SCR (Selective Catalytic Reduction) catalyst performance online evaluation method based on operation data

Abstract

The invention provides an SCR catalyst performance online evaluation method based on operation data, and relates to the technical field of computers.The method comprises the steps that firstly, a time sequence data set in an operation cycle of a selective catalytic reduction reactor is obtained, and the time sequence data set comprises an inlet flue gas parameter sequence, an outlet flue gas parameter sequence and an ammonia spraying device control parameter sequence; an inlet reactant flux distribution coefficient and an outlet product generation rate distribution coefficient of each partition of the catalyst layer are decoupled and are used as boundary constraint conditions to be input into a catalyst surface microcosmic reaction site occupation state deduction model; generating a catalyst surface active site occupancy rate distribution diagram and a sediment coverage proportion distribution diagram through iterative convergence treatment, calculating an intrinsic activity retention proportion and a through-flow resistance increase proportion of each partition of the catalyst based on the distribution diagrams, and finally generating a catalyst performance degradation spatial distribution thermodynamic diagram. And outputting the residual efficiency index of the catalyst and the positioning information of the maintenance area to realize online accurate evaluation of the performance of the SCR catalyst.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to an online evaluation method for the performance of SCR catalysts based on operational data. Background Technology

[0002] In industrial production, selective catalytic reduction (SCR) technology is a key means of controlling nitrogen oxide (NOx) emissions and is widely used in many fields such as thermal power generation, steel smelting, and chemical production. The core of SCR technology lies in the catalyst, whose performance directly determines the denitrification efficiency as well as the operational stability and economy of the entire system.

[0003] Currently, the evaluation of SCR catalyst performance mainly relies on periodic offline testing and laboratory analysis methods. Offline testing requires removing the catalyst from the reactor and sending it to a specialized laboratory for a series of complex tests, such as chemical analysis and physical structure characterization. These methods are not only time-consuming and labor-intensive, requiring downtime and affecting production continuity, but also, due to limitations in sampling locations, cannot comprehensively and accurately reflect the overall performance of the catalyst under actual operating conditions. While laboratory analysis can provide more accurate catalyst performance data, it also suffers from high costs and long cycles, and cannot promptly capture dynamic changes in catalyst performance, failing to meet the needs of real-time monitoring and rapid evaluation of catalyst performance in industrial production. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an online evaluation method for SCR catalyst performance based on operational data, the method comprising: Obtain a time-series dataset generated during the operating cycle of the selective catalytic reduction reactor, the time-series dataset including the inlet flue gas parameter sequence of the denitrification reactor, the outlet flue gas parameter sequence of the denitrification reactor, and the control parameter sequence of the ammonia injection device; Based on the time-series dataset, a reactant concentration distribution model for the denitrification reaction zone is constructed, and the inlet reactant flux distribution coefficient and outlet product generation rate distribution coefficient of each partition of the catalyst layer are decoupled. The inlet reactant flux distribution coefficient and the outlet product generation rate distribution coefficient are used as boundary constraints and input into the catalyst surface micro-reaction site occupancy state deduction model. Iterative convergence processing is initiated to generate a catalyst surface active site occupancy rate distribution map and a catalyst surface deposition coverage ratio distribution map. Based on the distribution map of the occupancy rate of active sites on the catalyst surface and the distribution map of the coverage ratio of deposits on the catalyst surface, the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst were calculated. A thermal map of the spatial distribution of catalyst performance degradation is generated based on the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst. Based on the thermal map of the spatial distribution of catalyst performance degradation, the catalyst remaining efficiency index and the location information of the catalyst maintenance area are output.

[0005] Furthermore, embodiments of the present invention also provide an online performance evaluation system for SCR catalysts based on operational data, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described online evaluation method for SCR catalyst performance based on runtime data by executing the machine-executable instructions.

[0006] Based on the above, by acquiring time-series datasets of the selective catalytic reduction reactor (SCR) during its operating cycle, covering information such as inlet flue gas parameters, outlet flue gas parameters, and ammonia injection control parameters, a reactant concentration distribution model for the denitrification reaction zone is constructed. The inlet reactant flux distribution coefficient and outlet product generation rate distribution coefficient of each catalyst zone are decoupled, enabling the characterization of reactant and product distribution in each catalyst zone during actual operation. These distribution coefficients are then used as boundary constraints in the catalyst surface micro-reaction site occupancy state deduction model. After iterative convergence processing, a catalyst surface active site occupancy rate distribution map and a catalyst surface deposit coverage ratio distribution map are generated, revealing the reaction state and deposition status of the catalyst surface. Based on these two distribution maps, the intrinsic activity retention ratio and flow resistance increase ratio of each catalyst zone are calculated, allowing for a quantitative assessment of catalyst activity and resistance changes. Finally, a spatial distribution heat map of catalyst performance degradation is generated, and the catalyst remaining efficiency index and catalyst maintenance area location information are output. This allows for timely detection of the region and degree of catalyst performance degradation, effectively improving the operating efficiency and stability of the SCR system and reducing operating costs. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the online performance evaluation method for SCR catalysts based on operational data provided in this embodiment of the invention.

[0008] Figure 2 This is a schematic diagram of exemplary hardware and software components of the online performance evaluation system for SCR catalysts based on operational data provided in this embodiment of the invention. Detailed Implementation

[0009] Figure 1 This is a flowchart illustrating an online performance evaluation method for SCR catalysts based on operational data, provided in one embodiment of the present invention. A detailed description follows.

[0010] The method described in this embodiment is applied to the denitrification system of a coal-fired power plant. This denitrification system employs a typical "high-dust arrangement," with a selective catalytic reduction reactor positioned between the economizer and the air preheater. The reactor is filled with a honeycomb vanadium-titanium-based catalyst. Online monitoring systems deployed at the inlet and outlet flues of the denitrification reactor continuously collect various operating parameters related to catalyst performance.

[0011] Step S110: Obtain the time-series dataset generated by the selective catalytic reduction reactor during its operating cycle. The time-series dataset includes the inlet flue gas parameter sequence of the denitrification reactor, the outlet flue gas parameter sequence of the denitrification reactor, and the control parameter sequence of the ammonia injection device.

[0012] In this embodiment, operational data related to the denitrification reactor is first acquired from the power plant's distributed control system. Specifically, data is continuously collected for a complete operating cycle via a data acquisition and monitoring control system interface at a preset sampling frequency, such as once per minute. This operating cycle can be a continuous operating period since the last catalyst soot blowing or offline cleaning, for example, a continuous 720 hours.

[0013] The collected data constitutes a multi-dimensional time-series dataset. The sequence of inlet flue gas parameters for the denitrification reactor includes at least the concentration of nitrogen oxides in the inlet flue gas, measured by an in-situ laser analyzer installed on the reactor inlet flue and recorded in parts per million (ppm) by volume; the oxygen concentration in the inlet flue gas, measured by a zirconia oxygen analyzer; and the temperature, pressure, and flow rate of the inlet flue gas, measured by thermocouples, pressure transmitters, and differential pressure flow meters, respectively. All inlet parameters are arranged in chronological order to form a corresponding time-series sequence.

[0014] The sequence of parameters for the flue gas at the outlet of the denitrification reactor includes at least the concentration of nitrogen oxides in the flue gas, which is measured by a laser analyzer based on the same principle installed on the flue gas duct at the reactor outlet; and the temperature, pressure and flow rate of the flue gas, which are measured by corresponding sensors. These parameters are also arranged in chronological order to form a corresponding time sequence.

[0015] The ammonia injection unit control parameter sequence includes the ammonia mass flow rate setpoint of the main ammonia injection pipe, the feedback values ​​of the regulating valve opening of each ammonia injection branch pipe, and data such as the dilution fan flow rate. This data reflects the real-time adjustment of the ammonia injection volume to control denitrification efficiency during the operating cycle. All the above parameters, including the nitrogen oxide concentration (NOx_in(t), oxygen concentration (O2_in(t), and inlet flue gas flow rate (FlueGas_in(t))) in the inlet flue gas parameter sequence; the nitrogen oxide concentration (NOx_out(t) and outlet flue gas flow rate (FlueGas_out(t)) in the outlet flue gas parameter sequence; and the ammonia injection mass flow rate (NH3_inj(t)) in the ammonia injection unit control parameter sequence, together constitute the basic time-series dataset for subsequent analysis. All data underwent preprocessing after acquisition, including removing obvious outliers caused by sensor malfunctions or communication interruptions, and timestamp alignment to ensure that data points at the same moment accurately reflect the instantaneous operating conditions of the reactor.

[0016] Step S120: Construct a reactant concentration distribution model for the denitrification reaction zone based on the time series dataset, and decouple the inlet reactant flux distribution coefficient and outlet product generation rate distribution coefficient for each partition of the catalyst layer.

[0017] This step aims to analyze the non-uniformity of flue gas flow within the catalyst layer from macroscopic operational data, and to quantify how this non-uniformity leads to differences in the distribution of reactants and products in different spatial regions of the catalyst layer.

[0018] Step S121: Extract the inlet nitrogen oxide concentration value and inlet oxygen concentration value contained in the inlet flue gas parameter sequence of the denitrification reactor from the time series dataset, and arrange the inlet nitrogen oxide concentration value and the inlet oxygen concentration value in the time sequence within the operating cycle to form an inlet reactant concentration time series matrix.

[0019] In this embodiment, the inlet nitrogen oxide concentration value NOx_in(t) and the inlet oxygen concentration value O2_in(t) are first extracted from the time-series dataset constructed in step S110. Both sequences are one-dimensional vectors that change with time t. To construct the subsequent model, they need to be organized into a two-dimensional matrix. Specifically, an inlet reactant concentration time-series matrix C_in is formed. The first column of this matrix is ​​the timestamp sequence, the second column is the inlet nitrogen oxide concentration value NOx_in(t) at the corresponding time, and the third column is the inlet oxygen concentration value O2_in(t) at the corresponding time. The number of rows in the matrix is ​​determined by the number of sampling points within the operating cycle. For example, if the operating cycle is 720 hours and sampling is performed once per minute, the inlet reactant concentration time-series matrix contains 43,200 rows of data.

[0020] Step S122: Extract the flue gas flow time series data of the inlet and outlet of the denitrification reactor from the time series dataset. Multiply the nitrogen oxide concentration value at the inlet by the inlet flue gas flow rate value at the corresponding time to obtain the inlet nitrogen oxide mass flow rate time series sequence. Multiply the nitrogen oxide concentration value at the outlet by the outlet flue gas flow rate value at the corresponding time to obtain the outlet nitrogen oxide mass flow rate time series sequence.

[0021] This step combines reactant concentration information with flue gas flow rate information to convert it into mass flow rate, thereby establishing the material balance relationship between the reactor inlet and outlet. The inlet flue gas flow rate values ​​FlueGas_in(t) and FlueGas_out(t) are extracted from the time-series dataset. Then, for each sampling time t, the following calculation is performed: the inlet nitrogen oxide mass flow rate Mass_NOx_in(t) equals the inlet nitrogen oxide concentration value NOx_in(t) multiplied by the inlet flue gas flow rate value FlueGas_in(t), and then multiplied by a unit conversion factor K_unit related to the molecular weight of the flue gas, i.e., Mass_NOx_in(t) = NOx_in(t) × FlueGas_in(t) × K_unit. Similarly, the outlet nitrogen oxide mass flow rate Mass_NOx_out(t) equals the outlet nitrogen oxide concentration value NOx_out(t) multiplied by the outlet flue gas flow rate value FlueGas_out(t), and then multiplied by the same unit conversion factor K_unit. Thus, two time-series sequences of mass flow rate that vary with time were obtained.

[0022] Step S123: Based on the inlet nitrogen oxide mass flow rate time series, the outlet nitrogen oxide mass flow rate time series, and the ammonia injection flow rate value contained in the ammonia injection device control parameter series, establish a set of mass conservation equations on the flue gas flow path inside the denitrification reactor. Each equation in the set of mass conservation equations corresponds to a differential unit cross section of the catalyst layer in the flue gas flow direction.

[0023] This step discretizes the physical space inside the reactor, dividing the catalyst layer into several micro-elements along the flue gas flow direction. According to the law of conservation of mass, for each micro-element, the total amount of reactants entering that micro-element minus the total amount of reactants leaving that micro-element should equal the total amount of reactants consumed by the chemical reaction within that micro-element. Considering that ammonia is injected by an ammonia injection device, its distribution is also affected by the flue gas flow. Therefore, for the i-th differential unit in the flue gas flow direction, the following mass conservation equation can be established: the sum of the inlet nitrogen oxide mass flow rate Mass_NOx_in_i and the inlet ammonia mass flow rate Mass_NH3_in_i equals the sum of the outlet nitrogen oxide mass flow rate Mass_NOx_out_i and the outlet ammonia mass flow rate Mass_NH3_out_i, plus the nitrogen oxide mass flow rate Mass_NOx_reacted_i consumed by the denitrification reaction within that micro-element. The reaction consumption term Mass_NOx_reacted_i is related to the reactant concentration, catalyst activity, and residence time within the micro-element. The above equations are combined across all micro-elements in the flue gas flow direction to form a system of equations. The known quantities in this system include the time series of nitrogen oxide mass flow rates at the reactor's total inlet and outlet calculated in step S122, and the ammonia injection mass flow rate value NH3_inj(t) obtained from the ammonia injection unit control parameter series; these serve as the boundary conditions for the system of equations.

