To study the relationship between the chemical composition and aggregation structure of windmill palm fibres, the Grey System theory, least-squares method and MATLAB® software were used to develop a grey model that relates the crystallinity index to the cellulose, hemicellulose and lignin content of the fibres after alkaline peroxide bleaching. The mean arithmetic error (=0.109) of the grey model (GM (1,4)) indicated that the model could predict the crystallinity of windmill palm fibre based on its chemical composition. Scanning electron microscopy, Fourier transform infrared spectroscopy and Raman microscopy were used to characterize the morphology and chemical composition of the windmill palm fibres before and after alkaline peroxide bleaching. The results indicated that silica, as well as most of the pectin, hemicellulose and lignin, were removed after peroxide bleaching, and thus, the crystallinity index of the peroxide-bleached samples increased. The GM (1,4) provided a reference for studying the influence of the chemical composition of windmill palm fibre on its aggregate structure.