|
ИСТИНА |
Войти в систему Регистрация |
ИПМех РАН |
||
Artificial Neural Network Applications to the space radiation environment modelling and forecastingстатья
- Авторы: Dmitriev A.V., Minaeva Yu S., Orlov Yu V., Persiantsev I.G., Suvorova A.V., Veselovsky I.S.
- Сборник: Proc. of ESA Workshop on Space Weather
- Год издания: 1999
- Место издания: ESTEC, Noordwijk, The Netherlands
- Первая страница: 393
- Последняя страница: 397
- Аннотация: The paper presents a short review of various directions of Artificial Neural Network (ANN) applications to modelling of near Earth space radiation distribution and dynamics and to the development of some methods of space weather forecasting. ANN models may be based on well-known physical laws or on some empirical rules. We can present four main directions of ANN applications. The first one is the development of strong non-linear quasi-stationary models with large number of input nodes. The examples of results in the first direction are: 3D model of the Earth's magnetopause and mapping of the near Earth high energy particle (electrons and protons) distribution. The second direction is modelling of cumulative and time shifted effects in the time series. The main problem in these models is searching the most significant measured physical parameters as input nodes and determination of the most appropriate time intervals for averaging or shifting parameter values. This direction permits to develop dynamical models of physical processes in multi-parametric time series. One of the models developed in the second direction is the model of the slot region of Earth's radiation electron belt dynamics depending on the solar wind conditions. The third direction is modelling of self-consistent time series by means of recurrent ANNs. These models take into account the information about prehistory of the system dynamics and hence they may be used for forecasting. The models forecasting sunspot number and average solar wind conditions are excellent examples of investigations in the third direction. The fourth direction is combination of the described directions by means of global ANN on the base of some classification rules which may be used in future for the development of an expert system for the space weather forecasting.
- Добавил в систему: Суворова Алла Васильевна