An intelligent regulation and control method and system for a dust removal system of a steel enterprise

Abstract

An intelligent regulation and control method and system for a dust removal system of a steel enterprise, comprising the following steps: collecting real-time data at a set frequency, including operating parameters at each dust generation point, valve opening, operating parameters of dust collectors, ash conveying equipment, fans and exhaust conditions, and dividing the data into operating variables and regulating variables; constructing a hybrid neural network model, including an LSTM-GPT-Neo main network, a DenseNet branch network and a meta-learning feature fusion correction layer, the LSTM-GPT-Neo main network combining LSTM and GPT-Neo architectures; inputting time series data for model training and optimization; inputting the data into the trained hybrid neural network model for calculation and analysis, and outputting the optimal operating parameters of the valve opening and the fan. The present application can significantly reduce the power consumption of the steel plant while ensuring the dust removal efficiency and stability, and avoid the problems of dust emission exceeding the standard due to insufficient air volume or equipment wear caused by excessive air volume.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, and particularly relates to a steel enterprise dust removal system intelligent regulation and control method and system. BACKGROUND

[0002] The steel industry is a key field of energy consumption and pollutant emission, and a large amount of dust is generated in each process of the steel production process. Improving the operation energy efficiency of the dust removal system is crucial for energy saving and consumption reduction and environmental protection compliance of the steel enterprise. In the process of the double carbon target work, improving the energy efficiency of the dust removal system is the key work of energy saving and consumption reduction of the steel plant.

[0003] The traditional dust removal system fan often adopts a fixed air volume operation mode, and often appears dust emission exceeding the standard due to insufficient air volume, or unnecessary energy consumption and equipment wear and tear due to excessive air volume. At present, the air volume regulation of the fan is more dependent on manual adjustment, which is often inaccurate due to the limitation of personnel experience, and cannot be adjusted in real time, causing unstable operation, emission not meeting the standard or waste of energy. In addition to the fan, the operation of the valve of the pipeline and the dust cleaning and ash conveying equipment also faces the problems of automation and precision control. In the intelligent era, production big data is a valuable asset, and various data in the actual operation of production can be a powerful basis for production control and potential tapping, especially in terms of energy saving and consumption reduction, real-time data collection and utilization are more important. According to the real-time air volume and pressure, dust production, dust point, dust collector resistance, dust cleaning and ash conveying operation state and other parameters in the production process of the steel plant, the valve opening, the operation parameters of the fan and the dust cleaning and ash conveying equipment are adjusted in real time, the operation efficiency of the dust removal system is improved, the power consumption cost of the steel plant is effectively reduced, and the carbon emission of the steel enterprise is reduced. SUMMARY

[0004] The purpose of the present application is to overcome the shortcomings of the prior art, provide a steel enterprise dust removal system intelligent regulation and control method and system, and solve the problems of energy waste in the traditional dust removal system of the steel industry, so as to realize efficient and energy-saving operation of the dust removal system.

[0005] The present application is realized by the following technical solutions:

[0006] A steel enterprise dust removal system intelligent regulation and control method, comprising the following steps:

[0007] S1, collecting real-time data according to a set frequency, and dividing the collected data into operation variables and adjustment variables; the collected data includes operation parameters at each dust point, valve opening K1 of the valve between each dust point and the dust collector, operation parameters of the dust collector, operation parameters of the ash conveying equipment, operation parameters of the fan and exhaust condition parameters;

[0008] S2, a hybrid neural network model is constructed, the hybrid neural network model comprising an LSTM-GPT-Neo main network, a DenseNet branch network and a meta-learning feature fusion correction layer; the LSTM-GPT-Neo main network is located at the front end of the model, and is used for calculating the optimal operating parameter distribution of the valve opening degree and the fan according to the input operating variable data, and the LSTM-GPT-Neo main network fuses the LSTM and GPT-Neo architecture, the LSTM is used to extract high-level time sequence features from the input operating variable data, the time sequence features contain key information of the equipment operating state, so that the GPT-Neo can perform more complex reasoning, and the GPT-Neo architecture can understand and analyze the operating variable data at a semantic level by introducing the idea of a large-scale pre-trained language model, and can enhance the ability to capture long-distance dependencies and complex patterns; the DenseNet branch network is used for calculating and analyzing the input adjustment variable data, and outputs a dust removal performance parameter, the dust removal performance parameter is used to represent the mapping relationship between the adjustment variable and the dust removal efficiency; the meta-learning feature fusion correction layer adopts a meta-learning strategy, and is used for dynamically adjusting the fusion weight and the fusion mode according to the outputs of the LSTM-GPT-Neo main network and the DenseNet branch network, so as to adapt to various complex and changeable situations;

[0009] S3, time sequence data of operating variables and adjustment variables are collected to form a data set, the data set is divided into a training set and a validation set in chronological order, the hybrid neural network model is trained through the training set, and the hybrid neural network model is optimized through the validation set;

[0010] S4, the operating variable data and the adjustment variable data are input into the trained hybrid neural network model for calculation and analysis, and the optimal operating parameters of the valve opening degree K1 and the fan are output, and the operation of the valve and the fan is real-time regulated and controlled according to the parameters; the optimal operating parameters of the fan include the speed r of the fan, the opening K2 of the inlet air valve and the opening K3 of the outlet air valve.

[0011] Further, the operating parameters of the dust generation point include air volume q1, air pressure p, temperature t1, humidity content wet and dust concentration m1, the operating parameters of the dust collector include dust collector resistance zu, ash amount c1 in the ash bucket, ash cleaning period T1, ash cleaning time T2 and ash cleaning intensity F, the operating parameters of the ash conveying equipment include ash conveying speed v and ash conveying amount c2, the operating parameters of the fan include the speed r of the fan, the opening K2 of the fan inlet air valve, the opening K3 of the fan outlet air valve, the temperature t2 and the current I, and the exhaust condition parameters include the exhaust air volume q2 discharged from the dust collector and the dust concentration m2 in the gas discharged from the dust collector;

[0012] The operation variables include air volume q1, air pressure p, temperature t1, humidity content wet, dust concentration m1, dust generation point s, resistance zu of the dust collector, ash content c1 in the ash bucket, temperature t2 of the fan, current I of the fan, air volume q2 of the dust collector, and dust concentration m2; the adjustment variables include rotating speed r of the fan, valve opening K1, opening K2 of the fan inlet valve, opening K3 of the fan outlet valve, ash cleaning period T1, ash cleaning time T2, ash cleaning intensity F, ash conveying speed v, and ash conveying amount c2.

[0013] Further, further, provided that there are n dust generation points, s dust collectors, and u fans in the dust removal system, and there are f collection time points, the processing and representation method of the operation variable data is:

[0014] The processing is performed by the function y=Y(q1, p, t1, wet, m1, zu, c1, t2, I, q2, m2), and the air volume q1, air pressure p, temperature t1, humidity content wet, and dust concentration m1 at the dust generation point can be represented as three-dimensional arrays, respectively q1[i][j][k], p[i][j][k], t1[i][j][k], wet[i][j][k], and m1[i][j][k], wherein i represents the dust generation point number, 0≤i≤n, j represents the collection time point number, 0≤j≤f, and k represents the data dimension;

[0015] The resistance zu of the dust collector, the ash content c1 in the ash bucket, the exhaust air volume q2 of the dust collector, and the dust concentration m2 are represented by three-dimensional arrays, respectively zu[h][j][k], c1[h][j][k], q2[h][j][k], and m2[h][j][k], wherein h represents the dust collector number, 0≤h≤s, j represents the collection time point number, 0≤j≤f, and k represents the data dimension;

[0016] The temperature t2 and current I of the fan are represented by three-dimensional arrays, respectively t2[g][j][k] and I[g][j][k], wherein g represents the fan number, 0≤g≤u, j represents the collection time point number, 0≤j≤f, and k represents the data dimension.

[0017] The processing and representation method of the adjustment variable data is:

[0018] The adjustment variables are processed through the function t=T(r,K1,K2,K3,T1,T2,F,v,c2). The rotational speed r of each fan, valve opening K1, fan inlet valve opening K2, fan outlet valve opening K3, cleaning cycle T1, cleaning time T2, cleaning intensity F, ash conveying speed v, and ash conveying amount c2 can be represented as three-dimensional arrays r[g][j][k], K1[g][j][k], K2[g][j][k], K3[g][j][k], T1[g][j][k], T2[g][j][k], F[g][j][k], v[g][j][k], and c2[g][j][k], respectively. Here, g is the fan number, 0≤g≤u. For ash conveying equipment, g is the fan number corresponding to the ash conveying equipment. j is the data collection time point, 0≤j≤f, and k is the data dimension.

