Abstract: To accurately simulate the spatiotemporal evolution of electron density in the lower ionosphere, this paper proposes a low-ionospheric ion chemistry model based on a physics-informed neural network (PINN). The ion continuity equations, initial conditions, and charge neutrality constraints are embedded into the loss function of the neural network. The network takes time, altitude, and spatiotemporal features as inputs, and outputs the densities of six positive ion species. The electron density is then obtained by summing these outputs. The study focuses on the hourly electron density data from lower ionosphere at Sanya station in May 2024, using IRI-2020 and NRLMSIS 2.1 model data for training and validation, with model performance evaluated using multiple metrics. The results demonstrate strong agreement between the predicted electron density and the IRI-2020 reference values under initial conditions. The R2 values exceed 0.99 for both training and validation sets, with MAPE values of 3.7140% and 4.0908%, respectively. Furthermore, the model produces direct predictions without iterative time-stepping, achieving a MAPE of approximately 6% after one minute of evolution. Error analysis further reveals that the prediction accuracy slightly decreases during sunrise and sunset periods, which consistent with the rapid variations in ionospheric photochemical processes. By integrating physical constraints with data-driven modeling, the proposed PINN-based model effectively simulates the evolution of electron density in the lower ionosphere. This study extends the application of PINN to low-ionospheric ion chemistry modeling, and provides a novel methodological framework and theoretical basis for electron density prediction in the lower ionosphere.