the 1st generation turbines, up to 45-130 variables on the 2nd generation. The other 10% turbine are a mix from the 3rd generation turbines and represent the latest industry developments in wind turbines technology and data variety, where it keeps a similar amount of measurement variables as the 2nd generation, because some systems have decreased their sensor base or increased the number of measurements, like the rotor (pitch), the gearbox and the converter.
Nevertheless, independently from the different turbine generations, there are two data quality issues that appear across the 3 different groups (SCADA data which exclude cumulative and counter measurements), which are:
