Abstract: To further improve energy efficiency, integrated energy system (IES) has realized closer interactions of various energy vectors and has become an important energy utilization way. IES load forecasting has become the primary prerequisite for IES planning and operation scheduling. It requires effective learning of multiple energy coupling information and accurate forecasting of multiple loads. This paper firstly analyzed energy consumption or demand characteristics in IES. This paper then provided a comprehensive overview on current technical works of IES load forecasting and particularly introduced the existing applications of data-driven learning techniques in this domain. Finally, the paper prospected future development of demand-side energy forecasting and envisaged constructing collaborative and multi-dimensional energy forecasting frameworks or systems in IES, taking advantages of the novel changes of multiple energy interactions of different energy vectors, multiple temporal horizons of load forecasts, multiple spatial aggregation levels of the energy systems, and multiple side combinations around energy.