SHORT-TERM PATTERN SEQUENCE FORECASTING FOR HOUSEHOLD ENERGY DISAGGREGATION

Abstract

Buildings have contributed significantly to world energy consumption and greenhouse gas emissions. Energy demand is expected to continue rising because of increased access to electricity in developing countries, greater ownership and use of energy-consuming household appliances, and rapid growth in the global buildings floor area. For this reason, it becomes necessary to propose efficient strategies to enhance energy efficiency in residential buildings or, perhaps more important to the occupants, by reducing their monthly electricity bill. Hence, this research paper entails accurate energy forecasting for efficient energy management and optimal decision making.

This study aims to propose the well-established Pattern Sequence-based Forecasting (PSF) algorithm and investigates its potential in predicting household energy consumption. A comparison study was made using Singapore and the UK household energy time-series data. Based on the periodic characteristics of the observations, some of the most common household appliances (e.g., refrigerator, television and washing machine) were given a day-ahead prediction and their energy usage patterns were identified. Additionally, factors that influence energy consumption were further discussed. Experimental results showed that the PSF algorithm had performed well with reasonable prediction accuracy and patterns of household appliances energy use is associated with outdoor temperatures.

These results suggest that the PSF algorithm can be useful in predicting electricity consumption to improve household energy efficiency, whether in Singapore or the United Kingdom (UK). On this basis, the concept of pattern sequence similarity-based technique should have a larger sample size for more accurate results, such as applying it to other types of dwellings in Singapore and non-residential premises like offices.