Pollutant emission dynamic prediction and intelligent optimization method, system and equipment

Abstract

The invention relates to a pollutant emission dynamic prediction and intelligent optimization method, system and equipment. The method comprises the steps of periodically obtaining emission data of a latest window in a window sliding mode, dynamically updating a real-time emission rate trend, and then performing index moving average to obtain a smoothed emission rate trend; calculating according to the accumulated actual emission amount and the smoothed emission rate trend to obtain a predicted total emission amount; calculating a discharge amount deviation value and a pollutant discharge concentration needing to be adjusted per hour; calculating according to the adjustment factor, the discharge amount deviation value and the pollutant discharge concentration needing to be adjusted per hour to obtain an hour allowable discharge mean value; calculating a feedforward control amount and a feedback control amount based on the predicted pollutant concentration and the hourly allowable discharge mean value; and the final emission control quantity is calculated in a dynamic weight distribution mode. According to the invention, accurate control of flue gas pollutants is realized, and it is ensured that the total emission reaches the standard in a control period.

Description

Technical Field

[0001] This invention relates to the field of environmental protection technology, and in particular to a method, system and equipment for dynamic prediction and intelligent optimization of pollutant emissions. Background Technology

[0002] Currently, flue gas pollutant emission control methods fall into two categories. The first is traditional PID control, which constructs a single feedback loop based on real-time monitored pollutant concentration data and adjusts the operating parameters of treatment equipment such as desulfurization, denitrification, and dust removal through proportional, integral, and derivative algorithms. However, PID control can only achieve short-term closed-loop regulation and lacks the ability to predict and plan for long-term emission targets. When operating conditions fluctuate, pollutant concentrations are prone to exceed limits due to lag in parameter adjustments, and it cannot correlate cumulative emission values ​​for proactive control. The second is manual experience control, which relies on on-site operators to manually adjust equipment parameters based on historical experience and real-time data. This method is greatly affected by differences in operator skill and experience, resulting in low control accuracy, slow response speed, and difficulty in coping with complex and changing operating conditions. Furthermore, it cannot achieve accurate calculation and dynamic control of long-term total emissions. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, system, and equipment for dynamic prediction and intelligent optimization of pollutant emissions that can achieve precise control of flue gas pollutants and ensure that the total emissions meet the standards within the control period, in order to address the above-mentioned technical problems.

[0004] A method for dynamic prediction and intelligent optimization of pollutant emissions, the method comprising: Determine the real-time emission rate trend, and construct a linear regression model based on the real-time emission rate trend; The emission data of the latest window is periodically acquired by using a sliding window method, and the real-time emission rate trend is dynamically updated based on the linear regression model to obtain the updated emission rate trend. An exponential moving average is applied to the updated emission rate trend to obtain a smoothed emission rate trend; the predicted total emissions are then calculated based on the cumulative actual emissions and the smoothed emission rate trend. The emission deviation value is calculated based on the total emission allowable value and the predicted total emission. The pollutant emission concentration that needs to be adjusted per hour is calculated based on the emission deviation value. An adjustment factor is set, and the hourly allowable average emission value is obtained by calculating based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. An LSTM model is trained using emission data, and then the trained LSTM model outputs predicted pollutant concentrations. Calculate the feedforward control quantity and the feedback control quantity based on the predicted pollutant concentration and the hourly allowable average emission value; Based on the feedforward control quantity and the feedback control quantity, the final emission control quantity is calculated by dynamically allocating weights.

[0005] On the other hand, a dynamic prediction and intelligent optimization system for pollutant emissions is also provided, including: The linear regression model construction module is used to determine the real-time emission rate trend and construct a linear regression model based on the real-time emission rate trend. The emission rate dynamic update module is used to periodically acquire the emission data of the latest window through a sliding window, and dynamically update the real-time emission rate trend based on the linear regression model to obtain the updated emission rate trend. The total emissions prediction module is used to perform an exponential moving average on the updated emissions rate trend to obtain a smoothed emissions rate trend; the predicted total emissions are calculated based on the cumulative actual emissions and the smoothed emissions rate trend. The emission deviation calculation module is used to calculate the emission deviation value based on the total emission allowable value and the predicted total emission. The emission concentration calculation module that needs adjustment is used to calculate the pollutant emission concentration that needs to be adjusted per hour based on the emission deviation value. The hourly allowable average emission calculation module is used to set an adjustment factor and calculate the hourly allowable average emission based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. The model training module is used to train an LSTM model using emission data, and then output the predicted pollutant concentrations using the trained LSTM model. The feedforward control quantity and feedback control quantity calculation module is used to calculate the feedforward control quantity and feedback control quantity based on the predicted pollutant concentration and the hourly average allowable emission value; The emission control quantity calculation module is used to calculate the final emission control quantity based on the feedforward control quantity and the feedback control quantity by dynamically allocating weights.

