{"id":{"repo_id":"nus","oai_identifier":"oai:scholarbank.nus.edu.sg:10635/246577"},"canonical_url":"https://search.dev.ndltd.org/etd/nus/oai:scholarbank.nus.edu.sg:10635/246577","repository":{"repo_id":"nus","name":"National University of Singapore","base_url":"https://scholarbank.nus.edu.sg/oai/request"},"display":{"title":"INTELLIGENT DYNAMIC THERMAL CONTROL USING DEEP LEARNING AND REINFORCEMENT LEARNING","abstract":"This thesis aims to develop practical and intelligent dynamic thermal control (DTC) methods for data centers (DCs) using deep learning and reinforcement learning (RL). The research investigates existing RL algorithms for DTC and proposes a comprehensive analysis framework considering the algorithm, control objective, and system dynamics. The challenges associated with implementing RL in real-world scenarios are identified, leading to a relaxed task formulation that considers the availability of a historical dataset. Subsequently, machine learning-based control-oriented modeling approaches, such as uncertainty quantification and a well-designed training framework, are explored to address challenges related to historical datasets. The thesis then develops two novel approaches: a safe model-free RL algorithm and an uncertainty-aware model-based control method. Rigorous evaluation through simulation experiments and theoretical analysis demonstrates their effectiveness. A physical scale-down testbed further validates their practicability in real-world systems. Overall, this thesis contributes to the field by advancing the implementation of DTC in DCs.","abstract_html":"This thesis aims to develop practical and intelligent dynamic thermal control (DTC) methods for data centers (DCs) using deep learning and reinforcement learning (RL). The research investigates existing RL algorithms for DTC and proposes a comprehensive analysis framework considering the algorithm, control objective, and system dynamics. The challenges associated with implementing RL in real-world scenarios are identified, leading to a relaxed task formulation that considers the availability of a historical dataset. Subsequently, machine learning-based control-oriented modeling approaches, such as uncertainty quantification and a well-designed training framework, are explored to address challenges related to historical datasets. The thesis then develops two novel approaches: a safe model-free RL algorithm and an uncertainty-aware model-based control method. Rigorous evaluation through simulation experiments and theoretical analysis demonstrates their effectiveness. A physical scale-down testbed further validates their practicability in real-world systems. Overall, this thesis contributes to the field by advancing the implementation of DTC in DCs.","abstract_has_math":false,"creators":["ZHANG QINGANG"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2023,"date_issued":"2023-07-10","date_published":"2023-07-10","updated_at":"2026-07-24T03:32:43Z","subjects":["Data Center, Thermal Control, Deep Learning, Reinforcement Learning, Building, Energy-Saving"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":null,"outbound_label":null,"outbound_source":null},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["ZHANG QINGANG"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2023-07-10"]},{"key":"dc:relation.isreferencedby","label":"Dc Relation Isreferencedby","values":["https://scholarbank.nus.edu.sg/handle/10635/246577"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Data Center, Thermal Control, Deep Learning, Reinforcement Learning, Building, Energy-Saving"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://scholarbank.nus.edu.sg/bitstreams/14ed71ee-91c1-449d-98f3-f58c24424fff/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This thesis aims to develop practical and intelligent dynamic thermal control (DTC) methods for data centers (DCs) using deep learning and reinforcement learning (RL). The research investigates existing RL algorithms for DTC and proposes a comprehensive analysis framework considering the algorithm, control objective, and system dynamics. The challenges associated with implementing RL in real-world scenarios are identified, leading to a relaxed task formulation that considers the availability of a historical dataset. Subsequently, machine learning-based control-oriented modeling approaches, such as uncertainty quantification and a well-designed training framework, are explored to address challenges related to historical datasets. The thesis then develops two novel approaches: a safe model-free RL algorithm and an uncertainty-aware model-based control method. Rigorous evaluation through simulation experiments and theoretical analysis demonstrates their effectiveness. A physical scale-down testbed further validates their practicability in real-world systems. Overall, this thesis contributes to the field by advancing the implementation of DTC in DCs."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["237fc8a7e24f56784dbc270c94b60488","aa21a33ebe2e274cfbbc50f2b89c58e9"]},{"key":"dc:title","label":"Title","values":["INTELLIGENT DYNAMIC THERMAL CONTROL USING DEEP LEARNING AND REINFORCEMENT LEARNING"]}]}],"canonical_facts":{"dc:creator":["ZHANG QINGANG"],"dc:date.issued":["2023-07-10"],"dc:description.abstract":["This thesis aims to develop practical and intelligent dynamic thermal control (DTC) methods for data centers (DCs) using deep learning and reinforcement learning (RL). The research investigates existing RL algorithms for DTC and proposes a comprehensive analysis framework considering the algorithm, control objective, and system dynamics. The challenges associated with implementing RL in real-world scenarios are identified, leading to a relaxed task formulation that considers the availability of a historical dataset. Subsequently, machine learning-based control-oriented modeling approaches, such as uncertainty quantification and a well-designed training framework, are explored to address challenges related to historical datasets. The thesis then develops two novel approaches: a safe model-free RL algorithm and an uncertainty-aware model-based control method. Rigorous evaluation through simulation experiments and theoretical analysis demonstrates their effectiveness. A physical scale-down testbed further validates their practicability in real-world systems. Overall, this thesis contributes to the field by advancing the implementation of DTC in DCs."],"dc:format.checksum.md5":["237fc8a7e24f56784dbc270c94b60488","aa21a33ebe2e274cfbbc50f2b89c58e9"],"dc:identifier.uri":["https://scholarbank.nus.edu.sg/bitstreams/14ed71ee-91c1-449d-98f3-f58c24424fff/download"],"dc:relation.isreferencedby":["https://scholarbank.nus.edu.sg/handle/10635/246577"],"dc:subject":["Data Center, Thermal Control, Deep Learning, Reinforcement Learning, Building, Energy-Saving"],"dc:title":["INTELLIGENT DYNAMIC THERMAL CONTROL USING DEEP LEARNING AND REINFORCEMENT LEARNING"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T03:32:43Z"}