Electricity Peak Demand Forecasting for Developing Countries

The current thesis aims to develop a peak demand forecast model suitable for developing countries based on their characteristic and availability of data. In this respect, we attempted to review a number of techniques used for energy forecasting and categorize them in terms of time ranges, the techniques used, and the cases in which they were employed. The advantages and disadvantages of each method were indicated and suitable approaches were devised to forecast the energy demand for small and large developing countries. We developed two different scenarios for small utilities depending on the availability of time series data. First, when considerable amount of time series data is available we proposed an econometric method to model the annual peak demand by which the key parameters affecting the electricity demand were discovered. The electricity demand was decomposed into weather sensitive demand and based demand to further examine the effect of extreme weather conditions on the peak demand. Second, when time series data is limited to merely annual peak demand records, an algorithm based on deterministic time series methods and fuzzy arithmetic was developed. These methods can be applied to forecast electricity demand of N. Cyprus and similar small islands. Thus, some advices were offered for electricity security plan of N. Cyprus. Finally, using the previously developed forecasting models, an approach was presented to forecast the peak demand of all developing countries based on their distinctive regional characteristic. The algorithm requires partitioning the country into smaller segments in which the previously developed forecast models for small utilities can be utilized. Keywords: Decomposition, Econometric Method, Extension Principle, Fuzzy Arithmetic, Peak Demand Forecasting, Time Series Method, Transformation Method