<p>Modeling nonstationary extreme wind speed events in the presence of seasonality and time-varying variability is crucial for building extreme wind-resilient offshore wind farms especially in coastal regions. In this paper, we propose and fit a nonstationary generalized extreme value model to extreme wind speed data with seasonality and time-evolving variability. We first diagnose nonstationarity across multiple temporal scales using wavelet power spectrum and maximum overlap discrete wavelet transform (MODWT) techniques. The wavelet-based analysis reveals pronounced variability at short-term, seasonal, and interannual scales, providing strong empirical justification for adopting a nonstationary extreme value framework. Guided by these findings, we fit a nonstationary Generalized Extreme Value (GEV) model in which seasonality is represented by sinusoidal covariates in the location parameter, while time-varying variability is captured through an exponential covariate in the scale parameter; the shape parameter is assumed constant. The performance of the proposed model is systematically evaluated using Monte Carlo simulation experiments, benchmarking it against stationary and simpler nonstationary alternatives. Results demonstrate that the proposed model consistently achieves superior goodness-of-fit according to the Akaike Information Criterion (AIC) and improved estimation accuracy, as measured by root mean squared error (RMSE), particularly as sample size increases. Using extreme wind speed data from selected wind stations in Kagoshima, results reveal that southern regions of Kagoshima, especially Kasari, face high return levels emphasizing their vulnerability to extreme winds. The proposed model provides valuable insight for wind-resistant infrastructure planning and management of offshore wind farm projects in Kagoshima.</p>