[0019] The data for the operating variables and adjustment variables need to be preprocessed. The preprocessing method is as follows:

[0020] The data is cleaned, obvious abnormal data is identified and removed, and missing values ​​are imputed. For cases where the proportion of missing data is small, it can be directly filled, while for cases where the proportion of missing data is large, it is directly discarded.

[0021] Then, the data with different dimensions and ranges are normalized and converted into dimensionless numbers;

[0022] Then represent the dimensionless data as a vector X = [x1, x2, ..., x...]. n ], where x i Let be the i-th input parameter, and n be the total number of input parameters.

[0023] Furthermore, it also includes: S5, by monitoring and controlling the air volume q1, air pressure p, dust concentration m1 at the dust generation point, the energy consumption of the dust collector, ash conveying equipment and fan, and exhaust parameters, evaluating the energy-saving effect of the control, the stability of system operation and the dust removal effect, and inputting the evaluation results into the hybrid neural network model for iterative optimization, so as to continuously improve the control strategy of the hybrid neural network model.

[0024] Furthermore, the LSTM-GPT-Neo backbone network is located at the front end of the model, and its output data represents the possible optimal values ​​of valve opening K1, fan speed r, fan inlet valve opening K2 and outlet valve opening K3 in the form of probability distribution, as well as confidence intervals and uncertainty estimates.

[0025] The LSTM unit employs a stacked structure of four LSTM layers, each containing 512 neurons, and incorporates an improved variant of the gated recurrent unit (GRU) to capture complex time-series features in running variable data. The computational analysis method of the LSTM unit is as follows:

[0026] Each LSTM layer is used to capture the time series characteristics of parameters, and residual connections are introduced between each LSTM layer, and a skip connection is added between two LSTM layers, the input of the previous LSTM layer is directly added to the output of the next LSTM layer element by element, and the output is:

[0027]

[0028] wherein, is the output hidden state after the residual connection and the skip connection operation, h t is the hidden state after processing by the current LSTM layer at time step t, h t-1 is the hidden state of the previous time step t-1;

[0029] The neurons in the LSTM layer are connected through weights, each neuron performs weighted summation on the input data and activation function conversion, and the output of the jth neuron h j is represented as: wherein w ij is the weight of the ith input parameter to the jth neuron, b j is the bias of the jth neuron, and f is the activation function ReLU, which is represented as: f(x) = max(0, x);

[0030] The DenseNet branch network includes four DenseBlock layers and a Transition layer between each two DenseBlock layers, each DenseBlock layer includes three convolutional layers and a skip connection, and each Transition layer is composed of a convolutional layer and a pooling layer; the meta-learning feature fusion correction layer includes a meta-learner MAML and a fusion module, the MAML obtains the adapted weight through gradient update according to the working condition characteristics, the fusion module performs weighted average on the dynamic weight, and finally outputs a corrected prediction value as the output data of the entire hybrid neural network model.

[0031] Further, in the S3 step, the training method of the hybrid neural network model is:

[0032] S3-1, collect historical running sequence data, including running parameters at each dust generating point, valve opening K1 of each dust generating point and dust collector, running parameters of the dust collector, running parameters of the ash conveying equipment, running parameters of the fan and exhaust condition parameters;

[0033] S3-2, after removing the running abnormal data (such as substandard dust removal effect, fan temperature or current exceeding the standard, dust collector running abnormally, etc.), the historical running data is evaluated, and the historical running data with excellent energy saving effect, system running stability and dust removal effect is selected.

[0034] S3-3, the screened data are taken as a data set, and are randomly divided into a training set and a validation set according to time training at a certain ratio; the data in the data set are input into a mixed neural network model after being preprocessed, a supervised learning method is adopted, valve opening K1 and the optimal operation parameter of the fan are taken as target variables, weights are updated through an Adam optimization algorithm, and a parameter update formula is:

[0035]

[0036] wherein, θ is a parameter to be optimized, η is a learning rate, and are first-order and second-order moment estimates after bias correction, and ∈ is a very small value (generally 10 -8 ) added to prevent division by zero.

[0037] Further, in the step S3-2, the method for data evaluation and screening is that: the average energy consumption of the system, the change rate of the air volume and air pressure at the dust removal point, and the dust removal effect (dust removal efficiency) in each time period are calculated, the data ranked after 40% are removed according to the dust removal effect, then the remaining data are sorted according to the average energy consumption, the data ranked after 30% are removed, then the remaining data are sorted according to the change rate of the air volume and air pressure, the data ranked after 20% are removed, and the screened data are obtained.

[0038] Further, in the step S3-3, the parameter setting for model training is that: the learning rate is set to 0.001, the maximum number of pre-training iterations is 500, the maximum number of fine-tuning iterations is 200; the sparse parameter is set to 0.4, which is helpful to reduce redundant information; the sparse penalty term parameter is 2, which is used to control the punishment degree of sparsity; the activation function adopts the ReLU function; and the fine-tuning loss function selects the mean square error loss function MSE.

[0039] Further, in the step S5, the method for inputting the evaluation result into the mixed neural network model for iterative optimization includes:

[0040] S5-1, the definitions of positive samples and negative samples are determined according to actual business requirements and evaluation targets;

[0041] S5-2, after the positive and negative samples are determined, a confusion matrix is constructed according to the evaluation data obtained through monitoring;

[0042] S5-3, based on the constructed confusion matrix, the evaluation indexes of the mixed neural network model are calculated, the evaluation indexes are input into the mixed neural network model, the current performance of the model is measured through a loss function, and then the regulation and control strategy is continuously improved.

[0043] Further, the evaluation result includes the following parameters:

[0044] Air volume tracking error rate v q : calculated according to the following formula:

[0045]

[0046] Wherein, q pred is the target air volume at the predicted dust generation point, q post is the measured air volume at the dust generation point after intelligent regulation;

[0047] Air pressure stability index s p : calculated according to the following formula:

[0048]

[0049] Wherein, σ(p post ) is the standard deviation of the air pressure at the dust generation point after intelligent regulation, μ(p post ) is the mean value of the air pressure at the dust generation point after intelligent regulation;

[0050] Dust concentration compliance rate R m : calculated according to the following formula:

[0051]

[0052] Wherein, T is the total sampling number in the statistical period, II() is an indicator function, m t is the dust concentration in the gas discharged from the dust collector at the tth sampling time.

[0053] Energy saving rate e: calculated according to the following formula:

[0054]

[0055] Wherein, E1 is the total energy consumption of the dust removal system before intelligent regulation, E2 is the total energy consumption of the dust removal system before and after intelligent regulation;

[0056] Air volume change rate Q: calculated according to the following formula:

[0057]

[0058] Wherein, q1 is the air volume at the dust removal point before intelligent regulation, q2 is the air volume at the dust removal point after intelligent regulation;

[0059] Air pressure change rate P: calculated according to the following formula:

[0060]

[0061] Wherein, p1 is the wind pressure at the dust removal point before intelligent regulation and control, and p2 is the wind pressure at the dust removal point after intelligent regulation and control.

[0062] Dust removal efficiency η: calculated according to the following formula:

[0063]

[0064] Wherein, m in is the dust concentration entering the dust remover, and m out is the dust concentration discharged from the dust remover.

[0065] Dust removal efficiency change: calculated according to the following formula:

[0066]

[0067] Wherein, η1 is the dust removal efficiency before intelligent regulation and control, and η2 is the dust removal efficiency after intelligent regulation and control.

[0068] An intelligent regulation and control system for a dust removal system of a steel enterprise, for realizing the above-mentioned intelligent regulation and control method for the dust removal system of the steel enterprise, comprising dust removal equipment, a data acquisition module, an intelligent regulation and control module, a running regulation and control module, and a monitoring and optimization module.

