UBC Theses and Dissertations
Algorithms design for localization and vehicle-to-grid control Wenbo, Shi
Location is the fundamental information in wireless networks to support a variety of applications. In the first part of the thesis, we focus on designing localization algorithms in wireless sensor networks and radio frequency identification (RFID) systems. For localization in wireless sensor networks, we study the problem of locating sensors in irregular areas. We formulate the localization problem as a constrained least-penalty problem. We then propose a two-phase algorithm to eliminate the impact of irregularities. Simulation results show that the two-phase algorithm outperforms some of the existing multihop localization algorithms in terms of a lower average localization error in both C-shaped and S-shaped topologies. For localization in RFID systems, we propose a novel approach named MDS-RFID to locate active RFID tags based on multidimensional scaling (MDS), an efficient data analysis technique. The approach has the advantage of fully utilizing the distance information in the network, and thus can achieve better localization results than previous methods. To evaluate the performance of the proposed MDS-RFID algorithm, we perform extensive simulations and experiments to compare it with existing RFID localization schemes. Simulation results show that the MDS-RFID algorithm can achieve a lower average localization error than multilateration and the LANDMARC system. The experimental results validate the simulations results and show the performance gain of the MDS-RFID algorithm over multilateration and LANDMARC in a real RFID system. In the second part of the thesis, we shift our focus from localization to vehicle-to-grid (V2G), an emerging system in future smart grid to enable the power flow from the electric vehicles (EVs) to the grid. We study the V2G control problem under price uncertainty brought up by the real-time pricing scheme. We model the electricity price as a Markov chain and formulate the problem as a Markov decision process (MDP). The Q-learning algorithm is then used to adapt the control operations to the hourly available electricity prices. Simulation results show that our proposed algorithm can work effectively in the real electricity market and it is able to increase the profit significantly compared with the conventional EV charging scheme.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International