The foreseen large-scale deployment of renewable energy sources and continued electrification of space heating and transport is expected to seriously affect the stability of the electricity grid. Nonetheless, by means of intelligent building energy management systems associated with building loads, supply and demand could better be matched and flexibility can be provided to the grid. Accurate short-term and small-scale (sub-loads) electricity load forecasting is key in exploiting this energy flexibility potential of the built environment. Therefore, this paper presents and compares two load forecasting methodologies for predicting the electricity demand of a chiller machine of a medium-sized office building. One technique comprehends linear regression using time series data of the outdoor temperature in order to predict the chiller's power demand. As the next technique, a new Clustering Algorithm for Prediction (CAP) was introduced that compares historic outdoor temperature profiles with forecasted temperature profiles. The energy demands corresponding to similar historic temperature profiles can then be used to make a prediction. The introduced CAP technique has shown better results than the regression technique with an obtained Coefficient of Variance of the Root Mean Square Error value of 23.4% compared to 27.6% of the regression technique.
|Title of host publication||Proceedings of 2020 IEEE PES Innovative Smart Grid Technologies Europe, ISGT-Europe 2020|
|Number of pages||5|
|Publication status||Published - 27 Oct 2020|
- Energy Load Forecasting
- Outdoor Temperature