Industrial Production Monitoring Method and Monitoring Device Applied to Pollution Reduction and Carbon Emission Reduction in the Ecological Environment

Abstract

An embodiment of the present invention discloses an industrial production monitoring method and monitoring device applied to pollution reduction and carbon emission reduction in the ecological environment. The method includes: firstly, obtaining the operation parameters of the target industrial production equipment and the set of environmental parameters in the associated area; secondly, processing the operation parameters and the set of environmental parameters through a pre-trained pollution reduction and carbon emission reduction prediction model to output the pollution emission feature vector and dynamic emission reduction strategy of the target industrial production equipment; then generating a multi-level control instruction set (equipment operation parameter adjustment threshold, energy switching priority, and pollution treatment node start-stop rule) for the target industrial production equipment according to the pollution emission feature vector and the dynamic emission reduction strategy; finally, transmitting the multi-level control instruction set to the control terminal of the target industrial production equipment through the Internet of Things interface, and dynamically adjusting the execution weight of the multi-level control instruction set based on the updated set of real-time feedback environmental parameters. In this way, pollution reduction and carbon emission reduction in industrial production can be achieved with high quality and intelligence.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of data analysis, and in particular to an industrial production monitoring method and monitoring device applied to pollution reduction and carbon emission reduction in the ecological environment. Background Art

[0002] In recent years, in order to avoid the further deterioration of the ecological environment, pollution control and carbon emission reduction technologies in the industrial production process are crucial.

[0003] However, there are certain defects in the existing pollution control and carbon emission reduction technologies. For example, traditional methods mostly adopt a single-dimensional monitoring and regulation mechanism, and usually discretely collect emission parameters based on fixed thresholds, lacking dynamic coupling analysis of equipment operating conditions and environmental factors. Some improved solutions attempt to introduce machine learning algorithms for emission prediction, but the model training data does not cover the operating states of the entire life cycle of the equipment, resulting in significant deviations in the prediction results when the operating conditions change suddenly.

[0004] In addition, the existing pollution treatment control technologies usually adopt a switch-type start-stop mechanism, lacking a collaborative calculation mechanism, which is extremely likely to cause a sharp increase in energy consumption caused by frequent start-stop of the treatment unit. At the same time, it is difficult for such pollution treatment control technologies to realize real-time optimization and iteration of control strategies. It can be seen that how to achieve high-quality and intelligent pollution reduction and carbon emission reduction in industrial production is a technical problem that urgently needs to be solved at present. Summary of the Invention

[0005] The embodiments of the present invention provide an industrial production monitoring method and monitoring device applied to pollution reduction and carbon emission reduction in the ecological environment, which are used to achieve high-quality and intelligent pollution reduction and carbon emission reduction in industrial production.

[0006] In a first aspect, the embodiments of the present invention provide an industrial production monitoring method applied to pollution reduction and carbon emission reduction in the ecological environment, which is applied to an industrial production monitoring device. The method includes: obtaining the operating parameters of a target industrial production device and a set of environmental parameters of an associated area, where the set of environmental parameters includes atmospheric pollutant concentration, carbon emission intensity, and energy consumption rate; processing the operating parameters and the set of environmental parameters through a pre-trained pollution reduction and carbon emission reduction prediction model to output a pollution emission feature vector and a dynamic emission reduction strategy of the target industrial production device; generating a multi-level control instruction set for the target industrial production device according to the pollution emission feature vector and the dynamic emission reduction strategy, where the multi-level control instruction set includes equipment operating parameter adjustment thresholds, energy switching priorities, and pollution treatment node start-stop rules; transmitting the multi-level control instruction set to a control terminal of the target industrial production device through an Internet of Things interface, and dynamically adjusting the execution weight of the multi-level control instruction set based on a dynamically updated set of real-time feedback environmental parameters.

[0007] In a second aspect, an industrial production monitoring device provided by an embodiment of the present invention includes:

[0008] a processor;

[0009] a storage device on which a computer program is stored,

[0010] when the computer program is executed by the processor, the processor implements any of the industrial production monitoring methods applied to pollution reduction and carbon emission reduction in the ecological environment.

[0011] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the industrial production monitoring method applied to pollution reduction and carbon emission reduction in the ecological environment are implemented.

[0012] It can be seen that the embodiments of the present invention have the following beneficial effects: First, obtain the operating parameters of the target industrial production equipment and the set of environmental parameters in the associated area; secondly, process the operating parameters and the set of environmental parameters through a pre-trained pollution reduction and carbon emission reduction prediction model to output the pollution emission feature vector and dynamic emission reduction strategy of the target industrial production equipment; then generate a multi-level control instruction set (equipment operating parameter adjustment threshold, energy switching priority, and pollution treatment node start-stop rule) for the target industrial production equipment according to the pollution emission feature vector and dynamic emission reduction strategy; finally, transmit the multi-level control instruction set to the control terminal of the target industrial production equipment through the Internet of Things interface, and dynamically adjust the execution weight of the multi-level control instruction set based on the updated set of real-time feedback environmental parameters. Designed in this way, it can achieve high-quality and intelligent pollution reduction and carbon emission reduction in industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a flowchart of an industrial production monitoring method applied to pollution reduction and carbon emission reduction in the ecological environment provided by an embodiment of the present invention.

[0014] Figure 2 is a schematic diagram of the basic structure of an industrial production monitoring device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] To make the above objects, features, and advantages of the present invention more obvious and understandable, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] See Figure 1 As shown, this figure is a flowchart of an industrial production monitoring method applied to pollution reduction and carbon emission reduction in the ecological environment provided by an embodiment of the present invention, and this method can be applied to an industrial production monitoring device. As Figure 1 shown, this method may include S102 - S108.

[0017] S102: Obtain the operating parameters of the target industrial production equipment and the set of environmental parameters of the associated area, where the set of environmental parameters includes atmospheric pollutant concentration, carbon emission intensity, and energy consumption rate.

[0018] In S102, the industrial production monitoring device collects the operating parameters of the target industrial production equipment and the set of environmental parameters of the associated area in real time through a multi-dimensional sensor network deployed on the coke oven body (target industrial production equipment) of the coking plant and the periphery of the plant area (associated area). Specifically, the operating parameters of the coke oven body include key operating condition indicators such as the internal temperature gradient distribution in the coke drum, the traveling speed of the coke pusher, and the opening degree of the gas injection valve. These parameters are obtained through measuring elements such as an embedded thermocouple array, a laser displacement sensor, and a pressure transmitter at a sampling frequency of several times per second.

[0019] The set of environmental parameters of the associated area is continuously obtained by environmental monitoring stations distributed within a radius of several hundred meters in the coking plant area. Among them, the atmospheric pollutant concentration synchronously measures the real-time concentrations of PM2.5, benzo[a]pyrene, and hydrogen sulfide through a beta-ray absorption particulate matter monitor and an ultraviolet differential absorption spectroscopy instrument; the carbon emission intensity uses tunable diode laser absorption spectroscopy technology to online analyze the concentrations of CO2 and CH4 in the exhaust gas flow from the coke oven chimney; the energy consumption rate records the power load curve of the coke oven heating system and the steam output rate of the waste heat boiler through an intelligent electricity meter group and a steam flow meter respectively. After the above parameter data is encapsulated by the industrial bus protocol, it forms a multi-modal data stream with time stamp alignment and is transmitted to the edge computing unit of the industrial production monitoring device.

[0020] S104: Process the operating parameters and the set of environmental parameters through a pre-trained pollution reduction and carbon emission reduction prediction model, and output the pollution emission feature vector and dynamic emission reduction strategy of the target industrial production equipment.

[0021] In S104, the industrial production monitoring device calls the pollution reduction and carbon emission reduction prediction model deployed in the GPU acceleration module to perform a coupling analysis on the coke oven operating parameters and the regional environmental parameters. The pollution reduction and carbon emission reduction prediction model is based on a deep temporal convolutional network architecture and is pre-trained through several sets of operating condition - environment mapping data accumulated during the historical production cycle. Its hidden layer contains more than 10 dilated causal convolutional modules, which can extract the non-linear correlation features between the coke oven combustion efficiency and pollutant diffusion.

[0022] When the coke pushing cycle parameters and the PM2.5 exceeding standard data on the southwest side of the plant area are collected in real time, the pollution reduction and carbon emission reduction prediction model first conducts a partial correlation analysis on the fluctuation of gas calorific value and the generation rate of benzo[a]pyrene, then locates the core variable group affecting the carbon emission peak through the attention mechanism, and finally outputs a pollution emission feature vector containing multi-dimensional features. For example, the third-dimensional eigenvalue reflects the dynamic correlation coefficient between the switching cycle of the coke oven regenerator chamber and the NOx emission, and the seventh-dimensional eigenvalue characterizes the quantitative relationship between the coke maturity and the volatile organic compound emission rate. Based on this pollution emission feature vector, the pollution reduction and carbon emission reduction prediction model synchronously generates a dynamic emission reduction strategy set, which includes the optimization scheme of the air-fuel ratio in the coke oven combustion chamber, the pressure regulation gradient of the raw gas purification system, and the coordinated operation schedule of the dry coke quenching device and the wet desulfurization tower. Each strategy item is accompanied by a confidence score and a response priority parameter.

[0023] S106: Generate a multi-level control instruction set for the target industrial production equipment according to the pollution emission feature vector and the dynamic emission reduction strategy. The multi-level control instruction set includes the adjustment threshold of equipment operation parameters, the priority of energy switching, and the start-stop rules of pollution treatment nodes.

[0024] In S106, the industrial production monitoring device constructs a multi-level control instruction system for the coking production line according to the modal decomposition result of the pollution emission feature vector. For example, for the coke oven body preparation layer, based on the VOCs control threshold corresponding to the seventh-dimensional eigenvalue, a positioning accuracy compensation instruction for the coke pusher is generated, requiring the automatic trigger of the riser pipe water seal height adjustment mechanism when the center temperature of the coke cake reaches 1050 ± 15 °C.

[0025] In the energy supply layer, according to the prediction of the carbon emission intensity trend, a priority matrix for the switching of dual fuels of coke oven gas and natural gas is formulated, stipulating that when the environmental monitoring station detects that the benzo[a]pyrene concentration in the downwind direction exceeds 2.5 ng / m³, the main pipeline cutoff device of the coke oven gas shall be immediately started, and it shall be gradually replaced with high-calorific value natural gas at a ratio of 1:0.7.

[0026] In the pollution treatment layer, combined with the equipment response delay parameter in the dynamic emission reduction strategy, the start-stop rules for the chemical product recovery section are generated, specifying that the crude benzene washing tower should start the emergency absorption mode 10 seconds in advance when the pressure fluctuation of the coke oven gas collecting pipe exceeds 150 Pa, while the desulfurization regeneration tower needs to perform variable-cycle regeneration operations according to the change of the tail gas H2S concentration gradient.

[0027] All the above control instructions can verify their logical completeness through Petri net modeling and establish a mapping relationship with the interlock protection mechanism of the DCS system.

[0028] S108: Transmit the multi-level control instruction set to the control terminal of the target industrial production equipment through the Internet of Things interface, and dynamically adjust the execution weight of the multi-level control instruction set based on the real-time feedback environmental parameter update set.

