A urea pyrolysis multi-objective optimization method, device, storage medium and program product

Abstract

The application discloses a urea pyrolysis multi-objective optimization method, device, storage medium and program product, belongs to the field of SCR denitration control, and aims to solve the dynamic balance problem of denitration efficiency, energy consumption and ammonia escape risk. The method comprises the following steps: identifying a current process scene type based on real-time operation data; according to the scene type, combining a weight adjustment signal for compensating time lag and an inference result of a fuzzy inference system, dynamically determining final optimization weights for balancing denitration efficiency, energy consumption and ammonia escape risk; using the weights to solve a target control scheme through a multi-objective optimization algorithm, and controlling the system according to the target control scheme. Through scene self-adaptation and dynamic weight adjustment, the application realizes intelligent control of complex working conditions, effectively reduces operation cost while ensuring emission compliance.

Description

Technical Field

[0001] This application relates to the field of SCR denitrification control, and in particular to a multi-objective optimization method, equipment, storage medium and program product for urea pyrolysis. Background Technology

[0002] Urea pyrolysis is the mainstream ammonia source supply technology for SCR denitrification systems in diesel engines and coal-fired boilers. It involves heating and decomposing urea solution into ammonia and carbon dioxide in a pyrolysis furnace, then supplying the generated ammonia to the SCR reactor for NOx reduction. This technology is widely used in both mobile and stationary denitrification systems because it eliminates the need for liquid ammonia storage.

[0003] Existing urea pyrolysis control methods typically employ a single-objective optimization strategy based on a PID controller. This involves setting the urea injection flow rate according to the NOx concentration at the SCR inlet while maintaining the pyrolysis furnace temperature within a preset range. Some technical solutions introduce a feedforward-feedback composite control, calculating the theoretical urea demand using feedforward signals of flue gas flow rate and NOx concentration, and then correcting it using feedback signals of the NOx concentration at the SCR outlet. Other technical solutions attempt to establish an empirical model of pyrolysis furnace temperature and urea conversion rate, determining the optimal pyrolysis temperature under different operating conditions using a lookup table method.

[0004] However, these methods present a multi-objective trade-off in actual operation. On the one hand, ensuring denitrification efficiency requires increasing the pyrolysis furnace temperature and urea injection rate, which leads to a significant increase in electric heater power consumption and urea consumption. On the other hand, excessive ammonia supply can cause ammonia escape exceeding the limit, while reducing ammonia supply will reduce denitrification efficiency. When boiler load fluctuates, existing control methods often struggle to maintain the expected balance between denitrification efficiency, energy consumption, and ammonia escape risk, resulting in excessively high operating costs for the system while meeting emission standards. Summary of the Invention

[0005] This application provides a multi-objective optimization method, equipment, storage medium, and program product for urea pyrolysis, which can reduce the risk of ammonia escape while improving denitrification efficiency.

[0006] Firstly, this application provides a multi-objective optimization method for urea pyrolysis. The method includes: calculating the state characteristics of the pyrolysis system based on real-time operating data, including pyrolysis parameters, denitrification parameters, and system operating parameters; inputting the pyrolysis system state characteristics into a scenario recognition decision tree to identify the current process scenario type; determining the corresponding initial weight vector based on the process scenario type, and adjusting the weights by combining a weight adjustment signal and a weight adjustment amount output by a fuzzy inference system to obtain the final optimized weights, which include denitrification efficiency weights, energy consumption weights, and ammonia escape risk weights; the fuzzy inference system infers the weight adjustment amount based on preset urea pyrolysis process coupling rules, and the weight adjustment signal is used to compensate for time delay effects and adapt to load changes; using the final optimized weights to solve a weighted multi-objective optimization problem using a multi-objective evolutionary algorithm to obtain a Pareto solution set, where the objective function of the weighted multi-objective optimization problem is related to the temperature dependence and thermal inertia characteristics of urea pyrolysis; and selecting a target control scheme from the Pareto solution set based on the process scenario type, where the target control scheme includes pyrolysis system control parameters and urea supply control parameters.

[0007] This embodiment constructs a scenario recognition decision tree and a dynamic weight adjustment mechanism, enabling the multi-objective optimization process to adaptively adjust the trade-off strategy between denitrification efficiency, energy consumption, and ammonia slip risk according to the current process scenario type. The system presets initial weight vectors for different process scenarios and compensates for time-delay effects through weight adjustment signals. It also captures the process coupling relationship between NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation through a fuzzy inference system. Both factors jointly correct the initial weights to obtain the final optimized weights. These weights drive a multi-objective evolutionary algorithm to generate a Pareto solution set, and then select the most suitable control scheme based on the scenario type. This scheme embeds process scenario cognition into the weight adjustment process, achieving adaptive tracking of the optimization strategy to operating conditions, reducing the system's operating costs while meeting emission standards, and improving the adaptability of control parameters to load changes and time-delay effects.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes: generating hierarchical control commands based on the response speed differences of the actuators, the hierarchical control commands including urea injection pump control commands for the fast layer, hot air fan control commands for the medium-speed layer, and electric heater control commands for the slow layer, the electric heater control commands for the slow layer being smoothed and filtered; issuing the hierarchical control commands to the field actuators, and monitoring the pyrolysis furnace temperature response curve, ammonia concentration change trend, NOx outlet concentration response, and ammonia escape value during execution; collecting actual effect data at a preset evaluation period after the execution of the hierarchical control commands, calculating compliance evaluation indicators, efficiency evaluation indicators, stability evaluation indicators, and time-delay response accuracy indicators based on the actual effect data; and updating the initial weight vectors corresponding to each process scenario type and the parameters of the fuzzy inference system based on the compliance evaluation indicators, efficiency evaluation indicators, stability evaluation indicators, and time-delay response accuracy indicators.

[0009] This embodiment introduces a closed-loop mechanism of execution monitoring and parameter self-updating after the optimization scheme is generated. The system generates hierarchical control commands based on the differences in the response speed of the actuators. Commands for the slow-speed electric heaters are smoothed and filtered to avoid frequent actions. During command execution, the actual responses of the pyrolysis furnace temperature, ammonia concentration, NOx outlet concentration, and ammonia escape value are monitored simultaneously. After the preset evaluation period, the system collects actual performance data to calculate four evaluation indicators: compliance, efficiency, stability, and time-delay response accuracy. Based on these indicators, the initial weight vectors and fuzzy inference system parameters corresponding to each process scenario are updated. This scheme, by feeding back the execution results to the weight configuration and inference rules, achieves continuous adaptation of the optimization model to the actual process characteristics, reducing the impact of initial parameter setting deviations on control performance.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the weight adjustment signal is obtained by the following method: A time-delay model is constructed, which characterizes the first time delay of adjusting the urea injection flow rate to ammonia production via pyrolysis, the second time delay of ammonia delivery from the pyrolysis furnace to the SCR reactor, and the third time delay of adjusting the pyrolysis furnace temperature to a stable state; based on historical flue gas flow data and historical NOx load data, a time series prediction model is used to predict the predicted values ​​of flue gas flow rate and NOx load within a future preset time period; the expected impact of the current urea injection flow rate and pyrolysis furnace temperature setpoint on denitrification efficiency, energy consumption, and ammonia escape risk within the future preset time period is calculated based on the time-delay model; the required ammonia supply change trend and pyrolysis furnace temperature adjustment trend within the future preset time period are calculated based on the flue gas flow rate prediction and NOx load prediction; based on the expected impact, the ammonia supply change trend, and the pyrolysis furnace temperature adjustment trend, combined with the degree to which the current load deviates from the steady-state operating point, a weight adjustment signal is generated to adjust the initial weight vector, the weight adjustment signal including a time-delay compensation component and a load adaptation component.

