Improved day-ahead predictions of load and renewable generation by optimally exploiting multi-scale dependencies

The increased contribution of renewable generation is substantially impacting the operation of power systems. In this context, it is essential to better characterize uncertainties by improving predictions so as to prevent the need to rely on costly oversized regulation reserves for alleviating system imbalances. This paper aims at developing a generic tool dedicated to the day-ahead forecasting of main sources of uncertainties in power grids, namely load as well as wind and photovoltaic generation. The objective is to overcome issues faced by current forecasting tools by using a recurrent neural network based on Gated- Feedback Long Short-Term Memory, an advanced architecture designed to process complicated time series with multi-scale characteristics. The results demonstrate the benefits of this method on forecasting complex and highly volatile variables. However, the architectural complexity of the neural network is more likely to lead to overfitting for variables with a strong deterministic component such as the load.

[1]  Ronald J. Williams,et al.  Gradient-based learning algorithms for recurrent networks and their computational complexity , 1995 .

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

[3]  Bart De Moor,et al.  Easy Hyperparameter Search Using Optunity , 2014, ArXiv.

[4]  Herbert Jaeger,et al.  Discovering multiscale dynamical features with hierarchical Echo State Networks , 2008 .

[5]  Jürgen Schmidhuber,et al.  Framewise phoneme classification with bidirectional LSTM and other neural network architectures , 2005, Neural Networks.

[6]  K. Ravikumar,et al.  On the use of wavelets packet decomposition for time series prediction , 2014 .

[7]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[8]  Jürgen Schmidhuber,et al.  Learning Precise Timing with LSTM Recurrent Networks , 2003, J. Mach. Learn. Res..

[9]  Geoffrey E. Hinton,et al.  Speech recognition with deep recurrent neural networks , 2013, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing.

[10]  Yoshua Bengio,et al.  Gated Feedback Recurrent Neural Networks , 2015, ICML.

[11]  Yoshua Bengio,et al.  Gradient Flow in Recurrent Nets: the Difficulty of Learning Long-Term Dependencies , 2001 .

[12]  Yu Peng,et al.  A novel hybridization of echo state networks and multiplicative seasonal ARIMA model for mobile communication traffic series forecasting , 2012, Neural Computing and Applications.

[13]  Reza Pulungan,et al.  Prediction by a Hybrid of Wavelet Transform and Long-Short-Term-Memory Neural Network , 2017 .

[14]  George Kurian,et al.  Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation , 2016, ArXiv.

[15]  Tek Tjing Lie,et al.  Wind Speed Forecasting Using Hybrid Wavelet Transform- ARMA Techniques , 2015 .

[16]  F. Vallée,et al.  Medium-Term Multimarket Optimization for Virtual Power Plants: A Stochastic-Based Decision Environment , 2018, IEEE Transactions on Power Systems.