[0029] In S108, the industrial production monitoring device transmits the multi-level control instruction set to the distributed control system in the central control room of the coking plant through the OPC UA protocol, and simultaneously establishes two-way data channels with the coke oven temperature measurement probe and the dust removal fan frequency converter.

[0030] For example, when executing the instruction to improve the sealing performance of the coke side oven door, the industrial production monitoring device continuously receives the real-time temperature field distribution data uploaded by dozens of infrared thermal imagers installed on the coke oven top. If an abnormal temperature rise area exceeding 50°C is detected in the southeast corner of Chamber No. 3, the execution weight of the instruction to adjust the pressing force of the equipment sealing strip is immediately increased from 0.7 to 0.9 according to the benzo[a]pyrene concentration growth rate in the updated environmental parameter set, and at the same time, the priority of the control item for the traveling speed of the coke pusher is reduced.

[0031] Another example is the instruction adjustment for the gas purification system. The industrial production monitoring device obtains the secondary voltage fluctuation data of the electrostatic tar precipitator every 200 milliseconds through the Modbus / TCP interface. When it is found that the deviation from the preset capture efficiency curve exceeds 12%, the online correction module of the dynamic emission reduction strategy is automatically triggered, the start-stop frequency of the washing liquid circulation pump is recalculated, and the updated control parameters are sent to the on-site actuator through the industrial wireless Mesh network. The whole process follows the safety integrity level requirements of the IEC 61508 standard to ensure that the transmission delay of the control instruction is always lower than the level of dozens of milliseconds.

[0032] Specifically, in the above technical solution, the equipment operation parameter adjustment threshold refers to the critical value or range of key parameters set according to the real-time working conditions and environmental feedback during the dynamic regulation process of industrial production equipment, which is used as the reference condition for triggering or terminating specific control actions. In complex industrial scenarios such as the coke oven body in the coking plant, the equipment operation parameter adjustment threshold can be obtained through the coupled analysis of multi-dimensional sensor data and prediction models, aiming to balance production efficiency and environmental protection goals.

[0033] For example, in the optimization scheme of the air-fuel ratio in the coke oven combustion chamber, the adjustment thresholds may include the upper and lower limits of the temperature gradient distribution inside the coke drum (such as 1050 ± 15 °C), the safe range of the opening degree of the gas injection valve (such as 30% - 70%), and the fluctuation tolerance of the traveling speed of the coke pusher (such as ±0.2 m / s). The setting of these thresholds is based on the correlation analysis of historical production data and pollutant generation rules. For example, when the maturity of the coke cake is insufficient, premature coke pushing will cause the emission rate of volatile organic compounds (VOCs) to exceed the standard. Therefore, the center temperature of the coke cake needs to be monitored in real time through an embedded thermocouple, and the sealing mechanism adjustment instruction is triggered when the threshold is reached.

[0034] In addition, the adjustment thresholds of equipment operating parameters have dynamic adaptability: when the environmental monitoring station detects a sudden increase in the PM2.5 concentration, the temperature fluctuation threshold can be temporarily tightened through an online learning algorithm, adjusting the original ±15 °C to ±10 °C to inhibit the generation of nitrogen oxides. This threshold system realizes logical verification through Petri net modeling to ensure compatibility with the interlock protection mechanism of the distributed control system (DCS). For example, when multiple thresholds trigger conflicts simultaneously, an arbitration strategy will be executed according to the preset priority parameters (such as safety > environmental protection > energy efficiency).

[0035] In the above technical solution, the energy switching priority refers to the fuel type switching sequence and ratio rule determined based on real-time environmental parameters, carbon emission constraints, and equipment status in a multi-energy supply system, which is used to optimize the energy structure and reduce pollution emissions.

[0036] In the coking scenario, the coke oven heating system can adopt a mixed energy supply mode of gas and natural gas. The construction of the switching priority matrix needs to comprehensively consider the pollutant diffusion model, energy cost, and equipment response characteristics.

[0037] For example, when the concentration of benzo[a]pyrene in the downwind direction of the plant exceeds 2.5 ng / m³, the coke oven gas supply will be cut off first according to the eigenvector output by the pollution reduction and carbon emission reduction prediction model, and the high-calorie natural gas replacement program will be started. The switching ratio increases in a gradient of 1:0.7 to ensure combustion stability. The design of the priority rule depends on multi-dimensional data analysis: the CO2 concentration in the chimney is monitored in real time through tunable diode laser absorption spectroscopy technology, and combined with the power load curve recorded by the smart meter, the carbon emission intensity and thermal efficiency ratio of different energy combinations are dynamically calculated.

[0038] For another example, during the low period of electricity demand, the waste heat boiler steam energy supply may be preferentially enabled to reduce the dependence on the power grid; while during the period when PM2.5 exceeds the standard, the proportion of natural gas is forced to be increased to more than 80%. In addition, when embedding the priority rules into the device control logic, physical delay needs to be considered. For example, the response time of the cut-off valve of the main gas pipeline is 2 seconds. Therefore, before issuing the instruction, it is necessary to predict the change trend of environmental parameters and trigger the switching action 1-3 sampling periods in advance through the temporal convolutional network to ensure the spatio-temporal matching between the instruction execution and the pollution diffusion rate.

[0039] Furthermore, the start-stop rules of pollution treatment nodes refer to the refined control strategies for the operation timing and trigger conditions of waste gas and wastewater treatment equipment, formulated based on real-time pollution characteristics and treatment efficiency models. In a coking plant, such rules cover the entire process from raw coke oven gas purification to desulfurization and denitrification. The core is to achieve precise start and stop of treatment equipment through multi-source data fusion.

[0040] For example, the emergency mode start rule of the crude benzene scrubbing tower stipulates that when the pressure fluctuation of the coke oven gas collecting pipe exceeds 150 Pa, the high-pressure spray system needs to be activated 10 seconds in advance to enhance the absorption efficiency; while the regeneration cycle of the desulfurization regeneration tower is dynamically adjusted according to the H2S concentration gradient. If the ultraviolet differential absorption spectrometer detects that the H2S concentration in the tail gas rises by more than 5 ppm per hour, the regeneration frequency will be increased from the default 4 hours / time to 2 hours / time. Rule formulation depends on the analysis results of pollution emission characteristic vectors. For example, the seventh-dimensional eigenvalue reflects the quantitative relationship between the VOCs emission rate and the coke maturity. When this value exceeds 0.85, the dry quenching coke device and the wet desulfurization tower will be linked to start and stop, the circulating water pump will be forced to start, and the waste gas residence time will be extended to 120 seconds.

[0041] In addition, equipment coordination and safety redundancy need to be considered during rule execution: the voltage data of the electrocatalytic tar precipitator is collected every 200 milliseconds through the Modbus / TCP interface. If the deviation from the preset efficiency curve exceeds 12%, the current start-stop sequence will be immediately suspended and switched to the standby washing liquid supply pipeline. All rules are certified by the safety integrity level (SIL) of the IEC 61508 standard to ensure that the shutdown protection is preferentially executed rather than efficiency optimization under extreme working conditions. For example, when the bearing temperature of the dust removal fan is detected to exceed 90 °C, the global shutdown instruction will be immediately executed regardless of the pollution index.

[0042] The embodiment of the present invention can achieve high-quality and intelligent pollution reduction and carbon emission reduction in industrial production by constructing a dynamic regulation mechanism that integrates multi-dimensional parameters.

[0043] First, the embodiments of the present invention innovatively establish a coupled analysis model of the equipment operating state, environmental pollutants, carbon emissions, and energy flow, breaking through the limitations of traditional isolated monitoring parameters. Through a pre-trained model, it collaboratively analyzes multi-source heterogeneous parameters, accurately identifies the non-linear correlation characteristics between pollution emissions and energy consumption, and provides a full-factor diagnosis ability for complex industrial scenarios.

[0044] Secondly, the embodiments of the present invention are based on a multi-level instruction generation architecture of dynamic feature vectors. By decoupling the emission reduction strategy into three interlocked dimensions: equipment operation, energy scheduling, and pollution control, it realizes the dynamic balance optimization of environmental protection indicators and production efficiency, and has higher system adaptability compared with traditional single-dimensional control methods.

[0045] Finally, the embodiments of the present invention design a weight adjustment algorithm with self-learning ability, relying on the real-time feedback data of the Internet of Things to establish an environmental response closed-loop, enabling the control strategy to be autonomously optimized according to the pollutant diffusion trend and energy supply and demand fluctuations, and forming an intelligent emission reduction paradigm with spatio-temporal evolution adaptability.

[0046] In summary, the embodiments of the present invention significantly improve the coordination and sustainability of industrial emission reduction measures through mechanism integration and dynamic optimization.

[0047] In a preferred embodiment, the processing of the operating parameters and the set of environmental parameters by the pre-trained pollution reduction and carbon emission prediction model in S104 to output the pollution emission feature vector and dynamic emission reduction strategy of the target industrial production equipment includes:

[0048] S1041: Perform time series alignment processing on the set of environmental parameters, and extract the atmospheric pollutant concentration fluctuation curve, carbon emission intensity change gradient, and energy consumption rate distribution matrix within the same time window.

[0049] In S1041, the industrial production monitoring device performs time series alignment processing on the set of environmental parameters in the associated area of the coke oven body in the coking plant. Exemplarily, a β-ray absorption method particulate matter monitor deployed on the coke oven chimney collects PM2.5 concentration data at a frequency of 2 times per second, while a tunable diode laser absorption spectrometer installed in the southwest corner of the factory area measures CO2 concentration at a period of 1 time per 5 seconds.

[0050] The industrial production monitoring device can use the timestamp interpolation algorithm to unify parameters such as atmospheric pollutant concentration and carbon emission intensity with different sampling frequencies onto a time reference axis of 1 time per second. For example, within the time window from 10:00:00 to 10:05:00 in the morning, align the instantaneous value of 0.8 m / s recorded by the pusher travel speed sensor with the benzo[a]pyrene concentration of 1.8 ng / m³ measured by the ultraviolet differential absorption spectrometer at the same moment to form an atmospheric pollutant concentration fluctuation curve with 60 consecutive sampling points.

[0051] Meanwhile, the smart meter data at three energy monitoring points in the coking plant area are spatially aggregated, and the comprehensive power load value of the whole plant per second is calculated to construct an energy consumption rate distribution matrix. The row dimension of this matrix represents energy types such as electricity and steam, the column dimension corresponds to main energy-consuming equipment such as the coke oven heating system and dust removal fans, and the matrix element value represents the proportion of energy consumption of each equipment at the current moment.

[0052] S1042: Integrate the atmospheric pollutant concentration fluctuation curve, the carbon emission intensity change gradient, and the energy consumption rate distribution matrix into a multi-dimensional environmental state tensor, and generate the pollution emission feature vector based on the state characteristics of the multi-dimensional environmental state tensor and the operating parameters.