[0011] This embodiment constructs a time-delay model and a time-series prediction mechanism to enable the weight adjustment signal to compensate for the multi-stage time-delay effects in the urea pyrolysis process. The time-delay model constructed by the system characterizes the time delay characteristics of three stages: urea injection to ammonia production, ammonia delivery to the SCR reactor, and pyrolysis furnace temperature adjustment. Based on this, it calculates the expected impact of the current control action after each stage of time delay on future denitrification efficiency, energy consumption, and ammonia escape risk. Simultaneously, the system predicts future flue gas flow and NOx load using a time-series prediction model, and extrapolates the adjustment trends of ammonia supply and pyrolysis furnace temperature. The weight adjustment signal integrates the expected impact, future adjustment trends, and current load deviation to form a time-delay compensation component and a load adaptation component. This scheme improves the predictive response capability of the optimization scheme to load fluctuations and reduces control overshoot or response lag caused by time-delay effects by embedding future operating condition prediction and time-delay propagation path into the weight adjustment process.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the objective function of the weighted multi-objective optimization problem includes: a denitrification efficiency objective function, characterizing the ratio of SCR inlet NOx concentration to outlet NOx concentration; an energy consumption objective function, including the weighted sum of pyrolysis furnace heating power and urea consumption; and an ammonia slip risk objective function, characterizing the cumulative risk value of SCR outlet ammonia slip concentration exceeding the standard.

[0013] This embodiment explicitly defines the calculation forms of three types of objective functions, enabling the multi-objective optimization process to quantitatively evaluate the performance differences of different control schemes. The denitrification efficiency objective function, energy consumption objective function, and ammonia slip risk objective function respectively characterize pollutant reduction capacity, operating costs, and emission compliance pressure. By transforming the core performance indicators of the urea pyrolysis process into computable mathematical expressions, this scheme allows the multi-objective evolutionary algorithm to search for Pareto solutions within a defined optimization space, improving the fit between the optimization results and actual process requirements.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, when a process scenario type changes, a weight smoothing transition process is executed. The weight smoothing transition process includes: determining the urgency of the scenario change, which is related to the process scenario type; determining a corresponding preset weight transition time based on the urgency; during the preset weight transition time, using a first-order inertial filtering algorithm to continuously and smoothly transition the final optimized weight from the final optimized weight corresponding to the previous scenario to the initial weight vector corresponding to the current scenario, wherein the filtering time constant of the first-order inertial filtering algorithm is proportional to the weight transition time; during the weight transition process, the rate of change of control parameters is monitored in real time, and when the rate of change of urea injection flow exceeds a preset first change threshold or the rate of change of pyrolysis furnace temperature setpoint exceeds a preset second change threshold, the weight transition time is automatically extended by a preset multiple to avoid system oscillation caused by frequent actions of the actuator; after the weight transition process ends, the current process scenario type is locked for a preset locking period, and during the locking period, even if the scenario recognition decision tree identifies a new scenario type, the scenario change is not immediately triggered.

[0015] This embodiment introduces a weighted smooth transition process to avoid control parameter jumps caused by abrupt weight changes during process scenario switching. The system determines the weight transition time based on the urgency of the scenario switch and uses a first-order inertial filtering algorithm to ensure a smooth and continuous transition of the final optimized weights from the previous scenario to the initial weight vector corresponding to the current scenario. During the transition, the system monitors the rate of change of urea injection flow and pyrolysis furnace temperature setpoints in real time. When the rate of change exceeds a preset threshold, the transition time is automatically extended to avoid system oscillations caused by frequent actuator movements. After the transition, the system locks the current scenario for a preset period. During the lock period, even if a new scenario is identified, the switch is not immediately triggered to prevent repeated scenario type switching when the operating conditions fluctuate near the critical point. This solution improves the stability of the control system during scenario switching and reduces the impact of weight adjustments on the actuators by linking the transition time to the urgency, adjusting the transition speed in real time, and setting a lock period.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, the input variables of the fuzzy inference system include NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation, and the output variables are denitrification efficiency weight adjustment, energy consumption weight adjustment, and ammonia slip risk weight adjustment. The preset urea pyrolysis process coupling rules include: if the NOx concentration deviation is positive and the ammonia slip concentration is lower than a preset safety threshold, then the denitrification efficiency weight adjustment is increased and the energy consumption weight adjustment is decreased; if the ammonia slip concentration is higher than a preset upper limit and the pyrolysis furnace temperature deviation is positive, then the ammonia slip risk weight adjustment is increased and the energy consumption weight adjustment is decreased.

[0017] This embodiment clarifies the input and output variables and process coupling rules of the fuzzy inference system, enabling weight adjustment to respond to the nonlinear correlation between NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation. When the NOx concentration deviation is positive and the ammonia slip concentration is below the safety threshold, the system increases the weight adjustment of denitrification efficiency and decreases the weight adjustment of energy consumption, prioritizing the improvement of denitrification performance. When the ammonia slip concentration is above the upper limit and the pyrolysis furnace temperature deviation is positive, the system increases the weight adjustment of ammonia slip risk and decreases the weight adjustment of energy consumption, prioritizing the suppression of ammonia slip. This scheme transforms the process constraint relationship between denitrification efficiency requirements and ammonia slip risk into executable inference rules, enabling the weight adjustment process to capture the multivariate coupling characteristics in actual operation and improving the response accuracy of the optimization strategy to process safety boundaries and performance requirements.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, the process scenario types include cold furnace start-up scenario, steady-state operation scenario, load impact scenario, ammonia escape critical scenario, pyrolysis furnace overheating scenario, and deep energy saving scenario; wherein the ammonia escape critical scenario and the pyrolysis furnace overheating scenario are defined as high urgency scenarios, the cold furnace start-up scenario and the steady-state operation scenario are defined as low urgency scenarios, and the load impact scenario and the deep energy saving scenario are defined as medium urgency scenarios.

[0019] This embodiment defines six types of process scenarios and their urgency levels, enabling the system to adopt differentiated optimization strategies for different operating states. Ammonia escape criticality scenarios and pyrolysis furnace overheating scenarios are defined as high-urgency scenarios to prioritize responses to emission exceedances and equipment safety risks. Cold furnace start-up scenarios and steady-state operation scenarios are defined as low-urgency scenarios to maintain control stability. Load shock scenarios and deep energy-saving scenarios are defined as medium-urgency scenarios to balance performance requirements and stability. This scheme maps typical operating states of the urea pyrolysis system to scenario types with clearly defined urgency levels, allowing the output of the scenario identification decision tree to directly drive the selection of weighted transition times and optimization strategies, thus improving the system's adaptability to the process requirements of different operating stages.

[0020] In a second aspect, embodiments of this application provide a urea pyrolysis multi-objective optimization device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the urea pyrolysis multi-objective optimization device to perform the method as described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a urea pyrolysis multi-objective optimization device, cause the urea pyrolysis multi-objective optimization device to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a urea pyrolysis multi-objective optimization device, cause the urea pyrolysis multi-objective optimization device to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the urea pyrolysis multi-objective optimization device provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application constructs a scenario-based decision tree and a dynamic weight adjustment mechanism, enabling the multi-objective optimization process to adaptively adjust the trade-off strategy between denitrification efficiency, energy consumption, and ammonia escape risk according to the type of process scenario. This solution achieves adaptive tracking of the optimization strategy to the operating conditions, reducing the system's operating costs while meeting emission standards.

[0026] 2. This application constructs a time delay model and a time series prediction mechanism to enable the weight adjustment signal to compensate for urea injection to ammonia production, ammonia transportation and pyrolysis furnace temperature. This scheme improves the predictive response capability of the optimization scheme to load fluctuations.

