Interpreting chemiluminescence signals and ratios in hydrogenated biogas flames in a micro gas turbine combustor

Abstract

Biogas is a renewable and infrastructure-compatible fuel for distributed power generation, but its high CO₂ content lowers heating value and flame speed, complicating flame stabilization and narrowing the flammability range. Hydrogen enrichment widens this range, accelerates ignition, and promotes full oxidation, enabling efficient, low-CO operation. This experimental study investigates hydrogen-enriched biogas combustion in a stagnation-point reverse-flow (SPRF) combustor, a practical configuration featuring internal flue gas recirculation (iFGR). Experiments were performed at 18 kW thermal power under various operating conditions in terms of biogas composition (0–40% CO₂ by v/v.), hydrogen fraction (0–50% by v/v.), air-inlet diameters (40 and 24 mm), air preheating (500–900 K), and equivalence ratio (0.4–0.6). The dataset comprises 108 test cases combining dual-camera, two-line imaging of OH*, CH*, C₂*, and broadband CO₂* (100 Hz) with exhaust CO/NO measurements. Results show that OH* intensity increases monotonically with H₂ and serves as a robust tracer of reactivity, while CH* and C₂* exhibit non-monotonic behaviour due to competing thermal and carbon-depletion effects. The OH*/CH* and OH*/ C₂* ratios track H₂ enrichment in the wider-inlet geometry, whereas enhanced mixing in the compact inlet compresses their dynamic range. Comparison with published premixed and diffusion flame studies indicates that OH* intensity and the OH*/CH* ratio exhibit consistent trends, whereas other chemiluminescence signals and ratios are strongly configuration-dependent and require validation for each specific combustor geometry. Flame hysteresis occurs between attached and detached modes: the attached regime yields higher CO and NO due to incomplete oxidation, while the detached regime achieves low-CO operation through intensified air–fuel–recirculation interactions. Finally, Gaussian Process Regression (GPR) models trained on chemiluminescence intensities accurately predict emissions, illustrating the potential of virtual sensing for hydrogen-biogas combustors.