Analytical Development for Smart City Applications

Abstract

Smart City development has been identified as the key infrastructure for the future city blueprint. Having abundant impact on key applications such as energy, transportation, government and healthcare, smart city has been instilled into researchers and industrial engineers. Per the significance and impact of the smart city applications, the smart city system design is highly demanding. However, the system design in smart city development faces certain challenges. The formulation of the smart city developments usually belongs to the non-linear modeling, which perplexes the computation process (e.g. real-time control process, optimization process, etc.). Besides, requirement of different smart city applications varies which renders “smart” system design and analytical development based on the specific demand of the applications.

Therefore, the thesis addresses the following four contributions in the analytical development for smart city applications.

A new interference mitigated high traffic advanced metering infrastructure (AMI) is proposed. ZigBee based multi-Layer AMI infrastructure is proposed for handling a large scale network. An interference mitigated model is developed with the related routing control design and the corresponding protocol development. A multi-objective optimization formulation is designed for the optimization of the AMI latency, transmission rate and accuracy performance.

A new 2-Tier fuzzified control scheme is designed for improving the fuel efficiency of PHEVs. A fusion of genetic algorithm and fuzzy logic in Tier-1 is designed to aggregate the training independent characteristics of genetic algorithm and the high capability of handling uncertainty and nonlinearity of fuzzy logic. Meanwhile, the proposed Tier-1 structure prevents the “mature convergence” and tedious training process. Tier-2 is designed based on the fuzzification concept, which prevent a tedious rule base phenomenon in single tier structure to ensure real-time performance and to handle dynamic driving conditions. It is proved that 2-Tier control structure is sufficient for the case.

A new charging infrastructure planning for electric vehicles is proposed. The problem is formulated in complex network with its nature to be suitable for network related modeling. Two planning algorithms, fast and comprehensive planning algorithms, are designed based on the formulation to provide solutions for the determination of the number and the location of the charging stations in the charging infrastructure. New multi-objective optimization formulation for the customer experience and utilization investment efficiency and cost is designed. A new merit is designed to give an objective and comprehensive on the evaluation of the charging infrastructure planning domain.

An accurate sleep stage classifier has been developed. The classifier can be applied on the accurate classification of sleep stage, which is effective to assess sleep apnea. The classifier is based on the ECG signal feature extraction. Awake, Light sleep and Deep sleep can be accurately classified in real-time. A new kernel is designed for the classifier of sleep stage for sleep apnea, which is designed based on the cross-correlation and convolution to reveal the similarity of ECG samples. Trade-off solutions are obtained by multi-objective optimization formulation to tune the kernel related parameters such as Lagrangian multipliers, misclassification tolerance, fitting degree, etc. sampling frequency is also considered in the optimization process. The proposed classifier provides a real-time and accurate assessment of sleep apnea.

Date of Award	9 May 2018
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kim Fung TSANG (Supervisor)