[0027] 3. This application introduces a closed-loop mechanism of execution monitoring and parameter self-updating. After the evaluation period ends, the initial weight vector and fuzzy inference system parameters corresponding to each process scenario are updated based on compliance, efficiency, stability and time delay response accuracy indicators. This solution realizes the continuous adaptation of the optimization model to the actual process characteristics and reduces the impact of initial parameter setting deviation on control performance. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a multi-objective optimization method for urea pyrolysis in an embodiment of this application.

[0029] Figure 2 This is another flowchart illustrating the multi-objective optimization method for urea pyrolysis in this application embodiment;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a urea pyrolysis multi-objective optimization device in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] The following describes the process of the method provided in this implementation. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a multi-objective optimization method for urea pyrolysis in an embodiment of this application.

[0034] S101. Calculate the state characteristics of the pyrolysis system based on the real-time operating data of the urea pyrolysis system. The real-time operating data includes pyrolysis parameters, denitrification parameters, and system operating parameters.

[0035] Real-time operational data refers to all measured values ​​reflecting the operating status of the urea pyrolysis system collected at the current moment, including three categories: pyrolysis parameters, denitrification parameters, and system operating parameters. Pyrolysis parameters include pyrolysis furnace temperature, urea injection flow rate, ammonia production concentration, and pyrolysis furnace pressure. Denitrification parameters include SCR inlet NOx concentration, SCR outlet NOx concentration, ammonia slip concentration, and SCR reactor temperature. System operating parameters include flue gas flow rate, flue gas temperature, and boiler load rate. The pyrolysis system state characteristics are represented by feature vectors extracted from the real-time operational data that characterize the current process state.

[0036] This step is executed at the beginning of each control cycle, providing state input for subsequent scenario identification and multi-objective optimization. The system collects real-time measurements, preprocesses the raw data (removing outliers, low-pass filtering, and interpolating missing values), and then calculates derived features. For example, it calculates the current denitrification efficiency based on the NOx concentration at the SCR inlet and outlet, the pyrolysis conversion rate based on the urea injection flow rate and ammonia production concentration, and the temperature stability index based on the deviation of the pyrolysis furnace temperature from the set value. Finally, the preprocessed measurements and derived features are combined into a state feature vector, typically with 15-25 dimensions. This design ensures that the raw measurements only reflect the immediate state of a single variable, while the derived features can integrate the relationships between multiple variables, more accurately representing the process stage the system is in.

[0037] S102. Input the pyrolysis system state characteristics into the scenario recognition decision tree to identify the current process scenario type.

[0038] The scenario recognition decision tree is a classification model built based on the characteristics of the urea pyrolysis process. It divides the system into different process scenario types through multi-level judgment of state characteristics. The process scenario types include cold furnace start-up scenario, steady-state operation scenario, load impact scenario, ammonia escape critical scenario, pyrolysis furnace overheating scenario, and deep energy saving scenario.

[0039] This step is executed immediately after the state characteristics are calculated, determining the scenario context for subsequent weight adjustments. The decision tree adopts a multi-level judgment structure: the first layer judges cold furnace start-up based on pyrolysis furnace temperature and temperature change rate (temperature <200℃ and rise rate >5℃ / min); the second layer judges emission risk based on ammonia slip concentration (concentration >8ppm or close to 80% of the limit); the third layer judges overheating risk based on temperature deviation (exceeding the set value by 20℃); the fourth layer judges impact state based on load change rate (flue gas flow rate change rate >10% / min or NOx concentration change rate >15% / min); the fifth layer judges energy-saving conditions based on denitrification efficiency and energy consumption (efficiency >95% and ammonia slip <3ppm); the remaining cases are steady-state operation.

[0040] S103. Determine the corresponding initial weight vector according to the process scenario type, and adjust the weights by combining the weight adjustment signal and the weight adjustment amount output by the fuzzy inference system to obtain the final optimized weights. The final optimized weights include denitrification efficiency weights, energy consumption weights, and ammonia escape risk weights. The fuzzy inference system infers the weight adjustment amount according to the preset urea pyrolysis process coupling rules. The weight adjustment signal is used to compensate for time delay effects and adapt to load changes.

[0041] The initial weight vector is a pre-configured weighting of denitrification efficiency, energy consumption, and ammonia slip risk for a specific process scenario. The weight adjustment signal includes a time-delay compensation component and a load adaptation component, used to compensate for the multi-stage time-delay effects of the urea pyrolysis process and adapt to load change trends. The final optimized weights represent the comprehensively corrected weight vector, directly driving the multi-objective evolutionary algorithm.

[0042] This step is executed after the scenario type is identified. The system first queries the initial weight mapping table based on the scenario; for example, the critical ammonia slip scenario corresponds to [0.3, 0.1, 0.6], and the deep energy-saving scenario corresponds to [0.2, 0.7, 0.1]. Then, a time-delay model is constructed to characterize the first time delay (30-60 seconds) from urea injection to pyrolysis ammonia production, the second time delay (10-20 seconds) from ammonia delivery to the SCR, and the third time delay (5-10 minutes) from pyrolysis furnace temperature adjustment to stability. LSTM is used to predict the load change over the next 5 minutes, calculating the expected impact of the current control action on future performance after the time delay, and generating a weight adjustment signal based on the load deviation. Simultaneously, NOx concentration deviation, ammonia slip concentration, and temperature deviation are input into the fuzzy inference system, and the weight adjustment amount is inferred according to the process coupling rules. Finally, the optimized weights are calculated and normalized by a weighted sum of the initial weights, the weight adjustment signal, and the fuzzy inference output.

[0043] In some specific instances, the weight adjustment signal is obtained through the following methods:

[0044] A time-delay model was constructed to characterize the first time delay in adjusting the urea injection flow rate to pyrolysis ammonia production, the second time delay in transporting ammonia from the pyrolysis furnace to the SCR reactor, and the third time delay in adjusting the pyrolysis furnace temperature to a stable level.

[0045] The first time delay (from urea injection to pyrolysis and ammonia production) refers to the process after the urea solution is atomized from the nozzle, involving heating and evaporation, pyrolysis (urea decomposing into ammonia and carbon dioxide), and the mass transfer of ammonia from the liquid phase to the gas phase. This time delay is affected by the pyrolysis furnace temperature, urea concentration, and atomized particle size, typically ranging from 30 to 60 seconds. A first-order inertial plus pure delay model is used to describe this process, where the conversion gain reflects the ratio of ammonia production to urea injection rate, the time constant reflects the pyrolysis reaction rate, and the pure delay time reflects the physical transport delay. Parameters are identified through step response experiments or calculated based on a mechanistic model using the Arrhenius equation.

[0046] The second time delay is due to the ammonia gas produced needing to be transported to the SCR reactor inlet via pipeline. This delay is determined by the pipeline length, flue gas velocity, and ammonia diffusion and mixing, typically ranging from 10 to 20 seconds. A pure delay model is used, where the delay time equals the pipeline length divided by the flue gas velocity. Fluctuations in flue gas velocity with boiler load cause this delay to change dynamically, requiring online correction of the delay parameter based on real-time flue gas flow.

[0047] The third time delay involves the power changes of the electric heater or hot air blower, the energy accumulation of the furnace body's heat capacity, and the response delay of the temperature sensor during pyrolysis furnace temperature adjustment, typically ranging from 5 to 10 minutes. It is described using a second-order inertial plus pure delay model, where the temperature adjustment gain reflects the proportional relationship between heating power and temperature rise, and the two time constants represent the furnace body's heat capacity and the dynamic response of the heater, respectively.

