Control, model method, device, and storage medium based on industrial processes

Abstract

This application provides a control, modeling method, device, and storage medium based on industrial processes. The method includes: acquiring operating data and corresponding first master control parameters during the operation of an industrial boiler; amplifying the operating data based on the first master control parameters to obtain corresponding input data; inputting the input data into a prediction model to predict second master control parameters at a predetermined future time point, so as to adjust the operation of the industrial boiler based on the predicted master control parameters; when the predetermined future time point is reached, collecting corresponding third master control parameters based on the industrial boiler; when the difference between the third master control parameter and the second master control parameter is greater than a predetermined threshold, using the operating data and the third master control parameter to construct training samples and updating the prediction model. This method enables accurate prediction multiple steps in advance, improves production safety, and has adaptive update capabilities to reflect changes in system characteristics in a timely manner.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a control method based on industrial processes, a training method for a prediction model based on industrial processes, an electronic device, and a storage medium. Background Technology

[0002] Industrial boilers are boilers used in industry that utilize the combustion of fuels such as coal, oil, and gas to generate heat for processing. Examples of industrial boilers include oil-fired boilers, gas-fired boilers, and waste incineration boilers.

[0003] Currently, due to the location of measuring points and the lag in the measurement process, industrial boilers exhibit a significant delay between the measured values ​​of key parameters and the control quantities. A transition period is required for the key parameters to reach a new stable state after a change in the control quantity. For example, before reaching the superheater, the flue gas generated by combustion must pass through the main combustion zone, the first flue, the second flue, the superheater, and the flues between each section. Since the main steam flow measurement point is located after the superheater, changes in primary and secondary air control quantities will only produce a corresponding change in the main steam flow measurement after a certain period of flue gas flow and heat exchange, with this delay time being approximately 2-3 minutes. In other words, the measured values ​​of the boiler's key parameters are lagging, reflecting past combustion conditions rather than current combustion conditions. Control adjustments based on these measured values ​​will also exhibit lag, easily causing large fluctuations in key parameters and affecting production safety. Summary of the Invention

[0004] This application provides a control method based on industrial processes to improve production safety.

[0005] Accordingly, embodiments of this application also provide a training method for a prediction model based on industrial processes, an electronic device, and a storage medium to ensure the implementation and application of the above system.

[0006] To address the aforementioned problems, this application discloses a control method based on an industrial process, the method comprising:

[0007] During the operation of the industrial boiler, acquire operating data and corresponding primary control parameters;

[0008] The operating data is amplified based on the first master control parameter to obtain the corresponding input data;

[0009] The input data is fed into the prediction model to predict the second master control parameter at a future time point, so as to adjust the operation of the industrial boiler based on the predicted master control parameter;

[0010] When the set future time point is reached, the corresponding third master control parameters are collected based on the industrial boiler.

[0011] When the difference between the third master control parameter and the second master control parameter is greater than a set threshold, the running data and the third master control parameter are used to form training samples and update the prediction model.

[0012] Optionally, the step of amplifying the operating data based on the first master control parameters to obtain the corresponding input data includes:

[0013] The first master control parameters and operating data are normalized respectively;

[0014] The processed first master control parameter and the processed running data are concatenated to obtain the corresponding input data.

[0015] Optionally, the prediction model predicts the second master control parameter at a predetermined future time point through the following steps:

[0016] The input data is reduced in dimensionality based on the kernel function matrix to obtain the reduced-dimensional input data.

[0017] The weight coefficient vector is obtained by iterative processing of the input function based on dimensionality reduction.

[0018] The second master control parameter for setting future time points is determined based on the weight coefficient vector.

[0019] Optionally, the step of performing dimensionality reduction processing on the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes:

[0020] The kernel function matrix is ​​calculated using the input data;

[0021] The dimension reduction matrix is ​​determined based on the confidence threshold dimension;

[0022] The input data is dimensionality reduced based on the dimensionality reduction matrix and kernel function matrix to obtain dimensionality-reduced input data.

[0023] Optionally, it further includes: calculating the difference between the acquired master control parameter and the predicted master control parameter, and determining whether the difference is greater than a set threshold; if the difference is not greater than the set threshold, outputting the second master control parameter to the control system for subsequent control.

[0024] Optionally, it also includes: adjusting the operation of the industrial boiler based on the predicted master control parameters set for future time points.

[0025] Optionally, the prediction model is updated using the aforementioned operational data and collected master control parameters to construct training samples, including:

[0026] The operating data is amplified based on the main control parameters to obtain the corresponding input data;

[0027] Multiple input data points before a set time point are combined with the acquisition master control parameters corresponding to the set time point to form a many-to-one training sample.

[0028] The input model is input into the prediction model to obtain the prediction master parameters;

[0029] The loss function is determined based on the main control parameters for prediction and the main control parameters for acquisition, and the prediction model is adjusted based on the loss function.

[0030] Optionally, the operating data includes: control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0031] This application also discloses a training method for a prediction model based on industrial processes, the method comprising:

[0032] Acquire historical data of industrial boilers, and extract historical operating data and historical main control parameters from the historical data;

[0033] The historical operating data is amplified based on the historical master control parameters to obtain the corresponding input data;

[0034] Sample data is constructed based on the historical master control parameters corresponding to the specified time points of the input data.

[0035] The input data is fed into the prediction model to predict the master control parameters at a specified time point;

[0036] The predicted master control parameters are compared with the historical master control parameters at a specified time point to determine the corresponding loss function;

[0037] The prediction model is adjusted according to the loss function to obtain the trained prediction model.

