Multi-target normal behaviour models for wind farm condition monitoring

TL;DR

This paper proposes multi-target regression models for wind turbine condition monitoring, demonstrating they reduce maintenance effort and cost while maintaining accuracy, compared to traditional single-target models.

Contribution

It introduces the application of multi-target regression models to wind turbine monitoring, highlighting their advantages over single-target models in efficiency and cost-effectiveness.

Findings

Multi-target models outperform single-target models in accuracy.

Multi-target models reduce maintenance costs and effort.

The approach is validated through a wind farm case study.

Abstract

The trend towards larger wind turbines and remote locations of wind farms fuels the demand for automated condition monitoring strategies that can reduce the operating cost and avoid unplanned downtime. Normal behaviour modelling has been introduced to detect anomalous deviations from normal operation based on the turbine's SCADA data. A growing number of machine learning models of the normal behaviour of turbine subsystems are being developed by wind farm managers to this end. However, these models need to be kept track of, be maintained and require frequent updates. This research explores multi-target models as a new approach to capturing a wind turbine's normal behaviour. We present an overview of multi-target regression methods, motivate their application and benefits in wind turbine condition monitoring, and assess their performance in a wind farm case study. We find that multi-target models are advantageous in comparison to single-target modelling in that they can reduce the cost and effort of practical condition monitoring without compromising on the accuracy. We also outline some areas of future research.