[0048] When constructing the composite time-delay model, the three time-delay components are connected in series to obtain the transmission relationship from urea injection flow rate and temperature setpoint to denitrification efficiency, energy consumption, and ammonia escape risk. Optionally, the Padé approximation can be used to expand the exponential term into a rational fraction to facilitate state-space modeling. Optionally, based on historical operating data, system identification methods (such as least squares method or subspace identification) can be used to directly fit the dynamic relationship between input and output, avoiding complex mechanism modeling.

[0049] Based on historical flue gas flow data and historical NOx load data, the predicted values ​​of flue gas flow and NOx load for a future preset time period are predicted by a time series prediction model.

[0050] Time series forecasting models predict load changes over a predetermined time period (typically 5-15 minutes, covering the longest time lag) based on historical data. The reason for choosing an LSTM network as the forecasting model is that flue gas flow and NOx load are affected by boiler combustion conditions, coal quality changes, and load scheduling, exhibiting non-stationary, nonlinear, and multi-period coupling characteristics. LSTM, through its gating mechanism, can effectively capture long-term dependencies.

[0051] In terms of model structure, the input layer receives the flue gas flow rate and NOx concentration sequences from the past 30 minutes, along with auxiliary features (boiler load rate, coal quality parameters, and ambient temperature). The hidden layer contains two LSTM units (64 neurons per layer) to capture temporal dependencies. The output layer is a fully connected layer that generates predicted flue gas flow rate and NOx load for the next 15 minutes, where NOx load equals flue gas flow rate multiplied by NOx concentration.

[0052] Training used historical data from the past three months, divided into training, validation, and test sets in a 7:2:1 ratio. The loss function employed a mean squared error plus a temporal continuity penalty (penalizing excessively large differences between predicted values ​​at adjacent time points to prevent drastic fluctuations in the prediction curve). The Adam optimizer was used with a learning rate of 0.001, a batch size of 32, and training for 100 epochs, employing an early stopping strategy to prevent overfitting. The model was considered to have met accuracy requirements when its mean absolute percentage error on the validation set was below 5%.

[0053] Optionally, the Prophet model can be used to decompose the trend, periodic, and residual components, which is suitable for load data with obvious daily or weekly periods. Alternatively, the Transformer architecture can be used to capture long-distance dependencies using a self-attention mechanism, but this incurs higher computational costs. No specific limitations are imposed here.

[0054] Based on the time-delay model, the expected impact of the current urea injection flow rate and pyrolysis furnace temperature setpoint on denitrification efficiency, energy consumption, and ammonia escape risk in the future preset time period after the first, second, and third time delays is calculated. Based on the flue gas flow rate prediction and NOx load prediction, the trend of ammonia supply change and pyrolysis furnace temperature adjustment in the future preset time period is calculated.

[0055] The expected impact is quantified by assessing the effect of current control actions on future performance after time delays. Substituting the current urea injection flow rate and temperature setpoints into the time-delay model, the flow rate of ammonia reaching the SCR after the first and second time delays, and the stable pyrolysis furnace temperature after the third time delay, are calculated. Combining the predicted flue gas flow rate and NOx load, the expected performance indicators for future timeframes are calculated.

[0056] The expected denitrification efficiency is calculated based on ammonia supply, NOx load, and pyrolysis furnace temperature. An empirical or neural network model is used to establish the mapping relationship between these three factors and the denitrification efficiency; this model is trained using historical data. If current control actions lead to insufficient ammonia supply or lower temperatures in the future, the expected denitrification efficiency will decrease.

[0057] The expected energy consumption includes two parts: electric heater power and urea consumption. Electric heater power is determined by the temperature setpoint; higher temperatures require higher power. Urea consumption is determined by the injection flow rate; higher flow rates result in higher consumption. The total energy consumption at future moments is calculated using a time-delay model.

[0058] The expected ammonia slip risk is determined based on the ratio of ammonia supply to NOx load. When the ammonia supply exceeds the amount required for the theoretical ammonia-to-nitrogen ratio, excess ammonia leads to an increased slip risk. The ammonia slip risk function is typically designed to account for the amount of ammonia exceeding the theoretical value; the larger this value, the higher the risk.

[0059] The future adjustment trend is based on the required rate of change in ammonia supply and the adjustment rate of pyrolysis furnace temperature calculated from load forecasts. The ammonia supply needs to be adjusted synchronously with changes in NOx load; when a future increase in NOx load is predicted, the urea injection flow rate needs to be increased in advance. The pyrolysis furnace temperature needs to be adjusted synchronously with ammonia demand to maintain pyrolysis efficiency; excessively low temperatures will lead to incomplete pyrolysis and reduced ammonia production. System inertia is considered in the adjustment trend calculation to avoid overshoot or oscillation caused by excessively rapid adjustments.

[0060] Based on the expected impact, the trend of ammonia supply changes, and the trend of pyrolysis furnace temperature adjustment, combined with the degree to which the current load deviates from the steady-state operating point, a weight adjustment signal is generated to adjust the initial weight vector. The weight adjustment signal includes a time delay compensation component and a load adaptation component.

[0061] The time-delay compensation component is generated based on the deviation between the expected impact and the current performance. If the expected denitrification efficiency is lower than the target value (future efficiency is expected to decline), the weight of denitrification efficiency is increased, making multi-objective optimization focus more on efficiency improvement. If the expected ammonia slip risk is higher than the threshold (future ammonia slip risk is expected to increase), the weight of ammonia slip risk is increased, making optimization tend to reduce ammonia supply. If the expected energy consumption is higher than the current value (future energy consumption is expected to increase), the weight of energy consumption is increased, promoting energy-saving control.

[0062] The time delay compensation component is calculated using a proportional relationship; the larger the deviation, the greater the compensation. The compensation coefficient is determined through simulation optimization, requiring a balance between compensation strength and system stability. Excessive compensation may lead to frequent and significant fluctuations in weights, while insufficient compensation will fail to effectively offset the effects of time delay.

[0063] The load adaptation component is generated based on the degree to which the flue gas flow rate and NOx load deviate from the steady-state point. The load deviation is defined as the relative deviation between the predicted flue gas flow rate and the steady-state flue gas flow rate. When the load deviation exceeds 10% (significant load change), the weight of denitrification efficiency is increased and the weight of energy consumption is decreased to cope with the impact, ensuring that emission compliance takes priority over energy saving. When the load deviation is less than 5% (close to steady state), the weight of energy consumption is increased for deep energy saving, because efficiency is easily guaranteed under steady-state conditions.

[0064] The load adaptation component also considers the direction of load change. When ammonia demand is predicted to rise, the weight of ammonia slip risk is appropriately increased to prevent excessive ammonia supply. When ammonia demand is predicted to fall, the weight of ammonia slip risk is decreased, allowing for a moderate increase in ammonia supply margin to compensate for forecast errors.

[0065] The final weight adjustment signal is a weighted sum of the time-delay compensation component and the load adaptation component. The fusion coefficient of the two components is usually set to 60% for time-delay compensation and 40% for load adaptation. This ratio is determined through backtesting and optimization using historical data and may be adjusted in different scenarios. The fused adjustment signal is applied to the initial weight vector and normalized to obtain the final optimized weights.

[0066] In some specific examples, the input variables of the fuzzy inference system include NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation, while the output variables are the denitrification efficiency weight adjustment, energy consumption weight adjustment, and ammonia slip risk weight adjustment. The preset coupling rules for the urea pyrolysis process include:

[0067] If the NOx concentration deviation is positive and the ammonia slip concentration is below the preset safety threshold, then increase the denitrification efficiency weight adjustment amount and decrease the energy consumption weight adjustment amount.