[0038] Optionally, the prediction model predicts the master control parameters for a specified time point through the following steps:

[0039] The input data is reduced in dimensionality based on the kernel function matrix to obtain the reduced-dimensional input data.

[0040] The weight coefficient vector is obtained by iterative processing of the input function based on dimensionality reduction.

[0041] The main control parameters for prediction at a specified time point are determined based on the weight coefficient vector.

[0042] Optionally, the step of performing dimensionality reduction processing on the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes:

[0043] The kernel function matrix is ​​calculated using the input data;

[0044] The dimension reduction matrix is ​​determined based on the confidence threshold dimension;

[0045] The input data is dimensionality reduced based on the dimensionality reduction matrix and kernel function matrix to obtain dimensionality-reduced input data.

[0046] Optionally, the operating data includes: control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0047] This application also discloses an electronic device, including: a processor; and a memory storing executable code thereon, wherein when the executable code is executed by the processor, the method described in this application is performed.

[0048] This application also discloses one or more machine-readable media storing executable code thereon, which, when executed by a processor, performs the method described in this application.

[0049] Compared with the prior art, the embodiments of this application have the following advantages:

[0050] In this embodiment, during the operation of the industrial boiler, operating data and corresponding first master control parameters are acquired. The operating data is amplified based on the first master control parameters to obtain corresponding input data. The causal relationship between the output data and the input variables affecting its changes is constructed. This input data is then input into a prediction model to predict a second master control parameter at a predetermined future time point. This allows for adjustment of the industrial boiler operation based on the predicted master control parameters, enabling accurate predictions multiple steps in advance and improving production safety. When the predetermined future time point is reached, a third master control parameter is collected based on the industrial boiler. If the difference between the third master control parameter and the second master control parameter exceeds a predetermined threshold, the operating data and the third master control parameter are used to construct a training sample and update the prediction model. This provides adaptive updating capabilities and timely reflection of changes in system characteristics. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the steps of a training method for a prediction model based on an industrial process, according to an embodiment of this application.

[0052] Figure 2 This is a flowchart illustrating the steps of an embodiment of a control method based on an industrial process according to this application;

[0053] Figure 3 This is a schematic diagram illustrating an example of the predictive effect of an industrial process based on this application;

[0054] Figure 4 This is a flowchart of an optional embodiment of a control method based on an industrial process according to this application;

[0055] Figure 5 This is a flowchart illustrating the steps of another embodiment of the industrial process control method of this application.

[0056] Figure 6 This is a flowchart illustrating the steps of another embodiment of the training method for a prediction model based on an industrial process according to this application.

[0057] Figure 7 This is a schematic diagram of the structure of an exemplary device provided in one embodiment of this application. Detailed Implementation

[0058] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] This application's embodiments can be applied to industrial production scenarios such as industrial boilers. During industrial production, the measurement and adjustment of some parameters may exhibit lag, affecting production safety. This application's embodiments provide a predictive model for online updates of the boiler's main parameter prediction model, establishing an adaptive model to address the boiler's time-varying characteristics.

[0060] This application embodiment establishes a prediction model for industrial boilers to predict key control parameters such as main steam flow, main steam pressure, and steam drum pressure. An initial prediction model can be trained in advance based on historical data. Then, the trained prediction model is used in conjunction with real-time data to predict the operation process of the industrial boiler. The prediction results are compared with the actual output at the corresponding time. If the difference is significant, the prediction model can be updated again based on the prediction results and the actual output, thereby achieving adaptive online updates.

[0061] The prediction model can be built through the following training process:

[0062] Step 102: Obtain historical data of the industrial boiler, and extract historical operating data and historical main control parameters from the historical data.

[0063] Historical data from industrial boilers can be acquired, such as various historical operational data collected through a Supervisory Information System (SIS), including control data like primary air fan frequency, secondary air fan frequency, and feed grate cycle, as well as measurement data like furnace temperature, oxygen content after the economizer, and steam drum water level. This historical data can then be preprocessed, including data cleaning and filtering. For example, it can first remove null, outlier, and incorrect values, followed by time homogenization. Since the time ranges for different measurement and control data may differ (e.g., some data are measured every minute, while others are measured every four minutes), time homogenization can be performed. Data from non-normal operating conditions can also be eliminated based on main steam flow rate.

[0064] Then, historical operating data and historical main control parameters corresponding to each time point can be obtained from historical data. In this embodiment of the application, the operating data includes: control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0065] Step 104: Based on the historical master control parameters, the historical operation data is amplified to obtain the corresponding input data.

[0066] Because industrial boilers exhibit significant time lag, changes in current output measurements are influenced by past input values ​​rather than the current time. Therefore, to improve prediction accuracy, this embodiment of the application performs input amplification, incorporating sampled values ​​of input and output variables from a historical period as model input. That is, based on historical operating data as input variables, historical master control parameters from the same time point are obtained, and these same historical master control parameters are used to amplify the historical operating data, resulting in the input data.

[0067] The step of amplifying the historical operating data based on the historical master control parameters to obtain corresponding input data includes: normalizing the historical master control parameters and the historical operating data respectively; and concatenating the processed historical master control parameters and the processed historical operating data to obtain the corresponding input data. In this embodiment, the input data can also be normalized. The historical master control parameters and the historical operating data can be normalized separately, facilitating comparison and weighting of data of different magnitudes. Normalization transforms dimensional variables into dimensionless variables, and then the processed historical master control parameters and the processed historical operating data can be concatenated to obtain the corresponding input data. In one example, Min-Max Normalization can be used to normalize the data to the range [0,1], as shown in Formula 1:

[0068]

[0069] Where x,x * The values ​​of the variable before and after normalization are x, respectively. max x is the maximum value of the sample data. min This represents the minimum value of the sample data.

