Multiple Linear Regression Based Base Transceiver Station Power Consumption Prediction

Abstract

The development in wireless Technologies particularly cellular mobile networks is attracting people to ease their daily activities. In mobile networks, Base Transceiver Station (BTS) is the main infrastructure element that is performing the task of connecting the customer equipment with the cellular network. As there is a rapid growth of mobile subscribers and number of base stations demand, power consumption is also increasing. Thus studying the impact of factors related to power consumption at a base station has a significant advantage to save resources as well as providing sustainable service for telecom sectors. The fundamental aim of this thesis is, therefore, to analyze factors such as temperature, humidity and other factors to base transceiver station power consumption by implementing machine learning algorithm techniques to datasets of base station power system. Specifically, multiple linear regression technique is applied to deal with independent variables in respect of dependent variable. This technique was proposed as all the assumptions or prerequisites to use it in solving machine learning problems are satisfied by the behavior of the dataset on which this study is based. Furthermore, Python, Micro soft Access and Microsoft Excel are used as tools for data preprocessing. This research used datasets of 187484 rows having 4 variables which was split into 80 %, 10% and 10% training, validation and test sets respectively which has given best result of accuracy as compared to other split options. Hence, the experimental result confirms that temperature and DC load current have a positive relationship with power consumptions whereas voltage and humidity are negatively related with power consumption. The model which employed multiple regression techniques has a prediction performance accuracy of 99.5 %.