Blast furnace ironmaking process anomaly detection method based on score co-integration

Abstract

The invention provides a blast furnace ironmaking process anomaly detection method based on score co-integration, which comprises the following steps: acquiring key process variables in a blast furnace ironmaking process, and forming a time sequence by the key process variables; carrying out stationarity analysis on the time sequence by adopting an augmented Dickey-Fuller test method, and identifying non-stationary variables in the data to obtain a non-stationary time sequence; extracting a non-stationary trend part of the non-stationary time sequence by adopting a trend extraction algorithm; modeling the trend part of the non-stationary variable by using the FCVAR, and constructing an anomaly detection model based on the FCVAR; through the control limit detection of the statistical magnitude, the abnormality detection of the blast furnace ironmaking process is realized, and if the statistical magnitude exceeds the control limit, the abnormality exists in the blast furnace ironmaking process. The method solves the problem of non-stability of variables in the ironmaking process, is particularly suitable for variables with long memory characteristics, and can accurately capture the long-term equilibrium relation between the variables, find tiny abnormal changes in time and ensure the safety and stability of production.

Description

Technical Field

[0001] The present invention relates to the technical field of process monitoring in iron and steel production. Specifically, it particularly relates to a method for abnormal detection of the blast furnace ironmaking process based on fractional cointegration. Background Art

[0002] Blast furnace ironmaking is the core process of iron and steel production, which realizes ironmaking by reducing iron from iron ore and melting it into pig iron. This continuous production process involves the interaction of burden from top to bottom and gas from bottom to top, and complex chemical and physical changes occur during the process. The stable operation of the blast furnace is crucial for ensuring the quality of molten iron and production efficiency, which depends on the accurate monitoring and control of the ironmaking process.

[0003] During the blast furnace ironmaking process, due to factors such as changes in raw fuel conditions, operating conditions, and equipment failures, process fluctuations are often caused, which may lead to abnormal furnace conditions and seriously affect the quality of molten iron and production stability. Blast furnace ironmaking involves numerous process variables, including burden ratio, blast pressure, top pressure, temperature, gas flow rate, etc. These variables directly affect the quality of molten iron and the operating stability of the blast furnace. Due to the complexity of the ironmaking process, these variables usually exhibit significant non-stationarity, that is, their statistical characteristics (such as mean and variance) change over time. Non-stationarity poses great challenges to the modeling and monitoring of the blast furnace ironmaking process. If effective monitoring and control measures are not taken in a timely manner, the safety and stability of production may be seriously affected. Therefore, effective monitoring and abnormal detection of the blast furnace ironmaking process are crucial for ensuring the safety and stability of production.

[0004] Currently, various multivariate statistical methods are widely used in industrial process monitoring, such as principal component analysis (PCA), partial least squares (PLS), and kernel principal component analysis (KPCA), etc. When dealing with the blast furnace ironmaking process with significant non-stationary characteristics, if some traditional statistical methods are directly applied without considering the influence of non-stationarity, problems such as "spurious regression" may occur, which will affect the effectiveness of monitoring and control. Therefore, statistical methods capable of processing non-stationary data are needed to ensure that the model accurately captures the dynamic characteristics of process variables and improve the reliability of monitoring and abnormal detection.

[0005] Cointegration analysis is a method that can effectively process non-stationary data. When there is a certain long-term equilibrium relationship among multiple non-stationary variables, these variables are called cointegrated. For many variables in the blast furnace ironmaking process, such as top pressure and permeability, they may exhibit a certain equilibrium state in the long term. Traditional modeling methods are difficult to capture this equilibrium relationship, while cointegration analysis can effectively model it and provide a basis for abnormal detection.

[0006] However, the cointegration analysis monitoring method in the existing technology is based on integer-order cointegration. However, in actual industrial processes, the time series of many variables has long memory, which indicates strong autocorrelation in the time series. The traditional integer-order cointegration theory is established based on the characteristics of variables temporarily deviating from the equilibrium point. Therefore, when modeling variables with long memory characteristics, it is difficult to describe their inherent long-term equilibrium relationship using integer-order cointegration. Summary of the Invention

[0007] According to the above-mentioned technical problems, a blast furnace ironmaking process detection method based on the Fractionally Cointegrated Vector Autoregressive (FCVAR) model is provided. By establishing the long-term equilibrium relationship of non-stationary variables in the ironmaking process through FCVAR, timely detection of process anomalies can be achieved. The method of the present invention first analyzes the stationarity and long memory of key variables in the blast furnace ironmaking process, and extracts the non-stationary trend part from the non-stationary variables through a trend extraction algorithm. Subsequently, an anomaly detection model is established based on the fractional cointegration theory by constructing FCVAR. The present invention can effectively capture the cointegration relationship between non-stationary variables in the blast furnace ironmaking process, and timely detect abnormal situations by detecting changes in the stationarity of variable combinations, especially having good performance in detecting small and gradual anomalies.

