Hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method

Abstract

The invention discloses a hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method, which relates to the technical field of automatic control, and comprises the following steps: collecting monitoring data, and preprocessing; constructing an intelligent diagnosis monitoring transformer model, and generating a fault diagnosis result; according to a fault diagnosis result, operation parameters are automatically adjusted through multivariable sliding mode control, and the stable operation state of the blast furnace is achieved; after it is determined that the operation state of the blast furnace is stable, blast furnace gas is collected, dynamic gas separation and compression storage are conducted, residual gas is purified, and waste heat is recycled; by constructing the intelligent diagnosis monitoring transformer model, real-time monitoring and accurate prediction of the operation state of the blast furnace are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control, and particularly to a control method for recovering blast furnace gas in a hydrogen-rich carbon cycle oxygen blast furnace. Background Art

[0002] In the process of blast furnace ironmaking, the hydrogen-rich carbon cycle oxygen technology has become one of the key methods to improve production efficiency and energy utilization rate. This technology injects gases rich in hydrogen and carbon into the blast furnace in a cycle to support combustion reactions and improve reduction efficiency. However, there are still many deficiencies in the existing technology in terms of gas recovery and utilization. Traditional methods often cannot fully separate and recover hydrogen and carbon dioxide in the gas, resulting in energy waste and environmental pollution. In addition, the accuracy of existing blast furnace monitoring and fault diagnosis is limited, making it difficult to achieve real-time and accurate state prediction and fault warning.

[0003] The deficiencies of the existing technology are mainly reflected in the following aspects: Firstly, the separation efficiency of hydrogen and carbon dioxide in the gas is low, and the recovery and utilization rate is not high, which limits the effective utilization of resources. Secondly, traditional monitoring has lags in data collection and processing, making it difficult to achieve real-time monitoring and fault diagnosis of the blast furnace operation status. In addition, existing fault diagnosis methods mostly rely on fixed thresholds and empirical judgments, lacking flexibility and self-adaptability, and it is difficult to cope with the complex and changeable blast furnace operation environment. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a control method for recovering blast furnace gas in a hydrogen-rich carbon cycle oxygen blast furnace to solve the problems of insufficient blast furnace operation efficiency and energy utilization rate.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a control method for recovering blast furnace gas in a hydrogen-rich carbon cycle oxygen blast furnace, which includes collecting monitoring data and performing preprocessing; constructing an intelligent diagnostic monitoring transformer model to generate a fault diagnosis result; automatically adjusting operation parameters through multivariable sliding mode control according to the fault diagnosis result to achieve stable blast furnace operation status; after determining that the blast furnace operation status is stable, collecting blast furnace gas and performing dynamic gas separation, compression storage, purifying the remaining gas, and recovering waste heat.

[0008] As a preferred scheme of the control method for recovering blast furnace gas in a hydrogen-rich carbon cycle oxygen blast furnace according to the present invention, wherein: the monitoring data includes temperature data, pressure data, oxygen data, and hydrogen concentration data.

[0009] As a preferred embodiment of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method of the present invention, wherein: the preprocessing includes noise removal, data normalization, outlier processing, and feature extraction through principal component analysis.

[0010] As a preferred embodiment of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method of the present invention, wherein: the steps for constructing the intelligent diagnostic monitoring transformer model are as follows.

[0011] Define the input as the preprocessed monitoring data and the output as the fault diagnosis result.

[0012] Divide the monitoring data into a training set, a validation set, and a test set.

[0013] Encode the preprocessed monitoring data through an encoder, extract features, and introduce a multi-head self-attention mechanism to generate enhanced features.

[0014] Use a decoder to process the encoded enhanced features to generate a predicted value of the blast furnace operating state, with the expression:

[0015]

[0016] Wherein, is the predicted value of the blast furnace operating state at the next time, is the predicted value of the blast furnace operating state of the actual measured value y, a i is the historical data coefficient, y(t - i + 1) is the actual operating result at the previous i time, y is the actual measured value, b j is the control input coefficient, u(t - j) is the control input at the previous j time, u is the actual value of the control input, t is the time index, n is the time step of the historical data, m is the time step of the control input, i is the index of the historical data, and j is the index of the control input.

[0017] Generate a fault diagnosis result based on the predicted value of the blast furnace operating state.

[0018] Use the training set to train the intelligent diagnostic monitoring transformer model and adjust the parameters through the backpropagation algorithm and the adaptive optimization algorithm.

[0019] Evaluate the performance of the intelligent diagnostic monitoring transformer model through the validation set.