[0070] In other examples, normalization may be performed in other ways, and this application does not limit the implementation of such methods.

[0071] Step 106: Based on the historical master control parameters corresponding to the specified time points of the input data, sample data is constructed.

[0072] This application embodiment can also perform output matching to determine the time point of the input data, and determine a specified time point based on the time point, such as the time point T steps after the time point. The historical master control parameters corresponding to the time point T steps after the historical input data can be used to form a many-to-one training sample form to obtain sample data.

[0073] This embodiment of the application can match the amplified input with the corresponding output after T steps in the original dataset to form a many-to-one training sample format. Then, the sample data can be filtered to select samples under various working conditions, ensuring that the total number of samples matches the set model sample size M.

[0074] Step 108: Input the input data into the prediction model to predict the master control parameters at a specified time point.

[0075] After the sample data is determined, the input data from the sample data can be fed into the prediction model for processing to predict the master control parameters at a specified time point.

[0076] In this embodiment, dimensionality reduction can be performed using kernel functions, and the iterative chain can be established to reduce the running time of the iterative processing. In an optional embodiment, the prediction model predicts the master control parameters at a specified time point through the following steps: dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data; iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determination of the master control parameters at the specified time point based on the weight coefficient vector. The dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes: calculating the kernel function matrix using the input data; determining the dimensionality reduction matrix based on the confidence threshold dimension; and performing dimensionality reduction of the input data based on the dimensionality reduction matrix and the kernel function matrix to obtain dimensionality-reduced input data.

[0077] In one example, the kernel function can be Principal Component Analysis (PCA), a widely used data dimensionality reduction algorithm. PCA transforms the original data (n-dimensional features) into a set of linearly independent data (k-dimensional features), where the k-dimensional features are novel orthogonal features, also known as principal components. A predictive model can be built using Weighted Least Squares Support Vector Regression (WLS-SVR), and then dimensionality reduction can be achieved using the kernel function PCA. This allows for iterative chaining, extending the WLS-SVR method to nonlinear regression problems. Alternatively, the weighted least squares method can be used to ensure that each sample point satisfies the KKT conditions during the iteration process. The calculation can be performed using the following steps:

[0078] 1) Calculate the kernel function matrix K using training samples. N =Φ N Φ N T The singular value decomposition (SVD) is then performed on it; the Gaussian kernel function is used, and its calculation formula is as follows:

[0079]

[0080] Where, x i ,x j σ represents the input data in the dataset, and σ is the kernel parameter.

[0081] 2) Determine the dimension R of the principal components based on the set confidence threshold, and obtain the dimensionality reduction matrix V of the principal components. R ;

[0082] 3) Project the training samples onto the reduced-dimensional space to obtain...

[0083] Among them, K N V is the kernel function matrix obtained in step 1). RThis is the principal component dimensionality reduction matrix obtained in step 2).

[0084] 4) Begin IQP algorithm iteration and initialize γ +[k] =β,γ -[k] =0, k is 1 in the first iteration cycle;

[0085] Where γ is the ratio of the sample Lagrange multiplier to the relaxation factor, and its calculation formula is shown in Formulas 10 and 11, k is the current iteration number, and β is the configured initial value.

[0086] 5) Calculate the weight coefficient vector ω according to the following formula. [k] ;

[0087]

[0088]

[0089]

[0090] Among them, X Φ =[Φ(x1),Φ(x1),…,Φ(x) l )],1 l×l and 1 l×1 Let l×l and l×1 represent matrices where all elements are 1, respectively.

[0091] 6) Calculate the relaxation factor ξ corresponding to the new weight coefficients according to the following formula. i +[k+1] and ξ i -[k+1] :

[0092]

[0093]

[0094] Among them, y i represents the output data in the dataset, b represents the model residual parameter, and ε represents the insensitivity coefficient.

[0095] 7) Calculate the Lagrange multiplier λ using the following formula. i +[k] and λ i -[k] :

[0096]

[0097]

[0098] Where C is the penalty coefficient.

[0099] Update γ according to the following formula +[k] and γ -[k]:

[0100]

[0101]

[0102] 8) If k = 1 or ||ω [k] -ω [k-1] If 2 > κ1 (κ1 is the set stopping training threshold), let k = k + 1, and return to step 5 to start the next iteration; otherwise, end the loop and output the weight coefficient vector ω. [k] As the final model parameters, model f can be expressed in the following form:

[0103] f(x i )=ω T ·Φ(x i )+b

[0104] =ω T ·Φ N T V R V R T Φ N ·Φ(x i )+b

[0105] =ω R T ·(V R T Φ N ·Φ(x i Formula 12

[0106] Among them, V R T Φ N ·Φ(x i Let be an R×1 dimensional vector. Using the PCA method described above, a high-dimensional mapping function is constructed in inner product form, which can then be computed using a kernel function. The corresponding high-dimensional mapping function after dimensionality reduction yields the final weight coefficients ω. R It is a 1×R dimensional vector.

[0107] Thus, the predictive master parameters for time T can be obtained through the predictive model.

[0108] Step 110: Compare the predicted master control parameters with the historical master control parameters at a specified time point to determine the corresponding loss function.

