By: Vitor Lopes
At: Instituto de Investigação Interdisciplinar, Anfiteatro
Wind energy is becoming a top contributor to the renewable energy mix, which raises potential reliability issues for the grid due to the fluctuating and intermittent nature of its source. This presentation addresses the use of Markov chain models for the analysis of wind power generation systems and shows how to extract relevant statistical information from historical data, such as the long-term wind speed and power distributions or persistence of different levels of power production.
Wind turbine behavior depends, among other factors, on the wind speed, direction and power production level, which must be incorporated into the Markov chain model. We propose a new definition of the wind turbine states, which takes the intrinsic interdependencies of these three variables into account. The application of standard maximum likelihood estimators to small historical datasets introduces limitations in the determination of Markov chain transition probabilities, namely in those associated with infrequent transitions. Thus, to capture more information from the data, a novel maximum likelihood estimator based on multi-step transitions is presented.
The methodology is illustrated with the analysis of a 2 yr dataset obtained from a wind power turbine located in the Pinhal Interior region in Portugal.