Supply chain inventory forecasting and control is an integral part of supply chain management system, and it is a focus that industries must pay attention to in their operation and management. In this paper, the supply chain inventory demand forecasting model is constructed from the perspective of supply chain end, combined with the Transformer model in AIGC technology. The DL-Informer model is used to improve the Transformer model, realize the feature fusion of graph convolutional neural network, design and solve the feature graph adjacency matrix and complete the information fusion of each feature subgraph to improve the prediction accuracy. Aiming at the problems faced by supply chain inventory demand forecasting, the traditional algorithm with strong local optimization ability is combined with the genetic algorithm, and the hybrid genetic algorithm (HGA) is proposed to solve the nonlinear optimization problem. In the supply chain inventory forecasting practice, when the forecast length is 12, the MSE, MAE and RMSE index values of this paper’s forecasting model are 0.202, 0.174 and 0.416, respectively, which have more stable long-term forecasting performance compared with other models. And in the nonlinear simulation optimization experiments, the HGA algorithm shows good convergence and outstanding optimization effect in the nonlinear problem of supply chain inventory.