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.

“Physics o fAuroral Phenomena", Proc. XXXVII Annual Seminar, Apatity, pp. 75-77, 2014 © Kola Science Centre, Russian Academy of Science, 2014 Polar Geophysical Institute IDENTIFICATION OF PLASMA STREAMS IN THE SOLAR WIND BY NEURAL NETWORK CLASSIFICATION APPROACH N.A. Barkhatov, S.E. Revunov, A.B. Vinogradov (Nizhniy Novgorod State Pedagogical University, N izhniy Novgorod, Russia) The paper presents a new method for the identification of plasma streams in the solar wind by self-learning neural classification network on their spectral features in magnetic hydrodynamic range. To do this, the wavelet skeleton spectrums of solar wind parameters of the interplanetary magnetic field recorded in Earth orbit by patrol spacecraft is calculate. Algorithm for classification of wavelet skeleton spectral images for plasma streams in the solar wind on the Earth's orbit, based on neural network processing of compressed data on the main magnetic and dynamic parameters of the flow is propose. A neural network like Kohonen layer differentiated by frequency ranges performs the classification. Specifically, analyzes and establishes the spectral features of solar plasma flows in the form of magnetic clouds (MC), corotating interaction regions (CIR), shock waves (Shocks) and high-speed streams from coronal holes (HSS). 1. Introduction One of the problem of modem geliogeophysics is the problem of identification of geoeffective structures in near- Earth space, in the solar wind flow is presented. In turn, this identification of flows is not possible without their preliminary classification. Universal algorithm for determining the signs of both transient (like shock waves) and long (like magnetic clouds) processes is requires. In this paper an algorithm based on the compression of spectral information about the basic parameters of the MHD plasma flows and their representation in the form of wavelet skeleton spectral patterns we propose. Have repeatedly noted that the spectral composition of disturbances with periods ranging from 10 to 90 minutes, associated with the solar plasma flows, driven by their type [Wawrzaszek and Macek, 2010; Tessein et al., 2011; Steed et al., 2011 and references therein ]. However, most researches in this area on the study of the nature of the turbulence of the solar streams without regard to their membership of a certain type are focused. Thus, it is a persistent task differentiate types of flows of characteristic features by correlation-spectral method obtained. In the present study on the refinement of the plasma flow of the solar wind from the characteristic spectral features of wavelet variations of velocity, density, temperature, magnetic field we are focuses. We assume that each type of geoeffective flow not just a specific set of disturbed parameters (velocity, density, pressure, temperature, size of the field), and the characteristic fundamental links between them [. Barhatov et al., 2013] is characterized. These links in the synchronization of associated with perturbed parameters of the wave packets is performed. The proposed approach allows us to quantify the level of synchronization between the disturbed parameters of plasma formation, presented in the form of compressed wavelet images (skeletons) [Revunov et al, 2013]. According to the results of this assessment with classification neural network concludes supplies flow to a particular type is proposed. Classification feature in this case is the degree of coherence of oscillatory processes in the parameters of the solar wind flow in a specific frequency range is performed. 2. Correlation-spectral data processing method Works o f proposed algorithm for identification of plasma flows in the solar wind by classification neural network approach on 24 of these types o f events for plasma flows on ACE and Wind spacecraft during the period from 2000 to 2007 by catalogues NASA ( http://cdaw.gsfc.nasa.gov ) and NOAA ( http://ngdc.noaa.gov ) was performed. Within each event wavelet spectrums of one minute data for density N, velocity V, temperature T, pressure P, module and component of the interplanetary magnetic field | В |, Bx, By, Bz was received. Each type of the analyzed flow of six events is presented: MC (28.07.2000, 29.12.2000, 12.04.2001, 28.05.2001, 09.08.2001, 17.04.2002), CIR (27.07.2003, 05.04.2005, 07.05.2007, 20.09.2007, 27.09.2007, 25.10.2007), Shocks (19.12.2002, 27.02.2003, 14.07.2003,’ 17.07.2003, 12.04.2004, 22.07.2004), HSS (01.03.2000, 26.07.2003, 20.11.2004, 04.07.2006, 29.07.2007, 17.12.2007). The use of wavelet analysis allows spectrums at a specific frequency range is obtain. This usually gives the opportunity to limit the scope o f the search features of the original signal [Daubechies, 2001]. In the present study, as a basic wavelet in all numerical experiments Daubechies of the fourth order is selected. Large-scale wavelet transform coefficients in two ranges MHD periods: 2-30 min (8.3-0.6 mHz) and 31-60 min (0.6-0.3 mHz) were considered. Further specification o f information (postprocessing) by calculating the wavelet skeleton spectrums, which reflecting the internal dynamics of the processes of different types and sizes can be achieved. An example of the results of the postprocessing of the wavelet transform to obtain skeletons that characterize the process of changing mode of pulsation at a specific frequency at a specific time (calculate window with constant time interval and growing with a period) in Fig. 1 is shown. 75