[0053] In S1042, the industrial production monitoring device performs multi-modal data fusion on the time-aligned environmental parameters and the coke oven body operating parameters. First, the concentration values of three types of pollutants, namely PM2.5, benzo[a]pyrene, and hydrogen sulfide, in the atmospheric pollutant concentration fluctuation curve are respectively mapped to the three-dimensional space coordinate axes to form the spatial distribution of the pollutant concentration field. Then, the carbon emission intensity change gradient data is converted into a concentration attenuation gradient vector along the dominant wind direction of the plant area. Then, the energy consumption rate distribution matrix is unfolded into a heat map on a two-dimensional plane, where the horizontal axis represents the time series and the vertical axis represents different energy types.

[0054] In the embodiment of the present invention, through the tensor splicing technology, the above three types of environmental parameters and the operating parameters of the coke oven body (including the temperature gradient of the coke drum and the opening of the gas valve, etc.) are superimposed in the fourth dimension to generate a multi-dimensional environmental state tensor containing 12 channels. Each channel of this tensor corresponds to a specific type of working condition or environmental index. For example, the 5th channel stores the product feature of the traveling speed of the coke pusher and the PM2.5 concentration. Based on this multi-dimensional environmental state tensor, the industrial production monitoring device uses the feature cross algorithm to generate the pollution emission feature vector, where the seventh-dimensional eigenvalue is obtained by calculating the Pearson correlation coefficient between the coke maturity index and the volatile organic compound emission rate, and the third-dimensional eigenvalue reflects the dynamic coupling strength between the regenerator commutation cycle and the NOx emission.

[0055] S1043: Use the convolutional neural network layer in the pollution reduction and carbon emission reduction prediction model to extract the spatial correlation features of the multi-dimensional environmental state tensor, and process the spatial correlation features through the long short-term memory network layer to generate a time-dependent pollution diffusion path prediction result, and generate the dynamic emission reduction strategy based on the pollution diffusion path prediction result.

[0056] In S1043, the industrial production monitoring device calls the deep neural network in the pollution reduction and carbon emission reduction prediction model for feature extraction and prediction. The convolutional neural network layer first performs three-dimensional convolution operations on the multi-dimensional environmental state tensor to identify the spatial association pattern between the high-temperature area of the coke oven combustion chamber and the peak concentration of PM2.5 in the northeast corner of the plant area. For example, in the operation with a convolutional kernel size of 3×3×2, it is detected that when the opening of the gas valve exceeds 65%, the concentration of benzo[a]pyrene 800 meters away from the chimney will increase by 0.3 ng / m³ after 120 seconds. The long short-term memory network layer then processes the time series of this spatial association feature to predict the pollutant diffusion path within the next 15 minutes.

[0057] Exemplarily, when it is identified that the traveling speed of the coke pusher drops below 0.5 m / s and the center temperature of the coke cake reaches 1035 °C, it is predicted that the H2S concentration in the raw coke oven gas will exceed the plant boundary limit value after 8 minutes, and a dynamic emission reduction strategy including operations such as starting the dry quenching device in advance and increasing the spray intensity of the scrubber by 20% is generated. This strategy also stipulates that if the real-time monitored steam flow rate drops by more than 10% of the rated value, the standby emission reduction plan will be automatically switched.

[0058] S1044: Calculate the adjustment threshold of the equipment operation parameters in the dynamic emission reduction strategy according to the pollution diffusion path prediction result and the preset emission standard boundary conditions.

[0059] In S1044, the industrial production monitoring device performs matching calculations on the pollution diffusion prediction result and the environmental protection standard. When the model output shows that the PM2.5 concentration 200 meters northwest of the coke oven will exceed the national standard limit of 75 μg / m³ after 300 seconds, it is first identified that the pollution source corresponding to this exceeding standard area is the failure of the riser pipe seal of Coke Oven No. 3. By querying the equipment database, it is confirmed that the maximum allowable emission load of this pollution source equipment is 2.8 kg / h, and the current real-time emission rate is 3.1 kg / h. According to the adjustable parameter range specified in the equipment process manual, it is determined that the emission load can be reduced to 2.5 kg / h by raising the water seal height of the riser pipe from 1200 mm to 1350 mm.

[0060] Furthermore, the industrial production monitoring device generates a step-by-step load reduction strategy implemented in three stages based on the above content: in the first stage, the water seal height is raised by 50 mm within 60 seconds, in the second stage, it is raised by 70 mm after an interval of 90 seconds, and in the third stage, the remaining 30 mm of the lifting amplitude is dynamically adjusted according to the real-time monitored data. At the same time, in combination with the pump energy consumption data fed back by the smart electricity meter, the execution time interval of each stage is extended by 10 seconds to ensure the stability of the power grid, and finally the adjusted water seal height threshold and stage time parameters are written into the equipment control instruction.

[0061] In a preferred embodiment, calculating the adjustment threshold of the device operation parameters in the dynamic emission reduction strategy according to the pollution diffusion path prediction result and the preset emission standard boundary condition in S1044 includes:

[0062] S10441: Identify the target area that exceeds the emission standard boundary condition in the pollution diffusion path prediction result and the pollution source device identifier associated with the target area.

[0063] In S10441, the industrial production monitoring device locates the correspondence between the pollution exceeding standard area and the pollution source through the geographic information system. When the output of the pollution reduction and carbon emission reduction prediction model shows that the concentration of benzo[a]pyrene at 500 meters downwind of the factory area will exceed 2.5 ng / m³ in the next 10 minutes, combined with the real-time wind speed and wind direction data and the historical pollution event library, it is determined that the pollution source corresponding to this exceeding standard area is the aging of the furnace door sealing strip of the No. 2 coke oven. By analyzing the equipment topology structure, the pollution source device identifier is associated with "coke oven body - No. 2 carbonization chamber - south furnace door", and the maintenance records of the last three months recorded in the digital twin model of the device are extracted to confirm that its maximum allowable leakage rate is 0.8 m³ / min.

[0064] S10442: Extract the historical environmental parameter repair records of the target area, and determine the equipment type, maximum emission load, and adjustable operation parameter range corresponding to the pollution source device identifier.

[0065] In S10442, the industrial production monitoring device retrieves the historical regulation records of the target equipment for parameter optimization. For the identified leakage problem of the south furnace door of the No. 2 coke oven, it is found that in a similar working condition last month, by increasing the sealing strip pressing force from 12 MPa to 14 MPa, the leakage rate was successfully controlled within 0.6 m³ / min. Combining the technical specifications provided by the equipment manufacturer, it is confirmed that the adjustable pressure range of the hydraulic regulation system of this furnace door is 10 - 16 MPa, and the response time is 45 seconds / MPa. At the same time, obtain the coke pushing cycle parameters at the current moment, and confirm that there is no planned coke pushing operation in the next 15 minutes, and there is an operation window for implementing pressure adjustment.

[0066] S10443: Generate a step-by-step load reduction strategy for the pollution source device identifier based on the maximum emission load and the adjustable operation parameter range.

[0067] In S10443, the industrial production monitoring device formulates a step-by-step equipment control plan. Based on the identified leakage limit of 0.8m³ / min and the current actual value of 0.9m³ / min, it is calculated that the leakage needs to be reduced by 25%. According to the equipment adjustment characteristics, it is determined to implement pressure increase in three stages: in the first stage, the clamping force is increased from 12MPa to 13MPa within 60 seconds, and the leakage is expected to be reduced by 10%; in the second stage, it is increased to 14MPa after an interval of 120 seconds, and then reduced by 10%; the remaining 5% of the adjustment is completed dynamically in the third stage according to the real-time monitoring data. A pressure sensor calibration link is set in each stage to ensure that the single adjustment range does not exceed the equipment safety threshold.

[0068] S10444: Dynamically allocate the load reduction amplitude and time interval in the device operation parameter adjustment threshold according to the constraint relationship between the step-by-step load reduction strategy and the real-time energy consumption rate.

[0069] In S10444, the industrial production monitoring device can monitor the power load margin of the entire plant in real time in the process of coordinating energy consumption constraints and emission reduction requirements, taking into account that the hydraulic system boost operation will increase the power consumption by an additional 2.3kW. When it is found that the total load of the plant reaches 95% of the critical value when the first stage boost operation is executed, the execution interval of the second stage is automatically extended from 120 seconds to 180 seconds, and the pressure adjustment range of the third stage is reduced from 0.5MPa to 0.3MPa. Through this dynamic adjustment, the total power load is always controlled below 98% of the rated capacity.

[0070] S10445: Map the load reduction amplitude and the time interval to a device operation parameter adjustment threshold in the multi-level control instruction set.

[0071] In S10445, the industrial production monitoring device converts the optimized parameters into executable instructions. The parameters such as the pressure adjustment threshold of 13.8MPa and the stage interval of 150 seconds are encapsulated as OPC UA protocol instructions and sent to the hydraulic control system of the No. 2 coke oven through the industrial ring network. At the same time, the relevant threshold parameters are written into the interlocking protection module of the distributed control system. When the actual pressure is detected to deviate from the target value by more than 0.2MPa, the alarm is automatically triggered and the backup pressure pump is started.

[0072] In an optional embodiment, the step of generating a multi-level control instruction set for the target industrial production equipment according to the pollution emission characteristic vector and the dynamic emission reduction strategy in S106 includes:

[0073] S1061: Analyze the pollution emission characteristic vector and the dynamic emission reduction strategy to identify pollution emission equipment groups and energy consumption nodes that are in an energy-wasting state.

[0074] In this embodiment, the industrial production monitoring device performs multi-dimensional analysis on the pollution emission characteristic vectors and dynamic emission reduction strategies of the coke oven body. Specifically, by analyzing the quantitative relationship between the volatile organic compound emission rate and the coke maturity reflected by the seventh-dimensional eigenvalue, the main pollution emission equipment group composed of the rising pipe water seal device and the coke side oven door sealing system of the coke oven is identified.

[0075] Meanwhile, based on the no-load operation data of the dust removal fan during the non-pushing coke period in the energy consumption rate distribution matrix, it is determined that this equipment is in a state of efficiency waste. For example, when the pollution emission characteristic vector shows that the contribution degree of benzo[a]pyrene concentration in the rising pipe area of Coke Oven No. 3 reaches 68% of the total emissions, the system marks it as a key pollution source; while the power consumption ratio of the dust removal fan recorded by the intelligent electricity meter during the coke cake maturity stage exceeds the process standard value by 15%, it is identified as an energy consumption node that needs to be optimized.

[0076] S1062: Obtain the operation status data of the pollution emission equipment group, and determine the current start-stop status and treatment efficiency of the pollution treatment node associated with the pollution emission equipment group based on the operation status data and the energy consumption node.

[0077] In this embodiment, the industrial production monitoring device obtains the real-time operation status of the pollution emission equipment group through the industrial bus protocol. For the identified rising pipe water seal device, it collects the water level sensor data, hydraulic pressure value, and readings of the surrounding temperature monitoring points, and confirms that the current water seal height is maintained at 1200 mm with minute fluctuations three times per minute. Then, it synchronously retrieves the operation log of the associated crude benzene scrubbing tower and confirms that it is in continuous operation mode but the absorption efficiency is only 82% of the design value. By comparing the real-time treatment efficiency curve with the preset emission reduction target, it is found that when the pressure in the coke oven gas collecting pipe exceeds 4.8 kPa, the benzene series removal rate of the scrubbing tower drops below 75%, indicating that this pollution treatment node fails to effectively match the current working condition requirements.