[0024] Step S124: Introduce the flue gas flow rate distribution non-uniformity coefficient on the horizontal cross section of the catalyst layer as a variable to be solved in the mass conservation equation system. Solve the mass conservation equation system by numerical iteration method to obtain the flue gas flow rate distribution non-uniformity coefficient and the flue gas flow rate allocation weight vector corresponding to each partition of the catalyst layer.

[0025] To characterize the non-uniformity of flue gas distribution in the horizontal direction (i.e., the cross-section perpendicular to the flue gas flow direction) of the catalyst layer, a flue gas flow rate distribution non-uniformity coefficient γ_j is introduced into each differential unit cross-section in the equation system established in step S123. This coefficient represents the ratio of the actual flue gas flow rate to the theoretical flow rate uniformly distributed according to the cross-sectional area in the j-th partition of the horizontal cross-section. The aforementioned flue gas flow rate distribution non-uniformity coefficient γ_j constitutes a vector, serving as the unknown quantity to be solved. Simultaneously, the catalyst layer is divided into M partitions in the horizontal direction, each partition having an independent flue gas flow rate distribution weight w_j, where w_j is proportional to γ_j and the cross-sectional area A_j of that partition. Then, a numerical iterative method, such as the Newton-Raphson iterative method, is used to solve this equation system containing nonlinear terms (reaction consumption terms related to concentration and flow rate). During the iteration process, the values ​​of γ_j and w_j are continuously adjusted to minimize the sum of squared residuals between the calculated outlet nitrogen oxide concentration value and the measured outlet nitrogen oxide concentration value NOx_out(t) of each infinitesimal element in the equation system. When the residual is less than the preset convergence threshold, such as 1e-6, the iteration terminates. At this time, the w_j obtained is the flue gas flow distribution weight vector W_dist=[w_1, w_2, ..., w_M] corresponding to each partition of the catalyst layer.

[0026] Step S125: Based on the flue gas flow rate allocation weight vector and the inlet nitrogen oxide mass flow rate time sequence, calculate the inlet nitrogen oxide mass flow rate allocated to each partition of the catalyst layer; use the ratio of the inlet nitrogen oxide mass flow rate of each partition to the total inlet nitrogen oxide mass flow rate as the inlet reactant flux allocation coefficient of each partition of the catalyst layer.

[0027] After obtaining the flue gas flow allocation weight w_j for each zone, and combining it with the inlet total nitrogen oxide mass flow rate Mass_NOx_in_total(t) (i.e., Mass_NOx_in(t) calculated in step S122), the inlet nitrogen oxide mass flow rate Mass_NOx_in_j(t) allocated to each catalyst zone j can be calculated. The calculation process is as follows: Mass_NOx_in_j(t) = Mass_NOx_in_total(t) × w_j / Σ(w_1, w_2, ..., w_M). The denominator is the sum of the weights of all zones, used for normalization. Then, the inlet reactant flux allocation coefficient α_j(t) is defined as the ratio of Mass_NOx_in_j(t) to Mass_NOx_in_total(t), i.e., α_j(t) = w_j / Σ(w_1, w_2, ..., w_M). The inlet reactant flux distribution coefficient α_j(t) reflects the proportion of the reactant mass flow rate entering the j-th catalyst zone at time t to the total reactant mass flow rate entering the reactor. Essentially, it is a direct manifestation of the non-uniformity of the spatial distribution of flue gas flow.

[0028] Step S126: Substitute the inlet reactant flux distribution coefficient of each zone of the catalyst layer into the mass conservation equation system to solve the material balance relationship between reactant consumption rate and product formation rate, and extract the outlet product formation rate distribution coefficient of each zone of the catalyst layer from the material balance relationship.

[0029] The inlet reactant flux distribution coefficient α_j(t) of each zone calculated in step S125 is used as a known quantity and substituted again into the mass conservation equations established in step S123. At this point, the unknown quantity in the equations becomes the denitrification reaction rate within each zone. By solving this modified equation set, the nitrogen oxide consumption rate R_NOx_cons_j(t) and the nitrogen (main denitrification product) generation rate R_N2_gen_j(t) of each zone j at time t can be obtained. These rates are derived from material balance. Subsequently, the outlet product generation rate distribution coefficient β_j(t) is defined as the ratio of the nitrogen mass flow rate generated in the j-th zone to the sum of the total nitrogen mass flow rates generated in all zones. That is, β_j(t) = (R_N2_gen_j(t) × V_j) / Σ(R_N2_gen_j(t) × V_j), where V_j is the catalyst volume of the j-th zone. β_j(t) characterizes the contribution share of each catalyst zone to the total denitrification efficiency at time t, which reflects the spatial distribution of the actual chemical reactivity within the zone.

[0030] Step S130: Input the inlet reactant flux distribution coefficient and the outlet product generation rate distribution coefficient as boundary constraints into the catalyst surface micro-reaction site occupancy state deduction model, start iterative convergence processing, and generate catalyst surface active site occupancy rate distribution map and catalyst surface deposit coverage ratio distribution map.

[0031] The model for inferring the occupancy status of micro-reaction sites on the catalyst surface is constructed as follows.

[0032] The inference model employs a physically based graph neural network architecture, specifically named a Physical Information Graph Convolutional Network. The model architecture consists of three core modules: a spatial feature encoder, a temporal evolution operator, and a site-state decoder.

[0033] The spatial feature encoder consists of three stacked graph convolutional layers. The adjacency matrix of each graph convolutional layer is constructed based on the physical connectivity between catalyst channels. Each node represents a catalyst partition, and the node features include the inlet reactant flux allocation coefficient, the outlet product generation rate allocation coefficient, flue gas temperature, and flue gas pressure for that partition. The first graph convolutional layer maps the input node features to 64-dimensional hidden states, the second layer maps the 64-dimensional features to 128-dimensional features, and the third layer outputs 128-dimensional spatial aggregated features. Each graph convolutional layer is followed by a batch normalization layer and a modified linear unit activation function.

[0034] The time evolution operator employs a gated recurrent unit structure with a hidden state dimension of 128. The input to this operator is the spatial aggregated feature sequence output by the spatial feature encoder, expanded over time steps. The gated recurrent unit contains update and reset gates. The update gate controls the proportion of hidden states from the previous time step flowing into the current time step, while the reset gate controls the degree to which hidden states from the previous time step are ignored. These two gating mechanisms enable the learning of the evolution of the occupancy state of reaction sites on the catalyst surface over time.

[0035] The site state decoder consists of two independent fully connected branches that share the final hidden state output by the temporal evolution operator. The first branch is the active site occupancy decoder, containing two fully connected layers. The first layer maps the 128-dimensional hidden state to 64 dimensions, and the second layer maps the 64-dimensional state to 1 dimension, outputting the predicted active site occupancy value for that partition at the current time step. The second branch is the sediment cover ratio decoder, with the same network structure, outputting the predicted sediment cover ratio for that partition at the current time step.

[0036] The loss function of the model is a weighted mean squared error loss, with the weight ratio of the active site occupancy prediction loss to the sediment coverage ratio prediction loss being 1:1. At the same time, a physical constraint regularization term is added, which is constructed based on the catalyst surface material conservation equation to penalize the degree to which the model prediction results violate the law of mass conservation.

[0037] The training process of the model is as follows: The training dataset comes from historical operating data collected over a three-year operating cycle of denitrification systems in similar coal-fired power plants. It includes 500 complete catalyst lifecycle data sets, each containing continuous monitoring data from the commissioning of a new catalyst to performance degradation and eventual failure to meet denitrification requirements. Each data set is divided into multiple time windows, each containing 30 consecutive days of operating data. The input format is a four-dimensional tensor consisting of the time step multiplied by the number of nodes and the feature dimension. The time step is 720 hours, the number of nodes corresponds to the number of partitions in the catalyst layer, and the feature dimension is 6, corresponding to the inlet reactant flux distribution coefficient, the outlet product formation rate distribution coefficient, flue gas temperature, flue gas pressure, ammonia injection rate, and outlet nitrogen oxide concentration, respectively. The output format is a three-dimensional tensor consisting of the time step multiplied by the number of nodes and 2, where 2 corresponds to the active site occupancy rate and sediment coverage ratio.

[0038] Training was performed using the Adam optimizer with an initial learning rate of 0.0005. A cosine annealing learning rate scheduling strategy was employed, with the learning rate decaying to 0.9 times the current value every 10 training epochs. The batch size was 16, and the number of training epochs was 200. An early stopping mechanism was used, terminating training when the validation set loss did not decrease for 15 consecutive epochs. The evaluation metrics for the validation set loss were mean absolute percentage error (MASE) and coefficient of determination (COD). Model versions with an MASE below 5% and a COD above 0.95 were retained.

[0039] In inference applications, real-time acquired time-series data is preprocessed in the same format as the training data, including Z-score standardization for each feature dimension, using the mean and standard deviation of the training dataset as the standardization parameters. The preprocessed data tensor is input into the model, which outputs predicted values ​​for active site occupancy and sediment coverage ratio for each partition at each time step. The output results are then de-standardized to obtain physically meaningful values ​​for active site occupancy and sediment coverage ratio, using the mean and standard deviation of the training dataset.

[0040] Step S131: Obtain the active site reaction kinetic parameter library contained in the catalyst surface micro-reaction site occupancy state deduction model.

[0041] In this embodiment, a model for inferring the occupancy state of micro-reaction sites on the catalyst surface is pre-constructed. The core of this model is a database containing various kinetic parameters, obtained through laboratory characterization and microkinetic experiments on similar catalysts. The parameter database stores the inherent properties of different types of active sites on the catalyst surface (e.g., vanadium oxide sites V-OH, titanium oxide sites Ti-O, etc.), specifically including: the standard reaction rate constant k_VOx_i (unit: moles per liter per second) for each site reacting with nitrogen oxides, which changes with temperature according to the Arrhenius equation; the standard adsorption equilibrium constant K_NH3_i (unit: liters per mole) for each site reacting with ammonia; and the deactivation rate decay factor δ_i when each site is covered by deposits such as ammonium sulfate or ammonium bisulfate, which represents the degree of inhibition of site activity by the deposit coverage.

[0042] Step S132: Map the inlet reactant flux distribution coefficient according to the spatial coordinates of each partition of the catalyst layer to obtain the cumulative inlet reactant flux time series curve of each catalyst partition during the operating cycle, and extract the reactant flux fluctuation amplitude and reactant flux peak time of each catalyst partition from the inlet reactant flux time series curve.

[0043] The time-varying inlet reactant flux distribution coefficient α_j(t) obtained in step S125 (one curve for each partition j) is correlated with the spatial coordinates of the catalyst partition (e.g., two-dimensional coordinates (x_j, y_j) based on the reactor inlet cross-section). Thus, for each partition j, its inlet reactant flux time-series curve F_j(t) = α_j(t) × Mass_NOx_in_total(t) / A_j is obtained over the operating cycle, where A_j is the cross-sectional area of ​​partition j. From this time-series curve, dynamic features are extracted: the reactant flux fluctuation amplitude Amp_j, defined as the difference between the maximum and minimum values ​​of the curve within the cycle; and the reactant flux peak time T_peak_j, defined as the time point corresponding to when the curve reaches its maximum value.

[0044] Step S133: Map the distribution coefficient of the output product generation rate according to the spatial coordinate position of each partition of the catalyst layer to obtain the time series curve of the output product generation rate accumulated in each catalyst partition during the operation cycle, and extract the product generation rate response lag time and product generation rate steady-state deviation of each catalyst partition from the time series curve of the output product generation rate.

[0045] Similarly, the effluent product generation rate allocation coefficient β_j(t) obtained in step S126 is mapped to the spatial coordinates (x_j, y_j) to obtain the product generation rate time-series curve P_j(t) = β_j(t) × (total denitrification product generation) for each zone. Dynamic features are extracted from this curve: the product generation rate response lag time τ_j can be defined as the time delay required for the effluent product generation rate to change accordingly after a change in the inlet reactant flux. This can be obtained by calculating the cross-correlation function between the input signal F_j(t) and the output signal P_j(t) and finding the time shift corresponding to the maximum value of the cross-correlation function; the steady-state deviation of the product generation rate ΔP_j is defined as the difference between the average value of P_j(t) at the end of the operating cycle and the average value for the entire cycle, reflecting the activity deviation state of the zone at the end of the cycle.

[0046] Step S134: Using the reactant flux fluctuation amplitude, the reactant flux peak time, the product formation rate response lag time, and the product formation rate steady-state deviation, define and calibrate the dynamic input conditions or related parameters of the catalyst surface micro-reaction site occupancy state deduction model, as the basis for model iterative calculation, and trigger the catalyst surface micro-reaction site occupancy state deduction model to start the active site occupancy state iterative calculation process.

