Improved Echo State Network Based on Data-driven and Its Application to Prediction of Blast Furnace Gas Output

Based on the complex process of blast furnace gas (BFG) system in steel industry, a network forecasting method based on data-driven is established in this paper for the prediction problem on BFG output. Since the practical data include a diversity of noises, an empirical mode decomposition approach is employed to decompose the time series signal into a group of independent intrinsic mode functions, and the formed small-scale intrinsic mode functions are de- noised by low-pass fllter with an adaptive threshold. Then, the re-constructed signals are used to build the forecasting model, in which an improved echo state network is proposed and the network output weights are obtained by singular value decomposition. Therefore, the ill-conditioned problem of previous linear regression is overcome and the forecasting precision is increased. The prediction results using practical production data show the validity of the proposed method and also provide the scientiflc decision support for the gas resources scheduling.

[1]  Harald Haas,et al.  Harnessing Nonlinearity: Predicting Chaotic Systems and Saving Energy in Wireless Communication , 2004, Science.

[2]  Herbert Jaeger,et al.  Adaptive Nonlinear System Identification with Echo State Networks , 2002, NIPS.

[3]  Skander Soltani,et al.  On the use of the wavelet decomposition for time series prediction , 2002, ESANN.

[4]  Han Min,et al.  Ridge regression learning in ESN for chaotic time series prediction , 2007 .

[5]  Gabriel Rilling,et al.  Empirical mode decomposition as a filter bank , 2004, IEEE Signal Processing Letters.

[6]  D.P. Mandic,et al.  Multi-step forecasting using echo state networks , 2005, EUROCON 2005 - The International Conference on "Computer as a Tool".

[7]  N. Huang,et al.  A study of the characteristics of white noise using the empirical mode decomposition method , 2004, Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences.

[8]  D. Prokhorov,et al.  Echo state networks: appeal and challenges , 2005, Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005..

[9]  Gene H. Golub,et al.  Matrix computations , 1983 .

[10]  H. Jaeger,et al.  Reservoir riddles: suggestions for echo state network research , 2005, Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005..

[11]  N. Huang,et al.  The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis , 1998, Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences.