Premium
Detecting ramp events in wind energy generation using affinity evaluation on weather data
Author(s) -
Fan Ya Ju,
Kamath Chandrika
Publication year - 2016
Publication title -
statistical analysis and data mining: the asa data science journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.381
H-Index - 33
eISSN - 1932-1872
pISSN - 1932-1864
DOI - 10.1002/sam.11308
Subject(s) - wind power , interval (graph theory) , computer science , data mining , data set , grid , feature selection , wind speed , schedule , set (abstract data type) , energy (signal processing) , meteorology , artificial intelligence , statistics , engineering , geography , mathematics , geodesy , combinatorics , electrical engineering , programming language , operating system
Ramp events, which are significant changes in wind generation over a short interval, make it difficult to schedule wind energy on the power grid. Predicting the occurrences of these events can help control room operators ensure that the load and generation on the power grid are in balance at all times. In this paper, we focus on predicting up‐ramp events, which are large increases in generation in a short time interval. We propose a novel detection algorithm that uses historical data to detect incoming pre‐ramp events, which are defined as the part of the time series that occurs before ramp events. Using wind energy generation data from Bonneville Power Administration in the mid‐Columbia Basin region, and weather data from the nearby meteorological towers, we define the concept of affinity of weather data to the preramp events. This is used to identify important weather variables and predict the pre‐ramp events. A comparison of our approach with traditional feature selection and classification methods indicates that our method identifies a similar set of features as important and gives better detection accuracy. © 2016 Wiley Periodicals, Inc. Statistical Analysis and Data Mining: The ASA Data Science Journal, 2016