[0047] The four dynamic features (Amp_j, T_peak_j, τ_j, ΔP_j) extracted in steps S132 and S133 are used as inputs to instantiate and calibrate the model for extrapolating the occupancy status of micro-reaction sites on the catalyst surface. For example, based on the reactant flux fluctuation amplitude Amp_j, the dynamic range of reactant concentration changes in each zone during iterative calculations can be set; based on the response lag time τ_j, the resistance coefficient for reactant diffusion and mass transfer within the catalyst micropores can be adjusted. After calibration, the model possesses initial conditions for the actual operating conditions and spatial distribution characteristics of the current reactor, and then iterative calculations of the active site occupancy status are initiated.

[0048] Step S135: During the iterative calculation of the active site occupancy status, the catalyst surface micro-reaction site occupancy status deduction model repeatedly adjusts the proportion of reactants occupied and the proportion of deposits covered by different types of active sites in each catalyst partition according to the dynamic balance between the reactant arrival rate and the product departure rate of each catalyst partition, until the sum of the residuals between the calculated reactant consumption rate and the calculated product generation rate of each catalyst partition is lower than the preset convergence threshold, at which point the iterative calculation is terminated.

[0049] This simulation model is essentially a numerical simulator for solving a system of differential equations. For each catalyst partition, based on the parameters calibrated in step S134, the model calculates the adsorption rate of reactants (nitrogen oxides and ammonia) to the active sites on the catalyst surface, the rate of reaction at the active sites, and the rate of product desorption from the sites within each tiny time step. Simultaneously, the model also simulates the formation and accumulation of deposits such as ammonium bisulfate at the active sites, a process related to the sulfur trioxide concentration in the flue gas and the reaction temperature. The model solves these rate equations iteratively, continuously updating the proportion θ_reactant_i occupied by reactants and the proportion θ_deposit_i covered by deposits at each site within each time step, ensuring a dynamic equilibrium between the reactant flux entering the sites and the product flux leaving the sites at the end of each time step. The iteration process terminates when the sum of the residuals between the calculated reactant consumption rate (based on the current θ_reactant_i) and the calculated product formation rate (based on the current θ_deposit_i and the reaction rate) of all catalyst partitions is less than a preset convergence threshold, such as 1e-4.

[0050] Step S136: Extract the proportion of reactants occupying different types of active sites in each catalyst partition from the iterative calculation results output by the catalyst surface micro-reaction site occupancy state deduction model, and interpolate and fill the proportion of reactants occupying all catalyst partitions according to spatial coordinates to generate a catalyst surface active site occupancy rate distribution map.

[0051] After iterative convergence, the proportion of each type of active site occupied by reactants at the final time is extracted from the model calculation results for each partition j. For example, for the i-th type of active site, its occupancy proportion is θ_reactant_i_j. For each partition, a comprehensive active site occupancy rate φ_j can be calculated by weighted averaging, with the weights determined by the initial density of each type of site. Then, the φ_j values ​​of all partitions are combined with their spatial coordinates (x_j, y_j) to form a discrete point set. Using the Kriging interpolation method, spatial interpolation is performed on all points on the entire horizontal cross-section of the reactor to generate a continuously distributed two-dimensional field map, i.e., the catalyst surface active site occupancy rate distribution map. The gray value of each pixel in this catalyst surface active site occupancy rate distribution map represents the active site occupancy rate at that location; a higher value indicates a higher degree of reversible occupancy by reactants.

[0052] Step S137: Extract the deposition coverage ratio of different types of active sites in each catalyst partition from the iterative calculation results output by the catalyst surface micro-reaction site occupancy state deduction model, and interpolate and fill the deposition coverage ratio of all catalyst partitions according to spatial coordinates to generate a catalyst surface deposition coverage ratio distribution map.

[0053] Similarly, the deposition coverage ratio θ_deposit_i_j of various active sites in each partition j is extracted from the model calculation results, and a comprehensive deposition coverage ratio ψ_j is calculated. Then, using the same spatial interpolation method as in step S136, the ψ_j value is mapped to the entire reactor cross section to generate a catalyst surface deposition coverage ratio distribution map. This catalyst surface deposition coverage ratio distribution map reflects the spatial distribution of deposits (mainly ammonium bisulfate) accumulated on the catalyst surface; higher values ​​indicate more severe clogging and coverage.

[0054] Step S140: Based on the distribution map of the occupancy rate of active sites on the catalyst surface and the distribution map of the coverage ratio of deposits on the catalyst surface, calculate the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst.

[0055] Step S141: Extract the final occupancy rate of active sites for each catalyst partition at the end of the operating cycle from the catalyst surface active site occupancy rate distribution map, obtain the initial active site density of the catalyst in the unused state, and calculate the effective active site density of each catalyst partition in the current state based on the deposition coverage ratio distribution map of each catalyst partition and the initial active site density of the catalyst in the unused state. Calculate the ratio of the effective active site density of each catalyst partition in the current state to the initial active site density to obtain the effective number ratio coefficient of active sites for each catalyst partition.

[0056] First, from the active site occupancy distribution map generated in step S136, read the final active site occupancy rate φ_j_end of each partition j at the end of the operating cycle. Obtain the initial total active site density D_init (unit: sites per square meter) of the catalyst in its unused state from the catalyst factory data or laboratory test report. From the sediment coverage ratio distribution map generated in step S137, read the sediment coverage ratio ψ_j of each partition j. Assuming that sediment coverage is the main cause of permanent deactivation of active sites, the effective active site density D_eff_j of each partition j in the current state can be calculated as: D_eff_j = D_init × (1 - ψ_j) × (1 - φ_j_end). Here, (1 - ψ_j) represents the proportion of sites not covered by sediment, and multiplying by (1 - φ_j_end) represents the proportion of sites not reversibly occupied by reactants among the uncovered sites. Then, define the effective number ratio coefficient of active sites η_j = D_eff_j / D_init.

[0057] Step S142: Based on the effective number ratio coefficient of active sites and the deposition coverage ratio distribution map of catalyst surface, establish the active site availability coupling equation for each catalyst partition and the corresponding deposition coverage ratio. Solve the active site availability coupling equation to obtain the effective exposure area of ​​active sites in each catalyst partition.

[0058] The effective exposed area of ​​active sites depends not only on the number of sites but also on the morphology of sediment cover. Sediments may partially cover the pore entrance, preventing internal active sites from contacting reactants. Therefore, a coupled equation is established: effective exposed area A_exp_j = A_geo_j × f(η_j, ψ_j), where A_geo_j is the geometric surface area of ​​partition j. The function f is an empirical model, such as f = η_j × (1-ψ_j)^m, where the exponent m reflects the nonlinear effect of sediment on pore blockage and can be calibrated by mercury intrusion porosimetry on similar catalysts. By substituting η_j and ψ_j, A_exp_j for each partition can be calculated.

[0059] Step S143: Obtain the initial intrinsic reaction rate constant of the catalyst in the unused state, and multiply the initial intrinsic reaction rate constant with the effective number ratio coefficient of active sites in each catalyst partition to obtain the actual reaction rate potential value of each catalyst partition in the current state.

[0060] The initial intrinsic reaction rate constant k_init (in meters per second) is obtained from the catalyst manufacturing data. This initial intrinsic reaction rate constant represents the reaction rate per unit area of ​​effective active sites. Therefore, the actual reaction rate potential value of each partition j in the current state is k_potential_j = k_init × η_j.

[0061] Step S144: Calculate the ratio of the actual reaction rate potential value of each catalyst partition in the current state to the initial reaction rate potential value corresponding to the initial intrinsic reaction rate constant, and obtain the intrinsic activity retention ratio of each catalyst partition.

[0062] The initial reaction rate potential is k_init itself. Therefore, the intrinsic activity retention ratio of each catalyst zone is γ_j = k_potential_j / k_init = η_j. This intrinsic activity retention ratio is directly determined by the effective number ratio coefficient of active sites, characterizing the degree to which the current reactivity of that zone is maintained relative to its initial state.

[0063] Step S145: Extract the deposition coverage ratio of each catalyst zone from the deposition coverage ratio distribution map on the catalyst surface, and calculate the remaining flow cross-sectional area ratio of each catalyst zone based on the geometric mapping relationship between the deposition coverage ratio and the blockage area of ​​the catalyst channel cross-section.

[0064] Catalyst channel blockage is primarily caused by the accumulation of sediment on the channel walls and at corners. Based on the sediment coverage ratio ψ_j obtained in step S137, and combined with the geometry of the catalyst channel (e.g., square for a honeycomb catalyst), a geometric mapping model can be established. This model assumes that the sediment adheres uniformly to the inner wall of the channel, and that the thickness of the blockage layer increases linearly or exponentially with increasing ψ_j. Based on this, the proportion of the remaining flow cross-sectional area ξ_j for each zone j can be calculated. For example, ξ_j = (1 - δ_s × ψ_j)^2, where δ_s is a shape factor related to the channel hydraulic diameter and sediment density.

[0065] Step S146: Obtain the initial channel flow cross-sectional area of ​​the catalyst in the unused state. Based on the physical relationship between flow resistance and channel flow cross-sectional area in fluid mechanics, calculate the theoretical flow resistance of each catalyst partition in the current state by the proportion of the remaining channel flow cross-sectional area of ​​each catalyst partition.

[0066] According to the Hagen-Poiseuille law or the Darcy-Weisbach equation, for laminar flow, the channel resistance is inversely proportional to the square of the flow cross-sectional area. Obtain the initial channel cross-sectional area A_channel_init under unused catalyst conditions. Then, the theoretical flow resistance R_flow_j for each partition j in the current state can be expressed as: R_flow_j = R_init / (ξ_j^2), where R_init is the theoretical flow resistance under the initial state, which is a constant.

[0067] Step S147: Calculate the ratio of the theoretical flow resistance of each catalyst partition in the current state to the theoretical initial flow resistance corresponding to the initial channel flow cross-sectional area to obtain the increase ratio of the flow resistance of each catalyst partition.

[0068] The increase in flow resistance is calculated as ζ_j = (R_flow_j - R_init) / R_init = (1 / ξ_j^2) - 1. This increase in flow resistance quantifies the degree of resistance increase caused by sediment blockage; a positive value indicates an increase in resistance.

[0069] Step S150: Generate a thermal map of the spatial distribution of catalyst performance degradation based on the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst, and output the catalyst remaining efficiency index and catalyst maintenance area location information based on the thermal map of the spatial distribution of catalyst performance degradation.

[0070] Step S151: Divide the catalyst layer into catalyst partition grids with spatial coordinates according to the geometric structure. Assign the intrinsic activity retention ratio and flow resistance increase ratio corresponding to each catalyst partition to the corresponding grid nodes in the catalyst partition grid, forming a spatial distribution field of intrinsic activity retention ratio and a spatial distribution field of flow resistance increase ratio.

[0071] Based on the M partitions defined in step S124, a two-dimensional grid is constructed. Each node coordinate (x_j, y_j) in the two-dimensional grid corresponds to a catalyst partition. The intrinsic activity retention ratio γ_j calculated in step S144 is assigned to the corresponding node, forming the intrinsic activity retention ratio spatial distribution field Γ(x, y). The flow resistance increase ratio ζ_j calculated in step S147 is assigned to the corresponding node, forming the flow resistance increase ratio spatial distribution field Z(x, y).

[0072] Step S152: Perform spatial gradient calculation on the spatial distribution field of the intrinsic activity retention ratio to obtain the decay gradient direction and decay gradient magnitude of the intrinsic activity retention ratio in the horizontal direction of the catalyst layer. Select the grid node coordinates whose decay gradient magnitude exceeds the preset decay threshold from the decay gradient magnitude as the location points of the activity decay acceleration region.

[0073] Calculate the spatial gradient of the spatially distributed field Γ(x, y). In two-dimensional space, the gradient is a vector whose direction points towards the direction of the fastest increase in function value. Since the intrinsic activity retention ratio typically decays from the inlet to the outlet or from the edge to the center, the opposite direction of the gradient is the decay direction. The gradient magnitude G_Γ(x, y) = sqrt(( Γ / x)^2+( Γ / y)^2) represents the rate of activity decline. A decay threshold Th_gamma is set, for example, 0.01 per meter. All grid node coordinates with gradient magnitudes greater than Th_gamma are selected; these points represent the accelerated activity decay region, where activity decreases rapidly in space.

[0074] Step S153: Perform spatial gradient calculation on the spatial distribution field of the increase in flow resistance to obtain the blocking gradient direction and blocking gradient magnitude of the increase in flow resistance in the horizontal direction of the catalyst layer. Select the grid node coordinates that exceed the preset blocking threshold from the blocking gradient magnitude as the location points of the blocking acceleration region.

[0075] Similarly, the spatial gradient of the spatially distributed field Z(x, y) is calculated to obtain the blocking gradient magnitude G_Z(x, y) = sqrt(( Z / x)^2+( Z / y)^2). Set a blocking threshold Th_zeta, for example, 0.005 per meter. Filter out the coordinates of all grid nodes with gradient magnitudes greater than Th_zeta; these points are the blocking acceleration regions, where the drag increases at a faster rate.

[0076] Step S154: Input the spatial distribution field of the intrinsic activity retention ratio and the spatial distribution field of the increase in flow resistance into the catalyst performance comprehensive evaluation function for weighted fusion calculation to generate the comprehensive performance degradation index corresponding to each catalyst partition. The weight coefficients of the intrinsic activity retention ratio and the increase in flow resistance in the catalyst performance comprehensive evaluation function are dynamically adjusted according to the catalyst operation history data.