[0006] On the other hand, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned method for dynamic prediction and intelligent optimization of pollutant emissions.

[0007] Compared with existing technologies, the method, system, and equipment for dynamic prediction and intelligent optimization of pollutant emissions provided by this invention have the following beneficial effects: This invention utilizes a sliding window and linear regression model to periodically incorporate the latest emission data, ensuring that emission rate calculations are always based on actual emission patterns and avoiding prediction deviations caused by long-term invariance of initial model parameters. An exponential moving average is applied to the updated emission rate trend to effectively filter short-term random fluctuations. Combined with the cumulative actual emissions, the predicted total emissions are calculated, further narrowing the gap between predicted and actual values ​​and improving prediction accuracy. By calculating the emission deviation, the required hourly pollutant emission concentration adjustment is derived, clearly defining quantitative targets and avoiding reliance on manual experience, thus achieving precise emission control. The introduction of adjustment factors dynamically adjusts the hourly allowable average emission, enhancing flexibility and adaptability. An LSTM model is used to obtain the predicted pollutant concentration, which, combined with feedforward and feedback control variables, achieves precise control of the hourly average.

[0008] This invention enables precise control of flue gas pollutants, ensuring that total emissions meet standards within the control period while optimizing operating costs. It is particularly suitable for industrial enterprises and other fields with complex emission scenarios, large fluctuations in operating conditions, and the need for long-term target control. It provides a practical technical solution for precise and scientific pollution control, improving the level of precision in emission control management. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention, and those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating the dynamic prediction and intelligent optimization method for pollutant emissions provided in Example 1. Figure 2 This is a block diagram of the dynamic prediction and intelligent optimization system for pollutant emissions provided in Example 2; Figure 3 This is an internal structural diagram of the computer device provided in Example 3.

[0011] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0013] It is understood that the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0014] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0015] Example 1 like Figure 1 As shown in the figure, this embodiment provides a method for dynamic prediction and intelligent optimization of pollutant emissions, including the following steps: Step 201: Determine the real-time emission rate trend and construct a linear regression model based on the real-time emission rate trend.

[0016] Step 202: Periodically acquire the latest emission data of the window using a sliding window method, and dynamically update the real-time emission rate trend based on a linear regression model to obtain the updated emission rate trend.

[0017] Step 203: Perform an exponential moving average on the updated emission rate trend to obtain a smoothed emission rate trend; calculate the predicted total emissions based on the cumulative actual emissions and the smoothed emission rate trend.

[0018] Step 204: Calculate the emission deviation value based on the allowable total emission value and the predicted total emission value.

[0019] Step 205: Calculate the pollutant emission concentration that needs to be adjusted per hour based on the emission deviation value.

[0020] Step 206: Set the adjustment factor. Calculate the hourly allowable average emission value based on the adjustment factor, emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour.

[0021] Step 207: Train an LSTM model using emission data, and then output the predicted pollutant concentration using the trained LSTM model.

[0022] Step 208: Calculate the feedforward control quantity and the feedback control quantity based on the predicted pollutant concentration and the hourly allowable average emission.

[0023] Step 209: Based on the feedforward control quantity and the feedback control quantity, calculate the final emission control quantity by dynamically allocating weights.

[0024] In the specific implementation of step 201, emission data is first collected. The collected emission data must cover five dimensions, including flue gas characteristics, reaction conditions, denitrification parameters, desulfurization parameters, and emission indicators. Specific parameters for different processes are collected more rigorously. Table 1 shows a parameter illustration of the emission data.

[0025] Table 1 Emission Data Parameters

[0026] The system consists of sensors that collect real-time operational data from the equipment; edge gateways that run edge computing algorithms to filter and compress the data; and an OPC UA server that receives data from the edge gateways and provides a standardized interface. The collected data is stored in a database for later retrieval. Other input parameters, such as control objectives, control cycles, and production plans, are entered manually. It's worth noting that data acquisition methods include not only real-time online data collection via sensors but also offline sampling and laboratory analysis data.

[0027] Then, determine the window duration. Select the nearest window time length A linear regression model was constructed using the emission data.