[0069] The dust removal equipment comprises gas hoods, conveying pipelines, valves, dust removers, ash conveying equipment, fans, and chimneys. The gas hoods are arranged at dust generation points. The gas hoods are connected to the dust removers through the conveying pipelines. The valves are arranged on the conveying pipelines. The dust removers are arranged at the rear ends of the gas hoods. The dust removers are provided with ash removal equipment for cleaning dust accumulated in the dust removers. The ash conveying equipment is arranged at the outlets of the ash hoppers of the dust removers for conveying the accumulated dust. The air inlets of the fans are connected to the dust removers, and the air outlets of the fans are connected to the chimneys.

[0070] The data acquisition module comprises a plurality of sensors and a data collector. The sensors are arranged at the positions of the conveying pipelines, the valves, and the dust removers connected to the gas hoods at each dust generation point. The data collector is connected to each of the sensors, the valves, the dust removers, the ash conveying equipment, and the fans for collecting the running parameters at each dust generation point, the valve opening degrees of the valves, the running parameters of the dust removers, the running parameters of the ash conveying equipment, and the running parameters of the fans. The data acquisition module is connected to the intelligent regulation and control module and the monitoring and optimization module to transmit the collected data.

[0071] The intelligent regulation and control module is embedded with a hybrid neural network model for calculating and analyzing the collected data to obtain the valve opening degree K1 and the optimal running parameters of the fan.

[0072] The operation regulation module is connected with the intelligent regulation module, and is used for receiving the optimal operation parameters of the fan obtained by the intelligent regulation module; the operation regulation module is also connected with the fan, and is used for adjusting the rotating speed of the fan, the opening degree of the air inlet valve and the opening degree of the air outlet valve according to the optimal operation parameters of the fan obtained by the intelligent regulation module, so as to control the size of the air volume of the fan;

[0073] The monitoring optimization module comprises a data analysis unit, a closed-loop feedback unit and a data storage unit; the data analysis unit is used for performing calculation analysis according to the collected data, and evaluating the energy-saving effect of regulation, the stability of system operation and the dust removal effect; the closed-loop feedback unit is used for transmitting the evaluation result to the intelligent regulation module, and performing iterative optimization in the mixed neural network model to form a dynamic optimization closed loop; and the data storage unit is used for storing the collected data and the data obtained by calculation analysis.

[0074] Further, the sensors comprise an air volume measuring instrument, an air pressure gauge, a dust concentration sensor, a temperature and humidity meter, a pressure sensor and a material level meter; the air volume measuring instrument is arranged at the conveying pipeline and the dust collector at each dust generation point, and is used for measuring the air volume q1 at the dust generation point and the exhaust air volume q2 exhausted from the dust collector; the air pressure gauge is arranged at the conveying pipeline at the dust generation point, and is used for measuring the air pressure p at the dust generation point; the dust concentration sensor is arranged at the conveying pipeline and the dust collector at the dust generation point, and is used for measuring the dust concentration m1 at the dust generation point and the dust concentration m2 exhausted from the dust collector; the temperature and humidity meter is arranged at the conveying pipeline at the dust generation point, and is used for measuring the temperature t1 at the dust generation point and the moisture content wet; and the pressure sensor and the material level meter are both arranged in the dust collector, and are used for measuring the resistance zu of the dust collector and the amount c1 of ash in the ash bucket.

[0075] The dust generation points are coke oven charging and coke guiding and discharging sites, secondary dust removal sites of converters and blast furnace tapping sites; and the gas collecting hoods are arranged at positions above the coke oven furnace heads, coke blocking machines, coke pushing cars, converter furnace mouths and blast furnace tapping mouths.

[0076] The intelligent regulation module further comprises a data cleaning unit, a dust collector operation optimization unit and a ash conveying equipment operation optimization unit; the data cleaning unit is used for removing abnormal data by adopting the 3σ principle, and performing normalization processing on the data; the dust collector operation optimization unit is used for analyzing the equipment operation state of the dust collector based on historical data and real-time data by a fault diagnosis algorithm, so as to dynamically adjust the ash cleaning period T1, the ash cleaning time T2 and the ash cleaning intensity F of the dust collector; and the ash conveying equipment operation optimization unit is used for analyzing the equipment operation state of the ash conveying equipment by a fault diagnosis algorithm, so as to dynamically adjust the ash conveying speed v and the ash conveying amount c2 of the ash conveying equipment.

[0077] The operation regulation module is connected with the dust collector and the ash conveying device at the same time, so as to dynamically adjust the ash cleaning period T1, the ash cleaning time T2 and the ash cleaning intensity F of the dust collector, and the ash conveying speed v and the ash conveying amount c2 of the ash conveying device according to the data of the intelligent regulation module.

[0078] The application can collect real-time parameters such as real-time air volume and pressure, dust production, dust production point, dust collector resistance, ash cleaning and ash conveying operation state of each key node of the steel plant in real time, adjust valve opening and fan control parameters in real time, accurately calculate the best operation parameters of the fan by using an intelligent regulation algorithm model, and automatically adjust the speed and air volume output (inlet and outlet air valves) of the fan, so that the fan can always operate under the best working condition, the power consumption of the steel plant can be significantly reduced under the condition of ensuring the dust removal efficiency and stability, the dust removal system can be stably and efficiently operated under different working conditions, the problems of dust emission exceeding the standard due to insufficient air volume or equipment wear caused by excessive air volume can be avoided, the operation of the dust collector and the ash conveying device can be intelligently regulated, the operation of these devices is more reasonable, the wear of the dust removal filter material, the dust removal air pipe, the ash conveying equipment can be greatly reduced, and the service life of the dust collector, the fan and other equipment can be prolonged. The training of the intelligent regulation algorithm model of the application can use historical operation data and data corrected from the historical operation data, the data can be automatically collected, a large amount of data can be provided for the data set of the model, and the operability of model training is improved. The intelligent regulation system of the application can automatically adjust with the change of the working condition, has strong adaptability to the working condition with unstable air volume, does not need frequent manual intervention, the flexibility and response speed of the system are improved, real-time monitoring, intelligent regulation and optimized management of the dust removal system are realized, and the needs of the green, low-carbon and high-quality development of the steel industry are met. BRIEF DESCRIPTION OF DRAWINGS

[0079] Figure 1 The structure block diagram of the dust removal equipment in the regulation system of the embodiment of the application.

[0080] Figure 2 The structure block diagram of the regulation system of the embodiment of the application.

[0081] Figure 3 The flowchart of the regulation method of the embodiment of the application.

[0082] Figure 4 The working principle block diagram of the regulation method of the embodiment of the application.

[0083] Figure 5 The framework structure diagram of the mixed neural network model in the embodiment of the application.

[0084] The figure shows that 1 is a gas collecting hood, 2 is a conveying pipeline, 3 is a valve, 4 is a dust collector, 5 is an ash cleaning device, 6 is an ash conveying device, 7 is a fan, and 8 is a chimney. Detailed Implementation

[0085] An intelligent control system for dust removal systems in steel enterprises, such as Figure 2 As shown, it includes dust removal equipment, a data acquisition module, an intelligent control module, an operation control module, and a monitoring and optimization module.

[0086] like Figure 1 As shown, the dust removal equipment includes a gas collection hood 1, a conveying pipe 2, a valve 3, a dust collector 4, an ash conveying device 6, a fan 7, and a chimney 8. Several gas collection hoods 1 are installed at the dust generation point. The gas collection hoods 1 are connected to the dust collector 4 through the conveying pipe 2. The valve 3 is installed on the conveying pipe 2. The dust collector 4 is equipped with a dust cleaning device 5, which is used to clean the dust accumulated in the dust collector 4. The ash conveying device 6 is installed below the dust collector 4 and is used to convey the accumulated dust. The air inlet of the fan 7 is connected to the dust collector 4, and the air outlet of the fan 7 is connected to the chimney 8. The gas after dust removal is discharged from the chimney 8.

[0087] One or more dust collection hoods 1 can be installed at a dust-generating point, depending on the area and dust concentration of the point. For steel enterprises, dust-generating points are mainly concentrated in processes such as coke oven coal charging and coke discharging, converter secondary dust removal, and blast furnace tapping. The dust collection hood 1 can be installed at locations such as the carbonization chamber furnace head, coke quencher, coke pusher, converter furnace mouth, and blast furnace tapping mouth. The dust collector 4 can be a bag filter 4, electrostatic precipitator 4, centrifugal dust collector 4, or a composite dust collector 4, etc., and is installed at the rear end of the dust collection hood 1. The corresponding dust removal equipment 5 can be a pulse jet cleaning system, wet scrubbing system, mechanical vibration system, or reverse air cleaning system. Currently, pulse jet bag filters 4 are more commonly used. The ash conveying equipment 6 can be conventional conveying equipment, such as screw conveyors, pneumatic conveying systems, and rotary valves. In steel enterprises, some dust has recycling value and can be transported to the corresponding section for further processing, such as being transported to the ash silo via ash conveying pipelines. The ash conveying device 6 is located at the ash hopper outlet of the dust collector 4. Its operation can be controlled according to the real-time ash volume to avoid continuous operation when there is no ash or the ash volume is low, which would cause energy waste and equipment wear. The fan 7 is an axial flow fan or a centrifugal fan, located between the outlet of the dust collector 4 and the chimney 8.

