The calorific value of coal mainly depends on the content of combustible organic elements and ash, and is a comprehensive indicator of coal quality. It is of great significance to quickly obtain the calorific value of coal entering the boiler for the optimal control of coal blending and combustion in thermal power generation. However, the measurement repeatability of LIBS for the coal calorific value has not yet met the requirement of a standard deviation (SD) of less than 120 J g⁻¹ in the Chinese national standard. This paper innovatively proposes an analytical method of XRF-assisted LIBS, which can not only analyze the combustible organic light elements by LIBS, but also can analyze the inorganic ash-forming elements by XRF with high stability, thus greatly improving the measurement repeatability of the coal calorific value. According to the stability of intensity, the elemental emission lines in the LIBS and XRF spectra are selected reasonably, and the prediction model is established through principal component analysis and multiple regression. The experimental results show that the SD of coal calorific values predicted is 72 J g⁻¹, which is one order of magnitude lower than the 700 J g⁻¹ of traditional LIBS, and fully meets the requirement of the national standard. It is found that the measurement instability of the calorific value mainly comes from the contribution of the C line, and the SD of the calorific value can be further reduced to 56 J g⁻¹ by means of spectral screening. This XRF-assisted LIBS technology with ultra-repeatability is expected to be applied in power plants, coal chemical industries and other industrial fields that need to pay close attention to coal quality.