Home > Publications > Doctoral dissertation & Master thesis
Domestic Conference
Title Applications of Mixed Optimization Models in Communications : From Mobile Computing to Smart Grids
Degree Ph.D.
Author Abbas Mehrabi
Advisor Kiseon Kim
Graduation Date 2017.02.24 File Link icon
    Date 2019-02-25 10:01
Mixed integer programming (MIP) optimization models have been used in sev eral research and practical disciplines. 
They have been particularly employed f or the scheduling and planning problems arisen in research fields such as comput er 
networks, wireless sensor networks (WSNs), mobile computing as well as the e merging interdisciplinary research areas. In this thesis, 
we investigate the ap plication of MIP models in the formulation of two research problems in the scope of energy harvesting wireless sensor networks (EH-WSNs) and smart grid. The pro posed problem formulations are in the form of integer linear programming (ILP) a nd mixed integer linear/nonlinear programming (MILP/MINLP) optimization models. Due to the intractability of the proposed models, we design efficient
heuristics to achieve suboptimal solutions for the problems within reasonable computation time. We also show through both theoretical and simulation analysis that the pro posed algorithms outperform their competitors in term of the quality of solution s returned by
the algorithm as well as the required computation time.

This thes is is divided into three major parts. In the first part, we investigate the issu e of network throughput maximization (NTM) in data collection in EH-WSNs using a mobile sink with path-constrained trajectory. Different from conventional WSNs, using a sink in
EH-WSNs introduces its own challenges due to the time varying c haracteristics of energy harvesting during different time intervals. 
To allevia te such complex scenario, we design concrete ILP optimization model to capture t he problem objective under the system and
energy harvesting models. The proposed ILP model takes into account the heterogeneous duration of sensor’s transmissio n as well as the periodically energy harvesting by sensor nodes during different time intervals. We design a simple yet efficient algorithm for NTM
problem whic h leads to significantly improvement in network throughput while reducing the co mputation time compared to the existing approaches. In order to further improve the throughput, different scenarios are proposed for NTM problem by optimizing s ome
practical system parameters in the second part of the thesis. The problem sc enarios are formulated as mixed integer linear/nonlinear programming (MILP/MINLP ) optimization models for which efficient low complexity heuristics are devised that result in further improvement 
in network throughput, improving the energy efficiency as well as exhibiting the same order of worst case complexity with re spect to their competitor. In the third part, we study the problem of maximizing the overall profit of both energy demand and supply entities in scheduling of P EVs under multiple charging stations multiple aggregators (MCSs-MAs) communicati on schema at large geographical scale of vehicle to grid (V2G) operation. A mixe d integer non-linear programming (MINLP) optimization model is proposed for the problem formulation which incorporates the dynamic arrival of PEVs, their energy requirement as well as the auxiliary costs associated with charging spots. Unde r the real-time scheduling, we propose an online and decentralized greedy alloca tion algorithm with internal updating heuristics which as shown through extensiv e simulations outperforms its alternative allocation competitor in term of obtai nable profit as well as the ancillary services. 
The effect of different model p arameters on the performance of V2G system are also studied through different si mulation case studies.
Finally, we conclude this thesis and leave the readers wi th some interesting future research directions.
광주과학기술원 한·러 MT-IT 융합기술연구센터 광주과학기술원정보통신공학부