METHODS AND MEANS OF AN AUTOMATED RESOURCE CONTROL AND ACCOUNTING SYSTEM FOR A UNIVERSITY

Abstract

Universities spend considerable resources on gas, water, and electricity, yet obtaining consumption data for each resource faces significant limitations due to the closed nature of gas distribution networks, the lack of granular water supply measurements at the level of individual pipeline sections, and the inability to localize bottlenecks within the electrical grid. Automated parameter collection tools make it possible to gather data on consumption processes and respond to abnormal situations in a timely manner. This paper describes the development and deployment of an automated resource control and accounting system at TNTU, which integrates all three resource types within a single platform. Since gas, water, and electricity are governed by different physical laws, the methods and tools for data collection, analysis, and processing cannot be the same for each, which makes the development of a unified mathematical model for all three resources impractical. The common criterion is economic feasibility – minimizing costs while maintaining normal infrastructure operation.

Since access to gas meter readings is restricted by gas distribution networks, the only way to retrieve data without physically modifying the metering device is to capture readings with an IP camera and recognize them using the TrOCR model. Water and electricity consumption data are received in real time from ESP32-based IoT controllers over LoRaWAN. The data acquisition approach for each resource reflects the unique nature of the organization's operational processes.

A two-layer method was developed for detecting water consumption anomalies. The first layer, based on a GRU autoencoder, identifies structural deviations of the daily consumption profile from the reference profile for the corresponding day type; the second layer, based on normalized deviation, detects nighttime leaks and sharp consumption spikes. Both layers account for the organization's operational schedule, which allows the system to distinguish an emergency from an expected rise in consumption. Deployment at TNTU reduced the time between an incident occurring and its detection.