[0088] The data acquisition module includes multiple sensors and a data collector. The sensors are respectively installed at the locations of the conveying pipe 2, valve 3, and dust collector 4 connected to the dust collection hood 1 at each dust generation point. The data collector is connected to the sensors, valve 3, dust collector 4, ash conveying equipment 6, and fan 7, and is used to collect the operating parameters at each dust generation point, the valve opening degree of valve 3, the operating parameters of dust collector 4, the operating parameters of ash conveying equipment 6, and the operating parameters of fan 7. The data acquisition module is connected to the intelligent control module and the monitoring and optimization module to transmit the collected data.

[0089] The intelligent control module is embedded with a hybrid neural network model, which is used to calculate and analyze the collected data to obtain the optimal operating parameters of the valve opening K1 and the fan.

[0090] The operation control module is connected with the intelligent control module, and is used to receive the optimal operating parameters of the fan obtained by the intelligent control module. The operation control module is also connected with the fan, and is used to adjust the rotating speed and air volume of the fan according to the optimal operating parameters of the fan obtained by the intelligent control module. The opening of the inlet and outlet air valves of the fan can be automatically adjusted by the actuator to control the air volume of the fan, so that the opening percentage required by the optimal operating parameters is reached. Meanwhile, the rotating speed of the fan is automatically adjusted by the frequency converter, so that the rotating speed required by the optimal operating parameters is reached, thereby controlling the air volume output.

[0091] The monitoring optimization module includes a data analysis unit, a closed-loop feedback unit and a data storage unit. The data analysis unit is used to calculate and analyze the collected data to evaluate the energy-saving effect of the control, the stability of the system operation and the dust removal effect, so as to ensure the effectiveness and reliability of the control method. The closed-loop feedback unit is used to transmit the evaluation results to the intelligent control module to iteratively optimize the hybrid neural network model and continuously improve the control strategy to form a dynamic optimization closed loop. The data storage unit is used to store the collected data and the data obtained by calculation and analysis.

[0092] Of course, a communication module should also be provided to realize data transmission between these modules through wired or wireless communication.

[0093] Specifically, the sensors include an air volume measuring instrument, an air pressure gauge, a dust concentration sensor, a temperature and humidity meter, a pressure sensor and a material level meter. The air volume measuring instrument is arranged at the conveying pipeline 2 and the dust collector 4 at each dust generation point, and is used to measure the air volume q1 at the dust generation point and the exhaust air volume q2 discharged from the dust collector 4. The air pressure gauge is arranged at the conveying pipeline 2 at the dust generation point, and is used to measure the air pressure p at the dust generation point. The dust concentration sensor is arranged at the conveying pipeline 2 and the dust collector 4 at the dust generation point, and is used to measure the dust concentration m1 at the dust generation point and the dust concentration m2 discharged from the dust collector 4. The temperature and humidity meter is arranged at the conveying pipeline 2 at the dust generation point, and is used to measure the temperature t1 at the dust generation point and the moisture content wet. The pressure sensor and the material level meter are both arranged in the dust collector 4, and are used to measure the resistance zu of the dust collector 4 and the amount of ash c1 in the ash hopper. Other data to be collected are directly connected with the corresponding devices, and the devices are provided with corresponding detection units to detect the data in real time. For example, the rotating speed, temperature and current value of the fan can be detected and controlled by the control unit of the fan, and the operating parameters of the dust collector and the ash conveying equipment can also be detected and controlled by the control unit, such as the ash cleaning period T1 and the ash cleaning time T2 of the dust collector.

[0094] The intelligent control module also includes a data cleaning unit, a dust collector operation optimization unit, and an ash conveying equipment operation optimization unit. The data cleaning unit is used to remove abnormal data using the 3σ principle, |x i If -μ|>3σ, the data is discarded and normalized.

[0095] The dust collector operation optimization unit is used to analyze the equipment operation status of the dust collector based on historical and real-time data and through fault diagnosis algorithms, so as to dynamically adjust the dust collector's cleaning cycle T1, cleaning time T2 and cleaning intensity F, so as to balance dust removal and energy saving.

[0096] The ash conveying equipment operation optimization unit is used to analyze the equipment operation status of the ash conveying equipment through fault diagnosis algorithms, promptly detect abnormal conditions of the equipment, and optimize the operation strategy of the ash conveying equipment based on the equipment operation status and the overall needs of the dust removal system. This mainly involves dynamically adjusting the ash conveying speed v and ash conveying amount c2 of the ash conveying equipment to improve the operating efficiency and reliability of the equipment and reduce energy consumption.

[0097] The fault diagnosis algorithm compares real-time data with the set normal range. If the data deviates from the normal range, it indicates that the equipment is operating abnormally. In this case, the operating status of the dust collector or ash conveying equipment needs to be adjusted according to the preset rules. The main operating status of the dust cleaning equipment and the ash conveying equipment is as follows: When the resistance zu of the dust collector 4 reaches a certain value, the dust cleaning equipment can be started in real time, and the dust cleaning time T2 and the dust cleaning intensity F can be dynamically adjusted according to the resistance zu. When the ash amount c1 in the ash hopper accumulates to a certain height, the ash conveying equipment can be started in real time, and the ash conveying speed v can be dynamically adjusted according to the ash amount c1.

[0098] The operation control module is connected to both the dust collector and the ash conveying equipment to dynamically adjust the dust collector's cleaning cycle T1, cleaning time T2, and cleaning intensity F, as well as the ash conveying speed v and ash conveying amount c2, based on data from the intelligent control module.

[0099] A method for intelligent control of dust removal systems in steel enterprises, such as Figure 3 It includes the following steps:

[0100] S1. Collect real-time data at a set frequency, which can be set to collect data once every 1 minute or 3 minutes. Figure 1 , Figure 4The collected data includes the operating parameters of each dust generation point, the valve opening degree K1 between each dust generation point s and the dust collector, the operating parameters of the dust collector, the operating parameters of the ash conveying equipment, the operating parameters of the fan and the exhaust condition parameters. The operating parameters of the dust generation point include air volume q1, air pressure p, temperature t1, humidity content wet and dust concentration m1. The operating parameters of the dust collector include dust collector resistance zu, ash amount c1 in the ash bucket, ash cleaning period T1, ash cleaning time T2 and ash cleaning intensity F. The operating parameters of the ash conveying equipment include ash conveying speed v and ash conveying amount c2. The operating parameters of the fan include the rotating speed r of the fan, the opening degree K2 of the fan inlet valve, the opening degree K3 of the fan outlet valve, temperature t2 and current I. The exhaust condition parameters include the exhaust air volume q2 of the dust collector and the dust concentration m2 in the gas exhausted from the dust collector.

[0101] The dust removal effect is affected by the temperature and humidity of the dust generation point. The dust collector resistance of the dust collector affects the dust removal effect and determines the operation of the ash cleaning and ash conveying equipment. The rotating speed of the fan can adjust the air volume and power output. The monitoring of the temperature t2 and current I of the fan can ensure that the fan works within a normal range. The collection of these data can comprehensively grasp the operation status of the entire dust removal system, so as to realize the monitoring of the corresponding equipment and environment, such as understanding the operation status of the dust collector and the ash conveying equipment, and controlling accordingly. Meanwhile, these data can also become historical operation data, serve as the training data set of the network model and become the basis of intelligent regulation and control.

