[Objective] With increasing renewable penetration, thermal power units are shifting from baseload generation to flexible regulation. For integrated energy parks containing coal -fired units, deep peak -regulation capability is essential for renewable accommodation and low-carbon operation. Oxy-fuel combustion power plants (OCPPs) offer advantages in low-load stable combustion and CO₂ capture, but their available downward regulation is constrained by the power consumption, ramping limits, and start –stop characteristics of the air separat ion unit (ASU), together with the boiler’s real-time oxygen demand. Assuming instantaneous oxygen supply may therefore overestimate feasible regulation capacity. This paper proposes a flexible peak -regulation dispatch method for an 2 oxy-fuel-combustion integrated energy system considering dynamic oxygen-supply constraints and liquid oxygen energy storage (LOES) buffering. 

[Methods] An electricity–oxygen–hydrogen–carbon coordinated dispatch model is established, explicitly representing OCPP operation, ASU oxy gen production, power -to-hydrogen (P2H) hydrogen production and by-product oxygen generation, LOES oxygen charging and discharging, hydrogen storage and utilization, and CO₂ capture. During deep peak-regulation periods with surplus renewable generation, additional electricity consumption is allocated to ASU, P2H, and LOES charging, converting surplus electricity into storable oxygen and hydrogen material flows. These materials are released in subsequent periods, thereby temporally decoupling oxygen and hydrogen production from their consumption. LOES is placed between ASU production and OCPP demand as an oxygen buffer, reducing the dependence of low-load OCPP operation on real-time ASU output. A grid–park bilevel optimization framework with iterative response feedback is further developed: the upper-level grid adjusts the dynamic peak-regulation compensation price according to regulation demand, while the lower-level park coordinates unit output, ASU, P2H, LOES, demand response, and CO₂ capture to minimize net operating cost and release feasible downward regulation capacity. 

[Results] Case studies under the specified parameters and operating conditions show that the proposed method improves deep peak-regulation performance while preserving oxygen-supply feasib ility. Compared with the traditional operating mode, the cumulative deep peak -regulation response reaches 596.8 MWh, an increase of 47.6%. Daily operating cost decreases by 15.1%, while direct carbon emissions, calculated from coal and natural -gas combustion net of captured and stored CO ₂, decrease by 77.7%. The results confirm that coordinated ASU –P2H–LOES operation transfers flexible electricity consumption to renewable-surplus periods and uses stored oxygen and hydrogen across time, supporting continuous oxygen supply and CO₂ capture. 

[Conclusion] Oxygen and hydrogen material-flow buffering mitigates the restriction imposed by dynamic ASU oxygen supply on OCPP downward regulation. The proposed method strengthens low -load stable combustion, continuous carb on capture, and low -carbon flexible operation in thermal -power integrated energy parks.

[Objective] With the comprehensive implementation of China ’s dual carbon goals and the full -scale construction of modern new power systems, Hainan ’s isolated island power grid has maintained a clean energy installed penetration rate steadily above 87.1%. Large -scale grid integration of onshore and offshore wind power, as well as centralized and distributed photovoltaic power, brings about prominent system challenges: insufficient peak shaving and frequency regulation capacity, continuous decline in system equivalent inerti a, volatile power flow and severe voltage fluctuations. These issues create major hidden risks to the secure, stable and economical operation of the provincial power grid. Restricted by local policies that ban the construction of new coal -fired generating units, existing in-service coal-fired units serve as the only multi-functional supporting power source for Hainan Power Grid. They can provide essential ancillary services including deep peak regulation, fast primary and secondary frequency modulation, dyn amic reactive power and voltage support, emergency spinning reserve, and system black -start capability. Against the backdrop of ultra -high renewable energy penetration, this paper systematically explores feasible flexible retrofitting modes for operational coal-fired units, defines their multi dimensional core functional positioning and quantifiable technical indicators, and constructs a complete low-carbon flexible transformation framework together with targeted grid security support strategies tailored to the unique operating characteristics of isolated island power grids. 

