Load forecasting is one of the most important things for the economic operation and optimal scheduling of integrated energy systems. However

with the diversification of energy demand on the load side and the increasing coupling between various loads

the load characteristics of the integrated energy system are becoming more and more complex. Therefore

this paper proposes a multivariate load prediction method for integrated energy systems based on variational modal decomposition (VMD) and combined deep neural networks. Firstly

VMD is used to decompose the multivariate load to reduce the non-stationarity of the original load data. Secondly

the multivariate load components obtained by VMD decomposition and external influences are designed as inputs. The temporal convolutional network (TCN) is used to extract the temporal features of multivariate data

and the graph convolutional network (GCN) is used to explore the coupling features between multivariate data. Then

long short-term memory (LSTM) is used to capture the long-term dependence of multivariate data. Finally

the integrated energy system of Arizona State University

Tempe Campus

is adopted as an example to verify the effectiveness and reasonableness of the method proposed in this paper.