To improve the identification accuracy of the static model of quartz flexible accelerometer, this study proposes a static model parameter identification method based on a Particle Swarm Optimization-Back Propagation (PSO-BP) neural network. This approach addresses the local optima susceptibility of Back Propagation (BP) neural networks through Particle Swarm Optimization (PSO) integration. The neural architecture is configured according to accelerometer input-output dimensions, where the PSO's global exploration capability optimizes the initial weight for the BP network. Precision centrifuge-based calibration experiments were conducted to validate the proposed method. Experimental results demonstrate that the PSO-BP neural network exhibits significantly enhanced capability in resolving nonlinear coefficients compared to the standard BP network, achieving a reduction of the mean squared error (MSE) by two orders of magnitude, which provides technical support for advancing the development of high-precision navigation technologies in airborne inertial navigation systems.