Shadows cause serious reduction of power generation in a photovoltaic (PV) power plant. This paper proposes a shadow detection method based on improved pulse coupled neural networks (PCNN) for partially shaded PV module images. Suitable initial parameters of unit-link PCNN are set. The ULPCNN is applied to the gray shading images, and the two-dimensional Otsu method is used to automatically determine the numbers of iterations. The segmentation that achieves the best threshold in the iteration is selected as an optimal result. Simulations verify that the test images are well segmented, and the method has better performance compared to the conventional PCNN and ULPCNN.