Method of Diagnostics of Operation Modes of Individual Heat Supply Units, Allowing to Detect Pre-Emergency Situations at an Early Stage

The use of machine learning methods to control heat supply systems is a pressing issue worldwide. The aim of the work is to develop methods for intellectualizing the operating modes of individual heat supply units (IHU) to improve their efficiency and reliability. To achieve this goal, the following tasks were solved: creating and debugging methods for diagnosing IH operating modes; using the cluster analysis method, in particular the K-means algorithm, to identify pre-emergency situations at an early stage; analyzing the relationship between outdoor air temperature data and the pressures of direct and return network water in IHU operating modes using Novosibirsk as an example. The most important results of the work include dividing the measured parameters into five clusters, each of which characterizes a certain IHS operating mode. This was confirmed by the "Elbow Method", which determined the optimal number, which made it possible to significantly improve the forecasting of emergency modes. Studies have shown that a sharp increase in outdoor temperature leads to an increase in the pressure of direct network water, which can cause accelerated wear of heating networks due to the peculiarities of weather-dependent automation regulation. Introducing additional parameters into the initial data, such as the service life of heating networks and weather conditions, can improve the accuracy of forecasts. The significance of the obtained results lies in the possibility of early detection of emergency and pre-emergency modes of IHU operation, which helps prevent accidents and reduce repair and maintenance costs.