Industrial Automatic Control Systems and Controllers Annotation << Back Development of A Methodology for Identification and Modeling of Dynamic Systems Based on Hybrid Neural Networks Temerbekova B.M., Mamanazarov U.B., Bekimbetov B.M. The article discusses the development of a methodology for the identification and modeling of dynamic systems based on neural networks in the context of polymetallic metallurgical production. The relevance of using modern approaches to account for the nonlinearity and non-stationarity of technological processes – including crushing, flotation, smelting, and refi ning – is described. A mathematical model based on a system of differential equations is presented, and a hybrid approach to their identification using recurrent architectures (LSTM/GRU) is proposed. The results demonstrate improved accuracy in forecasting dynamic parameters and determining response coefficients. Validation on real data shows that the proposed method reduces forecasting error to 3–4 %, making the model suitable for real-time process control. Keywords: structural-parametric synthesis, ore, crushing, flotation, technological parameters, neural networks, parameter identification, optimization, dynamic variable prediction, metallurgical production, neural network models. DOI: 10.25791/asu.8.2025.1605 Pp. 39-44.