Data-Driven Model For Heat Load Prediction In Buildings Connected To District Heating Networks

Abstract

In this study we investigate the heat load patterns in one building using multi-step forecasting model. We combine the Autoregressive models that use multiple eXogenous variables (ARX) with Seasonally adaptable Time of Week and Climate dependent models (S-TOW-C) (to correct model inaccuracies), to obtain a robust and accurate regression model that we called S-TOW-C-ARX used in time series forecasting. Based on the experiment results, it has been shown that the proposed model is suitable for short term heat load forecasting. The best forecasting performance is achieved in winter term where the prediction values are from 4 to 20 % away from the targets, which are commonly seen as very good values.