[0077] A comprehensive evaluation function is defined to combine the two dimensions of activity and resistance into a single comprehensive index. For each partition j, its comprehensive performance degradation index δ_j can be defined as: δ_j = w_γ × (1 - γ_j) + w_ζ × ζ_j. Here, w_γ and w_ζ are weighting coefficients, reflecting the relative importance of activity and resistance to the overall performance of the catalyst, respectively. These two weighting coefficients are not fixed values ​​but are dynamically adjusted based on the catalyst's historical operating data. For example, if historical data indicates that under current operating conditions, denitrification efficiency is more sensitive to changes in activity, then w_γ can be set to 0.7 and w_ζ to 0.3; conversely, if pressure drop is the main limiting factor, then w_ζ can be set to 0.7. Dynamic adjustment can be achieved by analyzing the correlation coefficients between denitrification efficiency and γ_j, and between pressure drop and ζ_j, in historical data.

[0078] Step S155: Convert the comprehensive performance degradation index corresponding to each catalyst partition into a color-coded value according to the preset color mapping rules, and perform color rendering processing on each grid node in the catalyst partition grid according to the color-coded value to generate a heat map of the spatial distribution of catalyst performance degradation based on spatial coordinates.

[0079] A color mapping rule is predefined, for example, mapping the overall performance degradation index δ_j from 0 to 1 to a gradient from blue (indicating good performance) to red (indicating severe performance degradation). For each grid node, its corresponding color code (such as RGB value) is found based on its δ_j value. Then, the grid is colored and rendered, and color interpolation is performed between adjacent nodes to generate a smooth heatmap reflecting the overall performance degradation of the catalyst layer. This heatmap shows the spatial distribution of performance degradation, with highlighted areas indicating areas of concern.

[0080] Step S156: Extract the spatial coordinates corresponding to all catalyst partitions whose comprehensive performance degradation index is lower than the preset lower limit threshold from the thermal map of catalyst performance degradation, merge all extracted spatial coordinates into connected regions, and generate catalyst maintenance area positioning information.

[0081] Set a performance threshold δ_limit, for example, 0.6. Filter out all grid nodes (or regions) with δ_j ≥ δ_limit from the heatmap. Then, perform connectivity analysis on the spatial coordinates of these nodes. Using a connected component labeling algorithm from image processing, merge adjacent (e.g., four-neighbor or eight-neighbor) degraded nodes into a single connected region. Each connected region represents a catalyst region requiring maintenance or repair. Finally, generate maintenance region location information, which can be a list containing the coordinates of multiple polygon vertices, to guide subsequent local soot blowing or ammonia injection optimization.

[0082] Step S157: Obtain the comprehensive performance degradation index of all catalyst zones in the catalyst layer, input the comprehensive performance degradation index of all catalyst zones into the overall catalyst efficiency evaluation model for weighted average calculation, and output the catalyst remaining efficiency index.

[0083] Obtain the overall performance degradation index δ_j for all M partitions. The overall catalyst efficiency assessment model can be a simple weighted average model or a more complex model that considers partition volume and importance. For example, the residual efficiency index θ_remain = 1 - Σ(δ_j × V_j) / Σ(V_j), where V_j is the catalyst volume of partition j. This residual efficiency index θ_remain is between 0 and 1, with a value closer to 1 indicating better overall performance retention and a value closer to 0 indicating severe overall performance degradation.

[0084] The overall performance evaluation model for catalysts is constructed as follows.

[0085] The model employs a multilayer perceptron architecture, comprising an input layer, two hidden layers, and an output layer. The input layer has M plus 1 neurons, where M is the number of partitions in the catalyst layer. Input features include the combined performance degradation index of the M partitions and one feature representing runtime. The first hidden layer contains 64 neurons, employing a modified linear unit activation function (MLU), followed by a dropout layer with a dropout rate of 0.2 to prevent overfitting. The second hidden layer contains 32 neurons, also employing MLU. The output layer contains one neuron, employing a sigmoid activation function, with an output value between 0 and 1 representing the residual efficiency index.

[0086] The training data for this model comes from historical operational data collected over a three-year operating cycle of denitrification systems in similar coal-fired power plants. It includes 500 complete catalyst lifecycle data sets, each labeled with a true remaining efficiency index at the end of its lifecycle. This index is obtained through offline catalyst sampling and detection. The input format is a two-dimensional matrix with rows equal to the number of samples and M+1 columns. The output format is a one-dimensional vector with a length equal to the number of samples.

[0087] Training was performed using the Adam optimizer with an initial learning rate of 0.001, a batch size of 32, and 100 training epochs. The loss function was the mean squared error loss. Five-fold cross-validation was used to evaluate model performance, and the model version with the smallest mean squared error on the validation set was selected.

[0088] In inference applications, the comprehensive performance degradation index δ_j of each partition calculated in step S154 and the current running time t are combined into an input vector. Before inputting the input vector into the model, the input vector is normalized to its minimum and maximum values, with the normalization parameters taking the range of values ​​from the training dataset. The model output is the catalyst remaining efficiency index θ_remain.

[0089] Step S210: Obtain the time-series variation curve of the inlet reactant flux distribution coefficient accumulated in each partition of the catalyst layer during the operating cycle, and extract the reactant flux fluctuation spectrum characteristics of each catalyst partition from the time-series variation curve of the inlet reactant flux distribution coefficient.

[0090] This step is a supplementary analysis of the catalyst's dynamic response characteristics. Based on the time-series curves F_j(t) of reactant flux at the inlet of each zone obtained in step S132, a Fast Fourier Transform is used to transform them from the time domain to the frequency domain, resulting in a spectrum. Features are extracted from the spectrum: the amplitude of the dominant wave frequency A_peak_j, i.e., the amplitude corresponding to the largest peak in the spectrum; the phase offset of the secondary wave frequency Φ_sub_j, i.e., the phase angle corresponding to the second largest peak in the spectrum; and the wave energy decay rate λ_j, i.e., the energy decay slope of the high-frequency part.

[0091] Step S220: Obtain the time-series variation curve of the cumulative outlet product generation rate distribution coefficient of each partition of the catalyst layer during the operating cycle, and extract the product generation rate response spectrum characteristics of each catalyst partition from the time-series variation curve of the outlet product generation rate distribution coefficient.

[0092] Based on the time-series curves P_j(t) of the product generation rate of each partition obtained in step S133, a fast Fourier transform is also used to obtain the spectrum. Features are extracted from these curves: the response main frequency delay time ΔT_j, which is the time delay corresponding to the phase difference between the input and output main frequencies; the response secondary frequency gain coefficient G_sub_j, which is the ratio of the output secondary frequency amplitude to the input secondary frequency amplitude; and the response bandwidth contraction ratio B_ratio_j, which is the ratio of the -3dB bandwidth of the output spectrum to the -3dB bandwidth of the input spectrum.

[0093] Step S230: Input the flux fluctuation spectrum characteristics of the reactants and the response spectrum characteristics of the product formation rate into the catalyst dynamic response characteristic analysis model for transfer function identification processing, and generate the reaction-mass transfer coupling transfer function expression for each catalyst partition.

[0094] A catalyst dynamic response characteristic analysis model is constructed, which is essentially a system identification tool. For each partition j, the spectral features obtained in step S210 (as input) and the spectral features obtained in step S220 (as output) are input into the model. The model fits a transfer function H_j(s) that describes the dynamic relationship from input to output using the least squares method or other system identification algorithms. This transfer function is usually expressed in rational fractional form, for example, H_j(s)=K_j*(s+z1) / (s^2+2*ζ*ω_n*s+ω_n^2). Where s is the Laplace operator, K_j is the gain, z1 is the zero, ζ is the damping ratio, ω_n is the undamped natural frequency, and the dominant root of the denominator polynomial is the reaction rate dominant pole parameter. The reaction rate-dominant pole parameter is the dominant root of the denominator polynomial of the transfer function, reflecting the dominant role of the chemical reaction rate on the catalyst surface on the system's dynamic response; the mass transfer rate-dominant zero parameter is the dominant root of the numerator polynomial of the transfer function, reflecting the dominant role of the diffusion mass transfer process of reactants within the catalyst micropores on the system's dynamic response. H_j(s) is the expression for the reaction-mass transfer coupling transfer function.

[0095] Step S240: Determine the type of performance degradation dominant mechanism for each catalyst partition based on the numerical relationship between the reaction rate-dominant pole parameter and the mass transfer rate-dominant zero parameter.

[0096] The poles and zeros of the transfer function H_j(s) reflect the dynamic characteristics of the system. Poles are typically related to reaction rates, while zeros are related to mass transfer processes. By analyzing the positions (real part absolute values) of the dominant poles and zeros, it is possible to determine which process is dominant. For example, if the real part absolute value of the dominant pole is much smaller than that of the dominant zero, it indicates that the system response is mainly limited by the reaction rate, i.e., it belongs to a chemical reaction rate-controlled decay mechanism; conversely, if the real part absolute value of the dominant pole is much larger than that of the dominant zero, it indicates that the system response is mainly limited by the mass transfer rate, i.e., it belongs to a mass transfer-diffusion rate-controlled decay mechanism.

[0097] Step S250: Mark the catalyst partitions belonging to the chemical reaction rate-controlled decay mechanism as the activity decay-dominant partitions, and mark the catalyst partitions belonging to the mass transfer diffusion rate-controlled decay mechanism as the blockage decay-dominant partitions, generating a catalyst layer partition decay mechanism classification map.

[0098] Based on the judgment result of step S240, each partition j is classified and labeled. All partitions determined to be of the chemical reaction rate control type are labeled as "activity decay dominant partitions"; all partitions determined to be of the mass transfer diffusion rate control type are labeled as "blockage decay dominant partitions". The labeling results are combined with the spatial coordinates of the partitions to generate a two-dimensional partition decay mechanism classification map, with different mechanism types marked with different colors, showing the spatial distribution of the dominant causes of catalyst performance decay.

[0099] Step S260: Based on the time-series change curve of the inlet reactant flux distribution coefficient corresponding to the active decay dominant partition in the catalyst layer partition decay mechanism classification diagram, extract the cumulative reactant processing total amount in the active decay dominant partition during the operating cycle, input the cumulative reactant processing total amount into the catalyst activity decay lifetime prediction model for extrapolation calculation, and output the first remaining operating time prediction value of the active decay dominant partition.

[0100] For regions marked as "dominantly degraded activity zones," the cumulative reactant throughput C_total_j over the operating cycle is extracted, which is achieved by integrating F_j(t) over time. This total throughput is then input into a pre-trained catalyst activity degradation lifetime prediction model. This model is a physics-based or data-driven model; for example, it could be an extension of the Arrhenius law, using the cumulative reactant throughput as a "damage" factor to calculate the remaining operating time T_remain_act_j for the zone at the current degradation rate, from its current state until its activity decreases to the point where it can no longer meet denitrification requirements.

[0101] The catalyst activity decay lifetime prediction model is constructed as follows.

[0102] The catalyst activity decay lifetime prediction model employs a Long Short-Term Memory (LSTM) network architecture, comprising an LSM layer, a discard layer, and a fully connected output layer. The LSM layer contains 128 hidden units, each containing a forget gate, an input gate, an output gate, and a memory unit. The forget gate controls the proportion of information retained from the previous time step, the input gate controls the proportion of new information added to the memory unit at the current time step, the output gate controls the output of the current hidden state, and the memory unit is responsible for long-term storage of important information. The LSM layer is followed by a discard layer with a discard rate of 0.3. The fully connected output layer contains one neuron, employs a linear activation function, and outputs the predicted remaining runtime.

[0103] The training data for this catalyst activity decay lifetime prediction model comes from historical operating data collected over a three-year operating cycle of denitrification systems in similar coal-fired power plants. This includes data from 500 complete catalyst lifecycle data sets, categorized into zones dominated by activity decay. Each sample's input sequence is 365 days long, and input features include the cumulative reactant treatment volume sequence, the average flue gas temperature of that zone within the corresponding time period, and the average ammonia-nitrogen molar ratio. The output is the actual remaining operating time for that zone from its current state until its activity decreases to a preset threshold.

[0104] Training was performed using the Adam optimizer with an initial learning rate of 0.0002, a batch size of 64, and 150 training epochs. The loss function was the mean absolute error loss. An early stopping mechanism was employed, terminating training when the validation set loss did not decrease for 20 consecutive epochs.

[0105] In inference applications, the time series sequence of the cumulative reactant treatment volume in the dominant reactive degradation zone during the operating cycle is normalized using the mean and standard deviation of the training dataset as normalization parameters. The normalized sequence is concatenated with the average flue gas temperature and the average ammonia-nitrogen molar ratio to form a three-dimensional input tensor, which is then input into the model. The model output is the predicted first remaining runtime for that zone. The output result is then denormalized to obtain the physical time value in hours.

