В е с т н и к алматинского университета

Түйін сөздер: ауытқуларды талдау, әдістер, қаланы жылумен жабдықтау, тиімділік. 
ANALYSIS OF DATA ANOMALIE DETECTION METHODS FOR USE 
IN ANALYTICS AND HEAT SUPPLY MANAGEMENT OF THE CITY 
 
I.T. Utepbergenov, A.M. Seitbekova* 
 
Non-profit JSC “Almaty University of Power Engineering and Telecommunications named after
Gumarbek Daukeyev”, Almaty, Kazakhstan 
E-mail: 
[e-mail]
 , 
[e-mail]
  
 
Annotation. This study is an analysis of modern publications. It is dedicated to improving and optimizing the 
operation of the heat supply system of city districts based on the detection and analysis of anomalous data in heat 
supply systems in terms of applicability in Kazakhstan. In the article, the choice of method for the task was carried out 
on the basis of a comparative analysis of modern research. It is based on the Unsupervised Variational Autoencoder 
(VAE). It belongs to machine learning methods for anomaly detection and gives the best results. The use of IoT sensors 
are considered and selected as a source of primary information and drift detection. They are available for analytics and 
control tasks in HVAC systems. Description and results of an effective anomaly detection method are given. They are 
based on adaptive background extraction during clustering. Buildings are the main consumers of the heat supply of 
city. Modern anomaly detection approaches in building automation systems in European countries are considered in 
detail. Conclusion: collected and processed by all autonomous subsystems using in the review methods for detecting 
anomalies in data for the tasks of analytics, control, optimization and forecasting of the heat supply of a city in 
Kazakhstan allows us to consider the entire network as an intelligent system and will certainly allow to get an effect.