[0109] Then, the predicted master control parameters at time T (i.e., the specified time point) can be compared with the historical master control parameters at that time point T to determine the corresponding loss function.

[0110] Step 112: Adjust the prediction model according to the loss function to obtain the trained prediction model.

[0111] The prediction model is then adjusted based on the loss function. Through multiple iterations, the main control parameters predicted by the prediction model meet the conditions, and the trained prediction model is obtained.

[0112] The prediction model is trained based on the above steps, and the calculation speed is relatively fast.

[0113] The model training and processing in this embodiment are computationally fast, therefore the method can also be used for full adaptive updates during online runtime, solving the problems of low computational efficiency and difficulty in online application of conventional full update methods. Through online full adaptive updates, model accuracy is monitored and updated in real time, ensuring that the model maintains high prediction accuracy even during long-term operation.

[0114] Reference Figure 2 The diagram shows a flowchart of an embodiment of a control method based on an industrial process according to this application.

[0115] Step 202: During the operation of the industrial boiler, acquire the operating data and the corresponding first main control parameters.

[0116] Various data can be collected during the operation of industrial boilers, including operational data and main control parameters. The operational data includes control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feed grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0117] Step 204: Amplify the operating data according to the first master control parameters to obtain the corresponding input data.

[0118] The operating data is amplified based on the master control parameters. For ease of distinction, the master control parameter used for input amplification is referred to as the first master control parameter, which is the master control parameter at the same time point as the input operating data. The predicted master control parameter at a future time point T is referred to as the second master control parameter, and the actual master control parameter at time point T is referred to as the third master control parameter. Therefore, the first master control parameter and the operating data at the same time point can constitute the input data.

[0119] In one optional embodiment, the step of amplifying the operating data based on the master control parameters to obtain corresponding input data includes: normalizing the first master control parameters and the operating data respectively; and concatenating the processed first master control parameters and the processed operating data to obtain the corresponding input data. Normalizing the first master control parameters and the operating data separately facilitates comparison and weighting of data of different magnitudes. Normalization transforms dimensional variables into dimensionless variables, and then the processed first master control parameters and the processed operating data can be concatenated to obtain the corresponding input data. In one example, a max-min normalization method can be used to normalize the data to the range [0,1], thereby obtaining the input data X(t) for the prediction model.

[0120] To address the large time delay characteristics of boilers, this scheme employs input amplification and output matching. Firstly, input amplification is performed by incorporating sampled values ​​of input variables X(tm), X(t-m+1), ..., X(tn) over a historical period, as well as sampled values ​​of output variables y(tm), y(t-m+1), ..., y(tn), into the model input. Secondly, output matching is performed by matching the amplified input with the corresponding output y(t+T) after T steps in the original dataset, creating a many-to-one training sample. Based on this approach, the causal relationship between the output y(t+T) and the input variables influencing its change is constructed, enabling predictions T steps in advance.

[0121] Step 206: Input the input data into the prediction model to predict and set the second master control parameters for future time points, so as to adjust the operation of the industrial boiler based on the predicted master control parameters.

[0122] Input the input data X(t) into the prediction model to predict the second master control parameter at a future time point T. And record it.

[0123] In one optional embodiment, the prediction model predicts the second master control parameter at a predetermined future time point through the following steps: dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data; iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determination of the second master control parameter at the predetermined future time point based on the weight coefficient vector. The dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes: calculating the kernel function matrix using the input data; determining the dimensionality reduction matrix based on the confidence threshold dimension; and performing dimensionality reduction processing on the input data based on the dimensionality reduction matrix and the kernel function matrix to obtain dimensionality-reduced input data. This process is similar to the corresponding process in the model training process described above, and therefore will not be repeated.

[0124] Taking the training and updating of a prediction model based on WLS-SVR as an example, compared with the conventional SVR solution based on quadratic programming, the WLS-SVR algorithm makes two improvements within the iterative framework: first, it uses the weighted least squares method to ensure that each sample point satisfies the KKT conditions during the iteration process; second, it uses kernel function PCA for dimensionality reduction and opens up the iteration chain, extending the WLS-SVR method to nonlinear regression problems. Therefore, the entire iterative algorithm has a shorter running time, allowing for a full update strategy during online runtime, i.e., reconstructing training samples based on the most recent data. Due to its fast computation speed, this method can be used for full adaptive updates during online runtime, solving the problems of low computational efficiency and difficulty in online application of conventional full update methods. Through online full adaptive updates, the model accuracy is monitored and updated in real time, ensuring that the model maintains high prediction accuracy even during long-term operation.

[0125] Step 208: When the set future time point is reached, the corresponding third master control parameters are collected based on the industrial boiler.

[0126] This application embodiment performs online adaptive updates. Therefore, when the set future time point T is reached, the third main control parameter y(t) corresponding to the industrial boiler can be collected, namely the primary fan frequency, secondary fan frequency, feeding grate cycle, etc.

[0127] Step 210: When the difference between the third master control parameter and the second master control parameter is greater than a set threshold, the running data and the third master control parameter are used to form training samples and update the prediction model.

[0128] The third master control parameter y(t) is compared with the second master control parameter predicted by the model T steps prior. The two are compared, and the difference is calculated. If the difference is greater than a set threshold, it is determined that the model is mismatched, and the prediction model is updated using the running data and the third master control parameter. If the difference is not greater than the set threshold, the model is determined to be accurate, and the second master control parameter is adjusted. The output is sent to the control system for subsequent control. Then the calculation for the next cycle continues.

