摘要：Widespread introduction of low energy buildings (LEBs), passive houses, and zero emission buildings (ZEBs) are national target in Norway. In order to achieve better energy performance in these types of buildings and successfully integrate them in energy system, reliable planning and prediction techniques for heat energy use are required. However, the issue of energy planning in LEBs currently remains challenging for district heating companies. This article proposed an improved methodology for planning and analysis of domestic hot water and heating energy use in LEBs based on energy signature method. The methodology was tested on a passive school in Oslo, Norway. In order to divide energy signature curve on temperature dependent and independent parts, it was proposed to use piecewise regression. Each of these parts were analyzed separately. The problem of dealing with outliers and selection of the factors that had impact of energy was considered. For temperature dependent part, the different methods of modelling were compared by statistical criteria. The investigation showed that linear multiple regression model resulted in better accuracy in the prediction than SVM, PLS, and LASSO models. In order to explain temperature independent part of energy signature the hourly profiles of energy use were developed.
其他摘要：Widespread introduction of low energy buildings (LEBs), passive houses, and zero emission buildings (ZEBs) are national target in Norway. In order to achieve better energy performance in these types of buildings and successfully integrate them in energy system, reliable planning and prediction techniques for heat energy use are required. However, the issue of energy planning in LEBs currently remains challenging for district heating companies. This article proposed an improved methodology for planning and analysis of domestic hot water and heating energy use in LEBs based on energy signature method. The methodology was tested on a passive school in Oslo, Norway. In order to divide energy signature curve on temperature dependent and independent parts, it was proposed to use piecewise regression. Each of these parts were analyzed separately. The problem of dealing with outliers and selection of the factors that had impact of energy was considered. For temperature dependent part, the different methods of modelling were compared by statistical criteria. The investigation showed that linear multiple regression model resulted in better accuracy in the prediction than SVM, PLS, and LASSO models. In order to explain temperature independent part of energy signature the hourly profiles of energy use were developed.
