Long-term broadband technology forecasting

The key word for the thesis is long-term demand forecasting which have been applied on telecommunications and especially on broadband accesses and traffic.The objective with the thesis has been to structure and present work on long-term broadband forecasting, to evaluate the forecasting results and to extract the learning. Each main chapter ends with a section called experiences and conclusions.The thesis is organized in seven main parts.The first part addresses application of the Delphi technique for long term forecasting broadband accesses. Three Delphi surveys, which have been conducted during a long period, have been evaluated. All three Delphi surveys have used similar procedures in carrying out the survey, except that two of the Delphi surveys were postal surveys, while one was carried out on site. The applied procedure is evaluated based on an important reference article on Delphi surveys and also based on the long-term forecasting results. The Delphi surveys are not very often used. Hence, the description of the way to conduct the surveys and the experiences and also the evaluations of the results are given specific attention in the thesis.The second part of the thesis has the title “Long-term broadband technology forecasting”. Results from three papers are presented and evaluated. The papers show the evolution of the forecasting modelling. The first forecasts for the broadband evolution in Western Europe were made before broadband was introduced in the residential market in Western Europe. The long-term forecasts were developed based on Logistic models. The modelling also includes substitution effects between broadband technologies. Experiences have shown that technological knowledge and techno-economic evaluations are crucial for making long-term broadband forecasts. Some attention is also put on available broadband accesses statistics and an approach to separate aggregated broadband statistics to access statistics for the business market and for the residential market.“Long-term forecasting models for cost components and technologies” is the third part in the thesis. To be able to evaluate broadband technologies, techno-economic calculations of the “economic” value of the relevant broadband technologies are very important. The extended learning curve model invented by Borgar T Olsen and Kjell Stordahl is presented. The model is much more powerful than the simple exponential learning curve. The extended learning curve makes long-term forecasts of component costs and has the ability to be used directly on technoeconomic calculations, as opposed to the traditional learning curve model, which does not predict the cost as a function of time. In addition the extended learning curve model has interpretable parameters. It is shown that the model may utilize a priori information in cases where too few observations are available.The fourth part addresses long-term traffic forecasting. Three papers are enclosed. The chapter starts with a short overview of relevant forecasting models. Then attention is paid to forecasting and network planning. A comprehensive overview of the field is given together with numerous references in the enclosed paper “Forecasting – an important factor for network planning”. Longterm forecasts for the core network is analyzed and discussed. Also some figures for the total broadband traffic evolution in the Norwegian core network is presented. The last paper described in the chapter shows how long-term traffic forecasts on aggregated level can be used for traffic matrix forecasting by using the extended weighted least square method. The chapter ends by listing several important drivers for new and enhanced broadband traffic that are important in traffic forecasting models.