[Methods] This study sorts out Hainan ’s current power supply mix, seasonal and diurnal load variation patterns, main grid structure and annual peak-valley balance status. On this basis, it clarifies the technical requirements for retrofitted units and proposes six transformation pathways: thermoelectric decoupling, deep peak -load retrofitting, energy storage coupling, intelligent digital operation upgrading, complementary coordination wit h new energy stations, and low -carbon emission reduction reconstruction. A multi-layer collaborative control strategy is established to mitigate grid operational risks, and the economic performance of retrofitted thermal units is quantitatively analyzed un der the rules of the provincial electricity spot market and auxiliary service market. 

[Results] Statistics indicate that Hainan’s total social electricity consumption reached 54.473 billion kW·h in 2025, with a year-on-year growth of 4.97%. Forecasts show the grid’s maximum peak shaving deficit will expand to 1.83 million kW by 2027. Comprehensive flexible retrofits can lower the minimum technical output of coal-fired units to below 25% of their rated capacity. 

[Conclusion] When combined with capacity compensation p ricing and standardized auxiliary service trading mechanisms, renovated thermal power assets can achieve coordinated optimization of grid safety operation, low -carbon energy transition and sustainable economic profitability. The research conclusions and tr ansformation pathways provide valuable engineering references for new power system construction in other isolated island power grids with high renewable energy penetration worldwide.

[Objective] To overcome the carbon penalty of limestone‑gypsum desulfurization in coal‑fired power plants and the poor utilization of dilute SO2, this study proposes an integrated technology combining ammonia‑based absorption with SO 2‑depolarized electrolysis (SDE) for simultaneous flue gas desulfurization and hydrogen production. The goal is to achieve zero‑carbon‑increase desulfurization, low‑energy hydrogen generation, high‑value by‑product recovery, and flexibl e load regulation to support next‑generation coal power upgrading. 

[Methods] A laboratory‑scale coupled system consisting of an ammonia absorption tower and a membrane electrolyzer was constructed. The anodic performance of an acidic sulfate system (0.5 mol/L H 2SO4) and an ammoniacal system (1% ammonia water) was compared under constant‑voltage electrolysis at 1.2 V and 1.4 V 2 using real flue gas ( 2 290~11 400 mg/m3 SO2). Current efficiency and hydrogen production energy consumption were calculated via F araday’s law. Density functional theory (DFT) calculations were performed to reveal the reaction mechanism, and a techno‑economic assessment was carried out for a 400 000 m3/h flue gas treatment scale. 

[Results] In the acidic sulfate system, physical dissolution of SO2 is severely limited by Henry’s law; at 1.2 V the current density was only 0.4 mA/cm2, and voltage surged rapidly due to reactant depletion, indicating an insurmountable mass‑transfer bottleneck. In contrast, the ammoniacal system chemicall y absorbs SO 2 to form ammonium sulfite, greatly increasing the concentration of electroactive species. Constant‑voltage electrolysis in the ammoniacal system at 1.2 V gave an average current of 0.298 A, a current efficiency of 76.9%, and a hydrogen energy consumption of 3.45 kW·h/m3; at 1.4 V , the average current increased to 0.768 A, the current efficiency rose to 87.5%, and the energy consumption was 3.83 kW·h/m3. The higher current efficiency at 1.4 V , despite being above the theoretical water‑spl itting voltage, indicates that the increased overpotential accelerates kinetics and enhances bubble‑induced convective mass transfer, compensating for minor oxygen evolution. DFT calculations show that the activation energy for sulfite oxidation in the ammoniacal system is 0.62 eV , which is 32% lower than that in the acidic system (0.91 eV), owing to stronger chemisorption of sulfite on graphite. For the 400 000 m3/h scale, the system achieves a desulfurization efficiency ≥98.5%, producing 7.2 million m3 of hydrogen and 22 000 t of ammonium sulfate annually. The annual net profit is estimated at 20.6~31.5 million yuan with a payback period of 3~4 years, based on current market prices. Sensitivity analysis shows that hydrogen and ammonium sulfate prices are the most influential factors, while the electrolyzer’s fast load response (0~100% within 5 s) can participate in grid peak‑shaving, yielding additional revenue. 

