@article{ChenChuaKoch, author = {Chen, Ying and Chua, Wee Song and Koch, Thorsten}, title = {Forecasting day-ahead high-resolution natural-gas demand and supply in Germany}, series = {Applied Energy}, journal = {Applied Energy}, number = {Volume 228}, doi = {https://doi.org/10.1016/j.apenergy.2018.06.137}, pages = {1091 -- 1110}, abstract = {Forecasting natural gas demand and supply is essential for an efficient operation of the German gas distribution system and a basis for the operational decisions of the transmission system operators. The German gas market is moving towards more short-term planning, in particular, day-ahead contracts. This increases the difficulty that the operators in the dispatching centre are facing, as well as the necessity of accurate forecasts. This paper presents a novel predictive model that provides day-ahead forecasts of the high resolution gas flow by developing a Functional AutoRegressive model with eXogenous variables (FARX). The predictive model allows the dynamic patterns of hourly gas flows to be described in a wide range of historical profiles, while also taking the relevant determinants data into account. By taking into account a richer set of information, FARX provides stronger performance in real data analysis, with both accuracy and high computational efficiency. Compared to several alternative models in out-of-sample forecasts, the proposed model can improve forecast accuracy by at least 12\% and up to 5-fold for one node, 3\% to 2-fold and 2-fold to 4-fold for the other two nodes. The results show that lagged 1-day gas flow and nominations are important predictors, and with their presence in the forecast model, temperature becomes insignificant for short-term predictions.}, language = {en} }