[0102] The collected data is divided into operating variables and regulating variables. The operating variables include the air volume q1, air pressure p, temperature t1, humidity content wet, dust concentration m1, dust collector resistance zu, ash amount c1 in the ash bucket, temperature t2 of the fan, current I of the fan, exhaust air volume q2 of the dust collector and dust concentration m2. The regulating variables include the rotating speed r of the fan, valve opening degree K1, opening degree K2 of the fan inlet valve, opening degree K3 of the fan outlet valve, ash cleaning period T1, ash cleaning time T2, ash cleaning intensity F, ash conveying speed v and ash conveying amount c2. The operating variables can be processed through the function y=Y(q1, p, t1, wet, m1, zu, c1, t2, I, q2, m2). The regulating variables can be processed through the function t=T(r, K1, K2, K3, T1, T2, F, v, c2). The energy saving parameters can be processed through the function e=E(Y, T).

[0103] In the presence of multiple dust generation points, multiple dust collectors, multiple fans, and corresponding different collection times, arrays and matrices can be used to represent various variables. Assuming there are n dust generation points, s dust collectors, and u fans, and there are f collection time points, for operating variables, the air volume q1, air pressure p, temperature t1, humidity content wet, and dust concentration m1 at each dust generation point can be represented as a three-dimensional array, respectively q1[i][j][k], p[i][j][k], t1[i][j][k], wet[i][j][k], m1[i][j][k], where i (0≤i≤n) represents the dust generation point number, j (0≤j≤f) represents the collection time point number, and k represents the data dimension. The resistance zu of the dust collector, the amount of ash c1 in the ash bucket, the exhaust air volume q2 of the dust collector, and the dust concentration m2 are represented by a three-dimensional array, respectively zu[h][j][k], c1[h][j][k], q2[h][j][k], m2[h][j][k], where h (0≤h≤s) represents the dust collector number. The temperature t2 and current I of the fan are also represented by a three-dimensional array, respectively t2[g][j][k], I[g][j][k], and g (0≤g≤u) represents the fan number. The operating variable function can be represented as:

[0104] y[j]=Y(q1[j],p[j],t1[j],wet[j],m1[j],zu[j],c1[j],t2[j],I[j],q2[j],m2[j]), where j represents the collection time point.

[0105] For the adjusting variables, the speed r of the fan, the valve opening K1, the opening K2 of the fan inlet valve, the opening K3 of the fan outlet valve, the cleaning period T1, the cleaning time T2, the cleaning intensity F, the ash conveying speed v, and the ash conveying amount c2 can also be represented by arrays according to different fans, dust collectors, and collection times. If it is assumed that the adjusting variables of different devices can be independently controlled, the speed r of each fan can be represented as a three-dimensional array r[g][j][k], where g is the fan number, 0≤g≤u, j is the collection time point, and k is the data dimension; the valve opening K, the cleaning-related parameters, and the ash conveying-related parameters can also be similarly represented, respectively K1[g][j][k], K2[g][j][k], K3[g][j][k], T1[g][j][k], T2[g][j][k], F[g][j][k], v[g][j][k], c2[g][j][k], and g represents the fan number. The adjusting variable function can be represented as t[j]=T(r[j],K[j],T1[j],T2[j],F[j],v[j],c2[j]), where j represents the collection time point.

[0106] The energy saving parameter function can be expressed as e[j] = E(y[j], t[j]), that is, the corresponding energy saving parameter is calculated according to the operating variable and the adjusting variable at each collection time point. In this way, through the array and matrix form, the collected data and related variables under different working conditions and different times can be clearly managed and processed.

[0107] S2, a hybrid neural network model is constructed, wherein the hybrid neural network model is composed of an LSTM-GPT-Neo backbone network, a DenseNet branch network and a meta-learning feature fusion correction layer. Figure 5

[0108] The LSTM-GPT-Neo backbone network is located at the front end of the model and fuses LSTM and GPT-Neo architecture. The input is the operating variable data, and the task of LSTM is to extract high-level time sequence features from the operating variable data, which contains key information of the device running state, so that the subsequent GPT-Neo can perform more complex reasoning. LSTM adopts a four-layer stacked structure, each layer has 512 neurons, and combines the improved variant of the gated recurrent unit (GRU), which can better capture the complex time sequence features in the operating variable data. The GPT-Neo architecture introduces the idea of large-scale pre-training language model to understand and analyze the operating variable data at the semantic level, enhancing the ability to capture long-distance dependencies and complex patterns. The input of this backbone network is the operating variable data after deep preprocessing (including data cleaning, normalization, feature engineering, outlier detection and missing value interpolation, etc.), which covers the air volume q1, air pressure p, temperature t1, humidity content wet, dust concentration m1, dust point s, resistance zu of the dust collector, ash content c1 in the ash bucket, temperature t2 of the fan, current I of the fan, air volume q2 and dust concentration m2 of the dust collector, etc. After processing, mathematical transformation and feature combination are used in the hidden layer to calculate the high-order statistical features, nonlinear relationship features and comparison features with historical data of each parameter, and these features are spliced to form a new feature array. The final output layer calculates and outputs the optimal operating parameter distribution of the valve and the fan according to these feature arrays, and gives the possible optimal values of the valve opening, the speed, the air volume, the current, the power and other parameters of the fan in the form of probability distribution, and outputs the confidence interval and uncertainty estimation of the parameters.

[0109] ​The DenseNet branch network comprises a plurality of DenseBlock layers (dense blocks) and Transition layers (transition layers) between each two DenseBlock layers, each DenseBlock layer comprises a plurality of convolution layers and a skip connection, and each Transition layer is composed of a convolution layer and a pooling layer. The DenseNet branch network inputs the regulation variable data, is used for calculating and analyzing the input regulation variable data, and outputs the dedusting performance parameter, which is used for representing the mapping relationship between the regulation variable and the dedusting efficiency.

[0110] The input of the meta-learning feature fusion correction layer comes from the outputs of the LSTM-GPT-Neo main network and the DenseNet branch network, the meta-learning feature fusion correction layer comprises a meta-learner and a fusion module, adopts a meta-learning strategy, dynamically adjusts the fusion weight and the fusion mode according to the outputs of the LSTM-GPT-Neo main network and the DenseNet branch network according to different working conditions and historical data, so as to adapt to various complex conditions. The fused data is further processed by a deep neural network, and a prediction value is output as the output data of the whole hybrid neural network model, which provides more accurate, intelligent and forward-looking decision basis for subsequent fan operation regulation and optimization of the whole dedusting system, ensures that the system can realize efficient and energy-saving dedusting effect under various complex working conditions, and provides strong support for the explainability and generalization ability of the model.

[0111] Meta-learning can quickly adapt to new tasks by learning the commonness of multiple tasks or data distributions. The hybrid neural network model of the application adds a meta-learner, which realizes rapid learning by improving the optimization process itself, and can quickly adjust the fusion weight according to the working condition. Model-agnostic meta-learning (MAML) can be used to optimize the initial parameters, so that the model can adapt to new tasks with only a small amount of gradient updates. For new working conditions, a small amount of samples are used for fine-tuning (Few-Shot Adaptation), and the fusion strategy is dynamically adjusted to correct the output value. The MAML optimized meta-learner can quickly adapt to new working conditions. The meta-learner inputs the working condition features and outputs the initial fusion weight, and then updates the weight through gradient to obtain the adapted weight. The fusion module performs weighted average on the dynamic weight, and finally outputs a corrected prediction value as the output data of the whole hybrid neural network model.

[0112] The hybrid neural network model is responsible for capturing the time series characteristics and long-term dependencies of the parameters. The model is trained using historical data, using a supervised learning method, with the best operating parameters of the fan (such as speed and air output) as the target variable, and adjusting the model parameters through optimization algorithms (such as gradient descent) to enable the system to accurately regulate the operating parameters of the fan. The model first starts from data input, goes through multiple LSTM layers, then goes through a fully connected layer to get the final prediction result, then calculates the loss and updates the model parameters to form a complete training cycle to continuously improve the accuracy of the model in regulating the best operating parameters of the fan.

[0113] S3, collect time series data of operating variables and adjusting variables to form a data set, divide the data set into training set and validation set in time sequence, train the hybrid neural network model through the training set, and optimize the hybrid neural network model through the validation set.

[0114] S4, input the operating variable and adjusting variable data into the trained hybrid neural network model for calculation and analysis, output the valve opening and the best operating parameters of the fan, and real-time regulate the operation of the valve and the fan according to the parameters. The operating parameters of the fan mainly include the speed of the fan, the opening of the inlet valve and the opening of the outlet valve.

