Physics of auroral phenomena : proceedings of the 37th Annual seminar, Apatity, 25 - 28 February, 2014 / [ed. board: A. G. Yahnin, N. V. Semenova]. - Апатиты : Изд-во Кольского научного центра РАН, 2014. - 125 с. : ил., табл.

Identification o fplasma streams in the solar wind by neural network classification approach Figure 3. Architecture of Kohonen classification artificial neural network ( n Input Concurent layer layer Training neural networks on total histograms obtained for the five events was performed. The sixth event always be reserved and used for testing the trained network. With classification self-learning neural network work consistently as input numerical series (additional histograms for training events №1-5 for each type of flow) is receives. In this information in a concise form information about the coherence of oscillatory processes in the solar wind parameters in a particular frequency range is contains. Then, to verify the trained neural network histogram test events №6 is used. Test results of neural networks, trained for two ranges of periods in the summary table 1 is presented. The numbers in the cells the number o f successes classification when you walk through the test events was indicate. Table 1 Specific period interval 2-30 min MC HSS Shocks CIR 3 (50%) 3 (50%) 4 (66%) 5 (83%) Specific period interval 30-61 min MC HSS Shocks CIR 6(100%) 4 (66%) 2 (33%) 3 (50%) Conclusion Performed by the neural network classification of spectral features of flows differentiated by frequency bands separation of skeleton images of disturbances on all analyzed parameters (N, V, T, P, | In |, Bx, By, Bz) showed. This confirms the existence o f the internal connections between the components of the wave dynamics in different plasma flows. It is shown that in the period range 2-30 min neural network performs reliable (in more than 50% of cases) the identification of the type of flows CIR (83%) and Shocks (66%). In the period range of 31-60 minutes - a reliable identification of the flow-type MC (100%) and HSS (66%). The proposed method in the online mode, monitoring of near space for the detection of geoeffective structures in the solar wind flow and forecasting global geomagnetic disturbances can be applied. Further development of methods of classification due to the improvement summation technique histograms in which each type of flow the most consistent set of parameters will be used. Acknow ledgements. RFBR grant 12-05-00425 and the program of the Ministry of Education and Science «Development of Scientific Potential of Higher Education, 2014-2016» supported this work. References Barkhatov N.A. Revunov S.E., Shadrukov D.V. Qualifying solar plasma flows by analysis and is within magnetospheric low- frequency oscillations during magnetic storms // Privolzhskiy scientific journal, NNGASU, № 1, p. 106-112, 2013 Daubechies I. Ten lectures on wavelets. -'Izhevsk: NITs «Regular and Chaotic Dynamics», 2001, 464 p. Revunov S E Shadrukov D.V., Serebryakova R.I. Analysis of spatio-temporal dynamics of low-frequency (2-8 mHz) magnetic disturbances during magnetic storms // Messenger of UNN N.I. Lobachevsky, 2013, № 5 (1), 83-91 Steed К С J Owen, P. Demoulin, and S. Dasso, Investigating the observational sig-natures of magnetic cloud substructure // J. Geophys. Res. -2011. -V .l 16, A01106, doi: 10.1029/2010JA015940. T e « e in J A С W Smith В J Vasquez, and R M. Skoug, Turbulence associated with corotating interaction regions at 1 AU: toertial and dissipation range magnetic field spectra// J. Geophys. Res. -2011. -V .l 16, A10104. doi-.10.1029/2011JA016647 A an d W M Macek, Observation of the multifractal spectrum in solar wind turbulence by Ulysses at high la“ //’j. Geophys. Res'. -2010. -V .l 15, A07104. doi:10.1029/2009JA0151763. 77