Coal analysis plays a crucial role in its efficient utilization and pollution control. Laser-induced breakdown spectroscopy (LIBS) exhibits promising application prospects in coal quality analysis because of its uniquely rapid and minimally destructive capabilities. However, the complex composition of coal induces severe matrix effects, which pose a formidable challenge to the accurate quantification in traditional LIBS analysis. In this study, a novel method named spectrum-ultrasound-image multi-modal fusion model (SUI-MM) was introduced to mitigate matrix effects in coal and improve the quantitative performance. SUI-MM extends conventional LIBS to a homologous multi-modal analysis framework combining spectra, ultrasound, and image, effectively compensating for the absence of physical structural and plasma spatial morphological information in spectra. By establishing a three-dimensional feature extraction and fusion strategy, it enables comprehensive characterization of coal plasma behavior and further improves quantitative performance. To verify the effectiveness of SUI-MM, experiments were performed on elemental and proximate analysis of coal. For the elemental analysis, the average R2p is improved to above 0.998, while RMSEp and AREp are reduced by 88% and 90%, respectively. For the proximate analysis, the average R2p is increased to over 0.999, and both RMSEp and AREp are decreased by 91% on average. Furthermore, ablation experiments confirm that image and ultrasonic signals contribute significantly to the improved analytical performance. These results demonstrate that the SUI-MM scheme can effectively mitigate matrix effect and significantly improve the quantitative accuracy in coal. In summary, SUI-MM is expected to further support the low-carbon transition of the energy industry.