Modelling energy demand – A systematic literature review

The assessment of transformation pathways for the energy system – crucial for all decisions towards climate neutrality – requires models of high temporal and spatial resolution. Information about energy demand serves as a critical input to infrastructure planning models. In this article, a systematic literature review of 342 recent publications on final energy demand modelling is provided with regards to the techniques and input data (demand drivers), the energy carrier and economic sector as well as the temporal and spatial granularity. This review sheds light into the variety of techniques and use cases of final energy demand forecasting; and it serves as a comprehensive literature overview and classification for researchers from both academia and private sector. It enables the readers to identify appropriate literature and data-model-combinations for jump-starting their projects. The analyzed articles have been compiled in comprehensive tables, which are structured by property, providing direct access to the respective articles. Electricity is the most forecasted energy carrier. The industrial sector is modelled the least often. Machine learning techniques are the most popular and are mainly used for electricity forecasting on short temporal horizons and building or regional level, however are dependent on historic load data. Engineering-based models are less dependent on historic consumption data and cover appliance-specific consumption on long temporal horizons. Metaheuristic and uncertainty techniques are used most frequently together with ML techniques to form hybrid models. Statistical well-proven techniques are used in various articles as benchmarks for novel approaches