[0078] S1063: Generate a forced shutdown instruction, a time-limited operation window, or an efficiency compensation mode in the start-stop rule of the pollution treatment node according to the difference between the treatment efficiency and the preset emission reduction target.

[0079] In this embodiment, the industrial production monitoring device formulates a refined control rule according to the efficiency deviation of the pollution treatment node. When it is detected that the circulating gas purification efficiency of the coke dry quenching device is 10% lower than the reference value, the system calculates that the carbon dioxide emission equivalent per unit ton of coke treated reaches 1.8 kg, exceeding the preset low-carbon treatment threshold of 1.5 kg. Immediately, a forced shutdown instruction is generated and the wet desulfurization tower is activated as a standby treatment node.

[0080] For the coke oven gas purification system in the buffer operation state, when the PM2.5 concentration shows an upward trend in the southeast corner of the factory area, dynamically allocate its limited-time operation window to the high-load production period from 08:00 to 12:00 every day, and switch to the low-power standby mode at other times. When the steam consumption of the chemical production and recovery section is lower than the design parameters but the treatment efficiency meets the standards, automatically enable the efficiency compensation mode and replace 20% of the steam power of the waste heat boiler with the clean power input of the factory photovoltaic energy storage system.

[0081] S1064: Embed the forced shutdown instruction, the limited-time operation window, or the efficiency compensation mode into the multi-level control instruction set according to the priority.

[0082] In this embodiment, the industrial production monitoring device constructs a priority execution system for multi-level instructions. Set the forced shutdown instruction for adjusting the water seal height of the riser to the highest priority and embed it into the safety interlock system of the coke oven body through hardwired direct connection; set the limited-time operation window instruction of the dust removal fan to the medium priority and align the time sequence with the production scheduling plan of the production scheduling system; the efficiency compensation mode is used as an interruptible instruction, allowing the temporary suspension of the clean energy switching operation during the power grid peak shaving period. All instructions ensure the mode switching of relevant equipment is completed 300 seconds before the start of the coke pushing operation through the time synchronization function of the OPC UA protocol.

[0083] In an alternative embodiment, the generating the forced shutdown instruction, the limited-time operation window, or the efficiency compensation mode in the start-stop rule of the pollution treatment node according to the difference between the treatment efficiency and the preset emission reduction target in S1063 includes:

[0084] S10631: Monitor the real-time treatment efficiency and energy consumption data of the pollution treatment node, and calculate the carbon emission equivalent per unit pollution treatment volume of the pollution treatment node.

[0085] In this embodiment, the industrial production monitoring device monitors the operation parameters of the coking wastewater treatment station in real time. By analyzing the COD removal rate of the biochemical treatment tank and the aerator energy consumption curve, it is calculated that 2.3 kg of carbon dioxide equivalent is generated per cubic meter of wastewater treated. This value is compared with the baseline carbon intensity of the coke oven chimney emissions to evaluate the environmental efficiency of the wastewater treatment node. For example, when it is monitored that the power consumption per unit denitrification of the denitrification reactor increases by 15% year-on-year, the carbon emission equivalent recalculation process is automatically triggered to update its environmental efficiency rating.

[0086] S10632: If the carbon emission equivalent exceeds the preset low-carbon treatment threshold, trigger the forced shutdown instruction and activate the standby treatment node.

[0087] In this embodiment, the industrial production monitoring device performs emergency control of high-carbon emission equipment. When it is detected that the regeneration cycle of the wet desulfurization tower is shortened to 2 hours per time and the carbon emission equivalent per unit sulfide treatment amount exceeds the 3.2 kg threshold, a forced shutdown instruction is immediately sent to the DCS system. At the same time, the standby processing node, the selective catalytic reduction device, is activated, and through the preset ammonia injection amount compensation algorithm, it is ensured that the flue gas nitrogen oxide concentration remains up to standard during the equipment switching process. The shutdown instruction includes a three-level confirmation mechanism: first, cut off the power supply of the desulfurization tower circulation pump, second, close the intake valve of the regeneration tower, and finally start the nitrogen purging program of the tower body.

[0088] S10633: If the carbon emission equivalent is within the preset buffer range, update the atmospheric pollutant concentration trend in the set according to the current environmental parameters, and dynamically allocate the start time and duration of the time-limited operation window.

[0089] In this embodiment, the industrial production monitoring device implements dynamic operation window regulation. For the medium-temperature naphthalene removal device in the coke oven gas purification system, when its carbon emission equivalent is in the buffer range of 1.6 - 1.8 kg / m³, the system combines the real-time wind speed and wind direction data transmitted back by the factory meteorological station, and allocates for this equipment to operate only under the condition of southeast wind, with the daily cumulative working time not exceeding 6 hours. For example, during the period when the prevailing wind direction in spring is northwest wind, the start time of the naphthalene removal device is automatically postponed to 14:00 in the afternoon, using the atmospheric diffusion conditions to reduce the risk of local pollutant accumulation.

[0090] S10634: If the carbon emission equivalent is lower than the low-carbon treatment threshold, but the treatment efficiency is lower than the preset reference value, increase the proportion of auxiliary clean energy input in the efficiency compensation mode.

[0091] In this embodiment, the industrial production monitoring device optimizes the energy structure of inefficient equipment. When the processing efficiency of the oil and gas recovery device in the crude benzene section is 5% lower than the design value, but the carbon emission equivalent remains at an excellent level of 1.2 kg / kg benzene series, the efficiency compensation mode is enabled. The specific measures include: switching 30% of the power of the drive motor of the washing liquid circulation pump to waste heat steam turbine power; introducing 15% of biomass gas to replace coke oven gas during the heating stage of the regeneration tower. Through the dynamic adjustment of the energy input ratio, the equipment processing efficiency is restored to more than 98% of the reference value.

[0092] S10635: Synchronize the forced shutdown instruction, the time-limited operation window, or the efficiency compensation mode with the production plan of the target industrial production equipment to generate a time-series start-stop rule instruction.

[0093] In this embodiment, the industrial production monitoring device realizes the deep coupling of control instructions and production plans. The forced shutdown instruction of the coke dry quenching device is matched with the coke oven pushing plan in terms of time sequence to ensure that the shutdown operation is executed after the coke quenching operation of the current furnace chamber is completed. At the same time, the limited-time operation window instruction of the wet desulfurization tower is decomposed into a micro-instruction sequence at 15-minute intervals, and is precisely synchronized with the gas switching cycle of the coke oven heating system. All start-stop rule instructions are verified through the work order management module of the manufacturing execution system (MES) to avoid conflicts with regular operations such as planned maintenance and equipment inspection.

[0094] In an alternative embodiment, the execution weights of the multi-level control instruction set are dynamically adjusted based on the real-time feedback-based updated set of environmental parameters in S108, including:

[0095] S1081: Receive the adjusted equipment operation parameters and the updated set of environmental parameters fed back by the control terminal.

[0096] In S1081, the industrial production monitoring device receives the adjusted equipment operation parameters uploaded by the coke oven body control terminal through the OPC UA protocol, including the correction value of the temperature gradient distribution of the No. 3 combustion chamber, the real-time measured value of the water seal height of the riser pipe, and the speed setting parameter of the dust removal fan inverter. At the same time, integrate the latest updated set of environmental parameters uploaded by the factory environmental monitoring station, which includes the PM2.5 concentration decline curve recorded by the β-ray absorption method particulate matter monitor on the southwest side, the slow decline trend of the CO2 emission intensity measured by the tunable diode laser absorption spectrometer, and the data on the increased steam production rate of the waste heat boiler fed back by the intelligent electricity meter group.

[0097] Based on this, the industrial production monitoring device uses the time series alignment algorithm to match the traveling speed of 0.78 m / s recorded by the pusher positioning sensor with the benzo[a]pyrene concentration of 1.2 ng / m³ detected by the ultraviolet differential absorption spectrometer at the same moment, forming a device-environment association data set containing time sequence tags.

[0098] S1082: Calculate the deviation degree between the updated set of environmental parameters and the expected emission reduction targets, and divide the weight adjustment interval according to the deviation degree.

[0099] In S1082, the industrial production monitoring device calculates the dynamic deviation degree between the updated set of environmental parameters and the preset emission reduction targets. By comparing the percentage difference between the current average PM2.5 concentration and the target value, and superimposing the second derivative of the carbon emission intensity change rate, a three-dimensional deviation degree evaluation matrix is constructed.

[0100] For example, when the PM2.5 concentration downwind of the coke oven chimney is still 12% higher than the target value after adjustment, and the decline rate of the carbon emission intensity is 30% lower than expected, the system determines that the deviation degree is in the second interval. The weight adjustment interval is divided using a fuzzy logic algorithm, and a multi-factor coupling analysis is performed by combining the coke maturity index and the energy consumption data of the dust removal fan to determine the adjustment amplitude of the priority of each control instruction.

[0101] S1083: If the deviation degree is in the first interval, increase the execution weight of the energy switching priority by a fixed step size.

[0102] In S1083, the industrial production monitoring device performs a weight increment operation for the deviation degree in the first interval. When it is detected that the over-standard range of the benzo[a]pyrene concentration at the factory boundary is within the range of 5% - 8% and lasts for 120 seconds, the switching priority of gas - natural gas is increased by a fixed step size of 0.05 per minute.

[0103] For example, during the operation of the coke oven heating system, if the updated set of environmental parameters shows that the pollutant diffusion speed at the monitoring point on the northeast side is insufficient, the execution weight of the natural gas substitution ratio is gradually increased from 0.65 to 0.8, forcing an increase in the supply of high - calorific value clean energy. This operation is implemented through the fuel distribution module of the DCS system, and the opening and closing sequence of the main gas pipeline cut - off valve and the output pressure of the natural gas pressure regulating station are adjusted synchronously.

[0104] S1084: If the deviation degree is in the second interval, dynamically scale the effective ratio of the device operation parameter adjustment threshold based on the change rate of the pollution emission eigenvector and the adjusted device operation parameters.

[0105] In S1084, the industrial production monitoring device performs dynamic scaling control on the deviation degree in the second interval. When there is a deviation of 15% - 20% between the actual sealing effect after the adjustment of the riser water seal height and the expected emission reduction target, based on the minute - level change trend of the volatile organic compound emission rate in the pollution emission eigenvector, the traveling speed threshold of the pusher car is dynamically adjusted from 0.8 ± 0.1 m / s to 0.75 ± 0.05 m / s.

[0106] Synchronously, according to the real - time monitoring data of the waste heat boiler steam flowmeter, the set value of the circulating gas flow of the coke dry quenching device can be reduced by a scaling factor of 0.9 to ensure that the effective ratio of the device operation parameter adjustment threshold matches the current working condition. This process verifies the integrity of the control logic through a Petri net model to prevent system oscillations caused by multi - parameter adjustments.

[0107] S1085: If the deviation degree is in the third interval, reset the multi - level control instruction set and re - trigger the strategy generation process of the pollution reduction and carbon emission reduction prediction model.