[0020] Use the test set to test the intelligent diagnostic monitoring transformer model.

[0021] As a preferred embodiment of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method of the present invention, wherein: the steps for encoding the preprocessed monitoring data, extracting features, and introducing a multi-head self-attention mechanism to generate enhanced features are as follows.

[0022] The encoder extracts the time series data from the monitored data and converts the time series data into time series feature vectors;

[0023] Perform positional encoding processing on the time series feature vectors;

[0024] In the multi-head attention mechanism, the time series feature vectors are converted into queries, keys, and values through linear transformation;

[0025] Calculate the similarity between the query and the key through dot product to obtain the attention weights;

[0026] Use multiple attention heads to calculate in parallel, splice the results of each attention head, and obtain the final time series feature vectors through linear transformation;

[0027] After completing the multi-head self-attention mechanism processing, the encoder converts the time series feature vectors into enhanced features.

[0028] As a preferred solution of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method described in the present invention, wherein: the generation of the fault diagnosis result specifically includes the following steps:

[0029] Set the normal and fault operation state thresholds of the blast furnace;

[0030] According to the predicted value of the blast furnace operation state, predict the expected state of each monitored parameter of the blast furnace at the next time;

[0031] If the predicted value of the blast furnace operation state is within the normal range of the operation state threshold, it is in a normal state;

[0032] If the predicted value of the blast furnace operation state exceeds the normal range of the operation state threshold, there is an abnormality;

[0033] According to the predicted value of the blast furnace operation state that exceeds the normal range, judge the severity of the fault, and compare the abnormal value predicted by the blast furnace operation state with the historical fault data to find the cause of the fault and generate the fault diagnosis result.

[0034] As a preferred solution of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method described in the present invention, wherein: the automatic adjustment of the operation parameters through multivariable sliding mode control specifically includes the following steps:

[0035] Define the multivariable sliding mode control input according to the sliding mode control theory, and the expression is:

[0036] u(t) = -k × sgn(s(z));

[0037] Among them, u(t) is the control input at the current time t, u is the actual value of the control input, sgn is the sign function, k is a positive constant, s(z) is the multivariable sliding mode surface function, and z is the vector of state variables;

[0038] According to the multivariable sliding mode surface function, select the optimal combination of state variables, automatically adjust the operation parameters of the oxygen injection amount, hydrogen injection amount, and the internal pressure of the blast furnace, and achieve the stable operation state of the blast furnace. The expression is:

[0039]

[0040] Among them, z 1 is the current temperature deviation, z 2 is the change rate of the state deviation, z 3 is the cumulative state deviation, c is the constant coefficient for adjusting the change rate z 2 of the state deviation, g is the constant coefficient for adjusting the cumulative state deviation z 3 and ∫ is the integral symbol, and et is the integral variable.

[0041] As a preferred solution of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method described in the present invention, among them: after determining that the operation state of the blast furnace is stable, collect the blast furnace gas and perform dynamic gas separation and compression storage, purify the remaining gas, and recover the waste heat. The specific steps are as follows:

[0042] Use the gas collection equipment installed at the top of the blast furnace to collect the blast furnace gas and transfer it to the central collector at the top of the blast furnace;

[0043] The central collector transports the blast furnace gas to the dynamic gas separation device through a pipeline;

[0044] In the dynamic gas separation device, separate hydrogen and carbon dioxide from the blast furnace gas;

[0045] Transport the separated hydrogen to the hydrogen recycling equipment and store it through a compressor and a gas storage tank;

[0046] Compress the separated carbon dioxide through a compressor and convert the carbon dioxide into a high-pressure state and store it in a high-pressure storage tank;

[0047] After separating hydrogen and carbon dioxide, the remaining gas enters the purification equipment, and the impurities and particulate matter in the remaining gas are removed by using a multi-stage filtration and washing method;

[0048] The purified remaining gas recovers waste heat through a heat exchanger to complete energy utilization.

[0049] In a second aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method described in the first aspect of the present invention is implemented.

[0050] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method described in the first aspect of the present invention is implemented.

[0051] The beneficial effects of the present invention are as follows: by constructing an intelligent diagnostic monitoring transformer model, the present invention realizes the real-time monitoring and accurate prediction of the blast furnace operation state. The intelligent diagnostic monitoring transformer model uses the multi-head self-attention mechanism to encode and extract features from the preprocessed monitoring data, generates the predicted value of the blast furnace operation state, improves the accuracy of the prediction result, and generates a fault diagnosis result by setting a threshold, realizing the early warning and analysis of potential faults in the blast furnace. Finally, the intelligent diagnostic monitoring transformer model effectively improves the safety and stability of the blast furnace operation, reduces the unplanned shutdown and maintenance costs, and improves the production efficiency and resource utilization rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0053] Figure 1 It is a flowchart of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method in Embodiment 1.

