Automatic Detection of Dynamic Wind Conditions in Mexican California: A Machine Learning-Driven Advancement in Wind Management

Abstract

Accurate detection and classification of wind states is crucial for accurately assessing and predicting wind energy production, fire spread behavior, air quality monitoring, and understanding complex meteorological phenomena. The complex nature of wind necessitates advanced data analysis techniques to extract meaningful patterns from meteorological datasets. This study presents a stochastic wind classification and identification analysis based on a Gaussian Mixture Model (GMM) clustering method applied to a case study in La Rumorosa'', Mexican California. By analyzing four meteorological variables (relative humidity, atmospheric pressure, wind speed, and wind direction) over five years, the method automatically identifies distinct wind conditions that can be defined as climate states, including well-known regional phenomena like Santa Ana winds and local orographic winds. Accurate detection of wind states enables better forecasting of wind energy potential at favorable sites, wildfire risk management through predicted fire behavior and monitoring pollutant/allergen dispersal patterns. The proposed approach offers a reliable, computationally efficient method for detecting wind patterns, extending to different geographical regions impacted by diverse wind phenomena.