[0028] In this embodiment, the nearest [location] is selected. Emission data is collected hourly (60 minutes) with one sampling point per minute. A linear regression model is constructed based on this data, as follows: ; In the formula, Indicates the first The emission rate corresponding to each time period; Indicates the trend of emission rates; Indicates the first One time period; This represents the intercept term.

[0029] In the specific implementation of step 202, the interception time period is set to... And determine the window time length. Then, by using a sliding window approach, the emission data of the latest window is periodically acquired. In this embodiment, the time period is set to 5 minutes. By acquiring the emission data of the latest window every 5 minutes (e.g., if the window time length is set to 60 minutes, one data point is acquired per second), the real-time emission rate trend is dynamically updated based on a linear regression model to obtain the updated emission rate trend, expressed as: ; In the formula, This indicates the updated emission rate trend; The amount of data representing emission data is taken in this embodiment as... .

[0030] In the specific implementation of step 203, an exponential moving average is applied to the updated emission rate trend to obtain a smoothed emission rate trend, expressed as: ; In the formula, This indicates the smoothed emission rate trend over the current time period; Indicates the smoothing factor; This represents the smoothed emission rate trend after the previous time period. It's worth noting that the smoothing factor... Control the weighting of historical data and current data. The larger the value, the more sensitive the model is to the latest data (faster response but greater fluctuations); The smaller the value, the smoother the model (slower response but more stable).

[0031] Then, the cumulative actual emissions are obtained every 5 minutes. Based on the cumulative actual emissions and the smoothed emission rate trend, the predicted total emissions are calculated, as expressed by: ; In the formula, This represents the projected total emissions within the target period; This indicates the cumulative actual emissions; This indicates the remaining time in the target period.

[0032] In the specific implementation of step 204, the emission deviation value is calculated based on the allowable total emission value and the predicted total emission value, and the expression is: ; In the formula, This indicates the deviation value of emissions; This indicates the permissible total emissions value. This represents the total amount of pollutants allowed to be emitted within the target period, and is a known value.

[0033] In the specific implementation of step 205, the pollutant emission concentration that needs to be adjusted per hour is calculated based on the emission deviation value, and the expression is as follows: ; In the formula, This indicates the pollutant emission concentration that needs to be adjusted every hour; This indicates the deviation value of emissions; Indicates the remaining time of the target period; This represents the average flue gas flow rate during the target period.

[0034] In the specific implementation of step 206, to avoid over-adjustment and achieve smooth fluctuations, an adjustment factor is introduced. The hourly allowable average emission value is calculated based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. The expression for the adjustment factor is: ; The expression for calculating the hourly average allowable emissions is: ; In the formula, K represents the adjustment factor; This represents the over-discharge penalty coefficient, which is a coefficient greater than 1. The larger the value, the smaller the degree of relaxation allowed by the system. This represents the energy-saving incentive coefficient, which is a coefficient between 0 and 1. The larger the value, the greater the reduction. This indicates the allowable total amount of pollutants emitted during the target period; This represents the average allowable emissions per hour; This represents the actual value of the current hourly average.

[0035] In the specific implementation of step 207, in order to predict future pollutant concentration changes based on historical data, a compensation signal is generated in advance to compensate for the time lag effect of the desulfurization and denitrification process. Optimized control commands are output by integrating LSTM feedforward and PID feedback correction.

[0036] Specifically, emissions data is acquired and used to train an LSTM model. However, to ensure model accuracy, input data with weak correlation to the output values ​​needs to be filtered out and removed. Influence factors with strong correlations are then identified and used to train the LSTM model. The filtering method employs the mutual information rule. Mutual information, a measure of information entropy in information theory, represents the relationship between two variables; a larger value indicates a more complex relationship, and vice versa. Influence factors with mutual information values ​​greater than 0.5 are retained as part of the dataset for model training.

[0037] Furthermore, to ensure high-quality input data for the model, data preprocessing is necessary. The steps are as follows: First, the original dataset is clearly aligned to eliminate invalid data and timestamps are standardized to ensure data integrity. Then, lag feature generation is performed to construct features reflecting the lag effects of the process, enabling the model to capture causal time-series relationships. Next, sliding window statistics are performed to extract time-series dynamic patterns, such as trends and volatility. Finally, standardization / encoding is performed to eliminate dimensional differences and adapt to the model's input requirements, completing the data preprocessing. Finally, the data is divided into training, validation, and test sets, with the division ratios set conventionally and not elaborated further. It is worth noting that in addition to the methods mentioned above, other data preprocessing methods include missing value handling, outlier detection and handling, normalization and standardization, logarithmic transformation, feature construction, and smoothing, which will not be detailed here.

