ObjectivesImproving the operational flexibility of coal-fired units is a major requirement for building a new power system, which puts forward higher requirements for stable combustion of coal-fired boilers at ultra-low loads.MethodsTaking a 660 MW ultra-supercritical swirling opposed-firing pulverized coal boiler as the research object, the concept of local critical heat load of the boiler is put forward. Using numerical simulation method, the combustion stability of the boiler under different loads is analyzed with the volume-averaged temperature of the burner region as the index of local regional heat load. Based on this, the stable combustion strategy of increasing the power of a single burner under ultra-low loads is put forward, and the local critical heat load of the boiler is obtained.ResultsUnder the conventional operation mode, when the load is reduced from 25% to 20%, combustion instability occurs in the boiler. At this time, the volume-averaged temperature drop in the main combustion area and the local heat load area behind the burner nozzle are 415 K and 174 K respectively which are much higher than 31.5 K and 30.5 K in the process of reducing from 30%Pe (Pe is the rated load) to 25%Pe. After adopting the steady combustion strategy of increasing the heat load of a single burner, the volume-averaged temperature in the local heat load area under 20%Pe is raised to the level under 30%Pe, and the combustion stability of the boiler is obviously improved.ConclusionAt 20%Pe, the local critical heat load to ensure the stable combustion of the studied boiler is that the operating power of a single burner reaches 60% of its rated power. The research results can provide an effective reference for the deep peak regulation and ultra-low load stable combustion of coal-fired units.

The quantification of components in mixed solid wastes is fundamental to waste-quality assessment and to informed decision-making for downstream utilization. Here, we develop a non-destructive quantitative framework that couples near-infrared hyperspectral imaging with machine learning for a five-component system comprising HDPE, PP, PS, pine shavings, and corn stalks. Standard normal variate (SNV) preprocessing and successive projections algorithm (SPA) feature selection were used to construct feature spectral inputs, and five machine learning models (PLSR, XGBoost, SVR, RF, and 1D-CNN) were benchmarked using sample-level grouped cross-validation. Results demonstrated that the 1D-CNN delivered consistently strong performance across all components (R2 = 0.977-0.989, MAE = 1.966-2.662 wt%). Guided by analyses of component crosstalk, agreement bias assessment, and error patterns, we further identified confusing component pairs and low-concentration regimes, and introduced targeted augmentation using decoupling and low-content datasets to reinforce the model. Shapley additive explanations (SHAP) analysis established interpretable links between key bands and component competition. Furthermore, the reinforced model demonstrated excellent stability in independent tests across days and batches. In leave-combination-out extrapolation test using five-component formulations, the model maintained high accuracy, with R2 ranging from 0.898 to 0.990 and MAE from 1.036 to 3.119 wt%, demonstrating transferability to unknown recipes and practical engineering potential. This study provides an interpretable and scalable route to rapid quantitative analysis for precision blending of mixed solid waste.

The coal combustion process is essentially a series of complex chain reactions driven by free radicals, and the chemical structure has a significant impact on the coal combustion process. To reveal the mechanism of Persistent Free Radicals (PFRs) in coal during combustion, 60 kinds of representative Chinese coals of varying coal ranks were selected and characterized by electron paramagnetic resonance (EPR) to determine the g value, PFRs concentration, and FWHM. Furthermore, proximate analysis, Raman spectroscopy and thermogravimetric analysis were employed to elucidate the influence of coal quality and chemical structure parameters on PFRs characteristics, and to develop quadratic regression models for predicting combustion temperatures. Results indicate that highly coalified coal samples exhibit increased PFRs concentration, reduced g value and FWHM, associated with condensed, highly cross-linked aromatic structures and low heteroatom content. In contrast, low rank coals present the opposite trend, indicating a predominance of aliphatic chains and less condensed aromatic domains, resulting in more reactive but less stable radicals. Combustion modeling achieved high predictive accuracy (R-2 > 0.85), the model revealed that PFRs concentration controlling the ignition process (T-i) while the activity of free radicals governing the peak (T-m) and burnout (T-b) combustion stage.

