MICROWAVE LINK PLANNIN & TIME SERIES BASED MICROWAVE LINK CAPACITY OPTIMIZATION

Abstract

For the last thirty years or so, microwave point to point link was the dominant backhaul technology in telecommunication companies. However, the development of mobile technology toward 4G & 5G and the expansion of cellular network in terms of coverage to provide mobile services for more people, has caused an unprecedented traffic growth in telecommunication companies. As a result of this unparalleled increase in traffic, telecommunication companies faced significant challenge in providing sufficient bandwidth in microwave transmission media. Cellular providers can overcome the issue of bandwidth constraint in microwave links by properly utilizing the available frequency resources and by carefully optimizing the capacity of existing microwave links in the network. While designing the microwave link usually frequency bandwidth which is a very limited and expensive resource is wasted due to lack of accurate knowledge about the traffic of the site. Therefore, in this project we first designed a microwave backhaul link for an LTE site using Ellipse software in order to explain the planning steps required for a microwave link. We also want to illustrate in this part that why the assigned frequency bandwidth during the planning phase doesn’t come out to be accurate enough and results in frequency waste. The second part of project is microwave link capacity optimization using time series forecasting method. As traditionally, microwave link capacity was analyzed and studied manually by transmission planning and optimization team which was a time consuming process and also a difficult task to accurately analyze link throughput or utilization of every individual link. [v] Recently, application of machine learning provide practical solution for several challenges in cellular networks. In fact with the help of machine learning technique network operators can bring advanced automation in different parts of the network. So in the second part of this project we suggest a time series forecasting method to automatically optimize capacity and frequency bandwidth in microwave link. Historical daily average throughput of every microwave link is used to forecast future capacity in the link based on time series prediction methods. Then considering the predicted capacity an appropriate frequency channel would be selected for every microwave link. Two different prediction model, LSTM and ARIMA is used to forecast future capacity in microwave links. Moreover, MSE, RMSE and MAPE criterion are used for performance evaluation. An MAPE of less than or equal to 6 and peak to average ratio of 20% is considered for frequency channel bandwidth optimization in microwave links.