Modeling-based Normalization Method for Heating Energy Consumption in District Heated Buildings

Abstract

As Climate Change continues to accelerate, energy efficiency as a means to reduce Green House Gas emissions is more relevant than ever. A significant share of energy is consumed by heating residential buildings. To be able to assess their energy efficiency, one must be able to compare energy consumptions at different times disregarding weather variation, and normalization is a method to do this. A new modeling-based normalization technique was proposed, where modeled heating demand during normal year and observed time would form the normalization coefficient. The aim was to have improved stability and accuracy compared to traditional normalization methods. Data analysis was carried out in cooperation with Leanheat Oy, mostly utilizing data from their database but also other sources. Three options were compared to be utilized as the model in the new method: Multiple Linear Regression, Generalized Additive Models (GAMs) and Random Forest. GAMs were chosen for the proposed method based on their performance. Modeling-based normalization was performed on 100 residential buildings in Helsinki. The results were approximately 7 % higher in consumption than with traditional Motiva method. The proposed method indicated improved stability in handling times with exceptionally warm outdoor temperatures. The proposed method produced promising results but should be studied further before practical implementation.