[0068] If the ammonia escape concentration is higher than the preset upper limit and the pyrolysis furnace temperature deviation is positive, then the ammonia escape risk weight adjustment amount is increased and the energy consumption weight adjustment amount is decreased.

[0069] Among them, NOx concentration deviation refers to the difference between the measured NOx concentration at the SCR outlet and the target emission value, which is used to characterize the degree of compliance of denitrification effect; ammonia slip concentration refers to the concentration of residual ammonia in the flue gas at the SCR outlet that has not participated in the denitrification reaction, which is used to characterize the rationality of ammonia supply; pyrolysis furnace temperature deviation indicates the deviation of the actual temperature from the set value, which is used to reflect the pyrolysis efficiency status; the preset safety threshold is usually 3-5 ppm, and the preset upper limit of concentration is generally 8 ppm. These thresholds are determined based on environmental emission standards and engineering practice experience.

[0070] This step is performed during the weight optimization process and aims to capture the coupling relationships between process variables through fuzzy reasoning. Specifically, the system calculates weight adjustments based on the real-time status of three key process indicators: NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation, using preset coupling rules. When NOx concentration exceeds the limit but ammonia slip is within a safe level, the system strengthens denitrification control by increasing the weight of denitrification efficiency and decreasing the weight of energy consumption. When ammonia slip approaches the limit and the pyrolysis temperature is too high, the system prioritizes controlling ammonia slip risk by increasing the weight of ammonia slip risk and decreasing the weight of energy consumption. This dynamic weight adjustment mechanism based on process coupling relationships can adaptively balance denitrification efficiency, energy consumption, and environmental constraints while ensuring the safe and stable operation of the system.

[0071] S104. Using the final optimization weights, solve the weighted multi-objective optimization problem through a multi-objective evolutionary algorithm to obtain the Pareto solution set. The objective function of the weighted multi-objective optimization problem is related to the temperature dependence and thermal inertia characteristics of urea pyrolysis.

[0072] Multi-objective evolutionary algorithms employ evolutionary computation principles to solve multi-objective optimization problems. They search for non-dominated solutions in the objective space through population iteration and survival of the fittest. Commonly used algorithms include NSGA-II and MOEA / D. Weighted multi-objective optimization problems sum the denitrification efficiency objective function, energy consumption objective function, and ammonia slip risk objective function according to their final optimization weights. The objective functions are related to the temperature dependence and thermal inertia characteristics of urea pyrolysis. The Pareto solution set represents the set of non-dominated solutions, where any solution is superior to all other solutions on at least one objective.

[0073] This step is executed after obtaining the final optimized weights. Specifically, the system defines decision variables including the pyrolysis furnace temperature setpoint (300-450℃), urea injection flow rate (50-500L / h), hot air fan frequency, electric heater power, etc., and sets upper and lower bound constraints. Objective functions are constructed: the denitrification efficiency objective is to maximize the ratio of SCR inlet to outlet NOx concentration; the energy consumption objective is to minimize the weighted sum of pyrolysis furnace heating power and urea consumption; and the ammonia escape risk objective is to minimize the cumulative risk of exceeding the SCR outlet ammonia escape concentration limit. Temperature-dependent characteristic models (Arrhenius relation) and thermal inertia characteristic models are embedded in the objective functions. A population of 100 individuals is initialized using the NSGA-II algorithm. Fitness is evaluated through fast non-dominated sorting and crowding calculation. Offspring are generated using simulated binary crossover and polynomial mutation, and the Pareto front is obtained after 50 generations of iteration. A Pareto optimal solution set is generated using a dynamic weight-driven multi-objective evolutionary algorithm, considering temperature dependence and thermal inertia constraints, satisfying the current scenario's weight preferences while retaining the flexibility of multi-objective trade-offs.

[0074] In some specific instances, the objective functions of the weighted multi-objective optimization problem include: a denitrification efficiency objective function, which characterizes the ratio of NOx concentration at the SCR inlet to NOx concentration at the outlet; an energy consumption objective function, which includes the weighted sum of the heating power of the pyrolysis furnace and the urea consumption; and an ammonia slip risk objective function, which characterizes the cumulative risk value of the SCR outlet ammonia slip concentration exceeding the standard.

[0075] Among them, the denitrification efficiency objective function refers to the pollutant removal capacity of the denitrification system quantitatively characterized by the ratio of NOx concentration at the inlet and outlet. The theoretical maximum value is 1, which indicates complete removal. The energy consumption objective function comprehensively considers the power consumption of the electric heater and the cost of urea reagent, and is used to characterize the economic efficiency of the system operation. The ammonia escape risk objective function is used to quantify the environmental risks of the system operation by cumulatively calculating the degree and duration of ammonia exceeding the standard at the SCR outlet.

[0076] Specifically, the denitrification efficiency objective function is expressed in the form η = 1 - Cout / Cin, where Cout and Cin are the inlet and outlet NOx concentrations, respectively; the energy consumption objective function is expressed in the form E = αP + βQ, where P is the heating power, Q is the urea flow rate, and α and β are conversion factors; the ammonia slip risk objective function is expressed in the form R = ∫max(0, CNH3 - Climit)dt, representing the integral of ammonia concentration exceeding the limit. This multi-objective function design considers both process performance indicators and operating costs and environmental constraints, enabling a comprehensive evaluation of the advantages and disadvantages of different control strategies and providing a scientific basis for finding the optimal control scheme.

[0077] S105. Select target control schemes from the Pareto solution set according to the process scenario type. The target control schemes include pyrolysis system control parameters and urea supply control parameters.

[0078] The target control scheme is the configuration of control parameters that best match the current process scenario, selected from the Pareto solution set. These parameters include pyrolysis system control parameters such as pyrolysis furnace temperature setpoint, urea injection flow rate setpoint, hot air fan frequency, and electric heater power, as well as urea supply control parameters such as urea solution supply pump speed and concentration setpoint.

[0079] This step is executed after the Pareto solution set is generated, providing executable parameter configuration for the control system. The system determines the screening criteria based on the scenario type: for critical ammonia slip scenarios, the solution with the lowest ammonia slip risk is prioritized; for deep energy-saving scenarios, the solution with the lowest energy consumption is prioritized; for steady-state operation scenarios, a weighted distance method is used to calculate the weighted Euclidean distance between each solution and the ideal point (100% denitrification efficiency, 0 energy consumption, 0 ammonia slip risk), and the solution with the smallest distance is selected. The system introduces a constraint filtering mechanism to eliminate solutions that may cause ammonia slip concentrations to exceed the 8 ppm limit or SCR outlet NOx concentrations to exceed the standard.

[0080] In some preferred embodiments, the method further includes:

[0081] Based on the difference in response speed of the actuators, hierarchical control commands are generated. The hierarchical control commands include urea injection pump control commands for the fast layer, hot air fan control commands for the medium-speed layer, and electric heater control commands for the slow layer. The electric heater control commands for the slow layer are processed by smoothing filtering.

[0082] Among them, the hierarchical control command refers to a set of multi-level control commands divided according to the response characteristics of the actuator, which is used to coordinate the action sequence of actuators with different dynamic characteristics; the fast-level urea injection pump control command refers to the injection flow rate adjustment command with a response time of less than 1 second, which is used to realize the rapid adjustment of ammonia supply; the medium-speed level hot air fan control command refers to the fan speed control command with a response time of 3-5 seconds, which is used to regulate the gas flow state of the pyrolysis furnace; the slow-speed level electric heater control command refers to the heating power control command with a response time of 30-60 seconds, which is used to maintain the pyrolysis temperature; the smoothing filter processing means that the control signal is low-pass filtered to suppress high-frequency oscillations, with a typical time constant of 60-120 seconds.