[Conclusion] The ammonia‑absorption‑ coupled SDE technology overcomes the mass‑transfer and kinetic limitations of dilute SO 2 electrolysis by combining chemical absorption with a low‑activation‑energy anodic reaction. It enables zero‑carbon‑increase desulfurization, low‑energy hydrogen production (about 20% lower than conventional water electrolysis), and high‑value fertilizer co‑production, while the flexible electrolysis load supports deep peak shaving and low‑load combustion stability via hydrogen co‑firing. This integrated approach offers a technically viable and economically attractive pathway for the clean and flexible transformation of coal‑fired power generation under China’s carbon‑neutral strategy.

[Objective] To address the cross-machine redundant fault-tolerance problem of virtual DCS controllers in thermal power plants under Hypervisor architecture, a VM -level cross -machine redundancy synchronization mechanism based on SR-IOV virtual links is proposed. In virtualized environments, controllers running as virtual machines on general-purpose servers must synchronize their control states across physical machines and Hypervisor layers, facing the challenges of virtualized I/O performance degradation, real-time master-slave role negotiation and fault detection, and bumpless control output transfer during switchover. 

[Methods] A VM -level paired cross machine redundancy architecture is designed, in which each VM is equipped with at least two VF redundant communication interfaces through SR-IOV passthrough technology. The VFs are bound to different physical ports of the 10 GbE network controller to achieve full-path physical redundancy from the VF level through the physical port level to the entire server lev el. A dual-task redundancy management protocol combining heartbeat and data synchronization is proposed. The heartbeat task runs as an independent thread and exchanges lightweight 64 -byte messages via dual A/B VFs at a 15 ms period to achieve millisecond -level fault detection and master -slave role 2 management. The consecutive loss threshold of N equals 4 is selected to tolerate one to two transient packet losses commonly observed on industrial Ethernet while maintaining fast detection speed. The data synchronization task is initiated by the primary VM and passively received by the standby VM through a dedicated server thread. IO data are fully synchronized through a dedicated thread, while FBD intermediate variables follow their respective logic execution task cycles ranging from 5 0 ms to 1 000 ms using a macro -level comprehensive yet micro-level tiered synchronization strategy—all intermediate variables are covered in their respective task cycles at the macro level, while the transmission is distributed acros s different time slots at the micro level to avoid instantaneous network congestion, thereby reducing master-standby computation deviation. 

[Results] A prototype system was built on the Phytium S5000C ARM processor and an embedded Type-1 Hypervisor platform, and experimentally verified using the actual control logic of a 600 MW ultra -supercritical thermal power unit as the workload. Under four fault injection scenarios —primary VM forced power -off (simulating server power supply failure), simultaneous disconnection of both A/B redundant link ports (simulating optical fiber interruption), vDCS runtime process abnormal exit, and Host single network disconnection —the master-standby switchover time is approximately 60 ms. The IO data end-to-end synchronization latency is approximately 85 μs for 2 KB of data transmitted in a 15 ms cycle. The FBD intermediate variable synchronization latency ranges from 76 μs to 185 μs, corresponding to data volumes from 0.2 KB to 18.6 KB, all substantially below their respective control task cycles of 5 ms to 1 000 ms. During the switchover process, the control output remains continuous and smooth, with no visible glitches on SCADA trend curves. A sinusoidal test signal generat ed by an FBD function block was monitored through the SCADA trend display, and no amplitude or phase discontinuity was observed at the switchover moment, providing intuitive visual verification of bumpless transfer. The FBD master -standby computation devia tion is controlled within a single task cycle. 