Biomass-derived carbon aerogels are attractive carbon supports, yet their structural evolution under coking environments remains poorly understood. Utilizing a waste-pomelo-peel derived Ni/carbon aerogel (Ni/CA) as a model system, this work investigates the co-evolution of deposited carbon species and the resulting reconstruction of the carbon framework during ethanol cracking to elucidate dynamic structural transformations. Time-on-stream characterization combined with density functional theory (DFT) calculations revealed the coexistence of two distinct carbon growth patterns. After 6 h at 700 degrees C, the total carbon deposition rate reached 0.503 g & sdot; g- cat 1 & sdot; h-1 with filamentous carbon (52.2%) ultimately exceeding amorphous carbon (47.8%). Initially, disordered amorphous carbon preferentially filled intrinsic micropores, whereas filamentous structures became increasingly dominant over time. Filamentous carbon grew via a Ni catalyzed tip growth mode, inducing pronounced framework reconstruction. DFT calculations suggested that representative carbon-containing states and C-C bonded configurations can be stabilized on Ni/carbon interfacial sites, providing energetic support for the experimentally observed persistence of deposited carbon. More importantly, filament entanglement generated secondary meso-/macroporous voids that partially compensated for the loss of intrinsic microporosity within the examined reaction window. These results demonstrate that carbon deposition on Ni/CA is not a purely deactivating process, but a dynamic carbon-on-carbon reconstruction involving competition between pore-blocking amorphous coking and filament-induced pore reorganization. The work provides a mechanistic understanding from a carbon materials perspective regarding deposited carbon evolution on biomass-derived carbon aerogels and clarifies how concurrent carbon growth pathways reshape pore hierarchy and interfacial accessibility within the examined reaction window under carbon-rich reaction conditions.

In recent years, the large-scale integration of renewable energy has posed major challenges to grid stability. Enhancing the flexibility of coal-fired power units has become a key measure to address the issue of renewable energy consumption. To address the ramp rate limitations caused by the large inertia and delays of conventional pulverizing systems, this study proposes a small pulverized coal silo (SPCS) based system that combines the advantages of direct-fired and intermediate storage systems, improving coal feeding and reduction rates. The system was implemented in engineering on a 350 MW once-through boiler-turbine (OTBT) unit and its effectiveness was verified, achieving a load ramp rate of no less than 3.5% Pe/min within a wide load range for the first time. The operational characteristics of the new system were systematically analyzed, and mass conservation equations were established with incorporation of system delays and inertia times. A four-input three-output mechanistic model of an OTBT unit under the wide-load operation conditions was developed. To further optimize the model parameters, we propose a chaotic enhanced adaptive black-winged kite optimization (CEABKA) method and employ it for the model parameter identification. The result indicates that an overall MAPE error below 2.6% for the data-identified model was achieved. The model was validated using operational and step disturbance data in SPCS field test. The results showed that the MAPE was less than 4.5% and 1%, confirming its high accuracy under real conditions. The developed high-fidelity model can provides a reliable foundation for implementing advanced control in SPCS-integrated units.