[0115] The automatic regulation of the operating parameters of the fan can match the current operation of the dust removal system, achieve energy saving effect, and the automatic control system can accurately adjust the speed and air output of the fan through the driving device connected with the fan, ensure that the fan can run efficiently under different working conditions, avoid excessive consumption of energy, and at the same time ensure the stability and reliability of the dust removal effect.

[0116] S5, through monitoring and controlling the air volume q1, air pressure p, dust concentration m1 at the dust production point, the energy consumption of the dust remover, ash conveying equipment and fan, and the exhaust condition parameters, evaluate the energy saving effect of the control, the stability of the system operation and the dust removal effect, and input the evaluation results into the hybrid neural network model for iterative optimization to continuously improve the control strategy of the hybrid neural network model.

[0117] In this embodiment, the LSTM unit adopts a four-layer LSTM stacking structure, each layer of LSTM includes 512 neurons; the DenseNet branch network includes four DenseBlock layers, each DenseBlock layer includes three convolution layers; the output dust removal performance vector can have five dimensions.

[0118] The method for calculating and analyzing by the LSTM unit is:

[0119] The LSTM backbone network comprises an input layer, a hidden layer and an output layer, the hidden layer comprises a plurality of LSTM layers, each LSTM layer is used to capture the time series characteristics of parameters, and residual connections are introduced between the LSTM layers, and a skip connection is added between two LSTM layers, the input of the previous LSTM layer is directly added to the output of the next LSTM layer element by element, and the output is:

[0120]

[0121] wherein, is the output hidden state after the residual connection and the skip connection operation, h t is the hidden state after processing by the current LSTM layer at time step t, h t-1 is the hidden state of the previous time step (time step t - 1).

[0122] The neurons in the LSTM layer are connected through weights, each neuron performs weighted summation on the input data and activation function conversion, and the output h j of the jth neuron is represented as: wherein w ij is the weight of the ith input parameter to the jth neuron, b j is the bias of the jth neuron, and f is the activation function ReLU, represented as: f(x) = max(0, x).

[0123] In the S3 step, the training method of the hybrid neural network model is:

[0124] (1) Collecting historical running sequence data, including running parameters at each dust generation point, valve opening K1 of each dust generation point and dust collector, running parameters of the dust collector, running parameters of the ash conveying equipment, running parameters of the fan and exhaust condition parameters.

[0125] (2) After removing abnormal running data (such as substandard dust removal effect, fan temperature or current exceeding the standard, dust collector running abnormally, etc.), the historical running data is evaluated, and the historical running data with excellent energy saving effect, system running stability and dust removal effect is selected. The method of data evaluation and selection can be: calculating the average energy consumption of the system in each time period, the change rate of air volume and pressure at the dust removal point, and the dust removal effect (dust removal efficiency), first sorting according to the dust removal effect, removing a certain percentage (such as 40%) of data at the back, then sorting the remaining data according to the average energy consumption, removing a certain percentage (such as 30%) of data at the back, then sorting the remaining data according to the change rate of air volume and pressure, removing a certain percentage (such as 20%) of data at the back, and obtaining the selected data.

[0126] (3) The selected data is used as the dataset, and randomly divided into training and validation sets according to a certain ratio; the data in the dataset is preprocessed and then input into the hybrid neural network model for training. Supervised learning method is adopted, and the learning rate is set to 0.001. A smaller learning rate can make the model converge more stably; the maximum number of pre-training iterations is 500; the maximum number of fine-tuning iterations is 200; the sparsity parameter is set to 0.4, which helps to reduce redundant information; the sparsity penalty term parameter is 2, which is used to control the penalty strength of sparsity; the activation function is ReLU (Rectified Linear Unit); the mean squared error (MSE) loss function is selected for fine-tuning. The valve opening K1 and the optimal operating parameters of the fan (such as speed and air volume output) are used as target variables, and the weights are updated by Adam optimization algorithm. The parameter update formula is:

[0127]

[0128] Where θ is the parameter to be optimized, and η is the learning rate. and These are the first and second moment estimates after bias correction, respectively. ∈ is a small value (typically 10) added to prevent division by zero. -8 ).

[0129] The data input into the model needs to be preprocessed. The data preprocessing method is as follows:

[0130] The data is cleaned to identify and remove obviously abnormal data, such as abnormally high or low values ​​caused by sensor malfunctions. Missing values ​​are imputed. If the proportion of missing data is small, it can be directly filled in. If the proportion of missing data is large, it is directly discarded.

[0131] Then, the data with different dimensions and ranges are normalized and converted into dimensionless numbers;

[0132] Then represent the dimensionless data as a vector X = [x1, x2, ..., x...]. n ], where x i Let be the i-th input parameter, and n be the total number of input parameters.

[0133] In step S5, the methods for inputting the evaluation results into the hybrid neural network model for iterative optimization include:

[0134] Firstly, the definitions of positive and negative samples need to be determined according to the actual business requirements and evaluation targets during the iterative optimization; after determining the positive and negative samples, the confusion matrix is constructed according to the evaluation data obtained by monitoring; based on the constructed confusion matrix, the recall rate and precision rate and other indicators are calculated, taking the energy saving effect as an example, the higher the recall rate, the stronger the model's ability to predict energy reduction, and more actual energy reduction cases can be captured; the calculated recall rate, precision rate and other indicators are input into the hybrid neural network model, and the current performance of the model is measured through the loss function, and then the regulation and control strategy is continuously improved, so that the intelligent regulation and control of the entire dust removal system is more accurate and efficient.

[0135] The hyperparameters of the model are adjusted, including learning rate, batch size, number of LSTM layers and number of neurons; if the data presents complex long-term dependence and the model cannot capture the features well, the number of LSTM layers can be increased, and if the model shows overfitting, i.e. the performance on the training set is good, but the performance on new data or test set is poor, and it is found through analysis that it is due to the complexity of the model, then the number of LSTM layers should be reduced.

[0136] The parameter combination that makes the model prediction performance optimal is found through grid search, random search or Bayesian optimization. Change the connection method between LSTM layers, such as skip connection, gate connection, etc.

[0137] In step S5, the effect monitoring module is used to monitor the air volume, air pressure and energy consumption and other parameters of the fan after adjustment in real time, and the system compares and analyzes the monitored parameters before and after adjustment, such as the change of air volume before and after adjustment, the difference of air pressure before and after adjustment, etc., calculates the energy consumption difference before and after adjustment of the fan, and evaluates the energy saving effect by comparing the electric energy consumption data before and after adjustment. At the same time, analyze the changes of air volume and air pressure, combine with the dust production data, evaluate the stability and dust removal effect of the dust removal system, such as the improvement of dust removal efficiency, the evaluation results can be used for model optimization.

[0138] The evaluation results include the following parameters:

[0139] Air volume tracking error rate ∈ q : calculated according to the following formula:

[0140]

[0141] Where q pred is the predicted target air volume, q post is the measured air volume after adjustment, and the air volume here refers to the air volume at the dust production point.

[0142] Air pressure stability index s p : calculated according to the following formula:

[0143]

[0144] wherein σ(p post ) is the standard deviation of the adjusted wind pressure, μ(p post ) is the mean value of the adjusted wind pressure, and the wind pressure here refers to the wind pressure at the dust generation point.

[0145] Dust concentration compliance rate R m : calculated according to the following formula:

[0146]

[0147] wherein T is the total number of sampling times in the statistical period, II() is an indicator function, m t is the dust concentration in the gas discharged from the dust collector at the tth sampling time.

[0148] Energy saving rate e: calculated according to the following formula:

[0149]

[0150] wherein E1 is the total energy consumption of the dust removal system before intelligent control, and E2 is the total energy consumption of the dust removal system before and after intelligent control.

[0151] Air volume change rate Q: calculated according to the following formula:

[0152]

[0153] wherein q1 is the air volume at the dust removal point before adjustment, and q2 is the air volume at the dust removal point after adjustment.

[0154] Wind pressure change rate P: calculated according to the following formula:

[0155]

[0156] wherein p1 is the wind pressure at the dust removal point before adjustment, and p2 is the wind pressure at the dust removal point after adjustment.

[0157] Dust removal efficiency η: calculated according to the following formula:

[0158]

[0159] wherein m in is the dust concentration entering the dust collector, and m out is the dust concentration discharged from the dust collector.

