Cold and warm air temperature spells during the winter and summer seasons and their impact on energy consumption in urban areas

The main objective of this study is to determine and analyze cold and warm air temperature spells in the last 6 years (2007-2012) and reveal their impact on electrical energy consumption in a small-sized city such as Sombor (Serbia) with less than 50,000 inhabitants. Hourly air temperature values and electrical energy consumption data have been used as database for all methods. Warm and cold temperature spells (during heat and cold waves) have had the increasing tendencies in the last 6 years and they reflect on additional electrical energy consumption. Detailed analysis showed that higher energy demands occur during workdays and daytime period. Monitoring of the amount of consumed energy showed a clear relationship during the winter cold temperature spells, when electrical energy demand was higher than 0.3 MW. In summer period, the relationship was weaker and consumption was higher than 0.15 MW only when temperature exceeded 30 °C. A small number of air condition devices in houses and companies and mainly one-store buildings with thick walls, which make a good insulation from the outside air temperatures, are probably the main reasons for the above-mentioned results in summer. This paper introduces a new method to resolve the problem of short-term load forecasting, based on the support vector machines (SVM) technology and particle swarm optimization that has been used to optimize the SVM parameters. Similar-day-based forecast has shown that similar days for training should be filtered also using classifier of temperature period (cooling degree-days or heating degree-days in a row). Forecasting error is smaller compared to solutions where similar days are found only on season and temperature.