[0129] In this embodiment, the main control parameters at future time points can be predicted using a predictive model, thus providing a more valuable reference for industrial production processes. Furthermore, based on the predicted main control parameters at future time points, the operation of the industrial boiler can be adjusted, such as determining the amount of subsequent fuel.

[0130] The method of updating the prediction model using the operational data and the third master control parameter includes: amplifying the operational data based on the first master control parameter to obtain corresponding input data; constructing a many-to-one training sample using the input data and the third master control parameter corresponding to a set time point; inputting the input model into the prediction model to obtain the fourth master control parameter; determining a loss function based on the fourth master control parameter and the third master control parameter; and adjusting the prediction model based on the loss function. In some other embodiments, the loss function can also be directly determined based on the third master control parameter and the second master control parameter, and the prediction model can be adjusted based on the loss function.

[0131] This process is similar to the model training process described above, so it will not be repeated here.

[0132] In summary, during the operation of an industrial boiler, operational data and corresponding first master control parameters are acquired. Based on these first master control parameters, the operational data is amplified to obtain corresponding input data. The causal relationship between the output data and the input variables affecting its changes is constructed. This input data is then input into a prediction model to predict a second master control parameter at a predetermined future time point. This allows for the adjustment of the industrial boiler operation based on the predicted master control parameters, enabling accurate predictions multiple steps in advance and improving production safety. When the predetermined future time point is reached, a third master control parameter is collected based on the industrial boiler. If the difference between the third master control parameter and the second master control parameter exceeds a predetermined threshold, the operational data and the third master control parameter are used to construct a training sample and update the prediction model. This requires adaptive update capabilities to reflect changes in system characteristics in a timely manner.

[0133] Compared to traditional single-step predictive modeling, this scheme employs input amplification and output matching for multi-step advance prediction. This overcomes the problem of traditional methods where prediction results follow the output of the previous moment, establishing a reasonable model that reflects the causal relationship between input and output. Furthermore, the multi-step advance prediction results provide more valuable references for the control system. This scheme was applied to the construction of a main steam flow prediction model for a certain plant. Using the plant's historical operating data (such as dozens of operating parameters including pusher stroke, pusher action, primary air volume, primary air pressure, secondary air volume, furnace temperature, flue gas oxygen content, feedwater flow rate, steam drum level, and main steam pressure), a steam flow prediction model was established. The prediction results are as follows: Figure 3 As shown, the predicted curve has a lead time of T=180s compared to the actual curve. This means that the prediction model can accurately predict the steam volume 180s later, providing an effective basis for subsequent steam volume trend forecasting and control decisions. This mitigates the impact of fuel calorific value uncertainty and large time lag on the large and difficult-to-control fluctuations in steam volume.

[0134] Based on the above embodiments, this application provides an adaptive online update-based industrial process control method, such as... Figure 4 As shown:

[0135] Step 402: During the operation of the industrial boiler, acquire the operating data and the corresponding first main control parameters.

[0136] Step 404: Normalize the first master control parameters and the running data respectively.

[0137] Step 406: The processed first master control parameters and the processed running data are concatenated to obtain the corresponding input data.

[0138] Step 408: Input the input data into the prediction model to predict the second master control parameter for a set future time point.

[0139] The prediction model predicts the second master control parameter at a predetermined future time point through the following steps: dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data; iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determination of the second master control parameter at the predetermined future time point based on the weight coefficient vector. The dimensionality reduction of the input data based on the kernel function matrix includes: calculating the kernel function matrix using the input data; determining the dimensionality reduction matrix based on the confidence threshold dimension; and performing dimensionality reduction processing on the input data based on the dimensionality reduction matrix and the kernel function matrix to obtain the dimensionality-reduced input data.

[0140] Step 410: When the set future time point is reached, the corresponding third master control parameters are collected based on the industrial boiler.

[0141] Step 412: Calculate the difference between the acquired master control parameter and the predicted master control parameter, and determine whether the difference is greater than a set threshold.

[0142] If not, proceed to step 414; if yes, proceed to step 416.

[0143] Step 414: Output the second master control parameter to the control system for subsequent control.

[0144] If the difference is not greater than a set threshold, the second master control parameter is output to the control system for subsequent control. Based on the predicted master control parameters for the set future time points, the operation of the industrial boiler is adjusted.

[0145] Step 416: Use the running data and the third master control parameters to form training samples.

[0146] Specifically, the operating data is amplified based on the master control parameters to obtain corresponding input data; multiple input data before a set time point are combined with the acquisition master control parameters corresponding to the set time point to form a many-to-one training sample.

[0147] Step 418: Input the input model into the prediction model to obtain the fourth master control parameter.

[0148] Step 420: Determine the loss function based on the fourth master control parameter and the third master control parameter, and adjust the prediction model based on the loss function.

[0149] Taking a waste incinerator as an example, changes in waste properties, equipment aging, and ash accumulation on the heating surface all cause the characteristics of the boiler combustion system to change over time. Existing models suffer from model mismatch when applied to industrial sites for extended periods. Furthermore, these models are highly dependent on initial training data, exhibit low generalization ability, and weak robustness, making them unsuitable for long-term online operation. Therefore, this application establishes a predictive model with adaptive update capabilities that reflects changes in system characteristics in real time. This allows for the study of online update algorithms for predicting key boiler parameters, establishing an adaptive model to address the time-varying characteristics of the boiler.

