Тип публикации: научное издание
Год издания: 2021
Идентификатор DOI: 10.1007/978-3-030-67892-0_9
Ключевые слова: control theory, cyber-physical systems, dynamic system, forecast, neural network, special economic zone
Аннотация: The issues of utility and the complexity of the use (opposition) of neural networks in management, the use of their prognostic abilities using the example of a nonlinear dynamic system with unknown environmental parameters as one of the tasks of the cyber-physical system are considered. The experiment uses a layered network structuПоказать полностьюre with a teacher, an “anti-rotation” learning optimization algorithm, neurons with sigmoid nonlinearity, a fully connected network for predicting the state of a dynamic system. The activity of a special economic zone under the influence of environmental parameters was taken as a cyber-physical system. The parameters of the neural network were revealed: for which the weighting coefficients of the neurons closed the calculation at the saturation point and thereby led to its paralysis. Also, the modes of stalling the calculation into a local minimum or maximum of any local function of the system were revealed, depending on the period and the importance of this parameter in this period. As a result of the experiment, the cyber-physical system obtained a prediction of a set of output values from a set of input values based on the experience obtained by the neural network while minimizing the forecast discrepancy. A set of competencies is proposed to improve the accuracy of the neural network forecast and control the cyber-physical system (special economic zone). © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Журнал: Studies in Systems, Decision and Control
Выпуск журнала: Vol. 350
Номера страниц: 97-106
ISSN журнала: 21984182
Издатель: Springer Science and Business Media Deutschland GmbH