An Improved Generalized Likelihood Uncertainty Estimation (GLUE) Approach for Uncertainty Analysis of Capacity Factors in Operational Wind Turbines

Abstract

The generalized likelihood uncertainty estimation (GLUE) method is widely used in many engineering applications due to its ease of implementation. Nevertheless, it has several drawbacks due to the subjectivity inherent in the statistically informal likelihood function, the cutoff threshold for identifying the behavioral solutions, and the probability in the posterior distribution. Hence, although the standard GLUE-based approach for wind energy probabilistic forecasting includes more measured data within the confidence intervals than the Monte Carlo (MC)-based approach, its performance becomes less in months with apparent wind speed fluctuations. For this reason, the present study extends the standard GLUE-based approach with three novel ways to include more behavior data points in the posterior distribution so that the result can be more representative. The resultant improved GLUE-based approach is assessed and compared with the standard GLUE-based and MC-based approaches through the 15 scenario test cases. Based on the results, the 50% and 90% confidence intervals of the proposed approach both match measured data better than the standard GLUE-based and MC-based approaches. Also, the proposed approach gives better predictions under all training and validation timespans. Thus, the improved GLUE-based approach is proven to be more effective and robust in accessing wind energy uncertainties.