Building upon the traditional physics-informed neural network, two improved methods, EmPINN and DL-PINN, are presented by incorporating dimension expansion and diverse physics loss functions. EmPINN innovatively introduces a neural network structure with residual connections and a dimension-expanding mechanism. DL-PINN, based on EmPINN, combines the dimension-expanding mechanism with gradient enhancement and variational physical information to incorporate physical information more effectively and improve the fitting capability of the neural network. Experimental results demonstrate that the proposed methods outperform traditional PINN method, achieving up to two orders of magnitude improvement in the accuracy of solving different partial differential equations.