[0038] During model training, the input dataset includes time-series data (30-minute sliding window, 1-minute sampling interval), pollutant concentrations (NOx, SO2), control parameters (ammonia injection rate, desulfurizer flow rate), and operating parameters (flue gas flow rate, temperature, oxygen content). Based on a set mean squared error (MSE) loss function and prediction bias penalty term, the model is trained online to obtain a trained LSTM model. The trained LSTM model then outputs predicted pollutant concentrations. The online learning method involves incrementally updating model parameters every 24 hours to adapt to operating condition drift. The pollutant concentration prediction timeframe is relatively short and can be adjusted according to actual conditions; in this embodiment, the output is the predicted pollutant concentration for the next 5-10 minutes.

[0039] In the specific implementation of step 208, coordinated control of feedforward and feedback is carried out. The feedforward gain uses prediction information to compensate for time delay and disturbance in advance, while the PID feedback handles the residual error.

[0040] The expression for calculating the feedforward control quantity is as follows: ; ; In the formula, Indicates the feedforward control quantity; Indicates feedforward gain; This indicates the predicted pollutant concentration; The amount of data representing control variables; Indicates the first The control quantity corresponding to each time period; Indicates the mean of the control quantity; Indicates the first The pollutant concentration changes corresponding to each time period; This represents the average change in pollutant concentration.

[0041] The expression for calculating the feedback control quantity is: ; ; In the formula, Indicates the feedback control quantity; This represents the proportionality coefficient, used to control the strength of the action's response to the current error; Ti represents the integration time, which is used to eliminate steady-state errors (the smaller Ti is, the stronger the integration effect). It represents the differential time, used to predict the lead time of future trends; This indicates the actual pollutant concentration at the current moment; This represents the deviation value at the current moment; This represents the average allowable emissions for the hour at the current moment; Represents the integral time variable The deviation value at time; This indicates the lag time, which is the time from the start of the control action to a change in the feedback data.

[0042] Among them, the proportionality coefficient Integral Time Differential time The calculation expression is: ; ; ; In the formula, It represents the system gain, which is the static amplification factor of the change in the system input to the change in the output. This represents the time constant, i.e., the time it takes for the system to reach steady state.

[0043] In the specific implementation of step 209, the weights are dynamically allocated according to the degree of deviation between the current concentration and the center of the target interval through a formula to avoid the emission concentration from continuously deviating from the target interval and to ensure that the pollutant emission concentration always converges stably within the target interval.

[0044] Among them, dynamic weight allocation The expression is: ; The final emission control quantity expression is as follows: ; In the formula, This indicates the final emission control amount; Indicates the maximum feedforward weight; This represents key real-time state variables, which are determined based on on-site operating conditions, such as inlet pollutant concentration, boiler load, or coal feed rate. Indicates the midpoint of the target interval; This indicates the half-width of the interval.

[0045] Finally, based on the final emission control targets, optimized instructions (such as ammonia injection rate and desulfurizer dosage) are generated, and these instructions are translated into physical actions through the execution output module. The execution output module can be understood as the end-effector of the desulfurization and denitrification control system, consisting of signal analysis, drive circuits, actuators (high-precision regulating valves, variable frequency feeders), and a feedback unit. It receives optimized instructions in real time, drives valve opening or motor speed, and dynamically adjusts the dosage of environmental protection reagents.

[0046] This invention enables the system to automatically trigger an enhanced control mode during orange / red weather alerts, dynamically tightening emission limits and reducing pollution load in advance to achieve compliant emissions. Through intelligent optimization control, it reduces the excessive addition of environmental reagents (ammonia, limestone, etc.), thereby lowering energy consumption and material waste. The method proposed in this invention can effectively reduce manual labor intensity and minimize decision-making errors.

[0047] It should be understood that, although this embodiment Figure 1 The steps are shown sequentially as indicated by the arrows, but they are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order in which these steps are performed; they can be executed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0048] Example 2 Based on the dynamic prediction and intelligent optimization method for pollutant emissions in Example 1, this example discloses a dynamic prediction and intelligent optimization system for pollutant emissions, such as... Figure 2 As shown, the pollutant emission dynamic prediction and intelligent optimization system includes: a linear regression model construction module 401, an emission rate dynamic update module 402, a total emission prediction module 403, an emission deviation calculation module 404, an emission concentration adjustment calculation module 405, an hourly allowable average emission calculation module 406, a model training module 407, a feedforward control quantity and feedback control quantity calculation module 408, and an emission control quantity calculation module 409, wherein: The linear regression model building module 401 is used to determine the real-time emission rate trend and build a linear regression model based on the real-time emission rate trend.