[0083] Specifically, the system first identifies the response speed of each actuator: the urea injection pump relies on frequency converter speed regulation to achieve rapid flow adjustment, and is classified as a fast layer; the hot air blower adjusts its speed through a frequency converter, and is affected by blower inertia and pipeline delay, and is classified as a medium-speed layer; the electric heater is constrained by the furnace's heat capacity and heat transfer process, resulting in a slow temperature response, and is classified as a slow layer. The system directly converts the urea flow setpoint in the target control scheme into injection pump speed commands, converts the pyrolysis furnace temperature requirement and flue gas flow into hot air blower frequency commands, and converts the temperature setpoint into electric heater power commands, which are then processed by a low-pass filter with a cutoff frequency of 0.01-0.02Hz. The process characteristics of actuator response speeds differing by two orders of magnitude in the urea pyrolysis system mean that frequent adjustments to slow-speed layer commands would lead to repeated start-stop cycles of the electric heater, causing furnace thermal stress fatigue and power fluctuations; while smoothing out fast-speed layer commands would weaken the system's responsiveness to sudden load changes. Through a hierarchical control architecture, agile tracking of ammonia supply is achieved in the fast layer, while stable operation of the pyrolysis temperature is maintained in the slow layer.

[0084] The hierarchical control commands are issued to the field actuators, and the temperature response curve of the pyrolysis furnace, the trend of ammonia concentration change, the NOx outlet concentration response and ammonia slip value are monitored during the execution process.

[0085] Specifically, the system sends commands for urea injection pump speed, hot air fan frequency, and electric heater power to field programmable logic controllers (PLCs) or distributed control systems (DCS) via an industrial control network. Upon receiving the commands, the actuators adjust their outputs according to their respective dynamic characteristics. During command execution, the system continuously collects measurement data from the pyrolysis furnace temperature sensor, ammonia concentration analyzer, SCR outlet NOx analyzer, and ammonia slip monitor at a sampling period of 1-2 seconds, constructing a real-time data stream. The system performs online identification of the pyrolysis furnace temperature response curve, extracting characteristic parameters such as rise time, overshoot, and steady-state error; performs differential calculations on the ammonia concentration change trend to determine if the concentration change rate is consistent with the urea injection flow rate adjustment direction; performs delay correlation analysis on the NOx outlet concentration response to identify the actual time lag from urea injection to NOx concentration change; and monitors the ammonia slip value at a threshold, triggering an early warning signal when the concentration exceeds a preset safety threshold of 80% (approximately 6-7 ppm).

[0086] After the execution of the hierarchical control instructions, actual effect data is collected in a preset evaluation cycle, and compliance evaluation indicators, efficiency evaluation indicators, stability evaluation indicators, and time delay response accuracy indicators are calculated based on the actual effect data.

[0087] Specifically, the system first determines whether the assessment period meets steady-state conditions. This is confirmed by checking whether the rate of change of the pyrolysis furnace temperature in the most recent 5 minutes is less than 2℃ / min and whether the fluctuation range of ammonia concentration is less than 10% of the average value. After entering steady state, the system extracts all measurement data within the assessment period from the time-series database and calculates compliance assessment indicators: the percentage of sampling points with SCR outlet NOx concentration below 50mg / m³ is used as the compliance rate, and the percentage of sampling points with ammonia slip concentration below 8ppm is used as the pass rate. When both indicators are greater than 95%, compliance is deemed achieved. Efficiency assessment indicators are calculated: the time-weighted average of the denitrification efficiency within the assessment period is compared with the expected denitrification efficiency in the target control scheme; a deviation of less than 3% is considered high accuracy. The actual total energy consumption (cumulative power consumption of electric heaters + energy converted from cumulative urea consumption) is compared with the energy consumption value predicted by the optimization algorithm; a ratio between 0.9 and 1.1 is considered accurate prediction. Stability assessment metrics: Calculate the standard deviation of the pyrolysis furnace temperature series; a value less than 5℃ is considered temperature stable. Calculate the coefficient of variation (standard deviation / mean) of ammonia concentration; a value less than 0.15 indicates stable ammonia supply. Calculate the root mean square (RMS) values ​​of the rate of change of urea injection flow rate and pyrolysis furnace temperature setpoint; values ​​below the preset threshold indicate stable control. Time-delay response accuracy metrics: Identify the measured time delay in adjusting the urea injection flow rate to the ammonia concentration response through cross-correlation analysis. Measured time lag of the response from ammonia concentration change to NOx concentration change , and the predicted value of the time delay model ᵗ、 Compare and calculate the relative error .

[0088] The initial weight vectors and parameters of the fuzzy inference system corresponding to each process scenario type are updated based on compliance assessment indicators, efficiency assessment indicators, stability assessment indicators, and time-delay response accuracy indicators.

[0089] Specifically, the system first maps the current operating conditions to predefined process scenario types based on characteristic parameters such as load level, flue gas flow rate, inlet NOx concentration, and cumulative catalyst operating time. For example, when the boiler load is between 80% and 100% and the standard deviation of flue gas flow rate is less than 5%, it is classified as a high-load stable operating condition.

[0090] For compliance assessment indicators, if the NOx compliance rate is below 95% or the ammonia slip compliance rate is below 95%, the system determines that the initial weight vector of the current process scenario does not pay enough attention to environmental constraints. The penalty function method is used to increase the weight coefficients of the denitrification efficiency target and the ammonia slip risk target. The specific adjustment amount Δw is proportional to the compliance rate deviation: Δw=k×(0.95-compliance rate), where k is the learning rate parameter, typically ranging from 0.1 to 0.3.

[0091] For efficiency evaluation indicators, if the actual denitrification efficiency deviates from the target value by more than 3% or the energy consumption ratio deviates from the range of 0.9-1.1, the system determines that the coefficient settings in the objective function are inaccurate. The recursive least squares method is used to re-identify the urea consumption coefficient β and the pyrolysis furnace heating power coefficient α in the energy consumption objective function based on the measured data. The updated formula is: θ(k+1)=θ(k)+K(k)[y(k)-φᵀ(k)θ(k)], where θ is the parameter vector to be identified, y is the measured energy consumption, φ is the eigenvector, and K is the gain matrix.

[0092] For stability assessment indicators, if the temperature standard deviation exceeds 5℃ or the ammonia concentration variation coefficient exceeds 0.15, the system determines that the dynamic adjustment of the weights is too aggressive. To make the weight transition time constant T in the fuzzy inference system smoother, the update rule is: T_new=T_old×(1+0.2×stability deviation).

[0093] Regarding the accuracy index of time delay response, if the relative error of time delay exceeds 15%, the system triggers the re-identification of time delay model parameters. The cross-correlation analysis method is used to re-estimate the time delay τ1 of urea injection to ammonia response and the time delay τ2 of ammonia concentration to NOx response based on the data of the most recent 30 evaluation cycles, and update the consequent parameters of the time delay prediction rule in the fuzzy inference system.

[0094] After the parameters are updated, the system stores the new parameters in the configuration file for the corresponding process scenario. When the same type of working condition is encountered again, the updated parameters are directly called to avoid the same deviation from occurring again.

[0095] When the process scenario type changes, a weighted smooth transition process is executed.

[0096] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the multi-objective optimization method for urea pyrolysis in this application embodiment.

[0097] S201. Determine the urgency of the scene switching. The urgency is related to the type of process scene.

