Load forecasting is of great practical value for optimizing power resource allocation, reducing operating costs and preventing potential system security problems. With the increasing capacity of new energy sources such as photovoltaic (PV) into the grid, the difficulty of load forecasting is gradually increasing. Load integration data including PV power generation has a large amount of redundant information, and many current studies only focus on the load itself, and modelling based on the electrical characteristics of the load alone will reduce the forecasting accuracy. In this paper, a load forecasting model based on feature fusion is proposed, which combines meteorological factors such as wind speed, wind pressure, temperature, humidity, etc., based on using the load power at the same time. The load decomposition is first performed using the sequence-to-point model, and then the load decomposition results are applied to the TCN-BiGRU model for load forecasting. The experimental results show that the root mean square error (RMSE) and mean absolute percentage error (MAPE) in this paper have high accuracy, which verifies that the model has load forecasting capability.