In-situ CO2 adsorption-enhanced H2 production technology for the WGS reaction holds promising application potential in the fields of low-carbon development and clean energy production. This study innovatively developed engineered nanoscale Cu-CeO2-MgO adsorptive-catalytic bifunctional systems by regulating the sequence of Cu impregnation and MgO gelation. Through multi-dimensional characterization and reactivity tests, the catalytic and adsorptive performance of the composite systems were systematically investigated. The results indicate that the introduction of Cu and Ce components modulates the surface electronic environment of MgO aerogels, inducing the reconstruction of surface basic sites and accelerating carbonation chemisorption. The binary and ternary metal-oxide/oxide-oxide interfaces (Cu-Ce-O-Mg) within the composites induce lattice distortion in CeO2 and MgO, while regulating the oxidation states, dispersion, and redox behavior of CuxO species. They also facilitate interfacial electron transfer among the multiple components, as evidenced by modulated distribution and coordination environment of Cu/Ce species and altered infrared/ultraviolet absorption features. Within the temperature range of 330–410 °C, the composites exhibit a remarkable enhancement in H2 production: the H2 yields of Cu-CeO2/MgO and Cu/CeO2-MgO surpass those of the physically mixed Cu-CeO2 + MgO by 39.5% and 70.8%, respectively. The effective interfacial engineering of the composites simultaneously accelerates both CO conversion and CO2 capture, significantly enhancing the catalysis-adsorption synergy.

Bio-oil from renewable biomass pyrolysis holds great potential as a feedstock for upgrading into biofuels and value-added chemicals. However, its water-insoluble fraction presents challenges for upgrading due to its high molecular weight, compositional complexity, and tendency to coke under heat, often requiring harsh conditions and facing catalyst deactivation. Here, we report a mild and efficient electrochemical conversion system tailored for the bio-oil-water-insoluble fraction. By integrating methanol-assisted hydrogen-transfer and metal nanoparticle synergistic catalysis, we suppressed competitive reactions and promoted the diffusion and hydrogenation of the water-insoluble components, enabling their conversion into hydrocarbon fuels under ambient conditions. At a current density of 100 mA cm-2, the system achieved a peak Faradaic efficiency of 21.30% within the first 30 min, while hydrocarbons accounted for over 98% of the products detected by gas chromatography-mass spectrometer. The catalytic performance of the system remained stable within 7 cycles, and the recovery rate of the catalyst exceeded 94%. Combined experimental and theoretical insights elucidate the competitive reactions among the water-insoluble fraction. This work represents an important step toward the mild and efficient full-fraction conversion of bio-oil and its practical utilization.

Methanol steam reforming (MSR) is a promising reaction to produce H2, but conventional thermal catalytic approach is plagued by the high energy consumption and limited activity. Recently, photothermal catalysis offers an efficient way to drive MSR reaction under mild conditions. Despite the impressive progress attained in this emerging field, unraveling the underlying catalytic essence still remains challenges, since the ambiguity in the dynamic evolution characteristics of active sites. Herein, this study reveals a principal fact that the microenvironment of the Cu sites over Cu-MgO catalysts can be adequately modulated under irradiation of characteristic light, thus efficiently initiating the MSR reactions (H2 production of 184.1 mmol g-1 h-1). Detailed characterizations uncover the dynamic evolution of Cu active sites mediated by the photogenerated carriers and their polarization field. Such photoactivation effects contribute to the formation of unique Cu0-VO (oxygen vacancy) associates, which significantly enhance the electron feedback to the anti-bonding orbitals of reactants, thus promoting the adsorption/activation. Moreover, the MSR reaction kinetics are also significantly optimized with the benefit of photoactivation effects, where the C-H bond cleavage of intermediate CH3O* and O-H bond activation from H2O are improved, thus lowering the propensity for CO side-reaction. The findings provide new insights into the modification of active sites via photoactivation engineering.

