Abstract
We focus on deep learning algorithms, improving upon the weather research and forecasting (WRF) model, and we show that the combination of these methods produces day-ahead wind speed predictions of high accuracy, with no need for previous-day measurements. We also show that previous-day data offer a significant enhancement in a short-term neural network for hour-ahead predictions, assuming that they are available on a daily basis. Our main contribution is the design and testing of original neural networks that capture both spatial and temporal characteristics of the wind, by combining convolutional (CNN) as well as recurrent (RNN) neural networks. The input predictions are obtained by a WRF model that we appropriately parameterize; we also specify a grid adapted to each park so as to capture its topography. Training uses historical data from five wind farms in Greece, and the 5-month testing period includes winter months, which exhibit the highest wind speed values. Our models improve WRF accuracy on average by 19.4%, and the improvement occurs in every month; expectedly, the improvement is lowest for the park where WRF performs best. Our neural network is competitive to state-of-the-art models, achieving an average MAE of 1.75 m/s. Accuracy improves for speed values up to 20 m/s, which are important in wind energy prediction. We also develop an RNN model and show that MAE reduces to less than 1 m/s for short-term predictions if actual data is employed.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The largest company in the sector, with presence in USA and Southeastern Europe.
References
Balluff, S., Bendfeld, J., Krauter, S.: Short term wind and energy prediction for offshore wind farms using neural networks. In: Intern. Conf. Renewable Energy Research & Applications, pp. 379–382 (2015)
Barhmi, S., Elfatni, O., Belhaj, I.: Forecasting of wind speed using multiple linear regression and artificial neural networks. Energy Syst. (2020). https://doi.org/10.1007/s12667-019-00338-y
Christoforou, E., Emiris, I.Z., Florakis, A.: Neural networks for cryptocurrency evaluation and price fluctuation forecasting. In: Pardalos, P., Kotsireas, I., Guo, Y., Knottenbelt, W. (eds.) Mathematical research for blockchain economy, pp. 133–149. Springer International Publishing, Cham (2020)
Council, G.W.E.: 51.3 GW of global wind capacity installed in 2018. https://gwec.net/51-3-gw-of-global-wind-capacity-installed-in-2018/ (2019). Retrieved 18 May 2020
Díaz, D., Torres, A., Dorronsoro, J.R.: Deep neural networks for wind energy prediction. In: Advances. Computational Intelligence, pp. 430–443 (2015)
Finamore, A.R., Calderaro, V., Galdi, V., Piccolo, A., Conio, G.: A wind speed forecasting model based on artificial neural network and meteorological data. In: Proc. IEEE 16th Intern. Conf. Environment and Electrical Engineering, pp. 1–5 (2016)
Ghaderi, A., Sanandaji, B.M., Ghaderi, F.: Deep forecast: deep learning-based spatio-temporal forecasting. arXiv:abs/1707.08110 (2017)
Godinho, M., Castro, R.: Comparative performance of AI methods for wind power forecast in Portugal. Wind Energy. https://onlinelibrary.wiley.com/doi/pdf/10.1002/we.2556 (2020)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9, 1735–80 (1997). https://doi.org/10.1162/neco.1997.9.8.1735
Huang, C.J., Kuo, P.H.: A short-term wind speed forecasting model by using artificial neural networks with stochastic optimization for renewable energy systems. Energies 11, 2777 (2018)
Huang, J., Lu, X., McElroy, M.: Meteorologically defined limits to reduction in the variability of outputs from a coupled wind farm system in the central us. Renew. Energy 62, 331–340 (2014)
Mauch, B., Apt, J., Carvalho, P., Small, M.: An effective method for modeling wind power forecast uncertainty. Energy Syst. (2013). https://doi.org/10.1007/s12667-013-0083-3
Men, Z., Yee, E., Lien, F.S., Wen, D., Chen, Y.: Short-term wind speed and power forecasting using an ensemble of mixture density neural networks. Renew. Energy 87, 203–211 (2016)
Moustris, K.P., Zafirakis, D., Alamo, D.H., Nebot Medina, R.J., Kaldellis, J.K.: 24-h ahead wind speed prediction for the optimum operation of hybrid power stations with the use of artificial neural networks. In: Perspectives on Atmospheric Sciences, pp. 409–414 (2017)
Ng, J., Hausknecht, M., Vijayanarasimhan, S., Vinyals, O., Monga, R., Toderici, G.: Beyond short snippets: deep networks for video classification. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4694–4702 (2015)
Papoutsakis, A., Zacharia, S., Kouvara, F., Karagiannis, D., Katsoulis, G., Daikos, A.: Wind model uncertainty analysis and validation over operating wind farms in complex terrain. Eur. Wind Energy Conf. Exhib. EWEC 1, 384–393 (2013)
