Prediction of energy consumption particularly in micro level is of vital importance in terms of energy planning and also implementation of any Clean Development Mechanism (CDM) activities that has become the order of the world today. It may be difficult to model household energy consumption using conventional methods such as time series forecasting due to many influencing factors. This paper presents a step wise regression model for forecasting domestic energy consumption based on micro level household survey data collected from Kerala, a state in southern part of India. The analysis of the data reveals significant influence of socio-economic, demographic, geographic, and family attributes upon total household energy requirements. While a wide variation in the pattern of energy requirements across the domestic sector belonging to different expenditure classes, per capita income level can be identified as the most important explanatory variable influencing variation in energy requirements. The models developed also demonstrates the influence of per capita land area, residential area among the higher income group while average age and literacy forms significant variables among the lower income group.