[0106] Step S270: Based on the sediment-related operating parameters of the blockage attenuation dominant partition in the catalyst layer partition attenuation mechanism classification diagram during the operating cycle, extract the potential sediment accumulation estimate of the blockage attenuation dominant partition during the operating cycle, input the potential sediment accumulation estimate into the catalyst blockage attenuation lifetime prediction model for extrapolation calculation, and output the second remaining operating time prediction value of the blockage attenuation dominant partition.

[0107] For regions marked as "plugging decay dominant zones," sediment-related operating parameters are extracted, such as operating temperature, sulfur trioxide concentration, and ammonia slip rate. Based on these parameters and combined with a sediment formation kinetic model, the potential sediment accumulation S_total_j for each such zone during the operating cycle is estimated. This accumulation is then input into a catalyst plugging decay lifetime prediction model. This model, based on hydrodynamics and pore blockage models, calculates the remaining operating time T_remain_blk_j when the proportion of remaining channel cross-sectional area ξ_j decreases to a critical value (e.g., 0.3) at this accumulation level.

[0108] The catalyst blockage attenuation lifetime prediction model is constructed as follows.

[0109] This catalyst clogging attenuation lifetime prediction model employs a temporal convolutional network architecture, comprising four dilated convolutional layers with dilation coefficients of 1, 2, 4, and 8. Each dilated convolutional layer contains 64 convolutional kernels of size 3. The dilated convolutions expand the receptive field through skip connections, enabling the model to capture long-term dependencies in the sediment accumulation process. Each dilated convolutional layer is followed by a weight normalization layer, a modified linear unit activation function, and residual connections for stabilizing training. The network terminates with a global average pooling layer and a fully connected output layer containing one neuron using a linear activation function, outputting the predicted remaining runtime.

[0110] The training data for this catalyst clogging attenuation lifetime prediction model comes from historical operational data collected over a three-year operating cycle of denitrification systems in similar coal-fired power plants. This includes data from 500 complete catalyst lifecycle data sets, categorized as zones dominated by clogging attenuation. Each sample's input sequence is 365 days long, and input features include an estimated sequence of potential deposit accumulation, the average flue gas temperature for that zone within the corresponding time period, and the average flue gas velocity. The output is the actual remaining runtime experienced by that zone from its current state until the drag increases to a preset threshold.

[0111] Training was performed using the Adam optimizer with an initial learning rate of 0.0005, a batch size of 32, and 200 training epochs. The loss function was a weighted sum of the mean squared error loss and the mean absolute error loss, with a weight ratio of 1:1. A learning rate decay strategy was employed, multiplying the learning rate by 0.5 every 50 epochs.

[0112] In inference applications, the time series of estimated potential sediment accumulation in the blockage attenuation-dominant zone during the operating cycle is normalized using the mean and standard deviation of the training dataset as normalization parameters. The normalized sequence is concatenated with the average flue gas temperature and average flue gas velocity to form a three-dimensional input tensor, which is then input into the model. The model output is the predicted second remaining runtime for that zone. The output is then denormalized to obtain the physical time value in hours.

[0113] Step S280: Perform a minimum value selection operation between the first remaining runtime prediction value and the second remaining runtime prediction value to obtain the overall remaining runtime prediction result of the catalyst layer.

[0114] The overall remaining runtime T_remain_total of the catalyst bed is determined by the weakest link. Therefore, the minimum value among T_remain_act_j of all active decay-dominant zones and T_remain_blk_j of all blockage decay-dominant zones is taken as the predicted overall remaining runtime of the entire catalyst bed, i.e., T_remain_total = min(all T_remain_act_j, all T_remain_blk_j). This overall remaining runtime prediction result forecasts the time the catalyst can continue to operate reliably under the current operation and maintenance strategy.

[0115] Step S310: Extract the final occupancy rate of active sites for each catalyst partition at the end of the operating cycle from the catalyst surface active site occupancy rate distribution map, and extract the final deposition coverage ratio for each catalyst partition at the end of the operating cycle from the catalyst surface deposition coverage ratio distribution map.

[0116] This step marks the beginning of another parallel analysis path, designed to reconstruct the reaction rates of the catalyst partitions more accurately. From the distribution maps generated in steps S136 and S137, the final occupancy φ_j_end and final coverage ratio ψ_j for each partition j are extracted.

[0117] Step S320: Establish an active site availability correlation equation for each catalyst partition based on the final occupancy rate of the active sites and the final coverage ratio of the sediment. Obtain the effective exposure area decay coefficient of the active sites in each catalyst partition by solving the active site availability correlation equation.

[0118] A correlation equation is established to decouple the effects of site occupancy and cover on the reaction rate. This correlation equation can be expressed as: A_exp_j = A_geo_j × (1 - φ_j_end) × (1 - ε × ψ_j), where ε is a decay coefficient reflecting the intensity of the effect of sediment cover on the availability of active sites, which is not merely a simple linear blockage but may also include electronic effects on uncovered sites. By substituting the known A_exp_j (from step S142) and A_geo_j, φ_j_end, and ψ_j, the decay coefficient ε_j of the effective exposed area of ​​active sites for each partition j can be solved.

[0119] Step S330: Input the effective exposure area decay coefficient of the active site of each catalyst partition into the catalyst partition reaction rate reconstruction model for parameter inversion calculation, and generate the reaction rate constant correction factor of each catalyst partition in the current state.

[0120] The catalyst partitioned reaction rate reconstruction model is a model based on microscopic reaction kinetics. The inputs to this model are the attenuation coefficient ε_j calculated in step S320 and the kinetic parameter library from step S131. The model performs inversion calculations, that is, it extrapolates the change in the microscopic reaction rate constant from the macroscopic attenuation coefficient, and finally outputs a reaction rate constant correction factor κ_j. This factor represents the ratio of the current actual reaction rate constant to the initial reaction rate constant after comprehensively considering reversible adsorption and irreversible coverage.

[0121] Step S340: Obtain the set of initial reaction rate constants for each partition of the catalyst layer in the initial state, and multiply each initial reaction rate constant in the set of initial reaction rate constants with the reaction rate constant correction factor of the corresponding catalyst partition to obtain the actual reaction rate constant of each catalyst partition in the current state.

[0122] Obtain the reaction rate constant k_init_j for each partition j in the initial state from the factory data or database (this value is usually uniform across the entire layer due to catalyst manufacturing uniformity, denoted as k_init). Then, the actual reaction rate constant k_actual_j in the current state = k_init × κ_j.

[0123] Step S350: Arrange the actual reaction rate constants of each catalyst partition in the current state according to the spatial coordinates of each partition of the catalyst layer to generate a spatial distribution matrix of the reaction rate constants of the catalyst layer.

[0124] Bind the k_actual_j of each partition j to its spatial coordinates (x_j, y_j) to form an M×3 matrix, where each row represents a partition and the three columns are the x-coordinate, y-coordinate, and k_actual_j, respectively. This matrix is ​​the spatial distribution matrix K_actual_map of the catalyst layer reaction rate constant.

[0125] Step S360: Obtain the average operating parameters of the catalyst layer during the operating cycle. The operating parameters include at least the inlet flue gas flow rate, the inlet nitrogen oxide concentration, and the residence time of the flue gas in the catalyst layer. Input the spatial distribution matrix of the reaction rate constant of the catalyst layer and the average operating parameters into the overall denitrification efficiency model of the catalyst layer for calculation to obtain the predicted value of the overall denitrification efficiency of the catalyst layer under the current state.

[0126] Obtain the average inlet flue gas flow rate Q_avg, the average inlet nitrogen oxide concentration C_NOx_in_avg, and the average residence time τ_avg of the flue gas calculated based on the catalyst volume and average flow rate during the operating cycle. Input the above operating parameters and K_actual_map obtained in step S350 into a catalyst bed overall denitrification efficiency prediction model. This catalyst bed overall denitrification efficiency prediction model is a spatially distributed reactor model that predicts the overall efficiency by integrating or weighting the denitrification efficiency of each partition. For example, the model can be based on the "one-dimensional plug flow" assumption, where for each partition j, its denitrification efficiency η_j = 1 - exp(-k_actual_j × τ_avg), and then the overall efficiency of the entire catalyst bed η_overall_pred = Σ(η_j × V_j) / Σ(V_j).

[0127] Step S370: Obtain the initial overall denitrification efficiency of the catalyst layer in the initial state, and calculate the ratio between the predicted value of the overall denitrification efficiency and the initial overall denitrification efficiency to obtain the catalyst layer comprehensive efficiency retention rate index.

[0128] Obtain the initial overall denitrification efficiency η_initial from the catalyst design data or performance test reports during the initial operation phase. Then, the catalyst bed overall efficiency retention rate index θ_efficiency = η_overall_pred / η_initial. This index directly measures the level of retention of current denitrification performance relative to the initial state.

[0129] Step S410: Extract the inlet flue gas temperature time series data and inlet flue gas pressure time series data contained in the inlet flue gas parameter sequence of the denitrification reactor from the time series dataset, and extract the outlet flue gas temperature time series data and outlet flue gas pressure time series data contained in the outlet flue gas parameter sequence of the denitrification reactor from the time series dataset.

[0130] This step aims to utilize conventional flue gas temperature and pressure monitoring data to assist in evaluating the catalyst's operating status. From the time-series dataset of step S110, the time series of inlet flue gas temperature T_in(t), inlet flue gas pressure P_in(t), outlet flue gas temperature T_out(t), and outlet flue gas pressure P_out(t) are extracted.

[0131] Step S420: Calculate the catalyst bed temperature drop at each operating moment based on the inlet flue gas temperature time series data and the outlet flue gas temperature time series data, and calculate the catalyst bed pressure drop at each operating moment based on the inlet flue gas pressure time series data and the outlet flue gas pressure time series data, generating catalyst bed temperature drop time series curves and catalyst bed pressure drop time series curves.

[0132] For each sampling time t, calculate the catalyst layer temperature drop ΔT_cat(t) = T_in(t) - T_out(t). Calculate the catalyst layer pressure drop ΔP_cat(t) = P_in(t) - P_out(t). Plot the calculation results as two curves varying with operating time: the temperature drop time series curve ΔT_cat(t) and the pressure drop time series curve ΔP_cat(t).

[0133] Step S430: Calculate the average sediment coverage ratio of the catalyst layer based on the final sediment coverage ratio of all catalyst zones in the catalyst surface sediment coverage ratio distribution map; establish a first correlation regression model between the sediment coverage ratio and the final temperature drop value at the end of the operating cycle based on the average sediment coverage ratio and the temperature drop time series curve of the catalyst layer.

[0134] Calculate the average value ψ_avg of the final sediment cover ratio ψ_j for all partitions j. Extract the final temperature drop value ΔT_end at the end of the operating cycle from the temperature drop time series curve ΔT_cat(t). Using ψ_avg as the independent variable and ΔT_end as the dependent variable, together with other historical operating data points (such as measurements after different cleaning cycles), establish a first-order correlation regression model using linear or nonlinear regression methods, for example, ΔT = a_T × ψ + b_T, where a_T and b_T are the regression coefficients to be fitted. This first-order correlation regression model expresses the relationship between sediment cover and the overall decrease in heat transfer performance of the catalyst layer.

[0135] Step S440: Calculate the average active site occupancy rate of the catalyst layer based on the final active site occupancy rate of all catalyst zones in the catalyst surface active site occupancy rate distribution diagram; establish a second correlation regression model between the active site occupancy rate and the final pressure drop value at the end of the operating cycle based on the average active site occupancy rate and the pressure drop time series curve of the catalyst layer.

[0136] Calculate the average value φ_avg of the final occupancy rate φ_j_end of active sites for all partitions j. Extract the final pressure drop value ΔP_end at the end of the operating cycle from the pressure drop time-series curve ΔP_cat(t). Using φ_avg as the independent variable and ΔP_end as the dependent variable, together with other historical data points, establish a second correlation regression model, for example, ΔP=a_P×φ+b_P, where a_P and b_P are regression coefficients. This second correlation regression model expresses the relationship between the reversible occupancy of active sites by reactants (mainly ammonia) and the overall flow resistance of the catalyst bed.

[0137] Step S450: Calculate the temperature drop change corresponding to each unit increase in sediment coverage using the first correlation regression model, and calculate the pressure drop change corresponding to each unit increase in active site occupancy using the second correlation regression model.

[0138] According to the first correlation regression model ΔT=a_T×ψ+b_T, its slope a_T represents the change in temperature drop caused by each unit increase in sediment cover (e.g., 0.01, or 1%). According to the second correlation regression model ΔP=a_P×φ+b_P, its slope a_P represents the change in pressure drop caused by each unit increase in active site occupancy.

[0139] Step S460: Perform dimensionless processing on the temperature drop change and the pressure drop change respectively to obtain the dimensionless index of temperature drop change and the dimensionless index of pressure drop change, and input them into the catalyst operating status online monitoring index generation model for fusion calculation, and output the catalyst operating status real-time monitoring index curve.

