Analysis of Failures, Downtimes and Reliability of Wind Turbines From SCADA Data

ABSTRACT With the growing demand for safe, efficient and reliable sustainable energy sources, wind energy has witnessed rapid growth worldwide in recent years. Reliability is critical to the operational performance of the wind turbines. Reliability analysis allows for the critical subassemblies to be identified and better maintenance strategies to mitigate the risk of failures and power outages. This paper proposes two algorithms to extract relevant reliability data from unprocessed SCADA data, followed by a reliability analysis of the extracted data. The time between failures and downtimes is modelled and classified into exact and right censored data. The key metrics including downtime, failure rate and availability are estimated and analysed. The maximum likelihood estimators for the exponential, Weibull and modified Weibull distributions are developed. The parameters and distributions are evaluated and compared using the log‐likelihood function, Akaike information criterion (AIC) and Bayesian information criterion (BIC). SCADA data from a real‐world case study wind farm in Portugal is analysed, and it is observed that the generator is the subassembly with the highest downtime and failure rate. Additionally, our findings suggest that the Weibull and modified Weibull distributions offer better modelling options for wind turbine failures when compared with traditional conditional failure rate models commonly referenced in the literature.