Abstract: Raman spectroscopy is a rapid and non-destructive tool for carbon materials; its high sensitivity to structural order makes it ideal for coal characterization. To address the complexity of coal structure, optimized peak deconvolution is employed to reveal structural evolution and to support low-carbon utilization and standard development. Divergent interpretations of coal structure have led to variable peak numbers and functions in previous work, compromising data comparability and reliability. Here, the optimal peak number and function were identified by the coefficient of determination (R²) and literature consensus. Gaussian, Lorentzian, Gaussian-Lorentzian, Voigtian and Pearson Ⅶ functions were systematically compared for the first-order region (1000–1800cm⁻¹). Gaussian–Lorentzian and Voigtian functions yield R² 0.9964 for both symmetric and asymmetric peaks; Gaussian outperforms Lorentzian for symmetric peaks, whereas the reverse holds for asymmetric peaks. Optimal strategies vary with metamorphic grade sample: coal-measure graphite require three peaks (D1, G, D2), high-rank coals four peaks (D4, D1, D3, G), both best fitted by Gaussian–Lorentzian or Voigtian functions (R²＞0.9979); high-ash coals (low-rank coal) need four peaks and are best described by Pearson Ⅶ (R²＞0.9983); pyrolysed coals (≤900°C) require ten peaks and are optimally fitted by Voigtian or Gaussian–Lorentzian (R²＞0.9967). This study demonstrates that function choice depends on peak symmetry, while coal complexity dictates the optimal deconvolution strategy, providing a quantitative protocol for Raman peak fitting of coal.