Achieving rapid and accurate prediction of coal calorific value is of great significance for coal utilization. This study compiled three representative large-sample coal datasets: 6375 samples from the United States, 64,419 and 5323 samples from Chinese coal-fired power plants. Four typical machine learning algorithms were employed to develop lower heating value (LHV) prediction models based on proximate analysis parameters. Each algorithm was trained independently on each of the three datasets. All models achieved R2 values exceeding 0.95 on their respective training datasets. However, when evaluated on unseen datasets, model performance deteriorated markedly, revealing that conventional data-driven models suffer from dataset dependence and insufficient generalization capability. To address these limitations, this study proposes a transfer learning framework based on a fully connected neural network (FCNN) using a "Pre-training and Fine-tuning" approach. The effects of different transfer strategies on model performance were systematically investigated. The results indicate that the performance of the transfer model improves with increasing the amount of fine-tuning data, and 300 samples are recommended as the minimum reliable size for fine-tuning. At this sample size, the root mean square error (RMSE) is reduced by up to 45.10% compared to direct cross-sample dataset prediction without transfer learning. Fine-tuning samples that adequately represent the target domain further improve transfer performance. Furthermore, freezing one hidden layer is optimal, whereas freezing two layers hinders adaptation to the target domain. Finally, a DeepCoal system was developed for early warning and prediction of calorific value. Independent data tests reveal that, under the recommended transfer strategy, RMSE is reduced by approximately 32.66% and 11.67% compared to direct cross-sample dataset prediction and retraining, respectively. These results demonstrate that the proposed transfer learning framework enables knowledge transfer across datasets and improves cross-sample dataset applicability.

This study systematically investigates the nonlinear evolution of pore structure and its coupled mechanism with chemical structural changes of the char from the pyrolysis of six biomass and three coal samples across 350-1400 degrees C. The results revealed that the BET surface area (SBET) and pore volume (Vtotal) exhibit a distinct three-stage trend of "increase-decrease-increase" with increasing temperature, whereas the average pore diameter (Daverage) follows the opposite trend. The first inflection point (FIP) and second inflection point (SIP) are primarily observed at 600 degrees C and 1000 degrees C for biochar, whereas around 700 degrees C and 1100 degrees C for coal chars. FIP and SIP correlate closely with the sum of the raw sample's moisture and volatile contents, and the ratio of small to large aromatic rings of the char, respectively. In addition, in Stage I (350 degrees C-FIP), the increase in SBET is strongly governed by the decline in H/C atomic ratio, Raman parameter alpha and A(GR+VL+VR)/AD, indicating that volatile release and the breakdown of aliphatic structures dominate pore generation. In Stage II (FIP-SIP), the magnitude of SBET reduction correlates positively with A(GR+VL+VR)/AD, reflecting partial collapse of pores induced by carbon skeleton compression. In Stage III (SIP-1400 degrees C), SBET exhibits strong positive correlation with structural index A(GR+VL+VR)/AG & sdot;CRM, Mg and Ca volatilization disturbs carbon layers, creating inter-cluster dislocations, promoting new pores. Based on these findings, a three-stage mechanistic model was established. Furthermore, a unified pore structure prediction model across wide temperature range was constructed and validated, with SBET average absolute error of only 5.41 m2/g.

The precise detection of biochar characteristics serves as a critical determinant in both production process optimization and targeted application selection. In this study, interpretable machine learning prediction models based on Raman spectroscopy, including extreme gradient boosting, support vector regression, feedforward neural network, random forest, and ridge regression, were developed for accurately predicting the characteristics of biochar derived from six different biomass, across a pyrolysis temperature range of 350-1000 °C. Results demonstrated that the feedforward neural network achieved superior overall predictive performance for key biochar characteristics (R2 = 0.89-0.95), including fixed carbon, volatile, H, O, atomic ratio of H/C and O/C. Highly accurate prediction of ash (R2 = 0.95) was achieved by integrating the results of multibasic prediction of volatile matter and fixed carbon and establishing a quantitative relationship with ash. A tripartite analytical framework was developed to improve model interpretability by integrating CARS for spectral feature selection, SHAP analysis to quantify feature importance, and mechanistic correlation analysis of model predictions linking selected bands to biochar structure. The robustness of the models was evaluated through tests on various enhanced datasets, confirming their resilience under different perturbations. This approach, combining Raman spectroscopy with machine learning, offers a rapid and reliable means for predicting biochar characteristics, facilitating more efficient control of biomass pyrolysis processes, and supporting the development of online monitoring techniques.