[0054] Figure 2 It is a schematic diagram of generating a fault diagnosis result in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification.

[0056] Many specific details are set forth in the following description in order to fully understand the present invention, but the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0057] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive with other embodiments.

[0058] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a control method for hydrogen-rich carbon cycle oxygen blast furnace gas recovery, including the following steps:

[0059] S1. Collect monitoring data and perform preprocessing.

[0060] Furthermore, the monitoring data includes temperature data, pressure data, oxygen data, and hydrogen concentration data;

[0061] Specifically, start the oxygen injection equipment and inject oxygen into the lower part of the blast furnace to support the combustion reaction;

[0062] Install temperature sensors, pressure sensors, gas analysis sensors, etc. at key positions in the blast furnace to ensure that these sensors can collect data accurately and in real time;

[0063] Use sensors to monitor the temperature, pressure, and gas composition in the blast furnace in real time;

[0064] The sensors continuously collect various data, including the concentrations of temperature, pressure, oxygen, and hydrogen;

[0065] Transmit the collected monitoring data to the central control platform through a wired or wireless network to ensure the integrity and real-time nature of the operation data of the blast furnace.

[0066] S1.1. The preprocessing includes removing noise, data normalization, outlier processing, and feature extraction through principal component analysis;

[0067] Furthermore, after the data reaches the central control platform, perform preprocessing on it;

[0068] Use methods such as low-pass filtering, high-pass filtering, or band-pass filtering to remove noise in specific frequency bands;

[0069] For example, the moving average method, by smoothing the data, reduces the volatility;

[0070] Convert data of different magnitudes to the same range through maximum and minimum normalization;

[0071] By calculating the interquartile range of the data, it is determined whether the data is an outlier. Generally, the part of the data points below the first quartile (Q1) minus 1.5 times the IQR or above the third quartile (Q3) plus 1.5 times the IQR is considered an outlier;

[0072] Principal component analysis is used for feature extraction. Specifically,

[0073] The data is standardized so that its mean is 0 and the standard deviation is 1;

[0074] The covariance matrix of the data set is calculated to understand the linear relationship between variables;

[0075] The covariance matrix is eigen-decomposed to obtain eigenvalues and eigenvectors;

[0076] According to the magnitudes of the eigenvalues, the main eigenvectors are selected as the principal components, and the data is projected onto these principal components;

[0077] It should be noted that by starting the oxygen injection equipment and installing sensors, the uniform distribution of oxygen in the blast furnace is ensured and the temperature, pressure, and gas composition are monitored in real time. The data collected by the sensors is transmitted to the central control platform through the network for preprocessing such as noise removal, data normalization, and outlier processing, improving the accuracy and reliability of the data. The preprocessed data is used to extract features through principal component analysis, reduce the dimension, enhance the data representativeness, and provide high-quality basic data for subsequent prediction of the blast furnace operation status and fault diagnosis.

[0078] S2. Build an intelligent diagnostic monitoring transformer model to generate fault diagnosis results.

[0079] Furthermore, the input is defined as the preprocessed monitoring data, and the output is the predicted value of the blast furnace operation status and the fault diagnosis result;

[0080] The monitoring data is divided into a training set, a validation set, and a test set;

[0081] An encoder and a decoder are selected as the intelligent diagnostic monitoring transformer model architecture;

[0082] The preprocessed monitoring data is encoded by the encoder to extract features, and a multi-head self-attention mechanism is introduced to generate enhanced features;

[0083] Specifically,

[0084] The encoder extracts the time series data in the monitoring data and converts the time series data into time series feature vectors;

[0085] The encoder extracts time series data (such as temperature, pressure, oxygen, and hydrogen concentration) from the preprocessed monitoring data and converts this time series data into time series feature vectors; the time series feature vectors contain the change information of the monitoring data over time and are the basis for subsequent processing;

[0086] It should be noted that the entire monitoring data set is divided into multiple small segments in chronological order. For example, if the monitoring data is recorded once per second, it can be segmented according to different time periods such as minutes, hours, or days. These small segments of data can be regarded as a segment of the time series;

