Adaptivni sustav upravljanja temeljen na podacima

master's thesis

• Title (english): Data-based Adaptive Control System
• Author: Petra Matijević
• Mentor: Tamara Petrović (mentor)
• Committee member: Tamara Petrović (član povjerenstva)
• Committee member: Stjepan Bogdan (član povjerenstva)
• Committee member: Frano Petric (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-19, Croatia

Abstract

Ovaj diplomski rad bavi se primjenom prediktivnog upravljanja (Model Predictive Control, MPC) u upravljanju temperaturom unutar podatkovnih centara. Na početku je proces modeliran pomoću diferencijalnih jednadžbi koje opisuju toplinsku dinamiku prostora, uzimajući u obzir prijenos topline kroz kondukciju i konvekciju. Nakon toga provedene su simulacije u MATLAB-u kako bi se ispitao odziv sustava u različitim uvjetima, uključujući prisutnost šuma mjerenja. Za identifikaciju sustava korištena je neuronska mreža, čime je omogućen pristup temeljen na podacima za predviđanje temperature na temelju masenog protoka i snage. Dodatno, dizajniran je i testiran MPC regulator, pri čemu je uspješno upravljanje temperaturom prikazano kroz simulacije. Ponašanje sustava analizirano je i u prisutnosti šuma, gdje su mjerene vrijednosti filtrirane kako bi se poboljšala točnost upravljanja. Rezultati su pokazali da predloženi pristup uspješno održava temperaturu prostora blizu zadane referentne vrijednosti, čak i u prisutnosti smetnji.

Abstract (english)

This thesis deals with the application of Model Predictive Control (MPC) in temperature control within data centers. Initially, the process was modeled using differential equations that describe the thermal dynamics of the space, taking into account heat transfer through conduction and convection. Subsequently, simulations were conducted in MATLAB to examine the system’s response under various conditions, including the presence of measurement noise. A neural network was used for system identification, enabling a data-driven approach for predicting temperature based on mass flow and power. Additionally, an MPC controller was designed and tested, demonstrating successful temperature control through simulations. The system's behavior was analyzed in the presence of noise, where the measured values were filtered to improve control accuracy. The results showed that the proposed approach successfully maintains the room temperature close to the desired setpoint, even in the presence of disturbances.

• Keywords: upravljanje zasnovano na podacima; prediktivno upravljanje; matematički model procesa; neuronska mreža
• Keywords (english): data-driven control; predictive control; mathematical model of the process; neural network
• Language: croatian
• URN:NBN: urn:nbn:hr:168:162384
• Study programme: Title: Information and Communication Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije (magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 21:31:45