The global energy transition necessitates a higher share of renewable energy sources, requiring coal-fired power plants (CFPPs) to provide flexible operation to compensate for renewable volatility. Conventional CFPPs are constrained by fuel supply inertia, hindering their ability to meet dynamic grid demands. This paper proposes a novel solution by coupling a Small Pulverized Coal Silo (SPCS) system between the direct-firing pulverizing system and the boiler, breaking the fuel supply bottleneck to provide flexible and rapid fuel feeding. A dynamic simulation model was developed, comprising a 350 MW boiler integrated with the SPCS system. The boiler subsystem within this coupled model was rigorously validated against actual plant data. The results demonstrate that the SPCS system significantly enhances the boiler load ramp rate, with the average ramp rate increasing from 1.26%Pe/min (without SPCS) to 3.23%Pe/min (with SPCS) when supplying pulverized coal at 7 kg/s within 2 min during the 50%-75% load range of the boiler. Moreover, under an identical SPCS coal feeding strategy, the enhancement in load ramp rate was more pronounced in the lower load range (50%-75%) compared to the higher load range (75%-100%). Furthermore, to ensure the energy-mass balance during the rapid fuel feeding process enabled by the SPCS, a regulation strategy based on synchronized feedwater compensation is proposed. The modulating effect of the feedwater-to-fuel compensation ratio (WFRSPCS) on the boiler's dynamic characteristics was investigated. Specifically for the investigated 350 MW supercritical boiler during the 50%-75% load ramp-up process, an optimal ratio of WFRSPCS = 5 was identified. This optimal value successfully maintains the main steam temperature stability while simultaneously yielding an additional main steam flow increment of approximately 25.4 kg/s. Consequently, this strategy achieves a coordinated enhancement in both the load ramp rate and the thermal stability of the boiler unit.

Amid the global low-carbon transformation and increasing hydrogen demand, sorption-enhanced steam gasification (SESG) of biomass has emerged as a promising technology for sustainable hydrogen production. However, conventional Ca-based materials suffer from limitations including poor cyclic stability and challenges in separating fine powders from coke and ash. In this study, Ca-based bifunctional (sorption-catalytic) pellets with a designed diameter of 2.5 mm, synthesized by sol-gel and graphite casting methods, effectively resolved the separation challenges of bifunctional materials from coke and ash during in situ adsorption catalysis through structural design. The catalytic performance under varying operating conditions was systematically evaluated. Mechanistic analysis revealed that Ce doping lowers the activation energy for surface chemical adsorption via grain refinement and electronic modulation, while promoting the redox cycling of oxygen species and enhancing the mobility of oxygen species. Al improves structural stability, while Ni and Co synergistically enhance catalytic activity in tar cracking and water-gas shift reactions. Ca-based bifunctional pellets were innovatively applied to the SESG reaction of biomass, achieving hydrogen yield of 711.5 mL/g and hydrogen concentration of 88.19 vol %, demonstrating competitiveness comparable to that of powder catalysts. The bifunctional pellets exhibited excellent catalytic cyclic stability, with the hydrogen concentration remaining 1.7 times that of pure CaO and a decay rate of only 8.18 % after 10 cycles. Importantly, the pellets demonstrated outstanding mechanical strength and separability, with negligible mass loss during separation. This work provides new insights into the design of high-performance Ca-based materials and process optimization strategies for efficient hydrogen production from biomass via SESG.