[0108] In S1085, the industrial production monitoring device processes severe deviation events in the third interval. When the updated set of environmental parameters shows that the CO concentration around the No. 3 coke oven rises by more than 50 ppm within 300 seconds and the dynamic emission reduction strategy fails, a global control instruction set reset operation is executed. First, terminate the pressure regulation instruction of the gas purification system that is being executed. Subsequently, re-initialize the input interface of the pollution reduction and carbon emission reduction prediction model, and input the latest collected coke cake temperature distribution matrix and dust removal fan vibration spectrum data into the model for strategy regeneration. During the reset process, continuously monitor the basic safety parameters of the coke oven body to ensure that the old and new instruction sets are switched within 0.5 seconds.

[0109] As an optional embodiment, the method further includes: embedding an adaptive learning module in the pollution reduction and carbon emission reduction prediction model, and the adaptive learning module is used to perform the following operations:

[0110] S202: Collect the execution results of the dynamic emission reduction strategy and the corresponding environmental parameter change trajectories within the historical period.

[0111] In this embodiment, the adaptive learning module of the industrial production monitoring device collects the control strategy execution data within the historical period. Extract 368 coke pushing operation records within the past 24 hours from the coke oven body operation database, including the adjustment value of the traveling speed of the coke pushing car each time, the change curve of the gas valve opening degree, and the corresponding PM2.5 concentration response data of the environmental monitoring station. At the same time, integrate the energy consumption characteristic maps of the dust removal fan under different load conditions to construct a multi-dimensional training data set including timestamps, equipment status, and environmental responses.

[0112] S204: Extract the pollution control events that do not meet the standards and the abnormal segments of the associated equipment operation parameters in the execution results.

[0113] In this embodiment, the industrial production monitoring device analyzes the pollution control events that do not meet the standards. By comparing the preset benzo[a]pyrene concentration threshold with the actual monitoring value, 3 unqualified events that occur when the coke oven switching cycle is shortened to 18 minutes are identified. Through correlation analysis, it is found that the abnormal segments of the corresponding equipment operation parameters include the loss of monitoring data of the riser water seal height, the positioning deviation of the coke pushing car exceeding 5 cm, and other abnormal conditions. The industrial production monitoring device extracts the coke tower temperature gradient distribution data, raw gas flowmeter readings, and waste heat boiler steam pressure fluctuation curves within 5 minutes before and after these events to form an abnormal condition feature vector set.

[0114] S206: Convert the pollution control events and the abnormal segments of the equipment operation parameters into model training samples, and update the feature extraction weights in the pollution reduction and carbon emission reduction prediction model based on the model training samples.

[0115] In this embodiment, the industrial production monitoring device converts pollution events into model training samples. For the identified riser seal failure event, the pressure fluctuation spectrum of the water seal hydraulic system during the abnormal period is extracted, and the main harmonic components are extracted through Fourier transform as the key pollution factors. The feature vector containing 12 harmonic amplitudes is spatially and temporally aligned with the rising curve of benzo[a]pyrene concentration in the synchronous environmental monitoring to generate a labeled training sample set. This sample set is used to adjust the feature extraction mode of the temporal convolutional layer in the pollution reduction and carbon emission reduction prediction model.

[0116] S208: Re-evaluate the current set of environmental parameters according to the updated pollution reduction and carbon emission reduction prediction model and the trajectory of the change in the environmental parameters, and generate an optimized dynamic emission reduction strategy to overwrite the multi-level control instruction set.

[0117] In this embodiment, the industrial production monitoring device applies the optimized model to regenerate the control strategy. Based on the updated feature extraction weights, the efficiency of the raw gas purification system under the current coke oven operation state is evaluated, and a new type of dynamic emission reduction strategy including frequency fine-tuning of the dry coke quenching device and riser water seal pressure compensation is generated. The new strategy overwrites the original control instruction set and is sent to the on-site actuator through the industrial wireless Mesh network. At the same time, the interlock protection parameter database of the DCS system is updated to ensure the real-time matching of the control logic and the physical characteristics of the equipment.

[0118] Specifically, updating the feature extraction weights in the pollution reduction and carbon emission reduction prediction model based on the model training sample in S206 includes:

[0119] S2061: Conduct a root cause analysis of the pollution control event to determine the key pollution factors in the abnormal segments of the equipment operation parameters.

[0120] In S2061, the industrial production monitoring device performs root cause tracing of pollution events. A multi-dimensional analysis is carried out on the PM2.5 exceeding standard event in the southeast corner of the coke oven, and it is determined that the key pollution factor is that the instantaneous fluctuation of the air-fuel ratio in the No. 3 combustion chamber exceeds ±8%. By correlating and analyzing the equipment logs, it is found that this fluctuation is caused by the misoperation of the oxygen regulating valve due to the calibration lag of the gas calorimeter analyzer. The system extracts the air-fuel ratio change rate, the coke oven top temperature gradient, and the dust removal fan current fluctuation data during the abnormal period, and constructs a feature matrix including time causal relationships.

[0121] S2062: Generate an error backpropagation link between the key pollution factor and the model prediction output layer.

[0122] In S2062, the industrial production monitoring device establishes an error backpropagation link, maps the identified air-fuel ratio fluctuation factor to the PM2.5 concentration prediction error at the output layer of the pollution reduction and carbon emission reduction prediction model, and constructs a gradient transfer path including a 7-layer neural network. During the backpropagation process, the response sensitivity of the temporal convolutional layer to the sudden change characteristics of the gas calorific value is mainly corrected to enhance the model's ability to capture the correlation between the combustion chamber pressure pulsation and pollutant generation.

[0123] S2063: Adjust the convolution kernel parameters of the convolutional neural network layer and the forgetting gate weights of the long short-term memory network layer in the pollution reduction and carbon emission reduction prediction model through the gradient descent algorithm.

[0124] In S2063, the industrial production monitoring device implements model parameter optimization, uses the stochastic gradient descent algorithm to adjust the weight distribution of the 3×3×2 convolution kernel in the temporal convolutional layer, and improves the extraction accuracy of the correlation characteristics between the coke oven switching cycle and NOx emissions by 15%. Synchronously update the forgetting gate parameters of the long short-term memory network layer to reduce the memory intensity of the historical invalid vibration data of the dust removal fan. A momentum factor is introduced during the parameter adjustment process to prevent the model from falling into a local optimal solution during the extraction of coke maturity characteristics.

[0125] S2064: Based on the set of simulated environment parameters, determine the strategy generation accuracy rate of the adjusted pollution reduction and carbon emission reduction prediction model. If the strategy generation accuracy rate meets the standard, lock the target weights; otherwise, roll back to the previous version of the pollution reduction and carbon emission reduction prediction model; where the target weights include the convolution kernel parameters and the forgetting gate weights.

[0126] In S2064, the industrial production monitoring device can use the set of simulated environment parameters to verify the model update effect for strategy generation testing, which includes boundary conditions such as artificially set sudden changes in coke oven pressure and step changes in gas calorific value. When the accuracy rate of the updated model in judging the coke cake maturity is increased to 98% and the response time of the emission reduction strategy is shortened to 8 seconds, the system locks the current convolution kernel parameters and forgetting gate weights. If it is found through testing that the error rate of the start-stop rule generation of the scrubber exceeds the tolerance limit, it will automatically roll back to the previous version of the model and trigger an alarm to notify the maintenance personnel to intervene and calibrate.

[0127] In an optional embodiment, the determining the strategy generation accuracy rate of the adjusted pollution reduction and carbon emission reduction prediction model based on the set of simulated environment parameters in S2064, and if the strategy generation accuracy rate meets the standard, locking the target weights, otherwise rolling back to the previous version of the pollution reduction and carbon emission reduction prediction model, includes:

[0128] S20641: Obtain the set of simulated environment parameters associated with the adjusted pollution reduction and carbon emission reduction prediction model; where the set of simulated environment parameters includes simulated atmospheric pollutant concentration, simulated carbon emission intensity, and simulated energy consumption rate.

[0129] In an embodiment of the present invention, an industrial production monitoring device calls a simulation environment parameter generation engine to construct a test data set. This data set includes a CO2 concentration fluctuation curve simulating the emissions from a coke oven chimney, a PM2.5 diffusion model on the southwestern side of the factory area, and a combination of coke pushing operation cycle parameters. Among them, the simulated atmospheric pollutant concentration is generated through fluid dynamics simulation, covering the benzo[a]pyrene concentration gradient distribution within a radius of 500 meters around the coke oven body; the simulated carbon emission intensity is reconstructed by interpolating historical production data, reflecting the unit coke carbon emissions under different coke oven gas calorific value conditions; the simulated energy consumption rate is output by using an equipment digital twin model, including the power consumption characteristic spectrum of the circulation fan of the coke dry quenching device in an abnormal vibration state. All simulated parameters are kept in temporal consistency with the real working condition data through a timestamp alignment algorithm, forming a test case set containing 72 hours of continuous working conditions.

[0130] S20642: Input the set of simulated environment parameters into the adjusted pollution reduction and carbon emission reduction prediction model to generate a predicted result of the simulated pollution diffusion path and a simulated dynamic emission reduction strategy.

[0131] In an embodiment of the present invention, an industrial production monitoring device inputs a set of simulated environment parameters into an updated pollution reduction and carbon emission reduction prediction model for strategy deduction. The pollution reduction and carbon emission reduction prediction model processes the working condition of the simulated coke oven body when the commutation cycle is shortened to 15 minutes, generates a predicted result of the pollution diffusion path for the next 3 hours, showing that the PM2.5 concentration in the area covered by the dust removal fan on the northeast side will exceed the limit value of 90 μg / m³. The simultaneously output simulated dynamic emission reduction strategy includes 12 control instructions. The third instruction requires raising the height of the riser water seal to 1350 mm, and the fifth instruction stipulates that when the fluctuation of the coke oven gas calorific value exceeds ±5%, the natural gas emergency replacement program should be started. The model also predicts that after implementing this strategy, the benzo[a]pyrene emission can be reduced to 1.8 ng / m³, and the carbon emission intensity can be reduced by 18%.

[0132] S20643: Extract the simulated values of the equipment operation parameter adjustment thresholds and the simulated instructions for the start-stop rules of pollution treatment nodes in the simulated dynamic emission reduction strategy.

[0133] In an embodiment of the present invention, an industrial production monitoring device extracts the key parameters in the simulated dynamic emission reduction strategy, and analyzes a simulated threshold value of 0.72 m / s from the control instruction for the traveling speed of the coke pusher. This value has a 3% negative deviation compared with the standard operation parameter range of 0.75 - 0.85 m / s. The simulated instructions for the start-stop rules of pollution treatment nodes show that when the simulated main gas pipe pressure reaches 5.2 kPa, the crude benzene scrubbing tower needs to enter the high-pressure spraying mode 20 seconds in advance, and this requirement conflicts with the 15-second buffer period stipulated in the current safety operation regulations. The system records such parameter deviations as feature vectors for subsequent accuracy evaluation.

[0134] S20644: According to the preset emission standard boundary conditions, match the simulated value of the equipment operation parameter adjustment threshold with the standard operation parameter adjustment threshold range, and verify the conflict status between the start-stop rule simulation instruction of the pollution treatment node and the emission reduction compliance rule; count the matching success rate of the simulated value of the equipment operation parameter adjustment threshold and the conflict coverage rate of the start-stop rule simulation instruction of the pollution treatment node, and calculate the strategy generation accuracy of the adjusted pollution reduction and carbon emission reduction prediction model based on the matching success rate and the conflict coverage rate.

