Data Fusion Pipelines (Wind-16)

Business Need

There is significant interest in predictive maintenance for wind turbines, which can lead to reduced downtime, increased availability, and higher utilisation of components’ remaining useful life. The literature estimates cost savings in Operational Expenditures (OpEx) of up to 12.5% and a further reduction in lost production of up to 15.75%, ultimately contributing to a lower cost of wind energy. Typically, such approaches rely on normal behaviour, or remaining useful life models which leverage SCADA data and the context around it such as work orders, stoppages, etc. Incorporating vibration data from Condition Monitoring Systems could provide better fault detection at the incipient stage and by extension provide more actionable business intelligence. However, the inherently different capture frequency, and the associated volume and velocity of the two data sources, pose significant challenges in fusing them and developing models that would have access to the asset context from multiple fused different data streams at the same time.

Key partners

The solution

The project saw the development of a data fusion pipeline that successfully merges the two data streams. In-depth data cleaning is incorporated, sidestepping the need for aggressive decay rates and allowing for feature engineering to be revisited once the pipeline is meshed with a model training and predictive/inference step. Code flexibility in the sense of multiple alternatives have also been included to cater to different assets and sites, while also components of the code can be re-used for further feature extraction in collaboration with vibration specialists.

Next steps

  • A measurement campaign capturing functional failures, leveraging the incorporated data cleaning to reduce volume constraints.
  • Mesh with a predictive pipeline that will provide a complete solution from raw capture to inference.
  • Expand feature extraction in collaboration with vibration specialists.

Industry quotes

The project has successfully demonstrated the potential of data fusion pipelines for predictive maintenance in wind turbines. The flexibility of the developed software code allows it to be applied to various types of data and assets, making it a versatile tool. The collaboration with Strathclyde underscores the value of academic-industry partnerships in driving innovation and achieving tangible improvements in the renewable energy sector.

Dr Sofia Koukoura, Scottish Power

We are pleased to report another successful project completed by the Low Carbon Power and Energy partnership between SSE, SPR and the University of Strathclyde. In this project data processing methods have been developed to combine data from multiple sources to enable automated assessment of the turbine health. This technology could drive efficiency in maintaining our generating fleet, which would ultimately lead to a lower cost of electricity.

Dr Stuart Killbourn, Technology Engineer, SSE