Neural network and genetic algorithm techniques for energy efficient relay node placement in smart grid

• Main Author: Safaei, Mahmood
• Format: Thesis
• Language: English
• Published: 2014
• Subjects: QA76 Computer software
• Online Access: http://eprints.utm.my/id/eprint/48539/1/MahmoodSafaeiMFC2014.pdf; http://eprints.utm.my/id/eprint/48539/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:85893?queryType=vitalDismax&query=Neural+network+and+genetic+algorithm+techniques+for+energy+efficient+relay+node+placement+in+smart+grid&public=true
• Institution: Universiti Teknologi Malaysia

Smart grid (SG) is an intelligent combination of computer science and electricity system whose main characteristics are measurement and real-time monitoring for utility and consumer behavior. SG is made of three main parts: Home Area Network (HAN), Field Area Network (FAN) and Wide Area Network (WAN). There are several techniques used for monitoring SG such as fiber optic but very costly and difficult to maintain. One of the ways to solve the monitoring problem is use of Wireless Sensor Network (WSN). WSN is widely researched because of its easy deployment, low maintenance requirements, small hardware and low costs. However, SG is a harsh environment with high level of magnetic field and background noise and deploying WSN in this area is challenging since it has a direct effect on WSN link quality. An optimal relay node placement which has not yet worked in a smart grid can improve the link quality significantly. To solve the link quality problem and achieve optimum relay node placement, network life-time must be calculated because a longer life-time indicates better relay placement. To calculate this life-time, it is necessary to estimate packet reception rate (PRR). In this research, to achieve optimal relay node placement, firstly, a mathematical formula to measure link quality of the network in smart grid environment is proposed. Secondly, an algorithm based on neural network to estimate the network life-time has been developed. Thirdly, an algorithm based on genetic algorithm for efcient positioning of relay nodes under different conditions to increase the life-time of neural network has also been developed. Results from simulation showed that life-time prediction of neural network has a 91% accuracy. In addition, there was an 85% improvement of life-time compared to binary integer linear programming and weight binary integer linear programming. The research has shown that relay node placement based on the developed genetic algorithms have increased the network life-time, addressed the link quality problem and achieved optimum relay node placement.