Evaluation And Optimization Of Scheduling For Smart Building Air-Conditioning Systems Under Uncertain Environment

Due to the notable advantages like efficient resource utilization, low maintenance cost and comfortable internal environment, smart buildings have drawn a lot of atten-tions and have become a hot area of research in recent years. As a critical part of air ven-tilation system in a smart building, the air-conditioning system and its scheduling strate-gies strongly affect the overall energy-consuming of the smart building as well as the comfort experiences of its users. Generally, the surrounding environment of the smart building has lots of uncertain factors, which make the design, evaluation and optimiza-tion of the scheduling strategies of air-conditioning system more complex. Therefore, how to effectively and quickly obtain an optimal scheduling solution for air-conditioning systems with minimum cost of energy and maximum satisfaction of users is becoming a major challenge in the system-level design of air-conditioning systems.To address the above challenge, this thesis investigates the variation-aware eval-uation and optimization methods for smart building’s air-conditioning systems under uncertain environments. This thesis proposes a novel framework that enables accurate modeling of uncertainties, effective evaluation of scheduling strategies, and fast search-ing of optimal configuration parameters for smart building’s air-conditioning systems. This thesis makes three major contributions as follows:1. Based on network of priced timed automata, we present a comprehensive model-ing approach that supports the formal description of smart building’s air-conditioning systems under uncertain environments. This paper analyzes the weather varia-tions, power variations of heaters and human activities within an uncertain en-vironment of smart building in detail. By using various distribution models, our approach supports accurate uncertainty modeling of all the components of air-conditioning systems.2. We propose an evaluation method of scheduling strategies for air-conditioning system based on statistical model checking. This method supports the quantita-tive analysis of building’s energy consumption and user comfort while different strategies are applied into the system under uncertain environments.3. A parameter configuration optimization method for air-conditioning system is proposed based on supervised learning methods. This method can quickly find the best parameter configuration instance (PCI) for a given smart building under uncertain environment from a large number of configuration candidates. Based on the obtained PCI, the smart building can have lower energy consumption and high user satisfaction. By resorting to the supervised learning method, the search time and resources for the best PCI can be drastically reduced.Experimental results show that our proposed framework can not only help smart building designers to effectively evaluate and select design candidates with different strategies, but also enables the quick search of optimal configuration parameters for a given air-conditioning system, which make the building more energy efficient and comfortable.