Multi-dimensional Evaluation of Temporal Neural Networks on Solar Irradiance Forecasting

With recent developments and advances in machine learning methods, traditional time series analysis techniques, e.g., ARMA, ARIMA, ARIMAX, SARIMAX, etc. models are being replaced by deep learning models. The application of Recurrent Neural Network (RNN) and Long-Short Term Memory (LSTM) to the problem of solar irradiance prediction has been shown in the literature to achieve state-of-the-art performance in this domain. For sequence modeling, Temporal Convolutional Networks (TCN) are gaining increasing attention because of the excellent trade-off between performance accuracy and time in training the models. In this paper, an evaluation of these deep learning methods is completed on the application of short-term, one hour ahead solar irradiance forecasting. The results show that TCNs can reduce the resource usage, depending on the data-folding size and model complexity, to require only 5% of the training time with only a loss of 6% RMSE.

[1]  Vladlen Koltun,et al.  An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling , 2018, ArXiv.

[2]  Ilhami Colak,et al.  Multi-period Prediction of Solar Radiation Using ARMA and ARIMA Models , 2015, 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA).

[3]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[4]  Guillaume Lample,et al.  Neural Architectures for Named Entity Recognition , 2016, NAACL.

[5]  Alireza Inanlougani,et al.  Evaluation of Time-Series, Regression and Neural Network Models for Solar Forecasting: Part I: One-Hour Horizon , 2017 .

[6]  Mohamed Abuella,et al.  Solar power forecasting using artificial neural networks , 2015, 2015 North American Power Symposium (NAPS).

[7]  Gordon Reikard.  Predicting solar radiation at high resolutions: A comparison of time series forecasts , 2009 .

[8]  Özlem Durmaz Incel,et al.  Time series forecasting on solar irradiation using deep learning , 2017, 2017 10th International Conference on Electrical and Electronics Engineering (ELECO).

[9]  Gang Sun,et al.  Squeeze-and-Excitation Networks , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[10]  Jürgen Schmidhuber,et al.  Highway Networks , 2015, ArXiv.

[11]  Eduard H. Hovy,et al.  End-to-end Sequence Labeling via Bi-directional LSTM-CNNs-CRF , 2016, ACL.

[12]  Stefan Achleitner,et al.  SIPS: Solar Irradiance Prediction System , 2014, IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks.

[13]  Yugang Niu,et al.  Hourly day-ahead solar irradiance prediction using weather forecasts by LSTM , 2018 .

[14]  Lukasz Kaiser,et al.  Attention is All you Need , 2017, NIPS.

[15]  Gregory D. Hager,et al.  Temporal Convolutional Networks: A Unified Approach to Action Segmentation , 2016, ECCV Workshops.

[16]  Wei Xu,et al.  Bidirectional LSTM-CRF Models for Sequence Tagging , 2015, ArXiv.

[17]  Yoshua Bengio,et al.  Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling , 2014, ArXiv.

[18]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[19]  Stefan Lessmann,et al.  A comparative study of LSTM neural networks in forecasting day-ahead global horizontal irradiance with satellite data , 2018 .

[20]  Tek Tjing Lie,et al.  Hourly global solar irradiation forecasting for New Zealand , 2015 .

[21]  Razvan Pascanu,et al.  On the difficulty of training recurrent neural networks , 2012, ICML.

[22]  Jeffrey L. Elman,et al.  Finding Structure in Time , 1990, Cogn. Sci..