Presentations

Abstract

Operators and owners of large Renewable Energy (RE) portfolios require innovative tools to efficiently operate and maintain the assets, monitor their performance and improve reliability. The time and expertise requirements to perform manual checks of each individual asset becomes prohibitive to do so at scale - automation is required. To this effect, certain processes can be checked automatically using digital representations of assets using statistical methods - i.e. Machine Learning (ML) models, and readily available 10-min resolution SCADA data and condition monitoring system (CMS) vibration data (e.g. time series and pre-computed vibration frequency spectra). However, to increase efficacy, accuracy and buy-in from operators and owners, GreenPowerMonitor (GPM), a DNV company, enrich the ML findings by Human-In-The-Loop (HITL) expertise of both the data-science and physical modelled machinery allowing reliability focussed maintenance predictions. In this study, the authors present the outcomes and learnings of an integration of a Predictive Maintenance Analytics (PMA) system with the Swiss asset owner/operator BKW Energie AG (BKW). BKW is an international energy and infrastructure company offering integrated services in the field of energy, building and infrastructure to Clients in Switzerland and other European Countries (Italy, Germany, France, Norway, Sweden). BKW is actively involved in the ownership, Technical and Commercial management (TCM) and Operations & Maintenance (O&M) of its wind portfolio. The deployed PMA system is an automated diagnostic service offering. The study offers guidance on what did and didn't work in an effort to support wider deployment of such systems in the wind industry.