Deep Learning-Based Approach for Time Series Forecasting with Application to Electricity Load

This paper presents a novel method to predict times series using deep learning. In particular, the method can be used for arbitrary time horizons, dividing each predicted sample into a single problem. This fact allows easy parallelization and adaptation to the big data context. Deep learning implementation in H2O library is used for each subproblem. However, H2O does not permit multi-step regression, therefore the solution proposed consists in splitting into h forecasting subproblems, being h the number of samples to be predicted, and, each of one has been separately studied, getting the best prediction model for each subproblem. Additionally, Apache Spark is used to load in memory large datasets and speed up the execution time. This methodology has been tested on a real-world dataset composed of electricity consumption in Spain, with a ten minute frequency sampling rate, from 2007 to 2016. Reported results exhibit errors less than 2%.

[1]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[2]  Alicia Troncoso Lora,et al.  A Nearest Neighbours-Based Algorithm for Big Time Series Data Forecasting , 2016, HAIS.

[3]  Alicia Troncoso Lora,et al.  Finding Electric Energy Consumption Patterns in Big Time Series Data , 2016, DCAI.

[4]  Francisco Martínez-Álvarez,et al.  A Survey on Data Mining Techniques Applied to Electricity-Related Time Series Forecasting , 2015 .

[5]  Gang Wang,et al.  Deep Learning-Based Classification of Hyperspectral Data , 2014, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[6]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[7]  Yue Zhang,et al.  Deep Learning for Event-Driven Stock Prediction , 2015, IJCAI.

[8]  Miguel J. Prieto,et al.  Menéndez, R.P.; Martínez, J.A.; Prieto, M.J.; Barcia, L.A.; Sánchez, J.M.M. A Novel Modeling of Molten-Salt Heat Storage Systems in Thermal Solar Power Plants. Energies 2014, 7, 6721-6740 , 2015 .

[9]  Xiang Li,et al.  Deep learning architecture for air quality predictions , 2016, Environmental Science and Pollution Research.

[10]  Keemin Sohn,et al.  Deep-Learning Architectures to Forecast Bus Ridership at the Stop and Stop-To-Stop Levels for Dense and Crowded Bus Networks , 2016, Appl. Artif. Intell..

[11]  Miriam A. M. Capretz,et al.  Energy Forecasting for Event Venues: Big Data and Prediction Accuracy , 2016 .

[12]  Igor V. Tetko,et al.  Data modelling with neural networks: Advantages and limitations , 1997, J. Comput. Aided Mol. Des..

[13]  F. Martínez-Álvarez,et al.  A survey on data mining techniques applied to energy time series forecasting , 2016 .

[14]  Geoffrey E. Hinton,et al.  On the importance of initialization and momentum in deep learning , 2013, ICML.

[15]  Scott Shenker,et al.  Spark: Cluster Computing with Working Sets , 2010, HotCloud.

[16]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[17]  Ugur Halici,et al.  A novel deep learning approach for classification of EEG motor imagery signals , 2017, Journal of neural engineering.