[0150] This application, leveraging the advantages of data-driven approaches and incorporating industrial process principles, develops a method for constructing a predictive model for key operating parameters of industrial boilers. The established predictive model for key parameters of a solid waste incinerator can accurately predict the changing trends of key parameters 2-3 minutes later, including main steam flow, main steam pressure, and exhaust gas pollutant concentration, based on current operating parameters. The prediction results of this model are combined with the control algorithm to guide the control system to act in advance based on the future changing trends of key parameters. This alleviates problems such as poor control performance and large fluctuations in operating parameters caused by large delays, large inertia, and model uncertainties, thereby achieving intelligent automatic control of the waste incinerator.

[0151] The above embodiments mainly use the production process of industrial boilers as an example for discussion. In actual other industrial production processes, there are also the effects of measurement point position and measurement link lag. The measured values ​​of the main parameters have obvious delays relative to the control quantities, such as in the raw material production and processing process and energy production process, which leads to inaccurate prediction and affects production safety.

[0152] Reference Figure 5 The diagram illustrates a flowchart of another embodiment of the control method for an industrial process according to this application.

[0153] Step 502: During the operation of the industrial equipment, acquire the operating data and the corresponding first master control parameters.

[0154] Step 504: Amplify the operating data according to the first master control parameters to obtain the corresponding input data.

[0155] The step of amplifying the operating data based on the first master control parameter to obtain the corresponding input data includes: normalizing the first master control parameter and the operating data respectively; and concatenating the processed first master control parameter and the processed operating data to obtain the corresponding input data.

[0156] Step 506: Input the input data into the prediction model to predict and set the second master control parameters at future time points, so as to adjust the operation of the industrial boiler based on the predicted master control parameters.

[0157] The prediction model predicts the second master control parameter at a predetermined future time point through the following steps: dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data; iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determination of the second master control parameter at the predetermined future time point based on the weight coefficient vector. The dimensionality reduction of the input data based on the kernel function matrix includes: calculating the kernel function matrix using the input data; determining the dimensionality reduction matrix based on the confidence threshold dimension; and performing dimensionality reduction processing on the input data based on the dimensionality reduction matrix and the kernel function matrix to obtain the dimensionality-reduced input data.

[0158] Step 508: When the set future time point is reached, the corresponding third master control parameters are collected based on the industrial boiler.

[0159] The difference between the acquired master control parameter and the predicted master control parameter is calculated, and it is determined whether the difference is greater than a set threshold. If the difference is not greater than the set threshold, the second master control parameter is output to the control system for subsequent control. Based on the predicted master control parameter set for a future time point, the operation of the industrial boiler is adjusted.

[0160] Step 510: When the difference between the third master control parameter and the second master control parameter is greater than a set threshold, the running data and the third master control parameter are used to form training samples and update the prediction model.

[0161] The process of using the operational data and the third master control parameter to construct training samples and update the prediction model includes: amplifying the operational data based on the first master control parameter to obtain corresponding input data; constructing a many-to-one training sample using the input data and the third master control parameter corresponding to a set time point; inputting the input model into the prediction model to obtain a fourth master control parameter; determining a loss function based on the fourth master control parameter and the third master control parameter; and adjusting the prediction model based on the loss function.

[0162] The operating data includes control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0163] Reference Figure 6 The diagram illustrates a step flowchart of another embodiment of a training method for a prediction model based on an industrial process according to this application.

[0164] Step 602: Obtain historical data of the industrial equipment production process, and obtain historical operating data and historical master control parameters from the historical data.

[0165] Step 604: Based on the historical master control parameters, the historical operation data is amplified to obtain the corresponding input data.

[0166] Step 606: Based on the historical master control parameters corresponding to the specified time point of the input data, sample data is constructed.

[0167] Step 608: Input the input data into the prediction model to predict the master control parameters at a specified time point.

[0168] The prediction model predicts the master control parameters at a specified time point through the following steps: dimensionality reduction of the input data based on the kernel function matrix to obtain dimensionality-reduced input data; iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determination of the master control parameters at the specified time point based on the weight coefficient vector.

[0169] The step of performing dimensionality reduction processing on the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes: calculating the kernel function matrix using the input data; determining the dimensionality reduction matrix based on the confidence threshold dimension; and performing dimensionality reduction processing on the input data based on the dimensionality reduction matrix and the kernel function matrix to obtain dimensionality-reduced input data.

[0170] Step 610: Compare the predicted master control parameters with the historical master control parameters at a specified time point to determine the corresponding loss function.

[0171] Step 612: Adjust the prediction model according to the loss function to obtain the trained prediction model.

[0172] This application establishes a predictive model with adaptive update capability and timely reflection of changes in system characteristics. It can study the online update algorithm of the predictive model of the main parameters of industrial equipment and establish an adaptive model to cope with the time-varying characteristics of boilers.

[0173] In all embodiments of this application, if user information is involved, it is collected, used and stored only after obtaining the user's authorization and permission. All kinds of operations based on user information are also executed only after obtaining the user's authorization and permission.

[0174] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0175] Based on the above embodiments, this embodiment also provides a control device based on industrial processes, which is applied in server-side electronic devices.

[0176] The input processing module is used to acquire operating data and corresponding first main control parameters during the operation of the industrial boiler; and to amplify the operating data according to the first main control parameters to obtain corresponding input data.

[0177] The prediction module is used to input the input data into the prediction model to predict the second master control parameters at a future time point, so as to adjust the operation of the industrial boiler based on the predicted master control parameters.

[0178] The update judgment module is used to collect the corresponding third master control parameters of the industrial boiler when a set future time point is reached.

