Demand forecasting in DHC-network using machine learning models

District heating and cooling systems are becoming increasingly popular to serve the thermal demands of consumers. To leverage the use of a DHC system to set demand curtailment targets using techniques like demand response, it is important to accurately model and forecast thermal demand. The data analytics based modelling framework for forecasting energy consumption requires knowledge about the historical consumption behaviour of the user. However, it is more often found, that only limited data about the historical consumption is available in a grid: in some cases only a subset of the buildings in a grid are instrumented, and even for those buildings instrumented a detailed recording of the consumption is not available. In this work, we try to provide a mechanism to forecast the demand of a building of a grid (for which we lack the historical consumption data) by using the historical consumption data of similar buildings in the same grid. We provide an evaluation of the methodology on an experimental data set obtained from households in Luletta, Northern Sweden.