Wind power prediction using deep neural network based meta regression and transfer learning

• Published in: Applied soft computing Vol. 58; pp. 742 - 755
• Main Authors: Qureshi, Aqsa Saeed, Khan, Asifullah, Zameer, Aneela, Usman, Anila
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2017
• Subjects: Meta-regressor; Meteorological properties; Sparse denoising auto-encoders; Transfer learning; Wind power prediction; Wind power prediction; Sparse denoising auto-encoders; Meteorological properties; Meta-regressor; Transfer learning
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2017.05.031

[Display omitted] •Wind power prediction system using Deep Neural Network based Ensemble and Transfer Learning.•Ensemble learning is shown to enable collective and robust decision on unseen data.•Use of Transfer Learning is shown to enable quick learning on a new wind farm.•Deep Auto-encoders act as base, whereas Deep Belief Network acts as Meta regressor.•The proposed wind predictor has good generalization and thus robust to small variations in metrological properties. An innovative short term wind power prediction system is proposed which exploits the learning ability of deep neural network based ensemble technique and the concept of transfer learning. In the proposed DNN-MRT scheme, deep auto-encoders act as base-regressors, whereas Deep Belief Network is used as a meta-regressor. Employing the concept of ensemble learning facilitates robust and collective decision on test data, whereas deep base and meta-regressors ultimately enhance the performance of the proposed DNN-MRT approach. The concept of transfer learning not only saves time required during training of a base-regressor on each individual wind farm dataset from scratch but also stipulates good weight initialization points for each of the wind farm for training. The effectiveness of the proposed, DNN-MRT technique is expressed by comparing statistical performance measures in terms of root mean squared error (RMSE), mean absolute error (MAE), and standard deviation error (SDE) with other existing techniques.