[Conclusions] The proposed mechanism can provide reliable fault-tolerant capability for virtual DCS controllers, offering engineering reference value for promoting the reliability and engineering application of domestically-produced autonomous controllable virtual DCS technology.

[Objective] To meet the deep peak regulation requirements of the unit, the steam flow state and the blast heating characteristics under the condition of small volume and low flow rate after the zero-load transformation of the low-pressure cylinder are analyzed, and related research is carried out. 

[Methods] Taking a 320 MW unit steam turbine as the research object, the MSCAN global photogrammetry system is used to establish a three-dimensional 2 flow field model of the single-channel of the lower cylinder with 7 stages of single blades on one side. Eight sets of working conditions with gradually varying inlet total pressure are set. The H-O-H structured grid, CFX steady-state calculation, IAPWSIF97 water vapor model and SST k-ω turbulence model are adopted for calculation. The working performance, flow structure and forced heating characteristics of the lower cylinder under different flow rates are systematically analyzed. 

[Results] The numerical calculation results show that when the steam flow entering the low-pressure cylinder drops to 7.19% of the rated condition, the last stage begins to output negative power. When the flow continues to decrease to 2.91% o f the rated condition, the low -pressure cylinder stops outputting power and enters a zero -output state. At this time, the flow entering the low -pressure cylinder is extremely low, and the low-pressure cylinder is in a forced heating state. The steam flow p attern in the last stage blade passage includes the lower end wall separation zone, the stator blade separation zone, the last stage rotor blade vortex, and the casing circumferential vortex, etc. At the same time, the position of the top of the rotating a nd stationary gap of the last stage blade shows a strong forced heating phenomenon. The forced heating effect significantly increases with the decrease in flow. When the steam flow drops to 1.05% of the rated condition, the average temperature of the last stage stator blade surface is 312.75 K higher than the rated condition, and the average temperature of the last stage rotor blade surface is 331.88 K higher than the rated condition. The maximum temperatures on the blade surfaces of the last rotor and stat or blades reach 761.58 K and 698.99 K. 

[Conclusion] After the steam turbine enters the zero-output condition of the low -pressure cylinder, the working performance of the low -pressure cylinder significantly deteriorates, the temperature in the final stage b lade passage rises noticeably, and the working environment of the blades becomes severely deteriorated. The rapid increase in blade surface temperature can easily cause thermal fatigue damage to the materials, thereby reducing the blade strength. Therefore, power plants should strictly control the deep load -following operation of the units. During the unit maintenance period, regular inspections of the blades should be conducted to enhance the operational safety of the units.

[Objective] Natural draft wet cooling towe rs (NDWCTs) are key cold -end components in 1000 MW power units, owing to their buoyancy -driven ventilation, low auxiliary power consumption, low maintenance demand, and high operational reliability. However, ambient crosswinds can significantly deteriorate their thermal performance by disturbing the originally axisymmetric airflow field, inducing local flow separation and recirculation, and weakening air-water heat and mass transfer within the tower. Conventional optimization strategies for fill arrangement mainly rely on globally symmetric non -uniform configurations, which are unable to fully accommodate the inherently asymmetric aerodynamic structure formed under crosswind conditions. To overcome this limitation, this study proposes a windward -sector fill arrangement optimization method based on three -zone non-uniform fill heights, aiming to actively reconstruct the asymmetric flow field and improve the c rosswind resistance of super -large NDWCTs. 

[Methods] A super -large NDWCT with a water -spraying area of 13 000 m² 2 serving a 1 000 MW power unit was investigated. A three-dimensional numerical model was established by solving the conservation equations of ma ss, momentum, energy, and species transport, with the standard k –ε model used for turbulence closure. The model reliability was verified by grid independence analysis and field measurements. A mesh with 2.86 million cells yielded an outlet water temperatur e deviation of 0.04%, and validation under four typical operating conditions showed relative errors of 2.42%~3.78%. Based on the original three-zone fill heights of 1.5 m, 1.75 m, and 2.0 m for the inner, middle, and outer zones, respectively, the fill sheet spacing in the windward sector was reduced from 30 mm to 26 mm. The effects of windward-sector angle θ from 0° to 180° and inner radius Rs from 0 to 59 m were then examined using the control variable method. 