[0160] Dust removal efficiency change: calculated according to the following formula:

[0161]

[0162] wherein η1 is the dust removal efficiency before intelligent regulation, and η2 is the dust removal efficiency after intelligent regulation.

[0163] After one month of intelligent regulation of the dust removal system using the hybrid neural network model, significant results were achieved. In terms of energy consumption, the system as a whole saved 25% of energy consumption. Before regulation, the air volume fluctuated within ±10%, while after regulation, it stabilized within ±3%. The air pressure fluctuation decreased from ±15% before regulation to ±5% after regulation, greatly improving the stability of system operation. In terms of model evaluation parameters, the accuracy reached 88%, the loss value was 0.12, the recall rate was 85%, and the precision rate was 86%. The F1 score calculated based on the recall rate and the precision rate was 85.5%. These actual operation effect data and model evaluation parameters fully prove the effectiveness and superiority of the hybrid neural network model in intelligent regulation of the dust removal system.

[0164] The above detailed description is specific to the feasible embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent implementation or modification that does not deviate from the present application shall be included in the patent scope of the present application.

Claims

1. A method for intelligent regulation and control of a dust removal system in a steel enterprise, characterized in that, The method comprises the following steps: S1, collecting real-time data at a set frequency, and dividing the collected data into operating variables and adjusting variables; the collected data includes operating parameters at each dust generation point, valve opening degree K1 of the valve between each dust generation point and the dust collector, operating parameters of the dust collector, operating parameters of the ash conveying equipment, operating parameters of the fan, and exhaust condition parameters; S2, constructing a hybrid neural network model, which comprises an LSTM-GPT-Neo main network, a DenseNet branch network, and a meta-learning feature fusion correction layer; the LSTM-GPT-Neo main network is located at the front end of the model, is used for calculating and analyzing the input operating variable data, and outputs the valve opening degree K1 and the optimal operating parameter distribution of the fan, and integrates the LSTM and GPT-Neo architectures, wherein the LSTM is used for extracting high-level time sequence features from the input operating variable data, and the time sequence features contain equipment operating state information, and the GPT-Neo architecture introduces the idea of a large-scale pre-trained language model, understands and analyzes the operating variable data at a semantic level, and enhances the ability to capture long-distance dependencies and complex patterns; the DenseNet branch network is used for calculating and analyzing the input adjusting variable data, and outputs dust removal performance parameters, which are used to represent the mapping relationship between the adjusting variables and the dust removal efficiency; the meta-learning feature fusion correction layer adopts a meta-learning strategy, is used for outputting the LSTM-GPT-Neo main network and the DenseNet branch network, and dynamically adjusts the fusion weight and the fusion mode according to different working conditions and historical data to adapt to various complex and changeable situations; S3, collecting time sequence data of the operating variables and the adjusting variables to form a data set, dividing the data set into a training set and a validation set in chronological order, training the hybrid neural network model through the training set, and optimizing the hybrid neural network model through the validation set; S4, inputting the operating variable data and the adjusting variable data into the trained hybrid neural network model for calculation and analysis, outputting the valve opening degree K1 and the optimal operating parameters of the fan, and real-time regulating and controlling the operation of the valve and the fan according to the parameters; the optimal operating parameters of the fan include the rotational speed r of the fan, the opening degree K2 of the air inlet valve, and the opening degree K3 of the air outlet valve.

2. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 1, characterized in that, The operating parameters of the dust generation point include air volume q1, air pressure p, temperature t1, humidity content wet, and dust concentration m1; the operating parameters of the dust collector include dust collector resistance zu, ash amount c1 in the ash bucket, ash cleaning period T1, ash cleaning time T2, and ash cleaning intensity F; the operating parameters of the ash conveying equipment include ash conveying speed v and ash conveying amount c2; the operating parameters of the fan include the rotational speed r of the fan, the opening degree K2 of the fan air inlet valve, the opening degree K3 of the fan air outlet valve, temperature t2, and current I; and the exhaust condition parameters include exhaust air volume q2 discharged from the dust collector and dust concentration m2 in the gas discharged from the dust collector. The operation variables include air volume q1, air pressure p, temperature t1, humidity content wet, dust concentration m1, dust generation point s, resistance zu of the dust collector, ash content c1 in the ash bucket, temperature t2 of the fan, current I of the fan, air volume q2 of the dust collector, and dust concentration m2; the adjustment variables include rotating speed r of the fan, valve opening K1, opening K2 of the fan inlet valve, opening K3 of the fan outlet valve, ash cleaning period T1, ash cleaning time T2, ash cleaning intensity F, ash conveying speed v, and ash conveying amount c2.

3. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 2, characterized in that, Supposing that there are n dust generation points, s dust collectors, and u fans in the dust removal system, and there are f collection time points, the processing and representation method of the operation variable data is as follows: The operation variables of the dust generation points, including air volume q1, air pressure p, temperature t1, humidity content wet, and dust concentration m1, are processed by the function y=Y(q1, p, t1, wet, m1, zu, c1, t2, I, q2, m2), and can be represented by three-dimensional arrays, q1[i][j][k], p[i][j][k], t1[i][j][k], wet[i][j][k], and m1[i][j][k], wherein i represents the dust generation point number, 0≤i≤n, j represents the collection time point number, 0≤j≤f, and k represents the data dimension; The resistance zu of the dust collector, the ash content c1 in the ash bucket, the exhaust air volume q2 of the dust collector, and the dust concentration m2 are represented by three-dimensional arrays, zu[h][j][k], c1[h][j][k], q2[h][j][k], and m2[h][j][k], wherein h represents the dust collector number, 0≤h≤s, j represents the collection time point number, 0≤j≤f, and k represents the data dimension; The temperature t2 and the current I of the fan are represented by three-dimensional arrays, t2[g][j][k] and I[g][j][k], wherein g represents the fan number, 0≤g≤u, j represents the collection time point number, 0≤j≤f, and k represents the data dimension; The processing and representation method of the adjustment variable data is as follows: The adjustment variables are processed by the function t=T(r, K1, K2, K3, T1, T2, F, v, c2), and the rotating speed r, valve opening K1, opening K2 of the fan inlet valve, opening K3 of the fan outlet valve, ash cleaning period T1, ash cleaning time T2, ash cleaning intensity F, ash conveying speed v, and ash conveying amount c2 of each fan can be represented by three-dimensional arrays, r[g][j][k], K1[g][j][k], K2[g][j][k], K3[g][j][k], T1[g][j][k], T2[g][j][k], F[g][j][k], v[g][j][k], and c2[g][j][k], wherein g is the fan number, 0≤g≤u, for the ash conveying equipment, g is the fan number corresponding to the ash conveying equipment, j is the collection time point, 0≤j≤f, and k is the data dimension.

4. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 1, characterized in that, The LSTM-GPT-Neo trunk network is located at the front end of the model, and outputs data in the form of a probability distribution representing possible optimal values of the valve opening K1, the rotational speed r of the fan, the opening K2 of the inlet valve of the fan, and the opening K3 of the outlet valve, as well as a confidence interval and an uncertainty estimate; The LSTM unit adopts a four-LSTM-layer stacking structure, each LSTM layer includes 512 neurons, and combines an improved variant of the gated recurrent unit GRU to capture complex time series features in the operating variable data; The calculation and analysis method of the LSTM unit is: Each LSTM layer is used to capture the time series features of the parameters, and residual connections are introduced between the LSTM layers, and a skip connection is added between two LSTM layers to directly add the input of the previous LSTM layer to the output of the next LSTM layer element by element, and the output is: wherein, is the output hidden state after residual connection and skip connection operations, h t is the hidden state after processing by the current LSTM layer at time step t, h t-1 is the hidden state of the previous time step t-1; The neurons in the LSTM layer are connected by weights, each neuron performs a weighted sum of the input data and an activation function transformation, the output of the jth neuron h j is represented as: where w ij is the weight of the ith input parameter to the jth neuron, b j is the bias of the jth neuron, and f is the activation function ReLU, represented as: f(x) = max(0, x); The DenseNet branch network includes four DenseBlock layers and a Transition layer between each two DenseBlock layers, each DenseBlock layer includes three convolutional layers and a skip connection, and each Transition layer is composed of a convolutional layer and a pooling layer; The meta-learning feature fusion correction layer includes a meta-learner MAML and a fusion module, the MAML obtains adaptive weights through gradient update according to the working condition features, and the fusion module performs weighted average on the dynamic weights.

5. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 1, characterized in that, In the S3 step, the training method of the hybrid neural network model is: S3-1, collect historical operation sequence data, including operating parameters at each dust generation point, valve opening K1 of each valve between the dust generation point and the dust collector, operating parameters of the dust collector, operating parameters of the ash conveying equipment, operating parameters of the fan, and exhaust condition parameters; S3-2, after removing the abnormal operation data, evaluate the historical operation data, and select the historical operation data with excellent energy saving effect, system operation stability and dust removal effect; S3-3, the selected data is used as a data set, and is randomly divided into a training set and a validation set according to time training at a certain proportion; the data in the data set is preprocessed and input into the hybrid neural network model for training, a supervised learning method is used, the valve opening K1 and the optimal operating parameters of the fan are used as target variables, and the weights are updated through the Adam optimization algorithm, and the parameter update formula is: where θ is the parameter to be optimized, η is the learning rate, and are the bias-corrected first and second moment estimates, respectively, and ∈ is a very small number added to prevent division by zero.

6. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 5, characterized in that, In the S3-2 step, the data evaluation and selection method is: the average energy consumption of the system, the change rate of the air volume and air pressure at the dust removal point, and the dust removal effect in each time period are calculated, the data is first sorted according to the dust removal effect, then the data ranked after 40% is removed, then the remaining data is sorted according to the average energy consumption, then the data ranked after 30% is removed, then the remaining data is sorted according to the change rate of the air volume and air pressure, and then the data ranked after 20% is removed to obtain the selected data; In the S3-3 step, the parameter setting of the model training is: the learning rate is set to 0.001, the maximum number of pre-training iterations is 500, and the maximum number of fine-tuning iterations is 200; The sparse parameter is set to 0.4, which helps to reduce redundant information; the sparse penalty term parameter is 2, which is used to control the penalty degree of sparsity; the activation function adopts the ReLU function; and the fine-tuning loss function selects the mean square error loss function MSE.

7. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 1, characterized in that, Also includes: S5, by monitoring the wind volume q1, the wind pressure p, the dust concentration m1 at the dust generation point after regulation and control, the energy consumption of the dust remover, the ash conveying equipment and the fan, and the exhaust condition parameters, evaluating the energy saving effect of the regulation and control, the stability of the system operation and the dust removal effect, and inputting the evaluation results into the hybrid neural network model for iterative optimization, so as to continuously improve the regulation and control strategy of the hybrid neural network model; The method for inputting the evaluation results into the hybrid neural network model for iterative optimization comprises: S5-1, defining the definitions of positive samples and negative samples according to actual business requirements and evaluation targets; S5-2, after determining the positive and negative samples, constructing a confusion matrix according to the evaluation data obtained by monitoring; S5-3, based on the constructed confusion matrix, calculating the evaluation results of the hybrid neural network model, and then inputting the evaluation results into the hybrid neural network model, measuring the current performance of the model through the loss function, and continuously improving the regulation and control strategy.

8. The intelligent regulation and control method for a dust removal system in a steel enterprise according to claim 7, characterized in that, The evaluation results include the following parameters: Air volume tracking error rate ∈ q : is calculated according to the following formula: wherein q pred is the target air quantity at the predicted dusting point, q post is the measured air quantity at the dusting point after intelligent regulation. Wind pressure stability index s p : is calculated according to the following formula: wherein σ(p post ) is the standard deviation of the wind pressure at the dust generation point after intelligent regulation, and μ(p post ) is the mean value of the wind pressure at the dust generation point after intelligent regulation. Dust concentration compliance rate R m : is calculated according to the following formula: where T is the total number of samples in the statistical period, II() is the indicator function, m t is the dust concentration in the gas exiting the dust collector at the tth sample. Energy saving rate e: calculated according to the following formula: Wherein, E1 is the total energy consumption of the dust removal system before intelligent regulation and control, and E2 is the total energy consumption of the dust removal system before and after intelligent regulation and control; Air volume change rate Q: calculated according to the following formula: Wherein, q1 is the air volume at the dust removal point before intelligent regulation and control, and q2 is the air volume at the dust removal point after intelligent regulation and control; Air pressure change rate P: calculated according to the following formula: Wherein, p1 is the air pressure at the dust removal point before intelligent regulation and control, and p2 is the air pressure at the dust removal point after intelligent regulation and control; Dust removal efficiency η: calculated according to the following formula: wherein m in is the dust concentration entering the dust collector, m out is the dust concentration exiting the dust collector; Dust removal efficiency change: calculated according to the following formula: Wherein, η1 is the dust removal efficiency before intelligent regulation and control, and η2 is the dust removal efficiency after intelligent regulation and control.

9. An intelligent regulation system for a dust removal system of a steel enterprise, for implementing the intelligent regulation method for the dust removal system of the steel enterprise according to any one of claims 1 to 8, characterized in that, It comprises dust removal equipment, a data acquisition module, an intelligent regulation and control module, an operation regulation and control module and a monitoring and optimization module. The dust removal equipment comprises a hood, a conveying pipeline, a valve, a dust remover, an ash conveying device, a fan and a chimney, a plurality of hoods are arranged at dust generation points, the hood is connected with the dust remover in communication through the conveying pipeline, the valve is arranged on the conveying pipeline, the dust remover is arranged at the rear end of the hood, the dust remover is provided with a dust cleaning device, the dust cleaning device is used for cleaning the dust accumulated in the dust remover, the ash conveying device is arranged at the outlet of the dust remover hopper and is used for conveying the accumulated dust, the air inlet of the fan is connected with the dust remover, and the air outlet of the fan is connected with the chimney. The data acquisition module comprises a plurality of sensors and a data collector, the sensors are respectively arranged at the positions of the conveying pipelines, valves and dust collectors connected with the hoods at each dust generation point, the data collector is connected with each of the sensors, valves, dust collectors, ash conveying devices and fans, and is used for collecting the operating parameters at each dust generation point, the valve opening of the valve, the operating parameters of the dust collector, the operating parameters of the ash conveying device and the operating parameters of the fan; the data acquisition module is connected with the intelligent control module and the monitoring optimization module to transmit the collected data; The intelligent control module is embedded with a hybrid neural network model, which is used for calculating and analyzing the collected data to obtain the optimal operating parameters of the valve opening K1 and the fan; The operating control module is connected with the intelligent control module to receive the optimal operating parameters of the fan obtained by the intelligent control module, and is connected with the fan to adjust the rotating speed of the fan, the opening of the air inlet valve and the opening of the air outlet valve according to the optimal operating parameters of the fan obtained by the intelligent control module, so as to control the size of the air volume of the fan; The monitoring optimization module comprises a data analysis unit, a closed-loop feedback unit and a data storage unit, the data analysis unit is used for calculating and analyzing the collected data to evaluate the energy saving effect of the control, the stability of the system operation and the dust removal effect, the closed-loop feedback unit is used for transmitting the evaluation results to the intelligent control module to iteratively optimize the hybrid neural network model to form a dynamic optimization closed loop, and the data storage unit is used for storing the collected data and the data obtained by calculation and analysis. 10.The intelligent regulation and control system of a steel enterprise dust removal system according to claim 9, characterized in that, The sensors comprise air volume measuring instruments, air pressure gauges, dust concentration sensors, temperature and humidity meters, pressure sensors and level meters, the air volume measuring instruments are arranged at the conveying pipelines and dust collectors at each dust generation point to measure the air volume q1 at the dust generation point and the exhaust air volume q2 discharged from the dust collector, the air pressure gauges are arranged at the conveying pipelines at the dust generation points to measure the air pressure p at the dust generation points, the dust concentration sensors are arranged at the conveying pipelines and dust collectors at the dust generation points to measure the dust concentration m1 at the dust generation points and the dust concentration m2 discharged from the dust collectors, the temperature and humidity meters are arranged at the conveying pipelines at the dust generation points to measure the temperature t1 and the moisture content wet at the dust generation points, and the pressure sensors and level meters are arranged in the dust collectors to measure the resistance zu of the dust collector and the amount c1 of ash in the ash hopper; The dust generation points are coke oven charging and coke guiding and discharging sites, secondary dust removal sites of converters and blast furnace tapping sites, and the hoods are arranged at the positions above the coke oven furnace heads, coke blocking machines, coke pushing cars, converter furnace mouths and blast furnace tapping holes.

Images