A Multi-Agent Fault Detection System for Wind Turbine Defect Recognition and Diagnosis

This paper describes the use of a combination of anomaly detection and data-trending techniques encapsulated in a multi-agent framework for the development of a fault detection system for wind turbines. Its purpose is to provide early error or degradation detection and diagnosis for the internal mechanical components of the turbine with the aim of minimising overall maintenance costs for wind farm owners. The software is to be distributed and run partly on an embedded microprocessor mounted physically on the turbine and on a PC offsite. The software will corroborate events detected from the data sources on both platforms and provide information regarding incipient faults to the user through a convenient and easy to use interface.