AVS2 is the new generation of video coding standard developed by the Audio Video Coding Standard Working Group of China. Similar to HEVC, a flexible partition structure is adopted to improve coding performance in AVS2. For intra coding, a coding unit (CU) is recursively split into sub CUs based on the quad-tree structure. CUs should be further split into prediction unit (PU) with various square or non-square shapes. All these processes increase encoding complexity dramatically. In this paper, a fast intra coding algorithm for CU partition and PU partition is proposed to reduce the complexity in AVS2 intra coding. Specifically, the statistical analysis model between texture complexity and partition mode is established. According to the model, an adaptive online-offline threshold selection algorithm is proposed to determine the early skipping and early termination in CU and PU level. Experimental results show that the proposed fast intra coding algorithm achieves more than a 54% encoding time reduction on average with only a 0.77% BD-rate increase under all-intra configuration for the AVS2 reference software RD18.0.