Accurate multivariate load forecasting is a prerequisite for optimal scheduling and stable operation of integrated energy system (IES). Aiming at the complex coupling relationship between multiple loads and many factors affecting load forecasting in IES,a method for multivariate load prediction based on extreme gradient boosting (XGBoost) and multi task learning (MTL) is proposed. Firstly,the contribution of each influencing factor to the multivariate load is obtained by XGBoost importance ranking,and the key factors influencing the load prediction are selected as the input of the prediction model based on the contribution,which ensures the effective correction of the input features to the multivariate load prediction. Secondly,the gated recurrent unit (GRU) is used as a shared layer to build the MTL prediction model,and the sub-tasks share information with each other to effectively exploit the complex coupling relationship between the loads. Finally,the validity of the proposed model is verified by using the load data of an integrated energy station in Shanghai as an example,and the results show that the model can adapt to the changes of various types of loads in the actual integrated energy system,effectively improving the prediction accuracy and reducing the training time.