國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
5. Conclusion and outlook
An Ensemble Empirical Mode Decomposition based Back-propagation Neural Network learning paradigm has been presented for electricity load forecasting and gold price forecasting. In this research, the moving window method is applied to the prediction process. Moreover, cubic spline interpolation and extrapolation were used as our strategy for electricity load forecasting. The measures RMSE, MAE, standard deviation of error were used to judge the performance. By using the meaningful IMFs as training data, we can predict the electricity load for the next hour and the gold prices for next day, all with accurate results. However, for further improvement of the results, the ensemble average method was employed. The outcome shows that the performance was better when using the ensemble average method than if it were not in use.
For future research, there are several feasible improvements discussed as follows:
First, since we used MATLAB which only can be used on normal PC to process our ANN program, in the future we can try to write an ANN program in FORTRAN which can be used on computer cluster to promote the computing speed. Second, try other combinations between the length of moving window, the numbers of interpolation and extrapolation and different parameters of network. Third, and most important, try different algorithms or architectures of neural network. For the gold price prediction, we can design some trading strategies based on our forecasting results in the future, and try to make a profit from trading gold. En Tzu Li (2011) has used this algorithm to forecast TAIEX options, and designed a moving FK indicator for algorithm trading and resulting in efficient performance. Therefore, in the future maybe we can take another financial product time-series for forecasting and with good trading strategy; we can earn lot of money.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
57
APPENDIX
In this appendix we collated the predicted results of GCB10 from May 12 to June 6, 2008 except for weekends. We selected points, error percentage are over 5%, and sketched the statistical results in Figure A.1. The 20 days‘ mean error and number of error-percentage points over 5% are quantized in Table A.1. We found that there are 15 points over 5% of the total 20 points from 7:00 to 8:00. Since the air-condition units are often turned on during this time, the load value jumps suddenly and the performance of prediction are usually not satisfactory. The performance of peak hour is better than off-peak hour, but in Table A.1 we can find that the mean errors between these two time periods are not too different. Due to the baseline of off-peak hour is about 400-600KWH while peak hour is about 800-1000 KWH.
Figure A.1 Statistics of points over 5% of GCB10 from
May 12 to June 6, 2008 except for weekends 02 4 6 8 10 12 14 16
0~1 1~2 2~3 3~4 4~5 5~6 6~7 7~8 8~9 9~10 10~11 11~12 12~13 13~14 14~15 15~16 16~17 17~18 18~19 19~20 20~21 21~22 22~23 23~24
個 數
Hour
‧ Table A.1 The mean error and collation of points over 5% of GCB10 from
May 12 to June 6, 2008 except for weekends
GCB10 (2008.5.12-6.6 , weekdays)
‧
Hybrid Wavelet Models for Short-Term Load Forecasting. IEEE transactions on power systems, VOL. 25, NO. 3, August 2010.Akgiray, V., G.G. Booth, J.J. Hatem, and C. Mustafa, 1991. Conditional Dependence in Precious Metal Prices. The Financial Review, 26, 367-386.
Chen, M.-C., Wei, Y, 2010. Exploring time variants for short-term passenger flow. J.
Transp. Geogr. doi:10.1016/j.jtrangeo.2010.04.003
Cummings, D.A.T., Irizarry, R.A., Huang, N.E., Endy, T.P., Nisalak, A., Ungchusak, K., Burke, D.S., 2004. Travelling waves in the occurrence of dengue haemorrhagic fever in Thailand. Nature 427 (6972), 344–347.
En Tzu Li, 2011. TAIEX Option Trading by using EEMD-based Neural Network Learning Paradigm. Master Thesis of Graduate Institute of Applied Physics, College of Science NCCU.
FENG Ping, DING Zhi-hong, HAN Rui-guang, ZHANG Jian-wei. 2009.
Precipitation-runo forecasting ANN model based on EMD. Systems Engineering-Theory & Practice, Vol.29, No.1, Jan., 2009.
G.A. Adepoju, M.Sc., S.O.A. Ogunjuyigbe, M.Sc., and K.O. Alawode, B.Tech.
Application of Neural Network to Load Forecasting in Nigerian Electrical Power System. The Pacific Journal of Science and Technology Volume 8. Number 1. May 2007 (Spring).
Hwang, P.A., Huang, N.E., Wang, D.W., 2003. A note on analyzing nonlinear and non-stationary ocean wave data. Applied Ocean Research 25 (4), 187–193.
Huang, N.E., Shen, Z., Long, S.R., Wu, M.C., Shih, H.H., Zheng, Q., Yen, N.-C., Tung, C.C., Liu, H.H., 1998. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London A 454 (1971), 903–995.
Huang, N.E., Wu, M.L., Qu, W.D., Long, S.R., Shen, S.S.P., 2003b. Applications of Hilbert–Huang transform to nonstationary financial time series analysis. Applied Stochastic Models in Business and Industry 19, 245–268.
