Evaluation of Annual Maximum Wind Power Outage Capacity Induced by Extremely Low Temperature Using Generalized Extreme Value Distribution

Abstract

Severe wind turbine outage events due to the triggering of low-temperature protections greatly threatened the power supply safety. However, little is known on the occurrence pattern of such rare but extreme events, which is, however, important information for reserve planning to system operators. Focusing on annual maximum outage capacity (AMOC), this paper introduces return level and return period as key indicators to depict the consequence and the frequency respectively. A probabilistic model is necessary for calculating return level. However, the limited records challenge the distribution modeling. This paper proposes a generalized extreme value (GEV) distribution model with augmented samples as inputs. Augmented samples are obtained by combining the low-temperature protection settings of wind turbines with the spatially interpolated temperature sequences covering decades via Thiessen polygon division. The parameters of GEV distribution are inferred by Markov Chain Monte Carlo (MCMC)-based Bayes estimation method. Abundant parameter samples can be derived to facilitate the confidence interval evaluation of return levels of the AMOC, and the uncertainty in the parameter inference by small-size samples can be accounted for. Case studies based on actual data of China validate the effectiveness of the proposed method, and the findings fill the present knowledge gap.