[0140] To compare temperature and pressure drops on the same scale, a_T and a_P are dimensionless. For example, the dimensionless exponent for temperature drop change is I_T = a_T / ΔT_ref, where ΔT_ref is a reference temperature drop value (e.g., the temperature drop at the initial stage of catalyst operation). The dimensionless exponent for pressure drop change is I_P = a_P / ΔP_ref, where ΔP_ref is a reference pressure drop value. I_T and I_P are input into the online monitoring index generation model for catalyst operating status. This model can be a weighted sum model: State index S(t) = w_T × (ΔT_cat(t) - ΔT_base) / ΔT_base + w_P × (ΔP_cat(t) - ΔP_base) / ΔP_base, where w_T and w_P are weights, and ΔT_base and ΔP_base are baseline values ​​(e.g., values ​​at the initial stage of operation). Plotting S(t) as a curve over time yields the real-time monitoring index curve for catalyst operating status. The real-time monitoring index curve of the catalyst operating status reflects the real-time deviation of the overall operating status of the catalyst layer (considering both heat exchange performance and resistance characteristics). The larger the value, the more serious the deviation.

[0141] Step S510: Extract the active site occupancy rate of each partition of the catalyst layer at different operating times from the active site occupancy rate distribution map on the catalyst surface, arrange the active site occupancy rates of each partition of the catalyst layer at different operating times in chronological order, and generate the spatiotemporal evolution trajectory of the active site occupancy rate of each catalyst partition.

[0142] This step aims to track the evolution of performance degradation over time. From the active site occupancy distribution map generated in step S136, not only the data at the final moment is extracted, but also the distribution maps at multiple intermediate moments (e.g., every 24 hours) within the operating cycle. For each partition j, its occupancy rate φ_j(t_k) at different moments t_k is extracted to form a one-dimensional sequence that changes over time, which is the spatiotemporal evolution trajectory of the active site occupancy rate of that partition Φ_j(t).

[0143] Step S520: Extract the sediment coverage ratio of each partition of the catalyst layer at different operating times from the sediment coverage ratio distribution map on the catalyst surface, arrange the sediment coverage ratio of each partition of the catalyst layer at different operating times in chronological order, and generate the spatiotemporal evolution trajectory of sediment coverage ratio for each catalyst partition.

[0144] Similarly, from the sediment cover ratio distribution map generated in step S137, the cover ratio ψ_j(t_k) of each partition j at different times t_k is extracted to form the spatiotemporal evolution trajectory Ψ_j(t) of the sediment cover ratio of that partition.

[0145] Step S530: Calculate the growth rate of active site occupancy rate of each catalyst partition with operating time based on the spatiotemporal evolution trajectory of active site occupancy rate of each catalyst partition, and calculate the growth rate of sediment coverage ratio of each catalyst partition with operating time based on the spatiotemporal evolution trajectory of sediment coverage ratio of each catalyst partition.

[0146] For each partition j, numerical differentiation (e.g., using the central difference method) of its spatiotemporal evolution trajectory Φ_j(t) yields the growth rate of active site occupancy over time, dφ_j / dt. Similarly, differentiation of Ψ_j(t) yields the growth rate of sediment cover over time, dψ_j / dt.

[0147] Step S540: Input the growth rate of the active site occupancy rate with operating time and the growth rate of the sediment coverage ratio with operating time into the catalyst performance degradation rate evaluation model for comparative analysis, and generate the degradation rate dominant factor for each catalyst partition.

[0148] A catalyst performance degradation rate assessment model was constructed. For each partition j, dφ_j / dt and dψ_j / dt were input into the model. The model quantifies the contribution ratio of the two mechanisms to the degradation rate by calculating a degradation rate dominance factor ρ_j. For example, ρ_j = (dφ_j / dt) / (dφ_j / dt + dψ_j / dt). If ρ_j is close to 1, it indicates that the degradation rate is mainly determined by the rate at which active sites are reversibly occupied by reactants; if ρ_j is close to 0, it indicates that the degradation rate is mainly determined by the rate at which deposits irreversibly cover the active sites.

[0149] Step S550: Based on the degradation rate dominance factor of each catalyst zone, divide each zone of the catalyst layer into an activity decay rate dominance zone group and a blockage decay rate dominance zone group.

[0150] Set a partitioning threshold ρ_limit, for example, 0.5. If ρ_j > ρ_limit for a certain partition j, then classify it into the "active decay rate-dominated partition group"; otherwise, classify it into the "blockage decay rate-dominated partition group".

[0151] Step S560: Extract the spatial coordinate range of all catalyst partitions included in the active decay rate dominant partition group, generate the boundary information of the active decay concentration region, and extract the spatial coordinate range of all catalyst partitions included in the blockage decay rate dominant partition group, generate the boundary information of the blockage decay concentration region.

[0152] Cluster analysis is performed on the spatial coordinates of all partitions in the active decay rate-dominant partition group to find the regions they cover and generate boundary information for these regions (e.g., represented by a set of convex hulls or minimum bounding rectangle vertices). Similarly, boundary information is generated for the clogging decay rate-dominant partition group.

[0153] Step S570: Merge the boundary information of the active attenuation concentration region with the boundary information of the blockage attenuation concentration region to generate a spatial distribution map of catalyst layer degradation zones.

[0154] The two types of boundary information are superimposed onto the same spatial distribution map of the catalyst layer. Different colors or patterns are used to fill the two regions, ultimately generating a spatial distribution map of catalyst layer degradation zones. This spatial distribution map distinguishes the spatial locations of different dominant degradation mechanisms within the catalyst layer.

[0155] Step S610: Extract the spatial regions from the thermal map of the spatial distribution of catalyst performance degradation that have a comprehensive performance degradation index exceeding a preset degradation index threshold, and mark the spatial regions as a set of candidate intervention regions.

[0156] This step is a preliminary analysis for refined intervention in areas with severe performance degradation. From the performance degradation heatmap generated in step S155, a degradation index threshold δ_th is set, for example, 0.7. All grid nodes with a comprehensive performance degradation index δ_j ≥ δ_th are extracted and connected to form several candidate intervention regions, which are recorded as a set of candidate intervention regions.

[0157] Step S620: Obtain the time-series variation curve of the inlet reactant flux distribution coefficient of each candidate intervention region in the candidate intervention region set during the operating cycle, and perform frequency domain transformation processing on the time-series variation curve of the inlet reactant flux distribution coefficient to generate a reactant flux spectrum map of each candidate intervention region.

[0158] For each region R in the candidate intervention region set, the time-series variation curve of its inlet reactant flux distribution coefficient can be obtained by averaging the values ​​of all partitions within that region, yielding F_R(t). A fast Fourier transform is then performed on F_R(t) to obtain the reactant flux spectrum for that region.

[0159] Step S630: Extract the dominant frequency from the reactant flux spectrum of each candidate intervention region, obtain the physical structure parameters of the catalyst, input the dominant frequency and the physical structure parameters into the catalyst microchannel resonance risk analysis model, evaluate the matching degree between the flow fluctuation dominant frequency and the inherent frequency of the catalyst channel structure, and calculate the channel resonance risk index of each candidate intervention region.

[0160] The dominant frequency f_R is extracted from the spectrum. The physical structural parameters of the catalyst are obtained, including the hydraulic diameter d_h, wall thickness t_w, and elastic modulus E of the channel. These parameters are then input into the catalyst microchannel resonance risk analysis model. This model is based on structural dynamics and calculates the channel's natural frequency f_nat = f(d_h, t_w, E, flue gas density). The matching degree between f_R and f_nat is then evaluated, and the resonance risk index χ_R = exp(-|f_R - f_nat| / f_nat). This resonance risk index ranges from 0 to 1; when f_R equals f_nat, χ_R = 1, indicating the highest resonance risk.

[0161] Step S640: Based on the channel resonance risk index, sort all candidate intervention regions to generate a candidate intervention region sorting list. Select the top-ranked candidate intervention regions from the candidate intervention region sorting list as the target intervention region set. Obtain the spatial location coordinates of each target intervention region in the target intervention region set in the catalyst performance degradation spatial distribution thermogram.

[0162] All candidate intervention areas are sorted from high to low according to the resonance risk index χ_R. The top N areas (e.g., the top 3) are selected as the target intervention area set. The spatial coordinates of these target intervention areas are obtained from the heat map, such as the coordinates of their center point or the set of boundary points.

[0163] Step S650: Input the spatial coordinates of each target intervention region in the set of target intervention regions into the control model of the ammonia injection device of the selective catalytic reduction reactor, and recalculate the ammonia distribution weight flowing through each target intervention region through the ammonia injection device control model.

[0164] The ammonia injection unit control model is a model that can adjust the opening degree of the ammonia injection branch pipe according to the spatial location. By inputting the spatial coordinates of the target intervention area into the model, the model recalculates the ammonia flow rate ratio w_NH3_R that should be allocated to the ammonia injection branch pipes above or in front of these areas through interpolation or based on the position-valve mapping relationship.

[0165] Step S660: Generate ammonia injection correction control instructions for the target intervention area based on the recalculated ammonia allocation weights.

[0166] Based on the newly calculated ammonia distribution weight w_NH3_R, specific control instructions are generated. These instructions, in the format of standard industrial control protocols (such as Modbus or Profibus), include the identifier of the target ammonia injection branch and the opening value (or flow setpoint) to be adjusted.

[0167] Step S670: Based on the ammonia injection correction control command, trigger the ammonia injection device to perform zoned flow regulation operation so that the ammonia mass flow rate received in the target intervention area is increased compared with other areas, and generate the optimized flue gas flow path after zoned ammonia injection.

[0168] The command generated in step S660 is sent to the controller of the on-site ammonia injection unit. The controller adjusts the opening of the regulating valve on the corresponding ammonia injection branch pipe accordingly, increasing the amount of ammonia flowing to the target intervention area. This operation alters the spatial distribution of the ammonia / flue gas mixture behind the ammonia injection grid, thereby generating a new, optimized flue gas flow path (i.e., localized ammonia compensation for deteriorated areas).

[0169] For example, after step S140, the method may further include: step S710: performing spatial clustering analysis on the intrinsic activity retention ratio and the flow resistance increase ratio of each catalyst partition, identifying catalyst partitions with an intrinsic activity retention ratio lower than a preset activity retention threshold and a flow resistance increase ratio higher than a preset resistance increase threshold, and marking the identified catalyst partitions as a risk deterioration partition set.

[0170] This step identifies high-risk, potentially rapidly expanding deterioration regions. Spatial clustering analysis is performed on the partition data γ_j and ζ_j obtained in steps S144 and S147, for example, using the DBSCAN clustering algorithm. An activity retention threshold γ_limit (e.g., 0.5) and a resistance increase threshold ζ_limit (e.g., 0.8) are set. All partitions that simultaneously satisfy γ_j < γ_limit and ζ_j > ζ_limit are identified and marked as "risk deterioration partitions," forming a set.

[0171] Step S720: Obtain the spatial coordinates of each risk degradation partition in the risk degradation partition set in the catalyst performance degradation spatial distribution heat map, extract the comprehensive performance degradation index distribution within a preset radius around each risk degradation partition from the catalyst performance degradation spatial distribution heat map, and generate the degradation field distribution of the neighborhood of the risk degradation partition.

[0172] For each partition p in the risk degradation partition set, obtain its spatial coordinates (x_p, y_p). In the heat map generated in step S155, a circular region with a radius of R_neighborhood (e.g., 1 meter) is delineated with (x_p, y_p) as the center. The comprehensive performance degradation index δ value of all grid points in this region is extracted to form the neighborhood degradation field distribution of the risk degradation partition.

[0173] Step S730: Perform gradient vector field calculation on the neighborhood degradation field distribution of each risk degradation partition to obtain the degradation gradient direction and degradation gradient magnitude of the neighborhood degradation field of each risk degradation partition. Input the spatial coordinates of each risk degradation partition, the degradation gradient direction, the degradation gradient magnitude, and the comprehensive performance degradation index of the risk degradation partition at the current moment into the catalyst degradation expansion trend inference model for iterative simulation.

[0174] For the neighborhood field of each risk-deteriorated partition p, its gradient is calculated to obtain the degradation gradient direction (pointing to the direction of the fastest increase in δ) and gradient magnitude at each point. This information, along with the spatial coordinates of partition p and its own comprehensive performance degradation index δ_p, is input into the catalyst degradation propagation trend inference model. This catalyst degradation propagation trend inference model is a physics- or data-driven spatiotemporal evolution model; for example, it can simulate how the degradation state of the current high-risk area "propagates" or "expands" along the gradient direction to its neighborhood based on the principle of degradation gradient diffusion.

[0175] Step S740: Simulate the process of the degradation state of the risk degradation zone expanding to its neighboring catalyst zones along the degradation gradient direction using the catalyst degradation extension trend inference model, and generate a predicted activity distribution map containing the predicted intrinsic activity retention ratio of each catalyst zone within the future preset running time and a predicted resistance distribution map containing the predicted flow resistance increase ratio of each catalyst zone within the future preset running time.

[0176] After iterative simulation, the model outputs predictions for a predetermined runtime T_future (e.g., the next 1000 hours). The predictions include two spatial distribution maps: one is a predicted activity distribution map, where each point predicts the proportion of intrinsic activity retained at that location after the next T_future, γ_pred(x, y); the other is a predicted drag distribution map, which predicts the proportion of increase in flow drag, ζ_pred(x, y).

[0177] Step S750: Extract the predicted intrinsic activity retention ratio of each zone of the catalyst at the end of the future preset running time from the predicted activity distribution map, and extract the predicted flow resistance increase ratio of each zone of the catalyst at the end of the future preset running time from the predicted resistance distribution map.