[Results] Under the dominant crosswind speed of 3 m/s, the air flow rate G, water temperature drop Δt, and Merkel number N first increased and then decreased with increasing θ or decreasing R s. The optimal configuration was obtained at θ = 150° and R s = 19 m, where Δt, N, and G increased by 0.218°C, 0.14, and 75.01 kg/s, respectively, compared with the original design. Within crosswind speeds of 1 ~7 m/s, the optimized scheme consistently improved thermal performance, with Δt increasing by 0.233、0.221、0.218、0.215、0.213、0.192 and 0.163° C at the correspo nding wind speeds. The improvement gradually weakened as crossflow momentum became dominant. 

[Conclusion] The proposed windward-sector fill optimization method effectively mitigates crosswind -induced thermal performance degradation by targeting the asymmetric flow characteristics of NDWCTs. It provides a practical and cost-effective strategy for crosswind-resistant design and retrofit of super-large NDWCTs.

[Objective] With the increasing integration of intermittent renewable energy sources into the power grid, large-capacity thermal power units are frequently required to participate in deep peak -shaving operations. This operational transition leads to frequent and drastic c hanges in boiler load, which significantly intensifies the thermal stress on heating surfaces and increases the risk of boiler "four-tube" leakages. However, early-stage weak leakage diagnosis faces two critical challenges: feature distribution drift caused by variable-load conditions and an extreme scarcity of fault samples in industrial sites. Traditional deep learning-based diagnostic models, which rely on fixed classification heads, often suffer from feature manifold collapse and severe overfitting when transferred from base -load to deep peak -shaving scenarios. To address these issues, this paper proposes a cross -condition few-shot fault diagnosis framework named Proto-MixMamba, which integrates time-series multi-scale decoupling with prototypical metric learning. 

[Methods] The research utilizes high-precision simulation data from a 600 MW supercritical unit to construct comprehensive datasets covering various load conditions and leakage severities. To move beyond traditional single -scale sequence modelin g and parameter -solidified classification paradigms, a dual-branch backbone network comprising a One -Dimensional Position Attention Module (1D -PAM) and a Mamba State Space Model is established. Specifically, the 1D -PAM branch is designed to capture local w eak transient shocks and high -frequency fluctuations in thermal parameters, while the Mamba branch leverages its selective scanning mechanism to model global low -frequency evolution trends and long -term dependencies with linear computational complexity. Th ese multi -scale features are adaptively integrated through a gating fusion mechanism that balances the importance of local and global information. To resolve the feature confusion caused by amplitude baseline shifts during load variations, the framework in troduces an N -way K -shot episodic meta-learning strategy. A prototypical metric layer is employed to replace the conventional fully connected classifier. By calculating the relative Euclidean distance between query samples and class prototypes in a high-dimensional embedding space, the model constructs a robust decision boundary that is invariant to absolute parameter magnitudes. 

[Results] Extensive experimental evaluations demonstrate the superiority of the Proto-MixMamba model. The results indicate that t he proposed method effectively overcomes the long -standing industrial challenge of misjudging physically symmetrical leakage locations, such as the left and right economizers, which exhibit highly similar thermal responses under variable -load interference. In the challenging cross-condition diagnosis task transitioning from 600 MW to 480 MW (deep peak -shaving), and under the extreme constraint of the 1 -shot setting (only one reference sample per category), the proposed model achieves a diagnostic accuracy o f 88.4%. This performance significantly outperforms traditional baseline models such as Long Short-Term Memory (LSTM) and ResNet, whose accuracies drop below 50% due to severe overfitting and domain shift. Visualization via t -SNE confirms that the prototyp ical metric mechanism successfully maintains high intra -class compactness and inter -class separability in the target domain.