[0087] Using the sliding window technique, a time window with a fixed length is set and the window is slid over the monitoring data. The data segment within each time window is extracted as a time series segment;

[0088] The sliding windows can overlap or not overlap, depending on the specific application requirements;

[0089] For example, setting a 5-minute window that moves once per minute can generate a series of time series segments;

[0090] Calculate the basic statistical features for the data within each time window, such as mean, variance, maximum value, minimum value, etc. These statistical features can be used as components of the feature vector;

[0091] Calculate time domain features such as trends, seasonality, and periodicity of the time series data. For example, extract the trend of the time series through the moving average method or the exponential smoothing method;

[0092] Perform frequency domain analysis on the time series data to extract frequency domain features. Common methods include Fourier transform (FFT) and wavelet transform, etc. Through these transforms, the frequency components of the data can be extracted;

[0093] Calculate the autocorrelation coefficient and partial autocorrelation coefficient of the time series data. The autocorrelation coefficient measures the correlation of a time series at different time lags, while the partial autocorrelation coefficient measures the correlation of the time series after excluding the influence of intermediate variables;

[0094] Perform positional encoding processing on the time series feature vectors to retain the order information in the time series feature vectors; add positional encoding to the time series feature vectors. The positional encoding is generated through a specific mathematical function and added to the monitoring data, enabling the intelligent diagnostic monitoring transformer model to perceive the sequential relationship of time steps and thus preserve the time information of the time series data;

[0095] In the multi-head attention mechanism, the time series feature vectors are converted into queries, keys, and values through linear transformation; expressed as:

[0096] Q = x(t)W Q , K = x(t)W K , V = x(t)W V ;

[0097] Among them, Q is the query vector, which is obtained by the linear transformation of the input time - series feature vector x(t) through the weight matrix W Q . x(t) is the input time - series feature vector, and these time - series feature vectors contain the characteristic information of time - series data (such as temperature, pressure, oxygen concentration, and hydrogen concentration). t represents time. K is the key vector, which is obtained by the linear transformation of the input time - series feature vector x(t) through the weight matrix W K . V is the value vector, which is obtained by the linear transformation of the input time - series feature vector x(t) through the weight matrix W V . W Q is the query weight matrix, W K is the key weight matrix, W V is the value weight matrix, and W is the weight matrix;

[0098] Through linear transformation, the time - series feature vector is converted into query, key, and value, which helps to establish connections between different features in the intelligent diagnosis and monitoring transformer model;

[0099] By calculating the similarity between the query and the key through the dot - product, the attention weight is obtained; the expression is:

[0100]

[0101] Among them, Attention(Q l , K l , V l ) is the attention weight calculated by the l - th attention head. Q l is the l - th query vector, which represents the result after the linear transformation of the time - series feature vector at the current time step. K l is the l - th key vector, which represents the result after the linear transformation of the time - series feature vectors of all time steps. It performs a dot - product calculation with the query matrix. V l is the l - th value vector, which represents the result after the linear transformation of the time - series feature vectors of all time steps. It obtains the final attention output through weighted summation. Softmax is a normalization function used to convert the result of the dot - product calculation into an attention weight. Through the Softmax function, the sum of all weights is 1, making the result easy to interpret and use. is the l - th query vector Q l and the transpose l of the key vector K of the dot - product, and this dot - product calculates the similarity between the query and the key. is the l-th key vector K l in its transposed form, where T represents transpose, is the scaling factor used to avoid overly large dot product results, and d k represents the dimension of the key matrix. By dividing by it can prevent the vanishing gradient problem caused by overly large results, and V l is the multiplication of the attention weights after Softmax processing and the l-th value vector to obtain the final weighted time series feature vector, where l is the index of the attention head;

[0102] The attention weights help the intelligent diagnostic monitoring transformer model identify which features are more important at the current moment;

[0103] Using multiple attention heads to calculate in parallel, concatenating the results of each attention head, and obtaining the final time series feature vector through a linear transformation; this can capture various complex feature relationships in the input data and improve the effect of feature extraction; the expression is:

[0104] MultiHead(Q, K, V) = Concat(head 1 , …, head L )W O ;

[0105] where, MultiHead(Q, K, V) represents the parallel calculation of multiple attention heads, each head has its own query, key, and value, and these heads independently calculate the attention scores. Concat is the concatenation function that concatenates the results of all attention heads. (head 1 , …, head L ) is the calculation result of each attention head. L is the number of attention heads used in the multi-head self-attention mechanism, and W O is the weight matrix used to perform a linear transformation on the concatenated result;

