TY - GEN
T1 - Very Short-Term Wind Speed Prediction Techniques Using Machine Learning
AU - Mogos, Aman Samson
AU - Salauddin, Md
AU - Liang, Xiaodong
AU - Chung, Chi Yung
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021/9/12
Y1 - 2021/9/12
N2 - Wind speed prediction plays an essential role in planning and operation of wind power systems. An accurate wind speed prediction can reduce costs and enhance the proper use of resources. Wind speed series have high nonlinearity and volatility. In this paper, data-driven models using machine learning (ML) algorithms have been developed to predict a very short-term wind speed. Historical wind speeds lagging up to 20 minutes with 1 minute time interval are used to predict the current and future (up to 5 minutes with L-minnte interval) wind speed. A performance comparative analysis of four ML algorithms including Multiple-Layer Perception Regressor (MLPR), Random Forest Regressor (RFR), K-nearest Neighbors Regressor (KNNR), and Decision Tree Regressor (DTR), is conducted, and their accuracy is evaluated by their R2 values, mean absolute error (MAE), standard deviation (SD) of MAE, mean absolute percentage error (MAPE), SD of MAPE and root mean square error (RMSE). It is found that MLPR gives the best prediction accuracy of 95.3%.
AB - Wind speed prediction plays an essential role in planning and operation of wind power systems. An accurate wind speed prediction can reduce costs and enhance the proper use of resources. Wind speed series have high nonlinearity and volatility. In this paper, data-driven models using machine learning (ML) algorithms have been developed to predict a very short-term wind speed. Historical wind speeds lagging up to 20 minutes with 1 minute time interval are used to predict the current and future (up to 5 minutes with L-minnte interval) wind speed. A performance comparative analysis of four ML algorithms including Multiple-Layer Perception Regressor (MLPR), Random Forest Regressor (RFR), K-nearest Neighbors Regressor (KNNR), and Decision Tree Regressor (DTR), is conducted, and their accuracy is evaluated by their R2 values, mean absolute error (MAE), standard deviation (SD) of MAE, mean absolute percentage error (MAPE), SD of MAPE and root mean square error (RMSE). It is found that MLPR gives the best prediction accuracy of 95.3%.
KW - Data-driven
KW - Machine Learning
KW - Regression
KW - Wind Speed Prediction
UR - http://www.scopus.com/inward/record.url?scp=85118434540&partnerID=8YFLogxK
U2 - 10.1109/CCECE53047.2021.9569134
DO - 10.1109/CCECE53047.2021.9569134
M3 - Conference article published in proceeding or book
AN - SCOPUS:85118434540
T3 - Canadian Conference on Electrical and Computer Engineering
BT - 2021 IEEE Canadian Conference on Electrical and Computer Engineering, CCECE 2021
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2021 IEEE Canadian Conference on Electrical and Computer Engineering, CCECE 2021
Y2 - 12 September 2021 through 17 September 2021
ER -