[Conclusion] The proposed Proto-MixMamba framework demonstrates exceptional robustness and generalization c apabilities in cross-condition few-shot scenarios. By successfully extracting invariant fault features and establishing a reliable metric space, this research provides a technically feasible and highly effective route for the accurate early warning and diagnosis of boiler "four-tube" leakages in label-scarce and operationally complex industrial environments.

[Objective] Multi-objective optimization of coal-fired boiler combustion is a critical strategy to achieve high-efficiency and environmentally friendly boiler operation. However, due to the strong nonlinearity, multivariable coupling, and inherent conflict between boiler efficiency and NO x emission, it is particularly challenging to identify optimal operating parameters using traditional trial -and-error approaches. 

[Methods] This study systematically builds prediction models for boiler exhaust flue gas temperature, carbon content in fly ash and NOx emission mass concentration based on the XGBoost model. Subsequently, the various heat losses are calculated, and a boiler efficiency model is further constructed using the anti -balance method. The g lobal optimization of XGBoost hyperparameters is performed using Bayesian optimization (BO) . To enable a comparative analysis, the corresponding BO-LSSVM and BO -BP models are also established . Based on the traditional NSGA -Ⅱ, the hypercube Latin sampling strategy and NDX crossover operator are introduced to improve the algorithm. 2 Collaborative optimization is carried out aiming at both enhancing boiler efficiency and lowering NO x emission mass concentration. The optimization decision variables include the furnace oxygen content, primary air pressure, secondary air damper openings and over -fire air damper openings . 

[Results] The results demonstrate that BO-XGBoost achieves a lower mean absolute error (EMAE) and a higher coefficient of determination (R²) on the test set compared to the other two models, which reveal that the BO-XGBoost model yields superior prediction accuracy and generalization performance. Relative to the traditional NSGA-Ⅱ, the improved NSGA-Ⅱ yields superior initial populations, reinforces global exploration capability, and effectively avoids the problem of local optima. As a result, the quality of Pareto solutions is significantly improved . The improved NSGA -II algorithm can generate many operating schemes superior to the original working conditions. The TOPSIS decision-making is adopted to select a trade-off solution from the Pareto solutions that balances boiler efficiency and NO x emission mass concentration. Statistical analysis over 200 operating conditions shows that the average boiler efficiencies for the original condition, the NSGA‑II and improved NSGA‑II trade-off solutions are 94.036 1%, 94.356 9%, and 94.506 8%, respectively, with the corresponding average NOx emission mass concentration of 173.13, 150.16, and 139.32 mg/m3. Accordingly, the trade-off solution obtained by the TOPSIS decision-making method successfully realizes the dual objectives of improving boiler efficiency and reducing NO x emission mass concentration. 

[Conclusion] This study establishes an integrated “prediction-optimization-decision” framework, which effectively combines BO-XGBoost prediction, improved NSGA -II optimization, and TOPSIS decision -making to achieve the collaborative optimization of complex combustion processes in coal -fired boilers. Moreover, t he proposed framework provides a systematic and practical solution for multi -objective optimization of boiler combustion systems.

[Objective] Object detection in intelligent surveillance systems at thermal power plants generates frequent false alarms in scenes with ambiguous visual boundaries, such as a partially occluded safety helmet or a small flame that is easily confused with refle cted light. Because every alarm has to be confirmed by hand, these false positives subject control-room operators to chronic alarm fatigue and gradually increase the risk that a genuine hazard is overlooked. This paper aims to solve this problem. 

[Methods] This paper proposes an alarm re-examination method that combines an object detector with a vision -language model (VLM) for semantic reasoning. Unlike a detector, which classifies an isolated bounding box, a VLM can weigh the surrounding scene against the wording of the relevant safety rule, which makes it better suited to the borderline cases that produce most false alarms. In the proposed pipeline, the detector preserves a high recall rate for initial screening so that few real hazards are missed, and the alarms it raises are then passed to the VLM, which performs semantic verification against power industry safety codes before any alarm reaches the operator. The study addresses four typical risk categories in plant operation: safety helmet wearing, work -uniform compliance, smoke, and open flame. For these categories, a progressive prompt scheme was developed in four stages—a keyword baseline, domain-specific definitions, chainof-thought reasoning, and multimodal exemplars—with each stage supplying the model with more task knowledge than the previous one. A domain-specific evaluation set of 1 068 alarm images was collected from the production site of a thermal power plant, and three domestic VLMs were compared on this set so that the influence of both the 2 prompt design and the model could be assessed. 

