Abstract: Adaptive optical systems require the wavefront sensors to realize realtime dynamic measurement, and the wavefront curvature sensing technology can meet such requirement. A novel wavefront curvature sensor based on the twisted diffraction grating has a greater advantage in the realization of detecting device, and its wavefront reconstruction algorithm is used for the measurement of optical surfaces. The numerical simulation of the detection signal was implemented with the theory of diffraction optics, and the numerical simulation of the wavefront reconstruction was realized with Green function algorithm under Neumann boundary conditions. The results indicate that the speed of Green function algorithm is fast, it meets the requirement of the real-time wavefront reconstruction, and the result of wavefront reconstruction for low-order Zernike polynomials is good. The analysis shows that the major source of wavefront reconstruction error is the boundary noise in the gradient of light intensity.