Py-GC/FID Screening Framework for Coal–Biomass–Solid Recovered Fuel Blends for Co-Firing: Additivity and Surrogate-Assisted Screening for Indonesian Fuels

Screening coal–biomass–SRF blends requires a fast testing method. This study assessed whether Curie-point pyrolysis with gas chromatography and flame ionisation detection (Py-GC/FID) can detect non-additive volatile release in blends of an Indonesian sub-bituminous coal with sawdust, rice husk, and SRF BBJP. Twelve blends (20–80 wt.% non-coal) were pyrolysed at 590 °C, and a Shape Deviation Index (SDI, 0–2) was used to quantify the dissimilarity between area-normalised chromatograms. For blends containing sawdust, early peaks outside the 15.68 s analysis window are excluded, so their SDI is not directly comparable to that of blends without sawdust. Three variants are distinguished: SDI_rep, between duplicate runs (noise; median 0.89); SDI_add, the residual from the additive expectation with the coal profile estimated from the blends (1.00); and SDI_ref, the deviation with an external coal reference (1.59, largely reference mismatch). SDI_add exceeded SDI_rep in five of nine blends by a median of 0.05. With duplicate runs, no departure from additivity exceeding run-to-run variation was detected; interactions below about 0.3 in SDI cannot be ruled out. A surrogate with a genetic-algorithm search, trained on SDI_ref, described the measured shape space (highest near 0.8 coal with BBJP; not a recommendation). SDI captures shape, not yield. Screening needs replicated runs and retention-time alignment.