[Results] Under the optimal prompt configuration, review accuracy increased from a baseline of 46.73% to 81.66%, the F 1 value from 41.54% to 86.36%, and the alarm -filtering efficiency (AFE) from 0.242 to 0.631. At the operator terminal, the total alarm volume requiring handling decreased by 33.6%, and the terminal-side false-alarm ratio fell from 31.9% to 12.5%. The progressive prompt design accounts for most of this improvement, as accuracy rose stage by stag e from the keyword baseline, while the three models behaved consistently under the same prompts. 

[Conclusion] Overall, pairing a high-recall detector with VLM-based semantic verification effectively reduces false alarms in thermal power plant surveillance systems while preserving detection coverage; because it relies only on domestically available models and a modest amount of on -site data, the method offers a practical reference for similar industrial applications.

[Objective] Municipal solid waste (MSW) incineration exhibits pronounced time-varying thermal characteristics because of fluctuations in feed composition, moisture content, and operating conditions. Pure data-driven models may also suffer from error accumulation and tracking drift during continuous free-running prediction. This study aims to develop a grey-box furnace-temperature object model that combines physical interpretability with stable dynamic tracking performance under field operating conditions. 

[Methods] The grate bed was spatially discretized into a series of MSW elements, and the evolution of moisture, volatile matter, fixed carbon, and ash was described using mass-conservation and component-consumption relationships. A furnace-temperature mechanism model driven by operating commands, including primary-air distribution and grate velocity, was then established based on mass and energy conservation. To overcome the limitations of assuming a constant feed composition, an equivalent dynamic lower heating value (LHV) mechanism was introduced through element-state recursion and instantaneous spatial integration. The proposed equivalent dynamic LHV is not intended to directly estimate the actual LHV of the incoming waste; instead, it serves as an equivalent heat-source term that reflects changes in the overall composition state of the grate bed. Field DCS data were preprocessed using median absolute deviation (MAD)-based outlier detection and moving-average smoothing. A serial hybrid optimization procedure combining the Sparrow Search Algorithm and the Nelder–Mead simplex method (SSA-NM) was subsequently employed to identify eight lumped model parameters offline. SSA was used to search for a favorable parameter region over a wide solution space, whereas NM performed derivative-free local refinement using the SSA result as its initial point. 

[Results] The model was evaluated using 10 h of independent continuous DCS operating data collected after the parameter-identification period from a waste-to-energy plant. The validation interval included continuous variations in furnace temperature, air flow, flue-gas oxygen content, waste feeding rate, steam flow, and flue-gas flow, rather than a manually selected steady-state segment. The proposed model achieved a root mean square error (RMSE) of 11.91 °C and a mean absolute percentage error (MAPE) of 0.85%. Compared with the static-LHV model using empirical initial parameters, the RMSE was reduced by 54.15%. Compared with the static-LHV model after parameter identification, the RMSE was reduced by 41.13%. A conventional long short-term memory network was also established as a pure data-driven baseline under a closed-loop free-running prediction mode. The grey-box model exhibited lower prediction error and more stable trend tracking over the continuous validation interval. 

[Conclusion] The integration of an equivalent dynamic LHV mechanism with field-data-based parameter identification enables the proposed grey-box model to retain mass- and energy-conservation relationships while improving furnace-temperature tracking under time-varying operating conditions. The resulting model can provide an interpretable object-model basis for subsequent control-strategy development and operating-parameter optimization in MSW incineration systems.