In recent years, inductively coupled plasma (ICP) has proven highly valuable in integrated circuit etching and material processing due to its high electron density and process controllability. However, traditional experiments have high trial-and-error costs, while numerical simulations face challenges such as complex modeling and convergence difficulties. This study proposes a deep learning-based surrogate model that employs a deep neural network (DNN) to efficiently predict ICP discharge characteristics. The DNN establishes nonlinear relationships between process parameters (pressure, input power, temperature) and discharge characteristics (electron density and temperature), bypassing the grid generation and multiphysics coupling complexities inherent to traditional simulations. Results demonstrate that compared with the traditional fluid simulations that take thousands of seconds, the trained DNN can produce highly consistent predictions in only 2 s. The correlation coefficient between true and predicted values reaches 0.999, significantly improving computational efficiency while maintaining high prediction accuracy. This deep learning-assisted ICP fluid simulation overcomes the major limitations of traditional methods while validating the potential of deep learning for intelligent control of low-temperature plasma systems.