Coal-fired boilers continue to face the simultaneous challenges of enhancing efficiency, reducing emissions, and maintaining safe operation. This study proposes an integrated multi-objective optimization framework that combines real-time detection, accurate and interpretable predictive modeling, and economically oriented optimization. An online detection system was deployed to collect key operational parameters, including coal property, carbon content in fly ash, and CO concentration. Support vector regression (SVR) models were constructed to predict boiler efficiency, NOx, and H2S. SHAP (SHapley Additive exPlanations) was used to enhance model interpretability and identify effective optimization variables. A novel benefit-oriented evaluation index was further proposed to economically quantify optimization outcomes and assess the influence of market price. Results indicate a clear linear correlation between H2S and CO, enabling the acquisition of real-time H2S data, and thereby allowing H2S to be incorporated into the optimization framework. The SVR models maintained high predictive accuracy under complex boiler conditions, with relative errors of 0.077% for efficiency and 0.59% for NOx. SHAP identified power, volatiles, and total overfire air flow as the most impactful features for predicting efficiency, NOx, and H2S, respectively. The direct optimization of the benefit function achieved a notably higher overall economic return compared with the approach of conducting multi-objective optimization followed by benefit evaluation. Under high coal price conditions, even moderate improvements in efficiency can lead to considerable cost savings. This strengthens the economic rationale for adopting advanced combustion optimization and emission reduction technologies, providing important guidance for the operation of coal-fired power plants.

Selective recovery of phenolics from waste wind-turbine blade epoxy is limited by uncontrolled secondary reactions during pyrolysis. Here, choline chloride–ethylene glycol deep eutectic solvent (DES) swelling was coupled with iron(III) citrate-assisted Fe introduction. Final temperature (400–600 °C) and Fe level were varied to resolve changes in thermal behavior, phase yields, and normalized gas chromatography–mass spectrometry peak areas. Pretreatment retained 81.16–88.81% of the resin, and inductively coupled plasma optical emission spectrometry measured 0.629–1.406 wt% Fe in Fe-containing composites. The main decomposition region shifted to lower temperature, while the full width at half maximum of the main DTG peak narrowed from 42.51 to 35.48 °C. At 500 °C, Fe0.45 (0.45 mol L−1 Fe3+ in the treatment solution; 0.750 wt% Fe in the composite) gave a C9-substituted-phenol normalized peak area of 57.70%, 35.26% higher than that of Raw. These products represented 63.65% of total phenolics. Higher Fe levels or a final temperature of 600 °C increased gas formation and reduced oil recovery, consistent with stronger secondary conversion. Product-distribution, Raman, and textural trends indicate that Fe0.45 at 500 °C provided the most favorable observed balance between precursor conversion and downstream dealkylation and condensation. Coupling DES swelling with controlled Fe introduction therefore offers a tunable route for the relative enrichment of C9-substituted phenols during the thermochemical valorization of waste wind-turbine blade epoxy.

Co-combustion of ammonia (NH3) and coal is a key strategy for reducing carbon emissions in coal-fired power plants. However, its impact on the coal ignition, combustion rate, and char structure remain unclear. In this study, OH planar laser-induced fluorescence (PLIF) was applied to capture the spatial-temporal evolution of OH radicals. Char structure characterization and density functional theory (DFT) analysis were conducted to reveal the effects of NH3 on char structural evolution and oxidation energy barriers. The results indicate that as the NH3 blending ratio increased, the surface temperature of coal particles gradually decreased. Meanwhile, the initial appearance of OH radicals was delayed, the fluorescence signal intensity decreased, and the volatile combustion rate declined. These findings suggest that NH3 suppresses coal particle ignition and subsequent combustion. NHx radicals interact with oxygen-containing functional groups, leading to the formation of carbon-nitrogen triple bonds, thereby enhancing the thermal stability of the char. This structural change suppresses the early release of CO, and contributes to a delay in coal ignition. Raman results indicate that NH3 inhibits char aromatization by interfering with the growth of aromatic layers and introducing nitrogen-containing structures, which is consistent with the weakened combustion behavior. Furthermore, the presence of NH during co-combustion increased the energy barriers for coal conversion to CO and CO2 by 43.32 kJ center dot mol- 1 and 4.62 kJ center dot mol- 1, respectively, indicating that NH adsorption inhibits the surface oxidation reactions of char and thus reduces its reactivity. Meanwhile, NH3 weakens the gas-phase oxidation capacity by consuming OH radicals and lowering the flame temperature. As a result, the rate of OH-related reactions, such as H center dot + O2 -> center dot OH + O center dot, are reduced, and the synergistic inhibition of both gas-phase and surface reactions ultimately leads to a decrease in overall combustion intensity.