[0135] In the embodiment of the present invention, the industrial production monitoring device performs strategy compliance verification. For example, compare the gas valve opening threshold in the simulated dynamic emission reduction strategy with the process design specification, and it is found that the opening adjustment value in the fourth stage exceeds the mechanical limit of the equipment by 2.3 percentage points. At the same time, it is detected that the execution of the start-stop rule of the coke dry quenching device in the simulated environment will cause the load rate of the dust removal fan to exceed the rated value, triggering the safety interlock protection mechanism. The system calculates that the matching success rate of the simulated value of the equipment operation parameter adjustment threshold is 87.5%, and the conflict coverage rate of the start-stop rule simulation instruction of the pollution treatment node is 14.3%. The comprehensive strategy generation accuracy is 82.6%.

[0136] S20645: If the strategy generation accuracy exceeds the preset accuracy threshold, lock the target weight in the adjusted pollution reduction and carbon emission reduction prediction model to the current convolution kernel parameter and the current forget gate weight; if the strategy generation accuracy does not exceed the accuracy threshold, load the convolution kernel parameter and the forget gate weight of the previous version of the pollution reduction and carbon emission reduction prediction model from the model version library, and overwrite the parameters of the adjusted pollution reduction and carbon emission reduction prediction model.

[0137] In the embodiment of the present invention, the industrial production monitoring device performs model parameter management according to the accuracy threshold: when the measured accuracy of 82.6% exceeds the preset qualified line of 80%, lock the convolution kernel parameter of the current version, and the distribution mode of [0.78, -0.23, 0.45] in the weight matrix of the third-layer convolution kernel is permanently saved. The forget gate weight of the long short-term memory network layer is adjusted from 0.32 to 0.28 to enhance the memory ability for sudden changes in coke maturity. If the accuracy does not meet the standard, restore the previous model parameters from the version library, including rolling back the weight configuration of [0.65, 0.12, -0.09] of the second-time series convolution layer.

[0138] S20646: Synchronize the locked target weight or the rolled-back convolution kernel parameter and forget gate weight to the online prediction interface of the pollution reduction and carbon emission reduction prediction model, so that the subsequent prediction results of the pollution diffusion path and the generation and update of the dynamic emission reduction strategy are consistent with the updated model parameters.

[0139] In an embodiment of the present invention, the industrial production monitoring device completes the model parameter synchronization operation, writes the locked convolution kernel parameters into the GPU acceleration module of the prediction model through an encrypted channel, and simultaneously updates the weight configuration file of the online interface. The updated model is immediately applied to real-time data processing. When abnormal fluctuations in the temperature of the regenerator of Coke Oven No. 2 are detected, the new convolution kernel can accurately identify the correlation characteristics between this working condition and the increase in the concentration of H2S in raw coke oven gas, and generate an optimization strategy including compensation for the ammonia water injection volume.

[0140] In another optional embodiment, the step of obtaining the set of environmental parameters in S102 includes:

[0141] S1021: Real-time collect the original data of atmospheric pollutant concentrations through the gas sensor network deployed in the associated area.

[0142] In an embodiment of the present invention, the industrial production monitoring device collects environmental data through a distributed gas sensor network. For example, 8 sets of beta-ray absorption method particulate matter monitors deployed on the top of the coke oven body measure the instantaneous concentration of PM2.5 during the coke pushing process at a frequency of 6 times per minute. 12 ultraviolet differential absorption spectrometers set at the factory boundary continuously scan the concentration distribution of benzo[a]pyrene and generate a three-dimensional pollution diffusion heat map every 10 seconds. All monitoring devices transmit the original data through an industrial wireless Mesh network and use time division multiple access technology to avoid channel conflicts.

[0143] S1022: Extract the hourly carbon emission intensity and energy consumption rate logs of the target industrial production equipment from the energy management platform.

[0144] In an embodiment of the present invention, the industrial production monitoring device docks with the data interface of the energy management platform, extracts the cumulative gas flow value from the PLC of the coke oven heating system, and calculates the carbon emission intensity per minute in combination with the coke output. The intelligent electricity meter group records the power consumption curves of main equipment such as dust removal fans and circulating water pumps with an accuracy of 0.5 seconds, and the steam flowmeter synchronously uploads the output data of the waste heat boiler. The system establishes a three-dimensional matrix of energy consumption rate, with the horizontal axis being the equipment number, the vertical axis being the time series, and the depth axis distinguishing energy types such as electricity and steam.

[0145] S1023: Perform noise filtering and outlier removal on the original data of atmospheric pollutant concentrations, the hourly carbon emission intensity, and the energy consumption rate logs to generate a preprocessed set of environmental parameters.

[0146] In an embodiment of the present invention, the industrial production monitoring device performs data cleaning and preprocessing, uses a sliding window filtering algorithm to eliminate instantaneous pulse interference in PM2.5 monitoring data, and identifies and eliminates abnormal sampling points of gas calorific value through box plot analysis. For the high-frequency noise data misreported by the dust removal fan vibration sensor, wavelet transform is applied to reconstruct the signal and restore the true load fluctuation characteristics. The preprocessed environmental parameter set retains 98.7% of the valid data points, forming a continuous and complete time series data set.

[0147] S1024: Align the preprocessed environmental parameter set according to the time stamp and compress it into a target transmission format; wherein, the target transmission format matches the pollution reduction and carbon emission reduction prediction model.

[0148] In an embodiment of the present invention, the industrial production monitoring device completes data format standardization, unifies parameters such as PM2.5 concentration and carbon emission intensity with different sampling frequencies to a time base of once per second, and performs lossless compression in HDF5 format. Meta-information such as device numbers and check codes is embedded in the data header, and is encapsulated into a 128-dimensional feature vector that meets the input requirements of the pollution reduction and carbon emission reduction prediction model through an industrial bus protocol. The size of each data packet is controlled within 512 bytes to ensure real-time transmission efficiency.

[0149] In another optional embodiment, the method further includes: deploying a policy execution monitoring module in the control terminal, and the policy execution monitoring module performs the following operations:

[0150] S302: Analyze the received multi-level control instruction set, and disassemble the multi-level control instruction set into operation codes that meet the device execution conditions.

[0151] In S302, the policy execution monitoring module of the industrial production monitoring device analyzes the multi-level control instructions, and converts the air-fuel ratio optimization instruction of the coke oven combustion chamber into a PID parameter adjustment code executable by the PLC, specifically including 32 operation parameters such as the opening change gradient of the oxygen flow valve and the gas pressure compensation coefficient. The traveling speed control instruction of the coke pusher is disassembled into underlying execution codes such as the set value of the servo motor speed and the track positioning verification frequency, and each code is attached with a safety verification identifier.

[0152] S304: Monitor the actual parameter adjustment amplitude and energy consumption change curve of the operation code during the operation of the device.

[0153] In S304, the industrial production monitoring device monitors the execution effect of the instruction in real time, tracks the actual traveling speed curve of the coke pusher through a laser displacement sensor, compares it with the target value of 0.75 m / s required by the control instruction, and calculates the instantaneous speed deviation value. Synchronously collect the output current waveform of the dust removal fan frequency converter, and analyze whether the energy consumption change trend meets the expected energy-saving target. All monitoring data is refreshed at a period of 200 ms to form an execution effect evaluation matrix with timestamps.

[0154] S306: If it is detected that the deviation between the actual parameter adjustment range and the instruction requirement exceeds the fault tolerance threshold in combination with the energy consumption change curve, trigger the device operation state rollback mechanism.

[0155] In S306, the industrial production monitoring device processes the event of exceeding the execution deviation standard. When it is detected that the actual value of the adjustment of the riser water seal height lags behind the instruction target by 120 mm and lasts for 30 seconds, trigger a three-level rollback mechanism: first, restore the hydraulic system pressure to the safety reference value, second, reset the integral term of the water seal height control loop, and finally restart the intelligent positioning module. Continuously monitor the seal performance index during the rollback process to ensure that the pollutant emission rate does not exceed the emergency threshold.

[0156] S308: Feed back the cause code of the device operation state rollback mechanism and the corrected parameter adjustment value to the pollution reduction and carbon emission reduction prediction model.

[0157] In S308, the industrial production monitoring device realizes an execution feedback closed loop, packs the rollback cause code "E0452" of the riser seal failure event and the corrected hydraulic pressure setting value into a feedback data packet, and uploads it to the pollution reduction and carbon emission reduction prediction model through the OPC UA protocol. The adaptive learning module of the model adjusts the confidence weight of the water seal height control strategy accordingly, and increases the safety margin design of the pressure compensation coefficient when generating the next strategy.

[0158] As an alternative but non-limiting embodiment, the extracting the spatial correlation features of the multi-dimensional environmental state tensor by using the convolutional neural network layer in the pollution reduction and carbon emission reduction prediction model in S1043, and processing the spatial correlation features through a long short-term memory network layer to generate a time-dependent pollution diffusion path prediction result, and generating the dynamic emission reduction strategy based on the pollution diffusion path prediction result includes:

[0159] S10431: Perform a sliding scan of the multi-dimensional environmental state tensor with multi-scale convolutional kernels to generate a spatial feature map containing local pollutant distribution patterns.

[0160] In S10431, a multi-scale convolutional kernel scanning operation is performed on the multi-dimensional environmental state tensor constructed for the associated area of the coke oven body in the coking plant. This tensor integrates twelve dimensions of environmental and operating parameters such as the temperature gradient distribution in the coke oven combustion chamber, the time-series data of the traveling speed of the coke pusher, and the thermal map of the PM2.5 concentration at the plant boundary. A convolutional kernel with a size of 3×3×2 is slid along the time axis for scanning to identify the spatial correlation pattern between the abnormal temperature fluctuations in the riser pipe area of the coke oven and the peak concentration of benzo[a]pyrene on the southwest side. For example, when the convolutional kernel scans that the temperature of the No. 3 combustion chamber exceeds 1050 °C and the gas valve opening is at the critical value of 65% during the same period, the activation value at the corresponding position in the output feature map is significantly increased, and this area is marked as a hot spot for volatile organic compound emissions.

[0161] S10432: Perform a cross-channel max pooling operation on the spatial feature map to compress redundant noise and retain the spatial correlation features in the local pollutant distribution pattern.

[0162] In S10432, for the three-dimensional pollutant concentration field formed by the exhaust gas flow from the coke oven chimney, the pooling layer screens the peak response features in each channel to suppress the noise interference caused by the vibration of the dust removal fan. For example, during the coke oven switching cycle, the pooling operation retains the strong correlation features between the gas calorific value fluctuation and the CO2 emission gradient, while filtering out the transient outliers caused by the steam pressure fluctuation. After pooling compression, the dimension of the feature map is reduced to 30% of the original data, but the retention rate of the key spatial correlation features exceeds 95%.

[0163] S10433: Input the spatial correlation features into the long short-term memory network layer in the order of time windows, and capture the attenuation gradient of the pollutant concentration in adjacent time windows through the gating mechanism to generate a time-dependent pollutant diffusion intensity sequence.