[0008] The technical means adopted by the present invention are as follows:

[0009] An abnormal detection method for blast furnace ironmaking process based on fractional cointegration, comprising:

[0010] S1. Collect key process variables in the blast furnace ironmaking process, and form a time series {X t};

[0011] S2. Use the augmented Dickey-Fuller test method to analyze the stationarity of the time series {X t}, identify non-stationary variables in the data, and obtain a non-stationary time series {X t} ns ;

[0012] S3. Use a trend extraction algorithm to extract the non-stationary trend part of the non-stationary time series {X t} ns ;

[0013]

[0014] S4. Use FCVAR to model the trend part of non-stationary variables and construct an anomaly detection model based on FCVAR;

[0015] S5. Construction Statistic to test whether abnormal conditions occur in the blast furnace ironmaking process.

[0016] Furthermore, the key process variables in the blast furnace ironmaking process collected in step S1 include: coke burden, pellets, top pressure, permeability, ore batch, coke breeze, cold air flow rate, blast pressure, differential pressure, top pressure to air volume ratio, and hot blast temperature.

[0017] Furthermore, in step S2, the augmented Dickey-Fuller test method is used to test and judge the stationarity of the time series, specifically including:[[]]

[0018] When judging the stationarity of the time series through the unit root test, if the time series has a unit root, it indicates that the time series is non-stationary.

[0019] Furthermore, in step S3, the non-stationary trend part of the non-stationary time series is extracted, specifically as follows:[[]]

[0020]

[0021] where x trend represents the trend part of x; H ∈ R n×n , represents the observation matrix; and represents the estimated value of the regression parameter; where λ represents the regularization parameter, D d represents the discrete approximation of the d-th derivative operator, x = {x1, x2,..., x n} ∈ R n , represents a sample sequence of a non-stationary variable.

[0022] Furthermore, in step S4, the anomaly detection model based on FCVAR is constructed, specifically as follows:[[]]

[0023]

[0024] where α ∈ R p×r represents the adjustment or loading coefficient, used to characterize the adjustment speed of each variable towards equilibrium, where p represents the number of features, r represents the fractional cointegration rank, and 0 ≤ r ≤ p; β represents the fractional cointegration vector; Δ d represents the fractional difference operator; L b represents the fractional lag operator, L b = 1 - Δ b , where 0 < b < d and d - b represents the fraction; Γ i controls the short-term behavior of the variable; ε t represents the observation error; ρ represents the restricted constant term parameter.

[0025] Further, in step S5, the constructed statistic is as follows:

[0026] Construct the cointegration relationship of non-stationary variables as follows:

[0027]

[0028] where γ t represents the cointegration relationship of non-stationary variables;

[0029] According to the constructed cointegration relationship of non-stationary variables, construct the statistic as follows:

[0030]

[0031] where γ t,new represents the cointegration relationship of the new sample, represents the statistical control limit with a confidence level of α, Λ represents the covariance matrix of, F p,n-p；α represents the critical value of the F-distribution with p and 1-p degrees of freedom and a confidence level of α.

[0032] Further, in step S5, to check whether there is an abnormality in the blast furnace ironmaking process, the specific method is as follows:

[0033] If the statistic is below the corresponding control limit, then the current blast furnace ironmaking process is normal;

[0034] If the statistic exceeds the control limit, then there is an abnormality in the current blast furnace ironmaking process.

[0035] Compared with the prior art, the present invention has the following advantages:

[0036] 1. The abnormal detection method for the blast furnace ironmaking process based on fractional cointegration provided by the present invention can effectively process non-stationary time series in the blast furnace ironmaking process, and is particularly suitable for variables with long memory characteristics, thereby improving the accuracy and sensitivity of abnormal detection.

[0037] 2. The abnormal detection method for the blast furnace ironmaking process based on fractional cointegration provided by the present invention can capture the long-term equilibrium relationship between variables, and can identify small and gradually accumulating abnormal changes in advance, which is of great significance for the refined control and stable operation of the blast furnace ironmaking process.

[0038] 3. The abnormal detection method for blast furnace ironmaking process based on fractional cointegration provided by the present invention is not only applicable to the monitoring of blast furnace ironmaking process, but also has good adaptability and generality, and can be extended to other industrial processes with similar non-stationary characteristics. For any industrial process involving multi-variable non-stationary dynamic behavior, the accurate monitoring of the system state and timely early warning can be realized through the method of the present invention, so as to effectively improve production efficiency and product quality, and reduce production risks and energy consumption.

[0039] Based on the above reasons, the present invention can be widely promoted in the fields of process monitoring in iron and steel production, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is the flowchart of the method of the present invention.

[0042] Figure 2 It is the coke burden trend extraction diagram provided by the embodiment of the present invention.

[0043] Figure 3 It is the pellet trend extraction diagram provided by the embodiment of the present invention.

[0044] Figure 4 It is the top pressure trend extraction diagram provided by the embodiment of the present invention.

[0045] Figure 5 It is the permeability trend extraction diagram provided by the embodiment of the present invention.

[0046] Figure 6 The figure is the detection result diagram provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0048] It should be noted that the terms "first", "second", etc. in the description, claims and the above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily limit to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0049] As Figure 1 shown, the present invention provides an abnormal detection method for blast furnace ironmaking process based on fractional cointegration, including:

[0050] S1. Collect key process variables in the blast furnace ironmaking process, and form a time series {X t};

[0051] S2. Adopt the augmented Dickey-Fuller (ADF) test method to conduct a stationarity analysis on the time series {X t}, identify non-stationary variables in the data, and obtain a non-stationary time series {X t} ns ;

[0052] S3. Adopt a trend extraction algorithm to extract the non-stationary trend part of the non-stationary time series {X t} ns ;

[0053]

[0054] S4. Use FCVAR to model the trend part of the non-stationary variable to construct an abnormal detection model based on FCVAR;

[0055] S5. Construct a statistic to test whether an abnormality occurs in the blast furnace ironmaking process.

[0056] Specifically, as a preferred embodiment of the present invention, the key process variables collected in step S1 in the blast furnace ironmaking process include: coke burden, pellets, top pressure, permeability, ore batch, coke breeze, cold air flow rate, blast pressure, differential pressure, top pressure air volume ratio, hot blast temperature.

[0057] In specific implementation, as a preferred implementation manner of the present invention, in step S2, the augmented Dickey-Fuller (ADF) test method is used to test and judge the stationarity of the time series, specifically including:

[0058] When testing and judging the stationarity of the time series through the unit root test, if the time series has a unit root, it indicates that the time series is non-stationary.

[0059] In this embodiment, the values of the test statistics (adf) of the four variables of coke burden, pellets, top pressure, and permeability are -2.5617, -1.7835, -0.5770, and -2.6755 respectively, and they are all greater than the critical values (-3.4529, -2.8715, -2.5721) at the significance levels of 1%, 5%, and 10%, indicating that the null hypothesis cannot be rejected and the hypothesis of the existence of a unit root is accepted, indicating that these four sequences are non-stationary.

[0060] In specific implementation, as a preferred implementation manner of the present invention, in step S3, the non-stationary trend part of the non-stationary time series is extracted as follows:

[0061]

[0062] where x trend represents the trend part of x; H ∈ R n×n , represents the observation matrix; and represents the regression parameter estimate value; where λ represents the regularization parameter, D d represents the discrete approximation of the d-th derivative operator, x = {x1, x2,..., x n} ∈ R n , represents a sample sequence of a non-stationary variable.

[0063] In this embodiment, the identity matrix is selected as the observation matrix to avoid problems caused by the selection of the basis, that is, H = I ∈ R n×n , I is the identity matrix. The trend extraction of coke burden, pellets, top pressure, and permeability is shown respectively as Figures 2-5 shown.

[0064] In specific implementation, as a preferred implementation manner of the present invention, in step S4, the anomaly detection model based on FCVAR is constructed as follows:

[0065]

[0066] where α ∈ R p×r represents the adjustment or loading coefficient, which is used to characterize the adjustment speed of each variable towards equilibrium. Among them, p represents the number of features, r represents the fractional cointegration rank, and 0 ≤ r ≤ p; β represents the fractional cointegration vector; Δd denotes the fractional difference operator; L b denotes the fractional lag operator, L b = 1 - Δ b , where 0 < b < d and d - b represents a fraction; Γ i controls the short-term behavior of the control variables; ε t denotes the observation error; ρ represents the restricted constant term parameter. In this embodiment, p = 4 and r = 4.

[0067] Specifically, as a preferred embodiment of the present invention, in step S5, the statistic constructed is as follows:

[0068] Construct the cointegration relationship of non-stationary variables as follows:

[0069]

[0070] where γ t represents the cointegration relationship of non-stationary variables;

[0071] According to the constructed cointegration relationship of non-stationary variables, construct the statistic as follows:

[0072]

[0073] where γ t,new represents the cointegration relationship of the new sample, represents the statistical control limit with a confidence level of α, Λ represents the covariance matrix of, F p,n-p；α represents the critical value of the F-distribution with p and 1 - p degrees of freedom and a confidence level of α.

