Real time Markov chains: Wind states in anemometric data

TL;DR

This paper introduces a Markov chain-based method to analyze wind states from anemometric data, providing a new way to model wind dynamics for improved wind resource assessment and wind farm design.

Contribution

It proposes a novel discretization technique using wind velocity vectors and clustering to model wind behavior with Markov chains, enhancing understanding of wind dynamics.

Findings

Markov chain models accurately capture wind state transitions.

Transition probabilities and residence times serve as site-specific signatures.

Method applied successfully to Mexican wind sites.

Abstract

The description of wind phenomena is frequently based on data obtained from anemometers, which usually report the wind speed and direction only in a horizontal plane. Such measurements are commonly used either to develop wind generation farms or to forecast weather conditions in a geographical region. Beyond these standard applications, the information contained in the data may be richer than expected and may lead to a better understanding of the wind dynamics in a geographical area. In this work we propose a statistical analysis based on the wind velocity vectors, which we propose may be grouped in "wind states" associated to binormal distribution functions. We found that the velocity plane defined by the anemometric velocity data may be used as a phase space, where a finite number of states may be found and sorted using standard clustering methods. The main result is a discretization technique useful to model the wind with Markov chains. We applied such ideas in anemometric data for two different sites in Mexico where the wind resource is considered reliable. The approximated Markov chains of both places give a set of values for transition probabilities and residence times that may be regarded as a signature of the dynamics of the site. We consider the results may impact either the technical decisions for the design of wind generators or the detailed resource assessment in sites with non-standard features.