[0106] head l = Attention(Q l , K l , V l );

[0107] where, head l is the calculation result of the l-th attention head, and each attention head calculates independently;

[0108] After the multi-head self-attention mechanism processing is completed, the encoder converts the time series feature vectors into enhanced features for representation. The time series feature vectors after the multi-head self-attention mechanism processing contain the complex relationships between various features in the original monitoring data. The encoder converts these time series feature vectors into enhanced feature representations, enabling the intelligent diagnostic monitoring transformer model to better capture the changes in the blast furnace operating state;

[0109] Use the decoder to process the encoded enhanced features to generate the predicted values of the blast furnace operating state. The expression is:

[0110]

[0111] where, is the predicted value of the blast furnace operating state at the next time, is the predicted value of the blast furnace operating state of the actual measured value y, a i is the historical data coefficient, used to weight the historical operation results. y(t - i + 1) is the actual operation result at the previous i time. y is the actual measured value, used to provide historical data. b j is the control input coefficient, used to weight the influence of the control input. u(t - j) is the control input at the previous j time. u is the actual value of the control input, such as for adjusting the operation parameters of the blast furnace, such as oxygen injection volume, hydrogen injection volume, etc. t is the index of time, n is the time step of the historical data, m is the time step of the control input, i is the index of the historical data, from 1 to n, and j is the index of the control input, from 0 to m;

[0112] This formula is a dynamic prediction, used to predict the operating state of the blast furnace at the next moment. Compared with the basic linear regression model, it introduces the historical data and the influence of the control input in the time series, and adapts to the historical outputs and control inputs at different time steps by adjusting the historical data coefficient and the control input coefficient, making the prediction more accurate;

[0113] Generate the fault diagnosis result according to the predicted value of the blast furnace operating state;

[0114] Specifically,

[0115] Set the thresholds for the normal and faulty operating states of the blast furnace;

[0116] The specific operating state thresholds are defined as follows:

[0117] Temperature value: Normal: 1200°C ≤ R ≤ 1600°C, Fault: R < 1200°C or R > 1600°C;

[0118] Pressure value: Normal: 1.5 MPa ≤ P ≤ 2.5 MPa, Fault: P < 1.5 MPa or P > 2.5 MPa;

[0119] Oxygen concentration value: Normal: 20% ≤ F ≤ 25%, Fault: F < 20% or F > 25%;

[0120] Hydrogen concentration value: Normal: 5% ≤ H ≤ 10%, Fault: H < 5% or H > 10%;

[0121] Based on the predicted values of the blast furnace operating state, predict the expected states of various monitoring parameters of the blast furnace at the next time.

[0122] If the predicted value of the blast furnace operating state is within the normal range of the operating state threshold, then this parameter is in a normal state.

[0123] If the predicted value of the blast furnace operating state exceeds the normal range of the operating state threshold, then there is an abnormality in this parameter and further analysis is required.

[0124] Based on the predicted value of the blast furnace operating state that exceeds the normal range, judge the severity of the fault (for example, if the predicted temperature value exceeds the normal range but not by much, it is a minor fault; if the predicted temperature value exceeds the range by a large amount, it is a serious fault), and compare the abnormal value of the predicted blast furnace operating state with the historical fault data to find the cause of the fault and generate a fault diagnosis result.

[0125] Use the training set to train the intelligent diagnostic monitoring transformer model, and adjust the parameters through the backpropagation algorithm and the adaptive optimization algorithm.

[0126] Use the training set to train the intelligent diagnostic monitoring transformer model, apply the backpropagation algorithm and the adaptive optimization algorithm (such as AdamW) to adjust the parameters of the intelligent diagnostic monitoring transformer model, and minimize the prediction error.

[0127] Evaluate the performance of the intelligent diagnostic monitoring transformer model through the validation set to prevent overfitting, and adjust the parameters of the intelligent diagnostic monitoring transformer model according to the performance of the validation set for further optimization.

[0128] Use the test set to test the intelligent diagnostic monitoring transformer model, and the evaluation metrics include prediction accuracy, recall rate, and F1 score to comprehensively evaluate the prediction accuracy and stability of the intelligent diagnostic monitoring transformer model.