[0074] Specifically, as a preferred embodiment of the present invention, in step S5, to check whether an abnormality occurs in the blast furnace ironmaking process, it is as follows:

[0075] If the statistic is below the corresponding control limit, then the current blast furnace ironmaking process is normal;

[0076] If the statistic exceeds the control limit, then an abnormality occurs in the current blast furnace ironmaking process.

[0077] In this embodiment, the confidence level α = 0.99. The statistic control limit is When the system is in a normal state, the statistic is within the threshold. When an abnormality occurs in the system, the statistic exceeds the threshold to detect the occurrence of an abnormality and trigger an alarm. From Figure 6 it can be seen that at times 153 - 241, Exceed the control limit, and it is determined that an abnormality occurs at this time.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An abnormal detection method for blast furnace ironmaking process based on fractional cointegration, characterized in that, Including: S1. Collect the key process variables in the blast furnace ironmaking process, and form a time series {X t} with the collected key process variables; S2. Use the augmented Dickey-Fuller test method to perform a stationarity analysis on the time series {X t}, identify the non-stationary variables in the data, and obtain the non-stationary time series {X t} ns ; S3. Use the trend extraction algorithm to extract the non-stationary trend part of the non-stationary time series {X t} ns ​ S4. Use FCVAR to model the trend part of non-stationary variables to construct an anomaly detection model based on FCVAR; S5. Construction Statistic to test whether abnormalities occur in the blast furnace ironmaking process 2. The abnormal detection method for the blast furnace ironmaking process based on fractional cointegration according to claim 1, characterized in that, The key process variables in the blast furnace ironmaking process collected in step S1, including: coke load, pellets, top pressure, permeability, ore batch, coke fines, cold air flow rate, blast pressure, differential pressure, top pressure to air volume ratio, hot blast temperature.

3. The abnormal detection method for blast furnace ironmaking process based on fractional cointegration according to claim 1, wherein, In step S2, the augmented Dickey-Fuller test method is used to test and judge the stationarity of the time series, specifically including: When judging the stationarity of the time series through the unit root test, if the time series has a unit root, it means that the time series is non-stationary.

4. A method for abnormal detection in the blast furnace ironmaking process based on fractional cointegration according to claim 1, characterized in that, In step S3, the non-stationary trend part of the non-stationary time series is extracted, specifically as follows: Among them, x trend represents the trend part of x; H ∈ R n×n , represents the observation matrix; and represents the estimated value of the regression parameter; among them, λ represents the regularization parameter, D d represents the discrete approximation of the d-th derivative operator, x = {x1, x2, …, x n} ∈ R n , represents a sample sequence of a non-stationary variable.

5. The abnormal detection method for the blast furnace ironmaking process based on fractional cointegration according to claim 1, wherein In step S4, the anomaly detection model based on FCVAR is constructed, specifically as follows: where α ∈ R p×r denotes an adjustment or loading coefficient, which is used to characterize the adjustment speed of each variable towards equilibrium. Here, p represents the number of features, r represents the fractional cointegration rank, and 0 ≤ r ≤ p; β represents the fractional cointegration vector; Δ d denotes the fractional difference operator; L b denotes the fractional lag operator, L b = 1 - Δ b , where 0 < b < d and d - b represents a fraction; Γ i controls the short-term behavior of the control variables; ε t denotes the observation error; ρ represents the restricted constant term parameter.

6. The abnormal detection method for blast furnace ironmaking process based on fractional cointegration according to claim 1, wherein In step S5, the constructed statistic is as follows: Construct the cointegration relationship of non-stationary variables as follows: Among them, γ t represents the cointegration relationship of non-stationary variables; Based on the constructed cointegration relationship of non-stationary variables, construct statistic as follows: Among them, γ t,new represents the cointegration relationship of the new sample, represents the statistical control limit with a confidence level of α, Λ represents the covariance matrix of, F p,n-p；α represents the critical value of the F-distribution with p and 1-p degrees of freedom and a confidence level of α.

7. A method for abnormal detection in the blast furnace ironmaking process based on fractional cointegration according to claim 1, characterized in that, In step S5, check whether there is an anomaly in the blast furnace ironmaking process, specifically as follows: If the statistic is below the corresponding control limit, the current blast furnace ironmaking process is normal; If the statistic exceeds the control limit, an abnormality occurs in the current blast furnace ironmaking process.

Images