[0164] In S10433, the industrial production monitoring device inputs the spatial correlation features into the long short-term memory network layer for time-series processing. The network forget gate dynamically adjusts the influence weight of the historical operation data of the coke oven. When it is detected that the traveling speed of the coke pusher is lower than 0.7 m / s for three consecutive cycles, the memory intensity of the coke cake maturity parameters is enhanced. The update gate adjusts the feature transfer ratio according to the real-time relationship between the current load rate of the dust removal fan and the pollutant diffusion rate. The output gate generates a pollutant diffusion intensity sequence for the next 15-minute time window, accurately reflecting the attenuation gradient of the benzo[a]pyrene concentration under the prevailing wind direction in the plant area.

[0165] S10434: Perform a regional superposition calculation based on the pollutant diffusion intensity sequence and the preset diffusion coefficient matrix to predict the migration path and concentration accumulation area of pollutants in multiple subsequent monitoring cycles, and generate the time-dependent pollution diffusion path prediction result.

[0166] In S10434, the industrial production monitoring device combines the atmospheric diffusion coefficient matrix to predict the pollution path. The diffusion intensity sequence output by the long short-term memory network is subjected to a tensor product operation with the real-time wind speed and wind direction data of the meteorological station to draw the migration trajectory of PM2.5 particles within the next three hours. The prediction results show that when the rotational speed of the dust removal fan on the northeast side of the coke oven drops to 85% of the rated value, a concentration accumulation area of coking benzene series will be formed 800 meters downwind. The system synchronously generates a heat map of the pollution diffusion path containing longitude and latitude coordinates and concentration thresholds, and marks the red warning area exceeding the limit value of 50 μg / m³.

[0167] S10435: Extract the critical area in the pollution diffusion path prediction result that exceeds the preset concentration threshold, and reverse-match the adjustable parameter items in the operating parameters according to the pollution source equipment identifier associated with the critical area.

[0168] In S10435, when the prediction path shows that the concentration of benzo[a]pyrene will exceed 2.5 ng / m³ at 200 meters around the riser pipe of Coke Oven No. 3 in 8 minutes, the industrial production monitoring device traces back to the operating parameters of the electric tar precipitator in the gas purification system. Through the mapping of the equipment topology relationship, the adjustable parameter item with the secondary voltage fluctuation of the current electric tar precipitator exceeding ±5% is identified and associated with the gas flow mutation event caused by the compression of the coke pushing cycle.

[0169] S10436: Generate a set of candidate emission reduction strategies containing the parameter adjustment sequence and amplitude combination based on the operation constraint conditions and historical adjustment effectiveness data of the adjustable parameter items; perform multi-objective optimization ranking of the energy consumption rate and emission reduction efficiency for the set of candidate emission reduction strategies, and select the strategy that meets the preset carbon emission intensity reduction rate and has the highest energy switching priority as the dynamic emission reduction strategy.

[0170] In S10436, for the voltage regulation item of the electric tar precipitator, three candidate schemes are screened out in combination with the historical regulation records: immediately increase the voltage to 45 kV, adjust it in stages to 42 kV, or switch to the standby power supply line. The multi-objective optimization algorithm evaluates the comprehensive effectiveness of each scheme. When the energy consumption growth rate of Scheme 2 is controlled at 8% and the benzene series removal rate is increased by 23%, it is marked as the optimal strategy. The final dynamic emission reduction strategy includes voltage adjustment instructions, speed reduction compensation measures for the coke pusher, and frequency conversion coordination parameters for the dust removal fan.

[0171] S10437: Perform spatio-temporal alignment verification on the dynamic emission reduction strategy and the pollutant migration path in the pollution diffusion path prediction result. If there is an uncovered pollutant accumulation area, trigger the strategy supplement mechanism and append the regional collaborative emission reduction instruction to the dynamic emission reduction strategy.

[0172] In S10437, after the industrial production monitoring device executes the dynamic emission reduction strategy, it monitors in real time and shows that there is still an unexpected accumulation of VOCs concentration in the crude benzene section area on the southeast side. The system immediately triggers a supplementary mechanism and issues an additional instruction to require the oil and gas recovery device in the chemical production and recovery section to start the high-pressure absorption mode 30 seconds in advance and increase the circulating flow rate of the washing liquid to 115% of the design value. The supplementary instruction is synchronously sent to the on-site PLC controller through the OPC UA protocol to ensure that the newly added pollution accumulation area is effectively controlled within 120 seconds.

[0173] As another alternative but non-limiting embodiment, synchronizing the forced shutdown instruction, the time-limited operation window, or the efficiency compensation mode with the production scheduling plan of the target industrial production equipment in S10635 to generate a time-serialized start-stop rule instruction includes:

[0174] S106351: Analyze the process section nodes and production batch timestamps in the production scheduling plan of the target industrial production equipment, and extract the production scheduling plan node set; wherein, the production scheduling plan node set includes the equipment start time, the raw material feeding cycle, and the finished product output time interval.

[0175] In S106351, the industrial production monitoring device analyzes the coke oven production scheduling plan and extracts the current production batch data from the manufacturing execution system, including key nodes such as the planned coke pushing time of the No. 2 carbonization chamber at 10:15:00 and the combustion chamber commutation cycle of the No. 3 combustion chamber from 08:30 to 09:45. When constructing the production scheduling plan node set, synchronously integrate the coal charging timestamp of the coke drum, the marking of the raw gas export stage, and the finished coke output time interval to form a spatio-temporal matrix containing 78 process section nodes.

[0176] S106352: Traverse the forced shutdown instruction, the time-limited operation window, or the efficiency compensation mode in the start-stop rules of the pollution treatment nodes, and identify the pollution treatment node identifier and the start-stop action type corresponding to each start-stop rule.

[0177] In S106352, the industrial production monitoring device matches the start-stop rules with the equipment identifiers, and identifies that the equipment number corresponding to the forced shutdown instruction of the dry quenching coke device is DJQ-03, the time-limited operation window instruction is associated with the dust removal fan CF-07, and the efficiency compensation mode acts on the crude benzene scrubbing tower XB-12. Each start-stop rule is appended with an action type code, such as the forced shutdown instruction is marked as an L1-level interruption operation, and the time-limited operation window is defined as a T2-level delayable action.

[0178] S106353: Match the pollution treatment node identifier with the equipment start time in the production scheduling plan node set to determine the execution time constraint of the start-stop action type in the production scheduling plan node set; wherein, the execution time constraint includes the earliest trigger time and the latest completion time allowing start-stop.

[0179] In S106353, it was found that the start and stop of crude benzene washing tower XB-12 needs to be implemented during the stable pressure stage of the coke oven gas collecting pipe, which corresponds to the 09:00-09:15 process window in the production schedule. The system sets the earliest triggering time of this instruction to be 30 seconds after the pressure stabilizes (09:00:30), and the latest completion time shall not be later than 60 seconds before the next batch of coal loading operation (09:14:00).

[0180] S106354: Generate an execution time window for the forced shutdown instruction, time-limited operation window or efficiency compensation mode according to the execution time constraint and the priority order of the start / stop action type; wherein the start time of the execution time window is no earlier than the earliest trigger time of the allowed start / stop and the end time is no later than the latest completion time.

[0181] In S106354, the time-limited operation window instruction of the dust removal fan CF-07 is allocated to the execution period of 10:05-10:25 in combination with the finished product output gap period of the production schedule. The start time of this window is 10 seconds later than the coke pushing completion signal of the No. 2 coke oven (10:05:10), and the end time is 300 seconds earlier than the reversing preparation stage of the No. 3 combustion chamber (10:22:00), ensuring that there is no conflict with the key production process.

[0182] S106355: Extract the energy consumption peak interval corresponding to the raw material delivery cycle in the production scheduling node set, perform overlapping analysis on the execution time window and the energy consumption peak interval, and adjust the starting time of the execution time window to avoid conflict with the energy consumption peak interval.

[0183] In S106355, industrial production monitoring devices can avoid energy consumption peaks. For example, if it is detected that the power load of the entire plant will reach its peak during the raw material delivery cycle of 10:00-10:15, the forced shutdown instruction execution window of the dry coke quenching unit DJQ-03 is adjusted from the original 10:10-10:30 to 10:16-10:36. The start time of the adjusted window is delayed by 6 minutes to avoid the maximum power consumption period of the raw material conveyor belt unit, ensuring that the power grid stability indicators meet safety regulations.

[0184] S106356: Generate a timestamp instruction chain of the forced shutdown instruction, time-limited operation window or efficiency compensation mode according to the adjusted execution time window and the finished product output time interval in the production scheduling node set; wherein each instruction in the timestamp instruction chain includes a target pollution treatment node identifier, a start / stop action type and an execution timestamp.

[0185] In S106356, the industrial production monitoring device can construct a timestamp instruction chain. For example, the efficiency compensation mode of the crude benzene scrubbing tower XB-12 is decomposed into three sub-instructions: start the high-pressure pump at 09:00:30, adjust the pH value of the washing liquid at 09:05:00, and activate the standby absorption tower at 09:10:15. Each instruction is embedded in the process gap of the production scheduling plan. For example, the start time of the high-pressure pump is arranged 15 seconds after the temperature measurement operation of the coke tower is completed, avoiding resource competition with the core production process.

[0186] S106357: Map the timestamp instruction chain to the device operation parameter adjustment threshold of the production scheduling plan node set, and verify whether the execution timestamp in the timestamp instruction chain causes the device operation parameter to exceed the device operation parameter adjustment threshold; if it exceeds, reassign the start time of the execution time window.

[0187] It can be understood that this step involves the verification of parameter threshold compliance. For example, it is detected that the pH value adjustment instruction at 09:05:00 will cause the instantaneous pressure of the scrubbing tower to exceed the upper limit of the device operation parameter adjustment threshold by 0.5 MPa, and the industrial production monitoring device automatically delays the execution time to 09:06:30. When the reallocated instruction is triggered, the pressure of the corresponding coke oven gas main pipe is in the descending stage, reserving sufficient safety margin for parameter adjustment.

[0188] S106358: Generate a time-serialized start-stop rule instruction sequence according to the verified timestamp instruction chain and the production batch timestamp in the production scheduling plan node set; where the start-stop rule instruction sequence is arranged in the order of the production batch timestamp, and the trigger time of each instruction is embedded in the process section node gap of the production scheduling plan node set.

[0189] In S106358, the industrial production monitoring device executes the step of generating the final start-stop rule sequence. For example, the instructions are arranged according to the production batch time axis to ensure that the time-limited instruction of the dust removal fan at 10:16:00 is fully executed before the commutation preparation stage of the No. 3 combustion chamber. The timestamp of each instruction is accurately aligned with the clock reference of the DCS system, and the deviation is controlled within ±50 ms, and the time synchronization function of the industrial bus protocol is used to achieve coordinated operation of the whole plant equipment.

[0190] See Figure 2 As shown, this figure is a schematic diagram of the basic structure of an industrial production monitoring device 200 provided by an embodiment of the present invention. The industrial production monitoring device 200 includes:

[0191] A processor 201;

[0192] A storage device 202, on which a computer program 2020 is stored;

[0193] When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the industrial production monitoring methods applied to environmental pollution reduction and carbon emission reduction.