[0098] The process scenario types include cold furnace start-up scenario, steady-state operation scenario, load impact scenario, ammonia escape critical scenario, pyrolysis furnace overheating scenario, and deep energy saving scenario; among them, the ammonia escape critical scenario and the pyrolysis furnace overheating scenario are defined as high-urgency scenarios, the cold furnace start-up scenario and the steady-state operation scenario are defined as low-urgency scenarios, and the load impact scenario and the deep energy saving scenario are defined as medium-urgency scenarios.

[0099] The core of determining urgency is identifying the current process scenario type and mapping it to the corresponding urgency level. For example, the system can identify scenarios by monitoring the following key parameters in real time:

[0100] Critical scenario for ammonia escape: The outlet ammonia concentration is close to the warning value of 8 ppm;

[0101] Overheating scenario in pyrolysis furnace: Temperature approaches the safety limit of 420℃;

[0102] Load shock scenario: Boiler load change rate exceeds 2% / min;

[0103] Deep energy-saving scenario: Inlet NOx is below 200mg / m³ and the system is stable;

[0104] Cold furnace start-up scenario: Pyrolysis furnace temperature is below 250℃;

[0105] Steady-state operation scenario: Key parameter fluctuations are all within the normal range;

[0106] S202. Determine the corresponding preset weight transition time based on the level of urgency.

[0107] The preset weight transition time refers to the time required to switch from the final optimized weight of the previous process scenario to the initial weight vector of the current scenario. This time parameter directly determines the speed of control strategy adjustment. The urgency level classification system quantifies the response requirements of scenario switching into three levels: high, medium, and low. Different levels correspond to different weight transition time setting strategies. There is a trade-off between weight transition time and system stability and response speed. If the transition time is too short, it will cause drastic changes in control parameters and lead to system oscillation. If the transition time is too long, it will delay the response to emergency conditions. The typical weight transition time setting range is: 30-60 seconds for high urgency scenarios, 90-180 seconds for medium urgency scenarios, and 240-360 seconds for low urgency scenarios. The specific values ​​need to be determined based on the response characteristics of the field actuators and system inertia.

[0108] S203. Within the preset weight transition time, the final optimized weight is continuously and smoothly transitioned from the final optimized weight corresponding to the previous scenario to the initial weight vector corresponding to the current scenario through a first-order lazy filtering algorithm. The filtering time constant of the first-order lazy filtering algorithm is proportional to the weight transition time.

[0109] Among them, the first-order lazy filtering algorithm is a commonly used signal smoothing method, and its mathematical expression is: Where w(t) is the weight value at the current time. Δt is the target weight value (i.e., the initial weight vector of the current scene), T is the filtering time constant, and Δt is the sampling period. The filtering time constant T directly determines the speed of weight transition. The larger T is, the slower the weight change and the more stable the system response.

[0110] This step is executed after the weight transition time is determined in S202, ensuring the continuity and stability of the control strategy adjustment. Specifically, the system first obtains the final optimized weight w_old corresponding to the previous scenario. This weight vector contains multiple components, such as the denitrification efficiency target weight α1, the energy consumption target weight α2, and the ammonia escape risk target weight α3. These weight values ​​have been adapted to the characteristics of the previous operating condition through the previous dynamic adjustment process. At the same time, the system extracts the initial weight vector w_new from the configuration library of the current scenario type as the target value for weight transition. The system calculates the time constant of the first-order inertial filtering algorithm based on the weight transition time T_trans determined in S202, using a linear proportional relationship: T=k×T_trans, where k is a proportionality coefficient, typically ranging from 0.3 to 0.5, so that the filtering time constant is slightly smaller than the transition time to ensure that the transition is completed within the preset time.

[0111] During the weight transition, the system iteratively calculates the weight update with a fixed sampling period Δt (typically 5-10 seconds). At each sampling moment, a first-order inertial filtering formula is applied: w(t)=w(t-Δt)+[Δt / T]×[w_new-w(t-Δt)]. The physical meaning of this formula is that the speed at which the current weight approaches the target weight is proportional to the current deviation. The larger the deviation, the faster the adjustment speed. As the weight gradually approaches the target value, the adjustment speed automatically slows down, forming a natural transition curve. For example, when switching from a steady-state operation scenario (denitrification efficiency weight 0.4, energy consumption weight 0.3, ammonia slip weight 0.3) to a load shock scenario (denitrification efficiency weight 0.5, energy consumption weight 0.2, ammonia slip weight 0.3), if the weight transition time is set to 120 seconds, the filter time constant T = 50 seconds, and the sampling period Δt = 10 seconds, then the first update of the denitrification efficiency weight is: w1(10s) = 0.4 + (10 / 50) × (0.5 - 0.4) = 0.42, and the second update is: w1(20s) = 0.42 + (10 / 50) × (0.5 - 0.42) = 0.436, and so on until it approaches the target value of 0.5.

[0112] S204. During the weight transition process, the change rate of control parameters is monitored in real time. When the change rate of urea injection flow exceeds the preset first change threshold or the change rate of pyrolysis furnace temperature setpoint exceeds the preset second change threshold, the weight transition time is automatically extended by a preset multiple to avoid system oscillation caused by frequent actions of the actuator.

[0113] Among them, the rate of change of control parameters is the magnitude of change of control commands per unit time, reflecting the intensity of control system regulation. The rate of change of urea injection flow rate (typical threshold 5-10 L / min²) and the rate of change of pyrolysis furnace temperature setpoint (typical threshold 2-5℃ / min) are two key monitoring indicators. The preset ratio (typical value 1.5-2.0) is used to extend the transition time and avoid system oscillation caused by frequent actuator movements.

[0114] This step is executed in parallel during the weight transition process, monitoring the dynamic response of key control parameters in real time. The system calculates the instantaneous rate of change of urea injection flow rate and temperature setpoint according to the sampling period: , When any rate of change exceeds a preset threshold, the system immediately extends the weight transition time to a preset multiple of its original value and recalculates the filter time constant, thus slowing down the rate of weight change. For example, when the rate of change of urea injection flow increases from 3 L / min² to 8 L / min² (exceeding the limit), the system extends the transition time from 120 seconds to 180 seconds, effectively suppressing drastic changes in control commands. This negative feedback mechanism achieves adaptive adjustment by monitoring actual response characteristics, transitioning from open-loop control to closed-loop control.

[0115] S205. After the weight transition process is completed, the current process scenario type is locked for a preset locking period. During the locking period, even if the scenario recognition decision tree recognizes a new scenario type, the scenario switching will not be triggered immediately.

[0116] The preset lockout period refers to the duration of scene lockout, typically 3-10 minutes, depending on the time required for system stabilization and the characteristics of the scene. During the lockout period, the system continues to run the scene recognition decision tree, but the recognition results do not trigger actual scene switching actions, avoiding repeated adjustments to the control strategy due to frequent switching.

[0117] This step is executed immediately after the weight transition process in S203 is fully completed. The system records the time t_complete when the weight transition is complete and starts the scenario lock timer. During the lock period T_lock (from t_complete to t_complete+T_lock), the scenario identification decision tree continues to run according to the normal cycle, analyzing process parameters in real time and outputting scenario type judgment results. However, the system will compare the identification results with the currently locked scenario type. If a different scenario type is identified, the system will not execute the scenario switching process of S201-S204, but will record the identification results in the log for subsequent analysis. Only after the lock period ends will the system reopen the scenario switching function. At this time, if the decision tree identifies a new scenario, the complete switching process will be triggered. For example, if the system has just switched from a steady-state operation scenario to a load impact scenario, the weight transition is completed within 120 seconds, and then a 5-minute lock period begins. In the second minute of the lock period, the decision tree may identify the scenario as a steady-state operation scenario due to a temporary easing of load fluctuations, but the system will not immediately switch back, but will continue to maintain the control strategy for the load impact scenario until the lock period ends.