In this work, we innovatively developed a Cu-CeO2-MgO adsorptive-catalytic bifunctional system for in situ CO2 capture in the WGS reaction, to achieve efficient simultaneous hydrogen production and carbon fixation. Multi-scale in situ spectroscopic techniques were employed to track the dynamic evolution of surface active sites, electronic structures, and adsorbed intermediate species during the reaction, to elucidate the synergistic mechanism of spatially adjacent catalytic and adsorption sites regulated by metal–oxide interfaces. The results demonstrate that a MgO aerogel shell is conformally coated on the CeO2 surface, forming multiple heterointerface configurations (CuOMg and CeOMg). These heterointerfaces enable highly dispersed anchoring of Cu clusters, and inhibit excessive surface segregation and agglomeration of Cu species via the confinement effect. In the early stage, hydroxylation of low-coordination basic sites on the MgO surface accelerates CO2 adsorption and H2O dissociation. Meanwhile, the CuOCe interfacial interaction drives the rapid formation of interfacial active oxygen vacancies (Cu+-Ov-Ce3+), and the CeOMg interfacial effect further promotes the controllable interfacial reduction of Ce4+ and local lattice distortion. Unlike the Cu-CeO2 catalyst that follows a dominant redox reaction pathway, the composite enables and enhances a formate-dominant reaction pathway through heterointerfaces and bimetal-support interactions. The interfacial synergistic electron transfer effect optimizes the coordination environment and valence distribution of Cu species, promoting the deep reduction of Cu2+ while stabilizing the Cu+ active centers. In the middle and late stages, surface carbonate species gradually transform from weakly bound monodentate configurations to highly thermally stable multidentate forms. The structural stability of the MgO lattice originates from the kinetic limitation of surface carbonation rather than adsorption site saturation. The sustained CO2 capture effectively alleviates the electronic competition effect of the product CO2, inhibits excessive consumption of lattice oxygen and reoxidation of low-valent metal species, and thus maintains the long-range crystal structure stability and interfacial activity of the composites.

This study developed a Raman-mapping method to visualize individual char particles during oxidation, tracking the spatiotemporal evolution of their chemical structure. 3 series of experiments were conducted: (i) micro-scale mapping of 1000 degrees C chars before oxidation to quantify inherent heterogeneity of the char particles, (ii) Raman mapping at room temperature after stepwise oxidation to resolve the progressive structural evolution during the char oxidation, and (iii) high-temperature in situ Raman spectroscopy of single char particle and char particle clusters to follow real-time carbon framework evolution and functional group formation during the oxidation sequence. The results show that Hongshaquan coal char pyrolyzed at 1000 degrees C exhibits intra- and interparticle structural variation, with individual heterogeneity peaking when the particle's average structure approaches that of the bulk char. Besides, oxidation reactivity of the char correlates more strongly with char surface heterogeneity than with average structural parameters. During the char oxidation, its structures proceed in 3 stages: (i) O2 attacks side chains and small aromatic rings, cleaving large aromatic rings and raising the fraction of oxygen-containing functional groups and small aromatic rings. (ii) These species accumulate to a peak and then deplete. (iii) A mineral-rich ash shell forms in the later stage of the oxidation, hindering oxygen diffusion and slowing down the oxidation of residual carbon. Single-particle oxidation results can capture the complete oxidation process and its chemical structure evolution, whereas particle clusters burn layer-by-layer, easily overlooking the early activation process. This study can provide a deep understanding on the oxidation mechanism of char single-particles.