[0179] The update module is used to update the prediction model by using the running data and the third main control parameter to form training samples when the difference between the third main control parameter and the second main control parameter is greater than a set threshold.

[0180] In summary, during the operation of an industrial boiler, operational data and corresponding first master control parameters are acquired. Based on these first master control parameters, the operational data is amplified to obtain corresponding input data. The causal relationship between the output data and the input variables affecting its changes is constructed. This input data is then input into a prediction model to predict a second master control parameter at a predetermined future time point. This allows for the adjustment of the industrial boiler operation based on the predicted master control parameters, enabling accurate predictions multiple steps in advance and improving production safety. When the predetermined future time point is reached, a third master control parameter is collected based on the industrial boiler. If the difference between the third master control parameter and the second master control parameter exceeds a predetermined threshold, the operational data and the third master control parameter are used to construct a training sample and update the prediction model. This requires adaptive update capabilities to reflect changes in system characteristics in a timely manner.

[0181] The input processing module is used to normalize the first master control parameter and the running data respectively; and to concatenate the processed first master control parameter and the processed running data to obtain the corresponding input data.

[0182] The prediction module is used to perform dimensionality reduction processing on the input data based on the kernel function matrix to obtain dimensionality-reduced input data; perform iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determine the second master control parameter for a future time point based on the weight coefficient vector.

[0183] The prediction module is used to calculate a kernel function matrix using input data; determine a dimensionality reduction matrix based on the confidence threshold dimension; and perform dimensionality reduction processing on the input data according to the dimensionality reduction matrix and the kernel function matrix to obtain dimensionality-reduced input data.

[0184] The update judgment module is further used to calculate the difference between the third master control parameter and the second master control parameter, and to determine whether the difference is greater than a set threshold; if the difference is not greater than the set threshold, the second master control parameter is output to the control system for subsequent control.

[0185] The update judgment module is also used to adjust the operation of the industrial boiler based on the second master control parameter set for a future time point.

[0186] The update module is used to amplify the running data according to the first master control parameter to obtain the corresponding input data; to form a many-to-one training sample with the input data and the third master control parameter corresponding to the set time point; to input the input model into the prediction model to obtain the fourth master control parameter; to determine the loss function based on the fourth master control parameter and the third master control parameter; and to adjust the prediction model based on the loss function.

[0187] The operating data includes control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level. The main control parameters include: main steam flow rate, main steam pressure, and steam drum pressure.

[0188] Based on the above embodiments, this application also provides a training device for a prediction model based on an industrial process, which is applied to a server-side electronic device.

[0189] The sample determination module is used to acquire historical data of industrial boilers, and to obtain historical operating data and historical master control parameters from the historical data; to amplify the historical operating data based on the historical master control parameters to obtain corresponding input data; and to construct sample data based on the historical master control parameters corresponding to the input data at a specified time point.

[0190] The model processing module is used to input the input data into the prediction model and predict the main control parameters at a specified time point;

[0191] The model adjustment module is used to compare the predicted master control parameters with the historical master control parameters at a specified time point to determine the corresponding loss function; and adjust the prediction model according to the loss function to obtain the trained prediction model.

[0192] The model training and processing in this embodiment are computationally fast, therefore the method can also be used for full adaptive updates during online runtime, solving the problems of low computational efficiency and difficulty in online application of conventional full update methods. Through online full adaptive updates, model accuracy is monitored and updated in real time, ensuring that the model maintains high prediction accuracy even during long-term operation.

[0193] The model processing module is used to perform dimensionality reduction processing on the input data based on the kernel function matrix to obtain dimensionality-reduced input data; perform iterative processing based on the dimensionality-reduced input function to obtain a weight coefficient vector; and determine the prediction master control parameters at a specified time point based on the weight coefficient vector.

[0194] The model processing module is used to calculate the kernel function matrix using the input data; determine the dimensionality reduction matrix based on the confidence threshold dimension; and perform dimensionality reduction processing on the input data according to the dimensionality reduction matrix and the kernel function matrix to obtain dimensionality-reduced input data.

[0195] This application also provides a non-volatile readable storage medium storing one or more modules (programs). When these modules are applied to a device, they enable the device to execute the instructions for the method steps in this application.

[0196] This application provides one or more machine-readable media storing instructions that, when executed by one or more processors, cause an electronic device to perform one or more of the methods described in the above embodiments. In this application, the electronic device includes devices such as servers and terminal devices.

[0197] Embodiments of this disclosure can be implemented as an apparatus with any suitable hardware, firmware, software, or any combination thereof, configured as desired, and the apparatus may include electronic devices such as servers (clusters) and terminals. Figure 7 An exemplary apparatus 700 is schematically shown that can be used to implement the various embodiments described in this application.

[0198] In one embodiment, Figure 7 An exemplary device 700 is shown, which includes one or more processors 702, a control module (chipset) 704 coupled to at least one of the processors 702, a memory 706 coupled to the control module 704, a non-volatile memory (NVM) / storage device 708 coupled to the control module 704, one or more input / output devices 710 coupled to the control module 704, and a network interface 712 coupled to the control module 704.

[0199] Processor 702 may include one or more single-core or multi-core processors, and processor 702 may include any combination of general-purpose processors or special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In some embodiments, device 700 can serve as a server, terminal, or other device as described in the embodiments of this application.

[0200] In some embodiments, the apparatus 700 may include one or more computer-readable media (e.g., memory 706 or NVM / storage device 708) having instructions 714 and one or more processors 702 that are combined with the one or more computer-readable media and configured to execute the instructions 714 to implement the module and thus perform the actions described in this disclosure.