[0129] It should be noted that by constructing the intelligent diagnostic monitoring transformer model, it is possible to analyze the preprocessed monitoring data in real time, generate the predicted values of the blast furnace operating state and the fault diagnosis results. This model adopts the multi-head self-attention mechanism to extract and enhance features, improve the prediction accuracy. By setting the operating state threshold, the model can identify normal and fault states, provide early fault warnings and detailed diagnoses. Finally, it effectively improves the safety and stability of the blast furnace operation, reduces the unplanned downtime and maintenance costs, and significantly improves the production efficiency and resource utilization rate.

[0130] S3. According to the fault diagnosis results, automatically adjust the operating parameters through multivariable sliding mode control to achieve stable blast furnace operation status.

[0131] Furthermore, according to the fault diagnosis results generated by the intelligent diagnosis and monitoring transformer model, automatically adjust the operating parameters of oxygen injection amount, hydrogen injection amount, and blast furnace internal pressure through multivariable sliding mode control. The sliding mode control ensures the stable and robust movement of the blast furnace on the sliding mode surface by designing the sliding mode surface.

[0132] Define the multivariable sliding mode control input according to the sliding mode control theory. The expression is:

[0133] u(t) = -k × sgn(s(z));

[0134] Where, u(t) is the control input at the current time t, which is used to calculate and determine how to adjust the operating parameters at this moment. u represents the actual value of the control input, which is used to automatically adjust the operating parameters of the blast furnace, such as oxygen injection amount, hydrogen injection amount, and blast furnace internal pressure, etc. t is the time index, which represents the value of the control input at a specific time point. sgn is the sign function, which determines the control direction. k is a positive constant, which determines the intensity of the control input. The larger the sliding mode gain k, the more rapid the response of the control input and the greater the overall adjustment strength. s(z) is the multivariable sliding mode surface function, which is a function of the state variables. z is the vector of state variables, which can include multiple states, such as temperature deviation, the change rate of state deviation, and cumulative state deviation, etc. Each state variable represents a key feature of the whole, which is used to define the overall state. By designing the sliding mode surface, the overall state moves along the sliding mode surface, thus realizing stable and robust control.

[0135]

[0136] Among them, when s(z) > 0, sgn(s(z)) is +1, and the direction of the control input is positive, which is used to increase or add a certain operating parameter. When s(z) = 0, it means there is no change in the direction of the control input and the current state is maintained. When s(z) < 0, it means the direction of the control input is negative, which is used to decrease or reduce a certain operating parameter.

[0137] According to the multivariable sliding mode surface function, select the optimal combination of state variables (such as reducing or increasing the oxygen injection amount and hydrogen injection amount), and automatically adjust the operating parameters of oxygen injection amount, hydrogen injection amount, and blast furnace internal pressure to achieve stable blast furnace operation status. The expression is:

[0138]

[0139] Where, z 1 is the current temperature deviation, such as the temperature deviation, which is the difference between the current state and the target state. z2 is the rate of change of the state deviation, such as the rate of change of the temperature deviation, and is the rate of change of the state deviation z 1 which represents the dynamic change of the overall state, and z 3 is the cumulative state deviation, such as the cumulative temperature deviation, and is the overall state deviation z 1 which is the time integral of the overall state deviation, represents the cumulative error of the overall within a certain time, c is the rate of change of the adjusted state deviation z 2 which is the constant coefficient, and g is the constant coefficient for adjusting the cumulative state deviation z 3 where ∫ is the integral symbol for performing integral operation on the function, indicating that the integration is carried out from time 0 to time t, and et is the integration variable representing the tiny change of time;

[0140] By defining a multi-variable sliding mode surface function, combining the current temperature deviation, state change rate and cumulative deviation, accurately select the optimal combination of state variables, use the sign function to determine the control direction, and adjust the control strength through a positive constant, so as to ensure that the blast furnace operation parameters (such as oxygen injection volume, hydrogen injection volume and internal pressure) are quickly adjusted to a stable state. Compared with the basic method, this method can more effectively cope with uncertainties, improve the control accuracy and response speed, enhance the overall robustness and stability, so as to optimize the blast furnace operation efficiency and ensure safety;

[0141] It should be noted that through multi-variable sliding mode control, the operation parameters of oxygen injection volume, hydrogen injection volume and blast furnace internal pressure are automatically adjusted according to the fault diagnosis results to ensure the stable state of the blast furnace. Design the sliding mode surface and control input formula to achieve precise adjustment of operation parameters, ensure stability and robustness, effectively improve the blast furnace operation efficiency and production quality, reduce energy consumption and operation costs, and ensure the stable operation of the blast furnace under different working conditions.