‧
neural very-short-term load forecasting based on chaotic dynamics reconstruction.Chaos, Solitons & Fractals Volume 29, Issue 2, July 2006, Pages 462-469
Hamid S. A. and Iqbal Z., Using neural networks for forecasting volatility of S&P 500 Index futures prices, Journal of Business Research, 2004, 57: 1116-112
James W. Taylor. An evaluation of methods for very short-term load forecasting using minute-by-minute British data, 2008. International Journal of Forecasting, 24 (4). pp. 645-658. ISSN 0169-2070
K.Hornik, M.Stinchocombe, 1989. H.White, Multilayer feedforward networks are universal approximators,NeuralNetworks2 (1989) 359–366.
Li, Q.S., Wu, J.R., 2007. Time–frequency analysis of typhoon effects on a 79-storey tall building. Journal of Wind Engineering and Industrial Aerodynamics 95 (12), 1648–1666.
Liang, H., Lin, Q.-H., Chen, J.D.Z., 2005. Application of the empirical mode decomposition to the analysis of esophageal manometric data in gastro esophageal reflux disease. IEEE Transactions on Biomedical Engineering 52 (10), 1692–1701.
Lean Yu, Shouyang Wang, Kin Keung Lai., 2008. Forecasting crude oil price with an EMD-based neural network ensemble learning paradigm. Energy Economics 30 (2008) 2623–2635.
Lean Yu, ShouyangWanga, KinKeungLai, FenghuaWenc, 2010. A multiscale neural network learning paradigm for financial crisis forecasting. Neuro computing, 73:716-725
Liu, K. Subbarayan, S. Shoults, R.R. Manry, M.T. Kwan, C. Lewis, F.I. Naccarino, J.
Autom. & Robotics Res. Inst., Texas Univ., Arlington, TX, 1996. Comparison of very short-term load forecasting techniques. IEEE Transactions on Power Systems. Vol. 11.
No. 2. May 1996
Mirmirani, S. and H.C. L, 2004. Gold Price, Neural Networks and Genetic Algorithm, Computational Economics, 23, 193-200.
‧
Mohsen Hayati and Yazdan Shirvany, 2007. Artificial Neural Network Approach for Short Term Load Forecasting for Illam Region. World Academy of Science, Engineering and Technology 28 2007
N.X. Jia, R. Yokoyamaa, Y.C. Zhoub, Z.Y. Gaoc, 2001. A flexible long-term load forecasting approach based on new dynamic simulation theory — GSIM.
International Journal of Electrical Power & Energy Systems Volume 23, Issue 7, October 2001, Pages 549-556
Nahi Kandil, Rene´ Wamkeue, Maarouf Saad, Semaan Georges, 2006. An efficient approach for short term load forecasting using artificial neural networks. International Journal of Electrical Power & Energy Systems Volume 28, Issue 8, October 2006, Pages 525-530.
Ray Ruichong Zhang, M.ASCE; Shuo Ma; Erdal Safak, M.ASCE; and Stephen Hartzell., 2003. Hilbert-Huang Transform Analysis of Dynamic and earthquake motion recordings. Journal of Engineering Mechanics, Vol. 129, No. 8, pp. 861-875.
Ray Ruichong Zhang, Shuo Ma, and Stephen Hartzell., 2003. Signatures of the Seismic Source in EMD-Based Characterization of the 1994 Northridge, California, Earthquake Recordings. Bulletin of the Seismological Society of America; February 2003; v. 93; no. 1; p. 501-518.
Ruqiang Yan, Student Member, IEEE, and Robert X. Gao, Senior Member,IEEE., 2006. Hilbert–Huang Transform-Based Vibration Signal Analysis for Machine Health Monitoring. IEEE Transactions on instrumentation and measurement, Vol. 55, No. 6.
Ruey-Hsun Liang, Ching-Chi Cheng, 2002. Short-term load forecasting by a neuro-fuzzy based approach. International Journal of Electrical Power & Energy Systems Volume 24, Issue 2, February 2002, Pages 103-111
Shahriar Shafiee and ErkanTopal, 2010. An overview of global gold market and gold price forecasting. Resources Policy Volume 35, Issue 3, September 2010, Pages 178-189.
Stephen A. Baker and Roger C. van Tassel, 1985. Forecasting the price of gold: A fundamentalist approach Atlantic Economic Journal Volume 13, No. 4, 43-51
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
62
Wu, Z., and N. E Huang, 2009. Ensemble Empirical Mode Decomposition: a noise-assisted data analysis method. Advances in Adaptive Data Analysis. Vol.1, No.1. 1-41.
Wei SUN, 2010. Research on GA-SVM Model for Short Term Load Forecasting Based on LDM-PCA Technique. Journal of Computational Information Systems 6:10 (2010) 3183-3189.
Yen-Rue Chang, 2011. Non-stationary time series analysis by using Hilbert-Huang transform: electricity consumption and gold price volatility. Master Thesis of Graduate Institute of Applied Physics, College of Science NCCU.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
63