Neural Network Simulation of a Residential Building for a Data-Driven Thermal Consumption

P. 1–4.

Stukach O., Ershov I., Dvurechenskaya N.

The growing complexity of urban infrastructure increased the efficiency of heat supply, which must be ensured by the efficient and sustainable operation of supply organizations. Overspending of heating energy or decrease in thermal comfort follows to decrease of control quality. Systemically, the heat consumption of residential buildings on a whole city is a set of time series characterized by correlations not only with the outside temperature, but also a lot of building characteristics. Optimization of the temperature mode in the heating system is proposed to solve this problem. The model of heat consumption in a multi-storey building was designed using Long Short-Term Memory simulation. High accuracy of series reproduction has been achieved. The model is based on the characteristics of the building and meteorological factors as the temperature of the outside air. The data-driven model has been verified by commercial heat metering database. The results show the possibility and necessity of design a neural network weather-dependent regulator. Also it may be serve a basis for justifying housing renovation programs, as well as plans of the thermal energy supply organizations management.

Keywords: нейросеть, artificial neural network

In book

2025 Dynamics of Systems, Mechanisms and Machines (Dynamics). International scientific and technical conference Omsk State Technical University, Omsk, Russia

IEEE, 2025.