[0142] S4. After determining that the blast furnace operation state is stable, collect the blast furnace gas and conduct dynamic gas separation, compression and storage, purify the remaining gas, and recover the waste heat.

[0143] Furthermore, use the gas collection equipment installed at the top of the blast furnace to collect the blast furnace gas and transfer it to the central collector at the top of the blast furnace (these gases are collected by multiple exhaust pipes and transported to the central collector at the top of the blast furnace);

[0144] The central collector transports the blast furnace gas to the dynamic gas separation device through an orderly pipeline;

[0145] In the dynamic gas separation device, the gas separation membrane efficiently separates hydrogen and carbon dioxide from the blast furnace gas;

[0146] Transport the separated hydrogen to the hydrogen recycling equipment, store it through a compressor and a gas storage tank, and inject it into the blast furnace again when needed to improve the utilization efficiency of hydrogen;

[0147] The separated carbon dioxide is compressed by a compressor, converted into a high-pressure state, and stored in a high-pressure storage tank.

[0148] The compressed carbon dioxide is transported to the high-pressure storage tank for safe storage. These storage tanks are equipped with temperature and pressure sensors to monitor the storage status in real time to ensure safety.

[0149] After separating hydrogen and carbon dioxide, the remaining coal gas enters the purification equipment, and a multi-stage filtration and washing method is used to remove impurities and particulate matter in the remaining coal gas.

[0150] The purified remaining coal gas recovers waste heat through a heat exchanger to complete energy utilization, and the heat is used to preheat the blast furnace raw materials or for use in the plant area.

[0151] Part of the purified coal gas is used for power generation by a gas turbine in a power plant, and the other part is used for heating in the plant area to improve energy utilization efficiency.

[0152] It should be noted that by collecting blast furnace gas and performing dynamic gas separation and compression storage, efficient separation and recovery of hydrogen and carbon dioxide are achieved. After hydrogen is stored, it can be recycled and injected into the blast furnace to improve utilization efficiency; carbon dioxide is stored under high pressure to ensure safety. The remaining coal gas removes impurities through purification equipment, and the waste heat is recovered by a heat exchanger for preheating blast furnace raw materials or for use in the plant area, improving energy utilization efficiency, reducing environmental pollution, and optimizing resource management.

[0153] This embodiment also provides a computer device applicable to the case of the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method proposed in the above embodiment.

[0154] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, operator networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads provided on the outer shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0155] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for realizing the control of hydrogen-rich carbon cycle oxygen blast furnace gas recovery as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), Erasable Programmable Read-Only Memory (EPROM for short), Programmable Read-Only Memory (PROM for short), Read-Only Memory (ROM for short), magnetic memory, flash memory, magnetic disks, or optical discs.

[0156] In summary, through the construction of an intelligent diagnostic and monitoring transformer model, the present invention realizes the real-time monitoring and accurate prediction of the blast furnace operation state. The intelligent diagnostic and monitoring transformer model uses a multi-head self-attention mechanism to encode and extract features from the preprocessed monitoring data, generates a prediction value of the blast furnace operation state, improves the accuracy of the prediction result, and generates a fault diagnosis result by setting a threshold, realizing the early warning and analysis of potential faults in the blast furnace. Finally, the intelligent diagnostic and monitoring transformer model effectively improves the safety and stability of the blast furnace operation, reduces unplanned shutdowns and maintenance costs, and improves production efficiency and resource utilization rate.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered by the scope of the claims of the present invention.

Claims

1. A hydrogen-rich carbon cycle oxygen blast furnace gas recovery control method, characterized in that: include, Collect monitoring data and perform pre-processing; Build an intelligent diagnosis and monitoring transformer model to generate fault diagnosis results; According to the fault diagnosis results, the operating parameters are automatically adjusted through multivariable sliding mode control to achieve stable operation of the blast furnace; After confirming that the blast furnace is operating in a stable state, the blast furnace gas is collected and dynamically separated and compressed for storage, the remaining gas is purified, and the waste heat is recovered.

2. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 1, characterized in that: The monitoring data includes temperature data, pressure data, oxygen data and hydrogen concentration data.

3. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 2, characterized in that: The preprocessing includes noise removal, data normalization, outlier processing, and feature extraction through principal component analysis.

4. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 3, characterized in that: The specific steps of constructing the intelligent diagnosis and monitoring transformer model are as follows: The input is defined as the preprocessed monitoring data, and the output is the fault diagnosis result; Divide the monitoring data into training set, validation set and test set; The preprocessed monitoring data is encoded through an encoder to extract features, and a multi-head self-attention mechanism is introduced to generate enhanced features; The decoder is used to process the encoded enhanced features to generate the predicted value of the blast furnace operation status, which is expressed as: in, is the predicted value of the blast furnace operation status at the next time, is the predicted value of the blast furnace operation status of the actual measured value y, a i is the historical data coefficient, y(t-i+1) is the actual operation result of the previous i time, y is the actual measurement value, b j is the control input coefficient, u(tj) is the control input at the previous j time points, u is the actual value of the control input, t is the index of time, n is the time step of the historical data, m is the time step of the control input, i is the index of the historical data, and j is the index of the control input; Generate fault diagnosis results based on the predicted value of blast furnace operation status; Use the training set to train the intelligent diagnosis and monitoring transformer model, and adjust the parameters through the back propagation algorithm and adaptive optimization algorithm; Evaluate the performance of the intelligent diagnosis monitoring transformer model through the validation set; The intelligent diagnosis and monitoring transformer model is tested using the test set.

5. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 4, characterized in that: The preprocessed monitoring data is encoded, features are extracted, and a multi-head self-attention mechanism is introduced to generate enhanced features. The specific steps are: The encoder extracts the time series data from the monitoring data and converts the time series data into a time series feature vector; Perform position encoding on the time series feature vector; In the multi-head attention mechanism, the time series feature vector is converted into query, key and value through linear transformation; Calculate the similarity between query and key through dot product to get attention weight; Use multiple attention heads to perform parallel calculations, concatenate the results of each attention head, and obtain the final time series feature vector through linear transformation; After completing the multi-head self-attention mechanism processing, the encoder converts the time series feature vector into enhanced features.

6. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 5, characterized in that: The specific steps of generating the fault diagnosis result are as follows: Set the thresholds for normal and fault operation status of the blast furnace; According to the predicted value of blast furnace operation status, predict the expected status of each monitoring parameter of blast furnace at the next time; If the predicted value of the blast furnace operation status is within the normal range of the operation status threshold, it is in a normal state; If the predicted value of the blast furnace operation status exceeds the normal range of the operation status threshold, an abnormality exists; The severity of the fault is determined based on the predicted value of the blast furnace operation status that exceeds the normal range, and the abnormal value of the blast furnace operation status prediction is compared with the historical fault data to find the cause of the fault and generate the fault diagnosis result.

7. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 6, characterized in that: The automatic adjustment of operating parameters by multivariable sliding mode control comprises the following specific steps: According to the sliding mode control theory, the multivariable sliding mode control input is defined as follows: u(t)=-k×sgn(s(z)); Where u(t) is the control input at the current time t, u is the actual value of the control input, sgn is the sign function, k is a positive constant, s(z) is the multivariable sliding surface function, and z is the vector of state variables; According to the multivariable sliding surface function, the best combination of state variables is selected to automatically adjust the oxygen injection amount, hydrogen injection amount and blast furnace internal pressure operating parameters to achieve stable blast furnace operation. The expression is: s(z)=z1+c×z2+g×∫0 t z3et; Among them, z1 is the current temperature deviation, z2 is the rate of change of the state deviation, z3 is the cumulative state deviation, c is the constant coefficient for adjusting the rate of change of the state deviation z2, g is the constant coefficient for adjusting the cumulative state deviation z3, ∫ is the integration symbol, and et is the integration variable.

8. The hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to claim 7, characterized in that: After the blast furnace operation state is determined to be stable, the blast furnace gas is collected and dynamically separated and compressed for storage, the remaining gas is purified, and the waste heat is recovered. The specific steps are: The blast furnace gas is collected by the gas collecting equipment installed on the top of the blast furnace and transmitted to the central collector on the top of the blast furnace; The central collector transports the blast furnace gas to the dynamic gas separation unit through pipelines; In the dynamic gas separation unit, hydrogen and carbon dioxide are separated from blast furnace gas; The separated hydrogen is transported to the hydrogen circulation equipment and stored through the compressor and gas storage tank; The separated carbon dioxide is compressed by a compressor and converted into a high-pressure state and stored in a high-pressure storage tank; After separating the hydrogen and carbon dioxide, the remaining gas enters the purification equipment, where multi-stage filtration and washing methods are used to remove impurities and particulate matter in the remaining gas; The remaining gas after purification is recycled through a heat exchanger to achieve energy utilization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the hydrogen-rich carbon circulating oxygen blast furnace gas recovery control method according to any one of claims 1 to 8 are implemented.

Images