Data driven modeling and optimization of energy systems

• Main Author: Zhang, Chuan
• Other Authors: Alessandro Romagnoli
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Mechanical engineering::Energy conservation
• Online Access: https://hdl.handle.net/10356/100897; http://hdl.handle.net/10220/48563
• Institution: Nanyang Technological University

Recent advances in data science and machine learning bring new opportunities for the modeling and optimization of energy system. Applications of machine learning models in energy system modeling and optimization are explored in the thesis. It is found that through the combination of feature engineering and machine learning, high-fidelity yet fast-response surrogate model could be constructed (20\% increase in building energy forecast example). Such machine learning based models are further incorporated into mixed integer nonlinear programming optimization framework to optimize the energy efficiency, payback period, and environmental impact of energy system. By combining greedy search with mixed integer nonlinear programming, CO2 emission of industrial co-generation system is reduced from 7921tons to 5195tons. A domain ontology for energy system modeling and optimization is established, the whole modeling and optimization method is combined with the ontology to develop an intelligent system to enable ontology-based automatic optimization for Jurong Island eco-industrial park Singapore. The work of this thesis shows that machine learning models, together with existing optimization framework, can automatically harness the knowledge database, formulate optimization problem, facilitate the energy system design and optimization related decision-making efficiently.