dos Reis, M.M.L., Mazetto, B.M., da Silva, E.C.M.: Economic analysis for implantation of an offshore wind farm in the Brazilian coast. Sustain. Energy Technol. Assess. 43, 100955 (2021)
Salcedo-Sanz, S., Pérez-Bellido, Á.M., Ortíz-García, E., Portilla-Figueras, A., Prieto, L., Paredes, D.: Hybridizing the fifth generation mesoscale model with artificial neural networks for short-term wind speed prediction. Renew. Energy 34, 1451–1457 (2009)
Salfate, I., Marin, J.C., Cuevas, O., Montecinos, S.: Improving wind speed forecasts from the weather research and forecasting model at a wind farm in the semiarid Coquimbo region in central Chile. Wind Energy 23(10), 1939–1954 (2020)
Shi, X., Chen, Z., Wang, H., Yeung, D., Wong, W., Woo, W.: Convolutional lSTM network: a machine learning approach for precipitation nowcasting. In: NIPS (2015)
Shi, X., Gao, Z., Lausen, L., Wang, H., Yeung, D., Wong, W., Woo, W.: Deep learning for precipitation nowcasting: a benchmark and a new model. arXiv:abs/1706.03458 (2017)
Tzamos, C.: Ventusnet: deep learning for wind speed prediction. MSc Thesis, Dept Informatics & Telecoms, National and Kapodistrian University of Athens (Greece). https://pergamos.lib.uoa.gr/uoa/dl/object/2878206 (2019)
Wang, H., Li, G., Wang, G., Peng, J., Jiang, H., Liu, Y.: Deep learning based ensemble approach for probabilistic wind power forecasting. Appl. Energy 188, 56–70 (2017). http://www.sciencedirect.com/science/article/pii/S0306261916317421
Wang, J., Shanshan, Q., Zhou, Q., Jiang, H.: Medium-term wind speeds forecasting utilizing hybrid models for three different sites in Xinjiang, China. Renew. Energy 76, 91–101 (2014)
WindEurope.org: Wind energy in Europe in 2018. https://windeurope.org/about-wind/statistics/european/wind-energy-in-europe-in-2018 (2019). Retrieved 18 May 2020
Acknowledgements
Research partially funded by project “PeGASUS: Approximate geometric algorithms and clustering with applications in finance” (MIS 5047662) under call “Support for researchers with emphasis on young researchers: cycle B” (EDBM103), co-financed by Greece and the EU (European Social Fund) under the operational program Human Resources Development, Education & Lifelong Learning 2014-20. We thank Meteorologica P.C. for offering hardware infrastructure and guidance on meteorological aspects.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1: WRF information and grid construction
This section provides the meteorological variables predicted by WRF, the coordinates of the WRF rectangle per park, and pseudocode for computing these rectangles (see Tables 4, 5).
Park with three extremal (circle), and another nine (triangle) turbines. The rectangle, defined by four corners (star), includes 35 WRF points (cross). The axes show longitude and latitude
Let us present the construction of a rectangular grid per park, to include its turbines and an appropriate number of WRF points (Algorithm 1, Appendix): one procedure (findparkborders) specifies a minimum axis-aligned rectangle, namely parkBox, enclosing all park turbines. We define an outer rectangle, called WRFbox, to include parkBox, and parameterized by STEP, the margin in longitude and latitude around parkBox. The larger STEP is, the more comprehensive is the model but also the higher is the complexity; more importantly, if WRFbox is too large, it may include irrelevant geographical elements like the outer slope of a nearby hill. Another procedure (collectwrfgridpoints) collects the WRF points forming WRFbox. In the park shown in Fig. 9 there are three extremal out of 12 turbines, and 35 WRF grid points in WRFbox. Each of its edges is defined by moving away from the corresponding edge of parkBox by \(STEP = 0.04^{\circ }\). For the other parks, we experimentally set \(STEP = 0.04\) or \(STEP = 0.035\) hence including 2 or 3 rows and columns of WRF grid points beyond parkBox.
Appendix 2: Statistical results
This section provides further statistical results of our predictions (see Fig. 10; Tables 6, 7).
Forecasting (gray) of WRF (left) and DSTNN (right) against actual values (dark gray). DSTNN improves WRF in all parks: Krekeza (top-left), Louzes (top-right), Rachoula-2 (middle-left), Rachoula-3 (middle-right) and Saint Georgios (bottom)
Rights and permissions
About this article
Cite this article
Christoforou, E., Emiris, I.Z., Florakis, A. et al. Spatio-temporal deep learning for day-ahead wind speed forecasting relying on WRF predictions. Energy Syst 14, 473–493 (2023). https://doi.org/10.1007/s12667-021-00480-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12667-021-00480-6