The article presents the main features of modeling complex distributed processes, reflects the relevance of the research and the importance of modeling such processes. The development of the neighborhood approach and the works of domestic and foreign authors who have made a significant contribution to the development of mathematical modeling of complex dynamic systems are considered. The types of neighborhood models are presented and the position of the new direction of hierarchical dynamic neuro-neighborhood models in the class of neighborhood models is reflected. The advantages of developing this approach are presented, namely, improving the interpretability of the model while simultaneously ensuring sufficient accuracy with generalization ability and resistance to noise. The main stages of construction are highlighted and the areas of application of hierarchical dynamic neuro-neighborhood models are presented. There are 3 ways of representing their structure: graphical, set-theoretic and matrix. The graphical presentation method is based on graphs divided into two layers, which describe the connections between nodes by transitions and outputs, respectively. The diagrams of layers and the general diagram of the node of the model under study are shown according to transitions and outputs. The set-theoretic method describes the model in the form of sets of nodes and hierarchical neighborhood connections between them. The matrix method allows you to present the model in the form of adjacency matrices for transitions and outputs by state and by external influences, respectively. A detailed description of hierarchical dynamic neuro-neighborhood models and neural networks in nodes is provided. The identification algorithm of the developed approach is described, and a diagram of the identification algorithm is shown. An example is given of constructing a hierarchical dynamic neuro-neighborhood model for predicting the total energy consumption of household appliances in a house, taking into account heating and weather conditions in the implemented Python program with automatic selection of optimal model parameters. A description of the source data taken from the Kaggle website is provided. The data was prepared, on the basis of which the resulting model was trained and tested. A diagram of a hierarchical dynamic neuro-neighborhood model of the predicted process is shown. Conclusions are drawn from the research done.

Python, simulation models, neural networks, neighborhood models, hierarchical dynamic neuro-neighborhood models, Python.