[0118] The following describes the urea pyrolysis multi-objective optimization device in the embodiments of this invention from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the physical device structure of a urea pyrolysis multi-objective optimization device in the embodiments of this application.

[0119] It should be noted that, Figure 3 The structure of the urea pyrolysis multi-objective optimization device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0120] like Figure 3 As shown, the urea pyrolysis multi-objective optimization device includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 302 or a program loaded from storage section 308 into random access memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0121] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including hard disks, etc.; and communication section 309 including network interface cards such as LAN (Local Area Network) cards, modems, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0122] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0123] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0125] Specifically, the urea pyrolysis multi-objective optimization device of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the urea pyrolysis multi-objective optimization method provided in the above embodiment.

[0126] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the urea pyrolysis multi-objective optimization device described in the above embodiments; or it may exist independently and not assembled into the urea pyrolysis multi-objective optimization device. The storage medium carries one or more computer programs, which, when executed by a processor of the urea pyrolysis multi-objective optimization device, cause the urea pyrolysis multi-objective optimization device to implement the urea pyrolysis multi-objective optimization method provided in the above embodiments.

[0127] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0128] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0129] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A multi-objective optimization method for urea pyrolysis, characterized in that, The method includes: The state characteristics of the pyrolysis system are calculated based on the real-time operating data of the urea pyrolysis system, which includes pyrolysis parameters, denitrification parameters, and system operating parameters. The state characteristics of the pyrolysis system are input into the scene recognition decision tree to identify the current process scene type; The initial weight vector is determined according to the process scenario type, and the weight is adjusted by combining the weight adjustment signal and the weight adjustment amount output by the fuzzy inference system to obtain the final optimized weight. The final optimized weight includes denitrification efficiency weight, energy consumption weight and ammonia escape risk weight. The fuzzy inference system infers the weight adjustment amount according to the preset urea pyrolysis process coupling rules. The weight adjustment signal is used to compensate for time delay effect and adapt to load changes. The weighted multi-objective optimization problem is solved using the final optimization weights through a multi-objective evolutionary algorithm to obtain the Pareto solution set. The objective function of the weighted multi-objective optimization problem is related to the temperature dependence and thermal inertia characteristics of urea pyrolysis. Target control schemes are selected from the Pareto solution set based on the process scenario type. The target control schemes include pyrolysis system control parameters and urea supply control parameters.

2. The method according to claim 1, characterized in that, The method further includes: Based on the difference in response speed of the actuators, hierarchical control commands are generated. The hierarchical control commands include urea injection pump control commands for the fast layer, hot air fan control commands for the medium-speed layer, and electric heater control commands for the slow layer. The electric heater control commands for the slow layer are processed by smoothing filtering. The hierarchical control command is sent to the field actuator, and the temperature response curve of the pyrolysis furnace, the trend of ammonia concentration change, the NOx outlet concentration response and ammonia escape value are monitored during the execution process. After the execution of the hierarchical control command, actual effect data is collected during a preset evaluation period, and compliance evaluation index, efficiency evaluation index, stability evaluation index, and time delay response accuracy index are calculated based on the actual effect data. The initial weight vectors corresponding to each process scenario type and the parameters of the fuzzy inference system are updated based on the compliance assessment index, the efficiency assessment index, the stability assessment index, and the time-delay response accuracy index.

3. The method according to claim 1, characterized in that, The weight adjustment signal is obtained through the following method: A time-delay model is constructed, which characterizes the first time delay of adjusting the urea injection flow rate to pyrolysis ammonia production, the second time delay of transporting ammonia from the pyrolysis furnace to the SCR reactor, and the third time delay of adjusting the pyrolysis furnace temperature to a stable state. Based on historical flue gas flow data and historical NOx load data, the predicted values ​​of flue gas flow and NOx load for a future preset time period are predicted by a time series prediction model. Based on the time delay model, the expected impact of the current urea injection flow rate and pyrolysis furnace temperature setpoint on the denitrification efficiency, energy consumption, and ammonia escape risk in the future preset time period after the first, second, and third time delays is calculated. Based on the flue gas flow rate prediction value and the NOx load prediction value, the expected trend of the required ammonia supply and the pyrolysis furnace temperature adjustment in the future preset time period are calculated. Based on the expected impact, the trend of ammonia supply change, and the trend of pyrolysis furnace temperature adjustment, combined with the degree of current load deviation from steady-state operating point, a weight adjustment signal is generated to adjust the initial weight vector. The weight adjustment signal includes a time delay compensation component and a load adaptation component.

4. The method according to claim 1, characterized in that, The objective function of the weighted multi-objective optimization problem includes: The denitrification efficiency objective function characterizes the ratio of NOx concentration at the SCR inlet to NOx concentration at the outlet. The energy consumption objective function includes a weighted sum of the heating power of the pyrolysis furnace and the urea consumption. The objective function for ammonia slip risk characterizes the cumulative risk value of exceeding the ammonia slip concentration limit at the SCR outlet.

5. The method according to claim 1, characterized in that, When the process scenario type changes, a weighted smooth transition process is executed, which includes: Determine the urgency of the scene switching, whereby the urgency is related to the type of process scene; Determine the corresponding preset weighted transition time based on the level of urgency; During the preset weight transition time, a first-order lazy filtering algorithm is used to make the final optimized weight transition smoothly from the final optimized weight corresponding to the previous scenario to the initial weight vector corresponding to the current scenario. The filtering time constant of the first-order lazy filtering algorithm is proportional to the weight transition time. During the weight transition process, the rate of change of control parameters is monitored in real time. When the rate of change of urea injection flow exceeds the preset first change threshold or the rate of change of pyrolysis furnace temperature setpoint exceeds the preset second change threshold, the weight transition time is automatically extended by a preset multiple to avoid system oscillation caused by frequent actions of the actuator. After the weight transition process is completed, the current process scenario type is locked for a preset locking period. During the locking period, even if the scenario recognition decision tree recognizes a new scenario type, the scenario switching will not be triggered immediately.

6. The method according to claim 1, characterized in that, The input variables of the fuzzy inference system include NOx concentration deviation, ammonia slip concentration, and pyrolysis furnace temperature deviation. The output variables are denitrification efficiency weight adjustment, energy consumption weight adjustment, and ammonia slip risk weight adjustment. The preset urea pyrolysis process coupling rules include: If the NOx concentration deviation is positive and the ammonia slip concentration is lower than the preset safety threshold, then the denitrification efficiency weight adjustment amount is increased and the energy consumption weight adjustment amount is decreased. If the ammonia escape concentration is higher than the preset upper limit and the temperature deviation of the pyrolysis furnace is positive, then the adjustment amount of the ammonia escape risk weight is increased and the adjustment amount of the energy consumption weight is decreased.

7. The method according to claim 5, characterized in that, The process scenario types include cold furnace start-up scenario, steady-state operation scenario, load impact scenario, ammonia escape critical scenario, pyrolysis furnace overheating scenario, and deep energy saving scenario; Among them, the ammonia escape critical scenario and the pyrolysis furnace overheating scenario are defined as high-urgency scenarios, the cold furnace start-up scenario and the steady-state operation scenario are defined as low-urgency scenarios, and the load impact scenario and the deep energy saving scenario are defined as medium-urgency scenarios.

8. A multi-objective optimization device for urea pyrolysis, characterized in that, The urea pyrolysis multi-objective optimization device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the urea pyrolysis multi-objective optimization device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the urea pyrolysis multi-objective optimization device, the urea pyrolysis multi-objective optimization device performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the urea pyrolysis multi-objective optimization device, the urea pyrolysis multi-objective optimization device performs the method as described in any one of claims 1-7.

Images