AN ANALYSIS OF INFORMATIONAL CONTENT OF DATA WHEN CREATING A PREDICTIVE MODEL OF ELECTRICITY PRICE 

Petukhova Natal'ya Yur'evna, Candidate of physical and mathematical sciences, senior staff scientist, the laboratory of numerical modeling, «Bazovye technologii» LLC (2 3-ya Yamskogo Polya street, Moscow, Russia), E-mail: natpet967@gmail.com 

004.75, 004.047 

10.21685/2072-3059-2019-3-2 

Background. The problem of creation of an effective method of forecasting the indicators of the equilibrium price of the electric power in the market for the day ahead is considered. The work purpose is to give a preliminary analysis of predictive data to identify the most important predictive factors.
Materials and methods. The models with application of neural networks and gradient boosting over a composition of algorithms are the basis of forecast creation. A mathematical analysis of the input data on the basis of their informational content assessment and the method of allocation of main components in correlation dependences are carried out.
Results. To select the optimal set of predictive factors, the correlation and cluster analysis was carried out. The data dimensionality reduction was carried out by the method of principle components. The informational content of indicators was defined by selective assessment of their mutual entropy.
Conclusions. The considered methods of analysis of data allow to range the indicators by their informational content, to carry out the cluster analysis, in due time to add new factors to the forecast. All of these lead to the most optimal predictive model. 

energy market prediction, artificial neural networks, principle components method, informative forecasting, information criterion, Lipschitz constant 

1. Khaykin S. Neyronnye seti [Neural networks]. Moscow: Vil'yams, 2006, part 1, 2, 654 p. [In Russian]
2. Friedman H. Annals of Statistics & Data Analysis. 2002, vol. 38, no. 4, pp. 1189–1232.
3. Bidyuk P. I., Zvorygina T. F. Upravlyayushchie sistemy i mashiny [Control systems and machines]. 2003, no. 2, pp. 93–99. [In Russian]
4. Zagoruyko N. G. Prikladnye metody analiza dannykh i znaniy [Applied methods of data and knowledge analysis]. Novosibirsk: IM SO RAN, 1999, 271 p. [In Russian]
5. Pupkov K. A., Egupov N. D. Nestatsionarnye sistemy avtomaticheskogo upravleniya: analiz, sintez i optimizatsiya [Nonstationary systems of automatic control: analysis, synthesis and optimization]. Moscow: Izd-vo MGTU im. Baumana, 2007, 340 p. [In Russian]
6. Komartseva L. G., Maksimov A. V. Neyrokomp'yutery [Neurocomputer]. Moscow: Izd-vo MGTU im. Baumana, 2004, 410 p. [In Russian]
7. Available at: http://www.atsenergo.ru/
8. Osovskiy S. Neyronnye seti dlya obrabotki informatsii [Neural networks for information processing]. Moscow: Finansy i statistika, 2002, 485 p. [In Russian]
9. Bakhvalov N. S., Zhidkov N. P., Kobel'kov G. M. Chislennye metody [Numerical methods]. 7th ed. Moscow: BINOM. Laboratoriya znaniy, 2011, 630 p. [In Russian]