[0049] The emission rate dynamic update module 402 is used to periodically acquire the emission data of the latest window through a sliding window, and dynamically update the real-time emission rate trend based on a linear regression model to obtain the updated emission rate trend.

[0050] The total emissions prediction module 403 is used to perform an exponential moving average on the updated emissions rate trend to obtain a smoothed emissions rate trend; and to calculate the predicted total emissions based on the cumulative actual emissions and the smoothed emissions rate trend.

[0051] The emission deviation calculation module 404 is used to calculate the emission deviation value based on the allowable total emission value and the predicted total emission value.

[0052] The emission concentration calculation module 405 is used to calculate the pollutant emission concentration that needs to be adjusted per hour based on the emission deviation value.

[0053] The hourly allowable average emission calculation module 406 is used to set the adjustment factor. Based on the adjustment factor, the emission deviation value and the pollutant emission concentration that needs to be adjusted per hour, the hourly allowable average emission is calculated.

[0054] The model training module 407 is used to train an LSTM model using emission data, and then output the predicted pollutant concentrations using the trained LSTM model.

[0055] The feedforward control quantity and feedback control quantity calculation module 408 is used to calculate the feedforward control quantity and feedback control quantity based on the predicted pollutant concentration and the hourly allowable average emission.

[0056] The emission control quantity calculation module 409 is used to calculate the final emission control quantity based on the feedforward control quantity and the feedback control quantity by dynamically allocating weights.

[0057] In this embodiment, the specific working processes and principles of the linear regression model construction module 401, emission rate dynamic update module 402, total emission prediction module 403, emission deviation calculation module 404, emission concentration adjustment calculation module 405, hourly allowable average emission calculation module 406, model training module 407, feedforward control quantity and feedback control quantity calculation module 408, and emission control quantity calculation module 409 are the same as those in Embodiment 1, and therefore will not be described again in this embodiment. Each unit module can be implemented entirely or partially through software, hardware, or a combination thereof. Each unit module can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each of the above unit modules.

[0058] Example 3 like Figure 3 The diagram illustrates a terminal device disclosed in this embodiment, comprising a transmitter, a receiver, a memory, and a processor. The transmitter transmits instructions and data, the receiver receives instructions and data, the memory stores computer-executed instructions, and the processor executes the computer-executed instructions stored in the memory to implement the method described in Embodiment 1 above.

[0059] It is important to note that the aforementioned memory can be either standalone or integrated with the processor. When the memory is set up independently, the terminal device also includes a bus for connecting the memory and the processor.

[0060] Example 4 This embodiment discloses a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the method in Embodiment 1 above.

[0061] 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. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0062] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for dynamic prediction and intelligent optimization of pollutant emissions, characterized in that, The method includes: Determine the real-time emission rate trend, and construct a linear regression model based on the real-time emission rate trend; The emission data of the latest window is periodically acquired by using a sliding window method, and the real-time emission rate trend is dynamically updated based on the linear regression model to obtain the updated emission rate trend. An exponential moving average is applied to the updated emission rate trend to obtain a smoothed emission rate trend; the predicted total emissions are then calculated based on the cumulative actual emissions and the smoothed emission rate trend. The emission deviation value is calculated based on the total emission allowable value and the predicted total emission. The pollutant emission concentration that needs to be adjusted per hour is calculated based on the emission deviation value. An adjustment factor is set, and the hourly allowable average emission value is obtained by calculating based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. An LSTM model is trained using emission data, and then the trained LSTM model outputs predicted pollutant concentrations. Calculate the feedforward control quantity and the feedback control quantity based on the predicted pollutant concentration and the hourly allowable average emission value; Based on the feedforward control quantity and the feedback control quantity, the final emission control quantity is calculated by dynamically allocating weights.

2. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 1, characterized in that, The real-time emission rate trend is dynamically updated based on the linear regression model to obtain the updated emission rate trend, including: The constructed linear regression model is expressed as: ； The real-time emission rate trend is dynamically updated based on the linear regression model to obtain the updated emission rate trend, expressed as follows: ； In the formula, Indicates the first The emission rate corresponding to each time period; Indicates the trend of emission rates; Represents the intercept term; This indicates the updated emission rate trend; Indicates the amount of data involved in the calculation; Indicates the first Each time period.

3. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 2, characterized in that, The updated emission rate trend is then subjected to an exponential moving average to obtain a smoothed emission rate trend, expressed as follows: ； In the formula, This indicates the smoothed emission rate trend over the current time period; Indicates the smoothing factor; This indicates the trend of emission rates after smoothing over the previous time period.

4. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 3, characterized in that, The predicted total emissions are calculated based on the cumulative actual emissions and the smoothed emission rate trend, as expressed by: ； In the formula, This represents the projected total emissions within the target period; This indicates the cumulative actual emissions; This indicates the remaining time in the target period.

5. The method for dynamic prediction and intelligent optimization of pollutant emissions according to any one of claims 1 to 4, characterized in that, The pollutant emission concentration that needs to be adjusted per hour is calculated based on the emission deviation value, as expressed by: ； In the formula, This indicates the pollutant emission concentration that needs to be adjusted every hour; This indicates the deviation value of emissions; Indicates the remaining time of the target period; This represents the average flue gas flow rate during the target period.

6. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 5, characterized in that, An adjustment factor is set, and the hourly allowable average emission value is calculated based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour, including: Set the adjustment factor, expressed as: ； The hourly allowable average emission value is calculated based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. The calculation expression is as follows: ； In the formula, K represents the adjustment factor; Indicates the over-discharge penalty coefficient; Indicates the energy-saving incentive coefficient; This indicates the allowable total amount of pollutants emitted during the target period; This represents the average allowable emissions per hour; This represents the actual value of the current hourly average.

7. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 6, characterized in that, The feedforward control variable and the feedback control variable are calculated based on the predicted pollutant concentration and the hourly allowable average emission value, including: The feedforward control quantity is calculated using the following expression: ； The expression for calculating the feedback control variable is: ； ； In the formula, Indicates the feedforward control quantity; Indicates feedforward gain; This indicates the predicted pollutant concentration; Indicates the feedback control quantity; Indicates the proportionality coefficient; Indicates the integration time; Represents the differential time; This indicates the actual pollutant concentration at the current moment; This represents the deviation value at the current moment; This represents the average allowable emissions for the hour at the current moment; Represents the integral time variable The deviation value at time; Indicates the lag time.

8. The method for dynamic prediction and intelligent optimization of pollutant emissions according to claim 7, characterized in that, Based on the feedforward control quantity and the feedback control quantity, the final emission control quantity is calculated by dynamically allocating weights, and the expression is: ； Among them, dynamic weight allocation The expression is: ； In the formula, This indicates the final emission control amount; Indicates the maximum feedforward weight; Represents key real-time state variables; Indicates the midpoint of the target interval; This indicates the half-width of the interval.

9. A dynamic prediction and intelligent optimization system for pollutant emissions, characterized in that, The system includes: The linear regression model construction module is used to determine the real-time emission rate trend and construct a linear regression model based on the real-time emission rate trend. The emission rate dynamic update module is used to periodically acquire the emission data of the latest window through a sliding window, and dynamically update the real-time emission rate trend based on the linear regression model to obtain the updated emission rate trend. The total emissions prediction module is used to perform an exponential moving average on the updated emissions rate trend to obtain a smoothed emissions rate trend; the predicted total emissions are calculated based on the cumulative actual emissions and the smoothed emissions rate trend. The emission deviation calculation module is used to calculate the emission deviation value based on the total emission allowable value and the predicted total emission. The emission concentration calculation module that needs adjustment is used to calculate the pollutant emission concentration that needs to be adjusted per hour based on the emission deviation value. The hourly allowable average emission calculation module is used to set an adjustment factor and calculate the hourly allowable average emission based on the adjustment factor, the emission deviation value, and the pollutant emission concentration that needs to be adjusted per hour. The model training module is used to train an LSTM model using emission data, and then output the predicted pollutant concentrations using the trained LSTM model. The feedforward control quantity and feedback control quantity calculation module is used to calculate the feedforward control quantity and feedback control quantity based on the predicted pollutant concentration and the hourly average allowable emission value; The emission control quantity calculation module is used to calculate the final emission control quantity based on the feedforward control quantity and the feedback control quantity by dynamically allocating weights.

10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the dynamic prediction and intelligent optimization method for pollutant emissions according to any one of claims 1 to 8.

Images