[0194] On this basis, a readable storage medium is provided, and a program or instruction is stored on the readable storage medium. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0195] It should be noted that the various embodiments in this specification are described in a progressive manner. The key point of each embodiment is the difference from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the systems or devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

Claims

1. An industrial production monitoring method for ecological environmental pollution reduction and carbon reduction, characterized in that: include: Obtaining operating parameters of target industrial production equipment and a set of environmental parameters of associated areas, wherein the set of environmental parameters includes atmospheric pollutant concentration, carbon emission intensity, and energy consumption rate; The operating parameters and the environmental parameter set are processed by a pre-trained pollution reduction and carbon reduction prediction model to output the pollution emission characteristic vector and dynamic emission reduction strategy of the target industrial production equipment: the environmental parameter set is time-series aligned to extract the atmospheric pollutant concentration fluctuation curve, the carbon emission intensity change gradient and the energy consumption rate distribution matrix within the same time window; the atmospheric pollutant concentration fluctuation curve, the carbon emission intensity change gradient and the energy consumption rate distribution matrix are merged into a multi-dimensional environmental state tensor, and the pollution emission characteristic vector is generated based on the multi-dimensional environmental state tensor and the state characteristics of the operating parameters; The convolutional neural network layer in the pollution reduction and carbon reduction prediction model is used to extract the spatial correlation features of the multidimensional environmental state tensor, and the spatial correlation features are processed through the long short-term memory network layer to generate a time-dependent pollution diffusion path prediction result, and the dynamic emission reduction strategy is generated based on the pollution diffusion path prediction result; Calculating the equipment operating parameter adjustment threshold in the dynamic emission reduction strategy according to the pollution diffusion path prediction result and the preset emission standard boundary conditions; Generate a multi-level control instruction set for the target industrial production equipment according to the pollution emission characteristic vector and the dynamic emission reduction strategy, wherein the multi-level control instruction set includes equipment operation parameter adjustment thresholds, energy switching priorities, and pollution treatment node start and stop rules; The multi-level control instruction set is transmitted to the control terminal of the target industrial production equipment through the Internet of Things interface, and the execution weight of the multi-level control instruction set is dynamically adjusted based on the real-time feedback environmental parameter update set.

2. The method according to claim 1, characterized in that The calculating, according to the pollution diffusion path prediction result and the preset emission standard boundary condition, the equipment operation parameter adjustment threshold in the dynamic emission reduction strategy includes: Identifying a target area that exceeds the boundary conditions of the emission standard in the pollution diffusion path prediction result and a pollution source equipment identifier associated with the target area; Extract the historical environmental parameter restoration records of the target area, and determine the equipment type, maximum emission load and adjustable operating parameter range corresponding to the pollution source equipment identification; Based on the maximum emission load and the adjustable operating parameter range, generating a step-by-step load reduction strategy for the pollution source equipment identifier; According to the constraint relationship between the step-by-step load reduction strategy and the real-time energy consumption rate, dynamically allocating the load reduction amplitude and time interval in the device operation parameter adjustment threshold; The load reduction amplitude and the time interval are mapped to a device operation parameter adjustment threshold in the multi-level control instruction set.

3. The method according to claim 1, characterized in that The step of generating a multi-level control instruction set for the target industrial production equipment according to the pollution emission characteristic vector and the dynamic emission reduction strategy includes: Analyzing the pollution emission characteristic vector and the dynamic emission reduction strategy to identify pollution emission equipment groups and energy consumption nodes in an energy-wasting state; Acquire the operation status data of the pollution emission equipment group, and determine the current start / stop status and treatment efficiency of the pollution treatment node associated with the pollution emission equipment group based on the operation status data and the energy consumption node; According to the difference between the treatment efficiency and the preset emission reduction target, generating a forced shutdown instruction, a time-limited operation window or an efficiency compensation mode in the pollution treatment node start and stop rules; embedding the forced shutdown instruction, the time-limited operation window or the efficiency compensation mode into the multi-level control instruction set according to priority; The generating of a forced shutdown instruction, a limited time operation window or an efficiency compensation mode in the pollution treatment node start and stop rule according to the difference between the treatment efficiency and the preset emission reduction target includes: Monitor the real-time processing efficiency and energy consumption data of the pollution treatment node, and calculate the carbon emission equivalent of the unit pollution treatment amount of the pollution treatment node; If the carbon emission equivalent exceeds a preset low-carbon processing threshold, the forced shutdown instruction is triggered and the backup processing node is activated; If the carbon emission equivalent is within the preset buffer zone, the start time and duration of the time-limited operation window are dynamically allocated according to the trend of atmospheric pollutant concentration in the current environmental parameter update set; If the carbon emission equivalent is lower than the low-carbon treatment threshold, but the treatment efficiency is lower than a preset reference value, then the auxiliary clean energy input ratio is increased in the efficiency compensation mode; The forced shutdown instruction, the time-limited operation window or the efficiency compensation mode are synchronized with the production schedule of the target industrial production equipment to generate a time-serialized start-stop rule instruction.

4. The method according to claim 1, characterized in that: The updating set of environmental parameters based on real-time feedback dynamically adjusts the execution weight of the multi-level control instruction set, including: Receiving an updated set of adjusted equipment operating parameters and environmental parameters fed back by the control terminal; Calculating the deviation between the environmental parameter update set and the expected emission reduction index, and dividing the weight adjustment interval according to the deviation; If the deviation is in the first interval, increasing the execution weight of the energy switching priority by a fixed step size; If the degree of deviation is in the second interval, dynamically scaling the effective ratio of the device operating parameter adjustment threshold based on the rate of change of the pollution emission characteristic vector and the adjusted device operating parameter; If the degree of deviation is in the third interval, the multi-level control instruction set is reset and the strategy generation process of the pollution reduction and carbon reduction prediction model is re-triggered.

5. The method according to claim 1, characterized in that The method further comprises: An adaptive learning module is embedded in the pollution reduction and carbon reduction prediction model, and the adaptive learning module is used to perform the following operations: Collect the execution results of the dynamic emission reduction strategy and the corresponding environmental parameter change trajectory within the historical period; Extracting the pollution control events that do not meet the standards and the associated abnormal equipment operation parameter fragments in the execution results; Converting the pollution control events and the abnormal fragments of the equipment operation parameters into model training samples, and updating the feature extraction weights in the pollution reduction and carbon reduction prediction model based on the model training samples; Re-evaluate the current set of environmental parameters based on the updated pollution reduction and carbon reduction prediction model and the environmental parameter change trajectory, generate an optimized dynamic emission reduction strategy and cover the multi-level control instruction set; The updating of the feature extraction weights in the pollution reduction and carbon reduction prediction model based on the model training samples includes: Conduct root cause analysis on the pollution control events to determine key pollution factors in the abnormal equipment operating parameter segments; Generate an error back propagation link between the key pollution factor and the model prediction output layer; Adjusting the convolution kernel parameters of the convolutional neural network layer and the forget gate weights of the long short-term memory network layer in the pollution reduction and carbon reduction prediction model by a gradient descent algorithm; Based on a set of simulated environmental parameters, the strategy generation accuracy of the adjusted pollution reduction and carbon reduction prediction model is determined. If the strategy generation accuracy meets the standard, the target weight is locked, otherwise it is rolled back to the previous version of the pollution reduction and carbon reduction prediction model; wherein the target weight includes the convolution kernel parameters and the forget gate weight.

6. The method according to claim 5, characterized in that The method of determining the strategy generation accuracy of the adjusted pollution reduction and carbon reduction prediction model based on the simulated environmental parameter set, and locking the target weight if the strategy generation accuracy meets the standard, or rolling back to the previous version of the pollution reduction and carbon reduction prediction model, includes: Acquire a set of simulated environmental parameters associated with the adjusted pollution reduction and carbon reduction prediction model; wherein the set of simulated environmental parameters includes simulated atmospheric pollutant concentration, simulated carbon emission intensity and simulated energy consumption rate; Inputting the simulated environmental parameter set into the adjusted pollution reduction and carbon reduction prediction model to generate simulated pollution diffusion path prediction results and simulated dynamic emission reduction strategies; Extracting the equipment operation parameter adjustment threshold simulation value and the pollution treatment node start-stop rule simulation instruction in the simulated dynamic emission reduction strategy; According to the preset emission standard boundary conditions, the simulated value of the equipment operating parameter adjustment threshold is matched with the standard operating parameter adjustment threshold interval, and the conflict status between the pollution treatment node start and stop rule simulation instruction and the emission reduction compliance rule is verified; the matching success rate of the simulated value of the equipment operating parameter adjustment threshold and the conflict coverage rate of the pollution treatment node start and stop rule simulation instruction are counted, and the strategy generation accuracy of the adjusted pollution reduction and carbon reduction prediction model is calculated based on the matching success rate and the conflict coverage rate; If the strategy generation accuracy exceeds the preset accuracy threshold, the target weight in the adjusted pollution reduction and carbon reduction prediction model is locked to the current convolution kernel parameters and the current forget gate weight; if the strategy generation accuracy does not exceed the accuracy threshold, the convolution kernel parameters and forget gate weight of the previous version of the pollution reduction and carbon reduction prediction model are loaded from the model version library, and the parameters of the adjusted pollution reduction and carbon reduction prediction model are overwritten; The locked target weights or the rolled-back convolution kernel parameters and forget gate weights are synchronized to the online prediction interface of the pollution reduction and carbon reduction prediction model, so that the subsequent pollution diffusion path prediction results and the generation of dynamic emission reduction strategies are consistent with the updated model parameters.

7. The method according to claim 1, characterized in that The step of obtaining the environmental parameter set includes: Collecting raw data of atmospheric pollutant concentrations in real time through a gas sensor network deployed in the associated area; Extracting hourly carbon emission intensity and energy consumption rate logs of the target industrial production equipment from the energy management platform; Perform noise filtering and outlier removal on the raw data of atmospheric pollutant concentration, the hourly carbon emission intensity and the energy consumption rate log to generate a preprocessing environmental parameter set; The pre-processing environment parameter set is aligned by timestamp and compressed into a target transmission format; wherein the target transmission format matches the pollution reduction and carbon reduction prediction model.

8. The method according to claim 1, characterized in that The method further comprises: A policy execution monitoring module is deployed in the control terminal, and the policy execution monitoring module performs the following operations: Parsing the received multi-level control instruction set, and disassembling the multi-level control instruction set into operation codes that meet the device execution conditions; Monitor the actual parameter adjustment range and energy consumption change curve of the operation code during the operation of the device; If it is detected in combination with the energy consumption change curve that the deviation between the actual parameter adjustment range and the instruction requirement exceeds the fault tolerance threshold, the device operation status rollback mechanism is triggered; The reason code of the equipment operation status rollback mechanism and the corrected parameter adjustment value are fed back to the pollution reduction and carbon reduction prediction model.

9. An industrial production monitoring device, characterized in that: include: processor; A storage device having a computer program stored thereon, wherein when the computer program is executed by the processor, the processor implements the industrial production monitoring method for ecological environmental pollution reduction and carbon reduction as described in any one of claims 1 to 8.

Images