\r\nhigh volatility, nonlinearity and high frequency that make forecasting

\r\nquite difficult. Electricity price has a volatile and non-random

\r\ncharacter so that, it is possible to identify the patterns based on the

\r\nhistorical data. Intelligent decision-making requires accurate price

\r\nforecasting for market traders, retailers, and generation companies.

\r\nSo far, many shallow-ANN (artificial neural networks) models have

\r\nbeen published in the literature and showed adequate forecasting

\r\nresults. During the last years, neural networks with many hidden

\r\nlayers, which are referred to as DNN (deep neural networks) have

\r\nbeen using in the machine learning community. The goal of this

\r\nstudy is to investigate electricity price forecasting performance of the

\r\nshallow-ANN and DNN models for the Turkish day-ahead electricity

\r\nmarket. The forecasting accuracy of the models has been evaluated

\r\nwith publicly available data from the Turkish day-ahead electricity

\r\nmarket. Both shallow-ANN and DNN approach would give successful

\r\nresult in forecasting problems. Historical load, price and weather

\r\ntemperature data are used as the input variables for the models.

\r\nThe data set includes power consumption measurements gathered

\r\nbetween January 2016 and December 2017 with one-hour resolution.

\r\nIn this regard, forecasting studies have been carried out comparatively

\r\nwith shallow-ANN and DNN models for Turkish electricity markets

\r\nin the related time period. The main contribution of this study

\r\nis the investigation of different shallow-ANN and DNN models

\r\nin the field of electricity price forecast. All models are compared

\r\nregarding their MAE (Mean Absolute Error) and MSE (Mean Square)

\r\nresults. DNN models give better forecasting performance compare to

\r\nshallow-ANN. Best five MAE results for DNN models are 0.346,

\r\n0.372, 0.392, 0,402 and 0.409.","references":"[1] H.Y. Yamin, S.M. Shahidehpour, and Z. Li. Adaptive short-term\r\nelectricity price forecasting using artificial neural networks in the\r\nrestructured power markets. International Journal of Electrical Power\r\n& Energy Systems, 26(8):571 \u2013 581, 2004.\r\n[2] David Young, Stephen Poletti, and Oliver Browne. Can agent-based\r\nmodels forecast spot prices in electricity markets? evidence from the\r\nnew zealand electricity market. Energy Economics, 45:419 \u2013 434, 2014.\r\n[3] Fabio Genoese and Massimo Genoese. Assessing the value of storage\r\nin a future energy system with a high share of renewable electricity\r\ngeneration; an agent-based simulation approach with integrated\r\noptimization methods. Energy Systems, 5(1):19, 2014.\r\n[4] T Kristiansen. A time series spot price forecast model for the nord pool\r\nmarket. International Journal of Electrical Power & Energy Systems,\r\n61:20 \u2013 26, 2014.\r\n[5] Rafal Weron and Adam Misiorek. Forecasting spot electricity prices:\r\nA comparison of parametric and semiparametric time series models.\r\nInternational Journal of Forecasting, 24(Energy Forecasting):744 \u2013 763,\r\n2008.\r\n[6] Dogan Keles, Jonathan Scelle, Florentina Paraschiv, and Wolf Fichtner.\r\nExtended forecast methods for day-ahead electricity spot prices applying\r\nartificial neural networks. Applied Energy, 162:218 \u2013 230, 2016.\r\n[7] P Mandal, AK Srivastava, T Senjyu, and M Negnevitsky. A new\r\nrecursive neural network algorithm to forecast electricity price for pjm\r\nday-ahead market. International Journal of Energy Research, 34(6):507\r\n\u2013 522, 2009.\r\n[8] G.P. Girish. Spot electricity price forecasting in indian electricity market\r\nusing autoregressive-garch models. Energy Strategy Reviews, 2016.\r\n[9] Heping Liu and Jing Shi. Applying arma\u2013garch approaches to\r\nforecasting short-term electricity prices. Energy Economics, 37:152 \u2013\r\n166, 2013.\r\n[10] AJ Conejo, MA Plazas, R Espinola, and AB Molina. Day-ahead\r\nelectricity price forecasting using the wavelet transform and arima\r\nmodels. IEEE Transactions on Power Systems, 20(2):1035 \u2013 1042, 2005.\r\n[11] Hang T. Nguyen and Ian T. Nabney. Short-term electricity demand\r\nand gas price forecasts using wavelet transforms and adaptive models.\r\nEnergy, pages 3674 \u2013 3685, 2010.\r\n[12] S. Voronin and J. Partanen. Forecasting electricity price and demand\r\nusing a hybrid approach based on wavelet transform, arima and neural\r\nnetworks. International Journal of Energy Research, 38(5):626\u2013637,\r\n2014.\r\n[13] M. Shafie-khah, M. Parsa Moghaddam, and M.K. Sheikh-El-Eslami.\r\nPrice forecasting of day-ahead electricity markets using a hybrid forecast\r\nmethod. Energy Conversion and Management, 52:2165 \u2013 2169, 2011.\r\n[14] Suyi Li Mohammed Shahidehpour, Hatim Yamin. Market Operations\r\nin Electric Power Systems. John Wiley & Sons Ltd, 2002.\r\n[15] Kumar A Aggarwal S. K, Saini L. M. Electricity price forecasting in\r\nderegulated markets: A review and evaluation. International Journal of\r\nElectrical Power And Energy Systems, 2009.\r\n[16] Fazil Gokgoz Fahrettin Filiz. Electricity price forecasting in turkey with\r\nartificial neural network models. Investment Management and Financial\r\nInnovations, 2016.\r\n[17] Hu M. Y. Guoqiang Z, Patuwo. Forecasting with artificial neural\r\nnetworks: The state of the art. International Journal of Forecasting,\r\n1998.\r\n[18] Yoshua Bengio et al. Learning deep architectures for ai. Foundations\r\nand trends R \u0002 in Machine Learning, 2(1):1\u2013127, 2009.\r\n[19] Olivier Delalleau and Yoshua Bengio. Shallow vs. deep sum-product\r\nnetworks. In J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. Pereira, and\r\nK. Q. Weinberger, editors, Advances in Neural Information Processing\r\nSystems 24, pages 666\u2013674. Curran Associates, Inc., 2011.\r\n[20] Ping-Huan Kuo and Chiou-Jye Huang. An electricity price forecasting\r\nmodel by hybrid structured deep neural networks. Sustainability,\r\n10(4):1280, 2018.\r\n[21] Tomaso Poggio, Hrushikesh Mhaskar, Lorenzo Rosasco, Brando\r\nMiranda, and Qianli Liao. Why and when can deep-but not\r\nshallow-networks avoid the curse of dimensionality: A review.\r\nInternational Journal of Automation and Computing, 14(5):503\u2013519,\r\nOct 2017.\r\n[22] K. Greff, R. K. Srivastava, J. Koutn\u00edk, B. R. Steunebrink, and\r\nJ. Schmidhuber. Lstm: A search space odyssey. IEEE Transactions on\r\nNeural Networks and Learning Systems, 28(10):2222\u20132232, Oct 2017. [23] S. Hosein and P. Hosein. Load forecasting using deep neural networks.\r\nIn 2017 IEEE Power Energy Society Innovative Smart Grid Technologies\r\nConference (ISGT), pages 1\u20135, April 2017.\r\n[24] Hrushikesh Mhaskar, Qianli Liao, and Tomaso A. Poggio. Learning\r\nreal and boolean functions: When is deep better than shallow. CoRR,\r\nabs\/1603.00988, 2016.","publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 138, 2018"}