[0178] At the end of the prediction time T_future, the predicted intrinsic activity retention ratio γ_pred_j for each partition j is read from the predicted activity distribution map, and the predicted flow resistance increase ratio ζ_pred_j for each partition j is read from the predicted resistance distribution map.

[0179] Step S760: Input the predicted intrinsic activity retention ratio and the predicted flow resistance increase ratio of each catalyst zone into the catalyst life end determination model for evaluation and calculation. If the predicted intrinsic activity retention ratio of a catalyst zone is lower than the preset life end activity threshold or the predicted flow resistance increase ratio is higher than the preset life end resistance threshold, then the catalyst is determined to have reached the overall life end, and a catalyst overall replacement warning signal is generated.

[0180] The catalyst lifetime end determination model is a simple threshold checker. A lifetime end activity threshold γ_EOL (e.g., 0.2) and a lifetime end resistance threshold ζ_EOL (e.g., 2.0) are set. If any partition j exists such that γ_pred_j < γ_EOL or ζ_pred_j > ζ_EOL, then the entire catalyst layer is determined to have reached or will reach its overall lifetime end at the end of the future T_future. At this point, a "catalyst overall replacement early warning signal" is generated.

[0181] Step S770: Associate and encapsulate the overall catalyst replacement early warning signal with the catalyst remaining efficiency index and catalyst maintenance area location information to generate a comprehensive catalyst performance evaluation report containing life end prediction information.

[0182] Finally, the warning signal generated in step S760, the remaining efficiency index θ_remain calculated in step S157, and the maintenance area location information generated in step S156 are packaged and correlated to generate a comprehensive catalyst performance evaluation report. This comprehensive catalyst performance evaluation report not only includes an assessment of the current state but also a prediction of the future end of its lifespan.

[0183] In one exemplary embodiment, an online performance evaluation system for SCR catalysts based on operational data is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2As shown, it includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements an online performance evaluation method for SCR catalysts based on operational data. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the housing of the online evaluation system for SCR catalyst performance based on operating data, or an external keyboard, touchpad, or mouse, etc.

[0184] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for online evaluation of SCR catalyst performance based on operational data, characterized in that, The method includes: Obtain a time-series dataset generated during the operating cycle of the selective catalytic reduction reactor, the time-series dataset including the inlet flue gas parameter sequence of the denitrification reactor, the outlet flue gas parameter sequence of the denitrification reactor, and the control parameter sequence of the ammonia injection device; Based on the time-series dataset, a reactant concentration distribution model for the denitrification reaction zone is constructed, and the inlet reactant flux distribution coefficient and outlet product generation rate distribution coefficient of each partition of the catalyst layer are decoupled. The inlet reactant flux distribution coefficient and the outlet product generation rate distribution coefficient are used as boundary constraints and input into the catalyst surface micro-reaction site occupancy state deduction model. Iterative convergence processing is initiated to generate a catalyst surface active site occupancy rate distribution map and a catalyst surface deposition coverage ratio distribution map. Based on the distribution map of the occupancy rate of active sites on the catalyst surface and the distribution map of the coverage ratio of deposits on the catalyst surface, the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst were calculated. A thermal map of the spatial distribution of catalyst performance degradation is generated based on the intrinsic activity retention ratio of each zone of the catalyst and the increase ratio of the flow resistance of each zone of the catalyst. Based on the thermal map of the spatial distribution of catalyst performance degradation, the catalyst remaining efficiency index and the location information of the catalyst maintenance area are output.

2. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The step of constructing a reactant concentration distribution model for the denitrification reaction zone based on the time-series dataset, and decoupling the inlet reactant flux distribution coefficient and outlet product formation rate distribution coefficient for each partition of the catalyst layer, includes: Extract the inlet nitrogen oxide concentration value and inlet oxygen concentration value contained in the inlet flue gas parameter sequence of the denitrification reactor from the time series dataset, and arrange the inlet nitrogen oxide concentration value and the inlet oxygen concentration value in the time order within the operating cycle to form an inlet reactant concentration time series matrix; Extract the flue gas flow time series data of the inlet and outlet of the denitrification reactor from the time series dataset, and multiply the nitrogen oxide concentration value at the inlet with the inlet flue gas flow rate value at the corresponding time to obtain the inlet nitrogen oxide mass flow rate time series sequence. Multiply the nitrogen oxide concentration value at the outlet by the corresponding flue gas flow rate value at the outlet to obtain the time series sequence of nitrogen oxide mass flow rate at the outlet. Based on the time series sequence of nitrogen oxide mass flow rate at the inlet, the time series sequence of nitrogen oxide mass flow rate at the outlet, and the ammonia injection flow rate value contained in the control parameter sequence of the ammonia injection device, establish a set of mass conservation equations on the flue gas flow path inside the denitrification reactor. Each equation in the set of mass conservation equations corresponds to a differential unit cross section of the catalyst layer in the flue gas flow direction. The non-uniformity coefficient of flue gas flow distribution on the horizontal section of the catalyst layer is introduced as a variable to be solved in the mass conservation equation system. The mass conservation equation system is solved by numerical iteration method to obtain the non-uniformity coefficient of flue gas flow distribution and the flue gas flow distribution weight vector corresponding to each partition of the catalyst layer. Based on the flue gas flow rate allocation weight vector and the inlet nitrogen oxide mass flow rate time series, the inlet nitrogen oxide mass flow rate allocated to each zone of the catalyst layer is calculated; the ratio of the inlet nitrogen oxide mass flow rate of each zone to the total inlet nitrogen oxide mass flow rate is used as the inlet reactant flux allocation coefficient of each zone of the catalyst layer, and the inlet reactant flux allocation coefficient is used to characterize the proportion of reactant mass flow rate entering each catalyst zone to the total inlet reactant mass flow rate; The inlet reactant flux distribution coefficient of each zone of the catalyst layer is substituted into the mass conservation equations to solve the material balance relationship between reactant consumption rate and product formation rate. The outlet product formation rate distribution coefficient of each zone of the catalyst layer is extracted from the material balance relationship. The outlet product formation rate distribution coefficient is used to characterize the proportion of the denitrification product mass flow rate generated by each catalyst zone to the total outlet denitrification product mass flow rate.

3. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The process of inputting the inlet reactant flux distribution coefficient and the outlet product generation rate distribution coefficient as boundary constraints into the catalyst surface micro-reaction site occupancy state deduction model, initiating iterative convergence processing, and generating a catalyst surface active site occupancy rate distribution map and a catalyst surface deposit coverage ratio distribution map includes: A library of active site reaction kinetic parameters is obtained from the model of the occupancy state of micro-reaction sites on the catalyst surface. The library of active site reaction kinetic parameters stores the standard reaction rate constants of different types of active sites on the catalyst surface reacting with nitrogen oxides, the standard adsorption equilibrium constants of different types of active sites on the catalyst surface reacting with ammonia, and the deactivation rate decay factor when different types of active sites on the catalyst surface are covered by ammonium sulfate deposits. The inlet reactant flux distribution coefficient is mapped according to the spatial coordinate position of each partition of the catalyst layer to obtain the time series curve of the inlet reactant flux received by each catalyst partition during the operating cycle. The reactant flux fluctuation amplitude and the peak reactant flux time of each catalyst partition are extracted from the inlet reactant flux time series curve. The export product generation rate allocation coefficient is mapped according to the spatial coordinate position of each partition of the catalyst layer to obtain the time series curve of the cumulative export product generation rate of each catalyst partition during the operation cycle. The product generation rate response lag time and product generation rate steady-state deviation of each catalyst partition are extracted from the export product generation rate time series curve. The dynamic input conditions or related parameters of the catalyst surface micro-reaction site occupancy state inference model are defined and calibrated using the reactant flux fluctuation amplitude, the reactant flux peak time, the product formation rate response lag time, and the product formation rate steady-state deviation. These serve as the basis for model iterative calculation and trigger the catalyst surface micro-reaction site occupancy state inference model to start the active site occupancy state iterative calculation process. During the iterative calculation of the active site occupancy status, the catalyst surface micro-reaction site occupancy status deduction model repeatedly adjusts the proportion of reactants occupied and the proportion of deposits covered by different types of active sites in each catalyst partition according to the dynamic balance between the reactant arrival rate and the product departure rate of each catalyst partition, until the sum of the residuals between the calculated reactant consumption rate and the calculated product generation rate of each catalyst partition is lower than the preset convergence threshold, at which point the iterative calculation is terminated. The proportion of reactants occupying different types of active sites in each catalyst partition is extracted from the iterative calculation results output by the model for the occupancy status of micro-reaction sites on the catalyst surface. The proportion of reactants occupying all catalyst partitions is interpolated and filled according to spatial coordinates to generate a distribution map of active site occupancy rate on the catalyst surface. The deposition coverage ratio of different types of active sites in each catalyst partition is extracted from the iterative calculation results output by the micro-reaction site occupancy state inference model on the catalyst surface. The deposition coverage ratio of all catalyst partitions is interpolated and filled according to spatial coordinates to generate a deposition coverage ratio distribution map of the catalyst surface.

4. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The calculation of the intrinsic activity retention ratio and the increase in flow resistance of each catalyst zone based on the catalyst surface active site occupancy distribution map and the catalyst surface deposit coverage ratio distribution map includes: Extract the final occupancy rate of active sites for each catalyst partition at the end of the operating cycle from the catalyst surface active site occupancy rate distribution map, obtain the initial active site density of the catalyst in the unused state, and calculate the effective active site density of each catalyst partition in the current state based on the deposition coverage ratio distribution map of the catalyst surface and the initial active site density of the catalyst in the unused state. Calculate the ratio of the effective active site density of each catalyst partition in the current state to the initial active site density to obtain the effective number ratio coefficient of active sites for each catalyst partition. Based on the effective number ratio coefficient of active sites and the deposition coverage ratio distribution map of catalyst surface, an active site availability coupling equation is established for each catalyst partition. The effective exposure area of ​​active sites in each catalyst partition is obtained by solving the active site availability coupling equation. The initial intrinsic reaction rate constant of the catalyst in the unused state is obtained, and the initial intrinsic reaction rate constant is multiplied by the effective number ratio coefficient of active sites in each catalyst partition to obtain the actual reaction rate potential value of each catalyst partition in the current state. The intrinsic activity retention ratio of each catalyst partition is obtained by calculating the ratio of the actual reaction rate potential value of each catalyst partition in the current state to the initial reaction rate potential value corresponding to the initial intrinsic reaction rate constant. The intrinsic activity retention ratio is used to characterize the degree to which the current reaction capacity of the catalyst partition is maintained relative to the initial reaction capacity. Extract the deposition coverage ratio of each catalyst zone from the deposition coverage ratio distribution map on the catalyst surface, and calculate the remaining flow cross-sectional area ratio of each catalyst zone based on the geometric mapping relationship between the deposition coverage ratio and the blockage area of ​​the catalyst channel cross-section. The initial channel flow cross-sectional area of ​​the catalyst in its unused state is obtained. Based on the physical relationship between flow resistance and channel flow cross-sectional area in fluid mechanics, the theoretical flow resistance of each catalyst zone in its current state is calculated from the proportion of the remaining channel flow cross-sectional area of ​​each catalyst zone. The ratio of the theoretical flow resistance of each catalyst zone in its current state to the theoretical initial flow resistance corresponding to the initial channel cross-sectional area is calculated to obtain the increase ratio of the flow resistance of each catalyst zone. The increase ratio of the flow resistance is used to characterize the degree of increase of the current flow resistance of the catalyst zone relative to the initial flow resistance.

5. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The process of generating a spatial distribution heatmap of catalyst performance degradation based on the intrinsic activity retention ratio of each catalyst zone and the increase ratio of flow resistance in each catalyst zone, and outputting the catalyst remaining efficiency index and catalyst maintenance area location information based on the spatial distribution heatmap of catalyst performance degradation, includes: The catalyst layer is divided into catalyst partition grids with spatial coordinates according to its geometric structure. The intrinsic activity retention ratio and the flow resistance increase ratio corresponding to each catalyst partition are assigned to the corresponding grid nodes in the catalyst partition grid, forming a spatial distribution field of intrinsic activity retention ratio and a spatial distribution field of flow resistance increase ratio. Spatial gradient calculation is performed on the spatial distribution field of the intrinsic activity retention ratio to obtain the decay gradient direction and decay gradient magnitude of the intrinsic activity retention ratio in the horizontal direction of the catalyst layer. The coordinates of grid nodes whose decay gradient magnitude exceeds a preset decay threshold are selected from the decay gradient magnitude as the location points of the activity decay acceleration region. Spatial gradient calculation is performed on the spatial distribution field of the increase in flow resistance to obtain the blocking gradient direction and blocking gradient magnitude of the increase in flow resistance in the horizontal direction of the catalyst layer. The grid node coordinates with blocking gradient magnitudes exceeding a preset blocking threshold are selected from the blocking gradient magnitudes as the location points of the blocking acceleration region. The intrinsic activity retention ratio spatial distribution field and the flow resistance increase ratio spatial distribution field are input into the catalyst performance comprehensive evaluation function for weighted fusion calculation to generate a comprehensive performance degradation index corresponding to each catalyst partition. The weight coefficients of the intrinsic activity retention ratio and the flow resistance increase ratio in the catalyst performance comprehensive evaluation function are dynamically adjusted according to the catalyst operation history data. According to the preset color mapping rules, the comprehensive performance degradation index corresponding to each catalyst partition is converted into a color code value. Based on the color code value, each grid node in the catalyst partition grid is colored and rendered to generate a heat map of the spatial distribution of catalyst performance degradation based on spatial coordinates. Extract the spatial coordinates corresponding to all catalyst partitions whose comprehensive performance degradation index is lower than the preset efficiency lower limit threshold from the thermal map of the spatial distribution of catalyst performance degradation. Merge all extracted spatial coordinates into connected regions to generate catalyst maintenance area positioning information. The comprehensive performance degradation index of all catalyst zones in the catalyst layer is obtained. The comprehensive performance degradation index of all catalyst zones is input into the overall catalyst efficiency evaluation model for weighted average calculation, and the remaining catalyst efficiency index is output. The remaining catalyst efficiency index is used to characterize the overall performance retention level of the catalyst layer in the current state relative to the initial state.

6. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The method further includes: The time-series variation curves of the inlet reactant flux distribution coefficients accumulated in each zone of the catalyst layer during the operating cycle are obtained. The reactant flux fluctuation spectrum characteristics of each catalyst zone are extracted from the time-series variation curves of the inlet reactant flux distribution coefficients. The reactant flux fluctuation spectrum characteristics include the amplitude of the main frequency of fluctuation, the phase offset of the secondary frequency of fluctuation, and the energy decay rate of fluctuation. The time-series variation curves of the cumulative outlet product generation rate distribution coefficients of each partition of the catalyst layer during the operating cycle are obtained. The product generation rate response spectrum characteristics of each catalyst partition are extracted from the time-series variation curves of the outlet product generation rate distribution coefficients. The product generation rate response spectrum characteristics include the response main frequency delay time, the response secondary frequency gain coefficient, and the response bandwidth contraction ratio. The spectral characteristics of reactant flux fluctuations and the spectral characteristics of product formation rate response are input into the catalyst dynamic response characteristic analysis model for transfer function identification, generating a reaction-mass transfer coupling transfer function expression for each catalyst partition. The reaction-mass transfer coupling transfer function expression includes the reaction rate dominant pole parameter and the mass transfer rate dominant zero parameter. The reaction rate dominant pole parameter is the dominant root of the denominator polynomial of the transfer function, and the mass transfer rate dominant zero parameter is the dominant root of the numerator polynomial of the transfer function. The performance degradation dominant mechanism type of each catalyst partition is determined based on the numerical relationship between the reaction rate-dominant pole parameter and the mass transfer rate-dominant zero parameter. The performance degradation dominant mechanism type includes chemical reaction rate-controlled degradation mechanism and mass transfer-diffusion rate-controlled degradation mechanism. Catalyst partitions belonging to the chemical reaction rate-controlled decay mechanism are marked as activity decay-dominant partitions, and catalyst partitions belonging to the mass transfer diffusion rate-controlled decay mechanism are marked as blockage decay-dominant partitions, generating a catalyst layer partition decay mechanism classification map. Based on the time-series change curve of the inlet reactant flux distribution coefficient corresponding to the active decay dominant partition in the catalyst layer partition decay mechanism classification diagram, the cumulative reactant processing total of the active decay dominant partition during the operating cycle is extracted, and the cumulative reactant processing total is input into the catalyst active decay lifetime prediction model for extrapolation and calculation, and the first remaining operating time prediction value of the active decay dominant partition is output. Based on the sediment-related operating parameters of the blockage attenuation dominant partition in the catalyst layer partition attenuation mechanism classification diagram during the operating cycle, the potential sediment accumulation estimate of the blockage attenuation dominant partition during the operating cycle is extracted. The potential sediment accumulation estimate is input into the catalyst blockage attenuation lifetime prediction model for extrapolation and calculation, and the second remaining operating time prediction value of the blockage attenuation dominant partition is output. The minimum value of the first remaining runtime prediction value and the second remaining runtime prediction value is selected to obtain the overall remaining runtime prediction result of the catalyst layer. The overall remaining runtime prediction result of the catalyst layer is used to characterize the remaining time that the catalyst layer can continue to maintain its denitrification performance under the current operating state.

7. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The method further includes: The final occupancy rate of active sites for each catalyst partition at the end of the operating cycle is extracted from the catalyst surface active site occupancy rate distribution map, and the final deposition coverage ratio for each catalyst partition at the end of the operating cycle is extracted from the catalyst surface deposition coverage ratio distribution map. Based on the final occupancy rate of the active sites and the final coverage ratio of the sediment, an active site availability correlation equation is established for each catalyst partition. By solving the active site availability correlation equation, the effective exposure area decay coefficient of the active sites in each catalyst partition is obtained. The active site availability correlation equation contains a linear superposition relationship between the portion of the active sites reversibly occupied by the reactants and the portion of the active sites irreversibly covered by the sediment. The effective exposure area decay coefficient of the active site of each catalyst partition is input into the catalyst partition reaction rate reconstruction model for parameter inversion calculation, and a reaction rate constant correction factor for each catalyst partition in the current state is generated. The reaction rate constant correction factor is used to characterize the degree of deviation between the actual reaction rate of the catalyst partition and the initial reaction rate. Obtain the set of initial reaction rate constants for each partition of the catalyst layer in the initial state, and multiply each initial reaction rate constant in the set of initial reaction rate constants with the reaction rate constant correction factor of the corresponding catalyst partition to obtain the actual reaction rate constant of each catalyst partition in the current state. The actual reaction rate constant of each catalyst partition in the current state is arranged according to the spatial coordinates of each partition of the catalyst layer to generate a spatial distribution matrix of the reaction rate constant of the catalyst layer. The average operating parameters of the catalyst bed during the operating cycle are obtained. The operating parameters include at least the inlet flue gas flow rate, the inlet nitrogen oxide concentration, and the residence time of the flue gas in the catalyst bed. The spatial distribution matrix of the reaction rate constant of the catalyst bed and the average operating parameters are input into the overall denitrification efficiency model of the catalyst bed for calculation. The predicted value of the overall denitrification efficiency of the catalyst bed under the current state is obtained. The overall denitrification efficiency model of the catalyst bed includes a nonlinear mapping relationship between the spatial distribution of the reaction rate constant and the denitrification efficiency. The initial overall denitrification efficiency of the catalyst layer in its initial state is obtained. The ratio of the predicted overall denitrification efficiency to the initial overall denitrification efficiency is calculated to obtain the catalyst layer comprehensive performance retention rate index. The catalyst layer comprehensive performance retention rate index is used to characterize the retention level of the overall denitrification performance of the catalyst layer relative to the initial denitrification performance.

8. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The method further includes: Extract the inlet flue gas temperature time series data and inlet flue gas pressure time series data contained in the inlet flue gas parameter sequence of the denitrification reactor from the time series dataset; extract the outlet flue gas temperature time series data and outlet flue gas pressure time series data contained in the outlet flue gas parameter sequence of the denitrification reactor from the time series dataset. The catalyst layer temperature drop at each operating moment is calculated based on the inlet flue gas temperature time series data and the outlet flue gas temperature time series data. The catalyst layer pressure drop at each operating moment is calculated based on the inlet flue gas pressure time series data and the outlet flue gas pressure time series data, generating a catalyst layer temperature drop time series curve and a catalyst layer pressure drop time series curve. Based on the final deposition coverage ratio of all catalyst zones in the catalyst surface deposition coverage ratio distribution map, the average deposition coverage ratio of the catalyst layer is calculated; based on the average deposition coverage ratio and the final temperature drop value at the end of the operating cycle of the catalyst layer temperature drop time series curve, a first correlation regression model between the deposition coverage ratio and the temperature drop change is established. Based on the final occupancy rate of active sites in all catalyst zones in the catalyst surface active site occupancy rate distribution diagram, the average active site occupancy rate of the catalyst layer is calculated; based on the average active site occupancy rate and the final pressure drop value at the end of the operating cycle of the catalyst layer pressure drop time series curve, a second correlation regression model between the active site occupancy rate and the pressure drop change is established. The first correlation regression model is used to calculate the temperature drop corresponding to each unit increase in sediment coverage, and the second correlation regression model is used to calculate the pressure drop corresponding to each unit increase in active site occupancy. The temperature drop change and the pressure drop change are respectively processed into dimensionless values ​​to obtain the dimensionless index of temperature drop change and the dimensionless index of pressure drop change. These are then input into the catalyst operating status online monitoring index generation model for fusion calculation, and the catalyst operating status real-time monitoring index curve is output. The catalyst operating status real-time monitoring index curve is used to characterize the degree of deviation of the comprehensive operating status of the catalyst layer at each moment during the operating cycle.

9. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, The method further includes: The active site occupancy rate of each partition of the catalyst layer at different operating times is extracted from the active site occupancy rate distribution map on the catalyst surface. The active site occupancy rates of each partition of the catalyst layer at different operating times are arranged in chronological order to generate the spatiotemporal evolution trajectory of the active site occupancy rate of each catalyst partition. Extract the deposition coverage ratio of each zone of the catalyst layer at different operating times from the deposition coverage ratio distribution map on the catalyst surface, arrange the deposition coverage ratio of each zone of the catalyst layer at different operating times in chronological order, and generate the spatiotemporal evolution trajectory of the deposition coverage ratio of each catalyst zone. The growth rate of active site occupancy rate of each catalyst partition with operating time was calculated based on the spatiotemporal evolution trajectory of active site occupancy rate of each catalyst partition, and the growth rate of sediment coverage ratio of each catalyst partition with operating time was calculated based on the spatiotemporal evolution trajectory of sediment coverage ratio of each catalyst partition. The growth rate of the active site occupancy rate with operating time and the growth rate of the deposit coverage ratio with operating time are input into the catalyst performance degradation rate evaluation model for comparative analysis to generate the degradation rate dominant factor for each catalyst partition. The degradation rate dominant factor is used to characterize the contribution ratio of activity decay and blockage decay in the degradation process of catalyst partition. Based on the degradation rate dominance factor of each catalyst zone, the catalyst layer is divided into an activity decay rate dominance zone group and a blockage decay rate dominance zone group. Extract the spatial coordinate range of all catalyst partitions included in the active decay rate dominant partition group to generate the boundary information of the active decay concentration region; extract the spatial coordinate range of all catalyst partitions included in the blockage decay rate dominant partition group to generate the boundary information of the blockage decay concentration region. The boundary information of the concentrated area of ​​active degradation and the boundary information of the concentrated area of ​​blockage degradation are merged to generate a spatial distribution map of catalyst layer degradation. The spatial distribution map of catalyst layer degradation is used to indicate the spatial distribution location of different degradation dominant mechanisms in the catalyst layer.

10. The online performance evaluation method for SCR catalysts based on operational data according to claim 1, characterized in that, After generating the spatial distribution heatmap of catalyst performance degradation based on the intrinsic activity retention ratio and the increase ratio of the flow resistance of each catalyst zone, and before outputting the catalyst remaining efficiency index and catalyst maintenance area location information, the method further includes: From the thermal map of the spatial distribution of catalyst performance degradation, spatial regions whose comprehensive performance degradation index exceeds a preset degradation index threshold are extracted, and these spatial regions are marked as a set of candidate intervention regions. Obtain the time-series variation curve of the inlet reactant flux distribution coefficient for each candidate intervention region in the candidate intervention region set during the operating cycle, and perform frequency domain transformation on the time-series variation curve of the inlet reactant flux distribution coefficient to generate a reactant flux spectrum map for each candidate intervention region. The dominant frequency is extracted from the reactant flux spectrum of each candidate intervention region to obtain the physical structure parameters of the catalyst. The dominant frequency and the physical structure parameters are input into the catalyst microchannel resonance risk analysis model to evaluate the degree of matching between the flow fluctuation dominant frequency and the inherent frequency of the catalyst channel structure, and the channel resonance risk index of each candidate intervention region is calculated. Based on the channel resonance risk index, all candidate intervention regions are sorted to generate a candidate intervention region sorting list. The candidate intervention regions with the highest sorting are selected from the candidate intervention region sorting list as the target intervention region set. The spatial location coordinates of each target intervention region in the target intervention region set are obtained in the catalyst performance degradation spatial distribution thermogram. The spatial coordinates of each target intervention region in the set of target intervention regions are input into the control model of the ammonia injection device of the selective catalytic reduction reactor, and the ammonia distribution weight flowing through each target intervention region is recalculated through the ammonia injection device control model. Based on the recalculated ammonia distribution weight, an ammonia injection correction control command is generated for the target intervention area. The ammonia injection correction control command is used to adjust the ammonia distribution ratio in the branch pipe of the ammonia injection device that flows through the corresponding spatial position of the target intervention area. Based on the ammonia injection correction control command, the ammonia injection device is triggered to perform a zoned flow regulation operation, so that the mass flow rate of ammonia gas received in the target intervention area is increased compared with other areas, and a zoned ammonia injection optimized flue gas flow path is generated.

Images