[0201] In one embodiment, the control module 704 may include any suitable interface controller to provide any suitable interface to at least one of the processors 702 and / or any suitable device or component communicating with the control module 704.

[0202] The control module 704 may include a memory controller module to provide an interface to the memory 706. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0203] Memory 706 may be used, for example, to load and store data and / or instructions 714 for device 700. In one embodiment, memory 706 may include any suitable volatile memory, such as suitable DRAM. In some embodiments, memory 706 may include double data rate type 4 synchronous dynamic random access memory (DDR4 SDRAM).

[0204] In one embodiment, the control module 704 may include one or more input / output controllers to provide an interface to the NVM / storage device 708 and (one or more) input / output devices 710.

[0205] For example, NVM / storage device 708 may be used to store data and / or instructions 714. NVM / storage device 708 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (one or more) non-volatile storage devices (e.g., one or more hard disk drives (HDDs), one or more optical disc drives (CDs), and / or one or more digital universal optical disc (DVD) drives).

[0206] NVM / storage device 708 may include storage resources that are part of a device on which device 700 is mounted, or that are accessible to the device but do not necessarily have to be part of the device. For example, NVM / storage device 708 may be accessed via a network through one or more input / output devices 710.

[0207] One or more input / output devices 710 may provide an interface for device 700 to communicate with any other suitable device. Input / output devices 710 may include communication components, audio components, sensor components, etc. A network interface 712 may provide an interface for device 700 to communicate via one or more networks. Device 700 may wirelessly communicate with one or more components of a wireless network according to any of one or more wireless network standards and / or protocols, such as accessing wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G, 5G, etc., or combinations thereof.

[0208] In one embodiment, at least one of the processors 702 may be logically packaged with one or more controllers (e.g., memory controller modules) of the control module 704. In one embodiment, at least one of the processors 702 may be logically packaged with one or more controllers of the control module 704 to form a system-in-package (SiP). In one embodiment, at least one of the processors 702 may be integrated with the logic of one or more controllers of the control module 704 on the same die. In one embodiment, at least one of the processors 702 may be integrated with the logic of one or more controllers of the control module 704 on the same die to form a system-on-a-chip (SoC).

[0209] In various embodiments, device 700 may be, but is not limited to, a server, desktop computing device, or mobile computing device (e.g., laptop computing device, handheld computing device, tablet computer, netbook, etc.). In various embodiments, device 700 may have more or fewer components and / or different architectures. For example, in some embodiments, device 700 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touchscreen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0210] The detection device can use a main control chip as a processor or control module, and sensor data, position information, etc. can be stored in a memory or NVM / storage device. The sensor group can be used as an input / output device, and the communication interface can include a network interface.

[0211] This application also provides an electronic device, including: a processor; and a memory storing executable code thereon. When the executable code is executed, the processor performs one or more methods as described in this application embodiment. In this application embodiment, the memory can store various types of data, such as target files, file-application association data, and user behavior data, thereby providing a data foundation for various processing operations.

[0212] This application also provides one or more machine-readable media having executable code stored thereon, which, when executed, causes a processor to perform one or more of the methods described in this application.

[0213] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0214] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0215] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0216] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0217] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0218] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0219] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0220] The foregoing has provided a detailed description of a control method based on an industrial process, a training method for a prediction model based on an industrial process, an electronic device, and a storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A control method based on industrial processes, characterized in that, The method includes: During the operation of the industrial boiler, the operating data and corresponding primary control parameters are acquired. These primary control parameters include: main steam flow rate, main steam pressure, and steam drum pressure. The first master control parameters and operating data at the same point in time are normalized respectively; The processed first master control parameter and the processed running data are concatenated to obtain the corresponding input data; The input data is fed into the prediction model to predict the second master control parameter at a future time point, so as to adjust the operation of the industrial boiler based on the predicted master control parameter; When the set future time point is reached, the corresponding third master control parameters are collected based on the industrial boiler. When the difference between the third master control parameter and the second master control parameter is greater than a set threshold, the running data and the third master control parameter are used to form training samples and update the prediction model.

2. The method according to claim 1, characterized in that, The prediction model predicts the second master control parameter at a predetermined future time point through the following steps: The input data is reduced in dimensionality based on the kernel function matrix to obtain the reduced-dimensional input data. The weight coefficient vector is obtained by iterative processing of the input function based on dimensionality reduction. The second master control parameter for setting future time points is determined based on the weight coefficient vector.

3. The method according to claim 2, characterized in that, The dimensionality reduction processing of the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes: The kernel function matrix is ​​calculated using the input data; The dimension reduction matrix is ​​determined based on the confidence threshold dimension; The input data is dimensionality reduced based on the dimensionality reduction matrix and kernel function matrix to obtain dimensionality-reduced input data.

4. The method according to claim 1, characterized in that, Also includes: Calculate the difference between the third master control parameter and the second master control parameter, and determine whether the difference is greater than a set threshold. If the difference is not greater than the set threshold, the second master control parameter is output to the control system for subsequent control.

5. The method according to claim 1, characterized in that, Also includes: The operation of the industrial boiler is adjusted based on the second master control parameter set for the future time point.

6. The method according to claim 1, characterized in that, The process of using the aforementioned operational data and the third master control parameter to construct training samples and updating the prediction model includes: The operating data is amplified based on the first master control parameter to obtain the corresponding input data; The input data and the third master control parameters corresponding to the set time points are used to form a many-to-one training sample; The input model is input into the prediction model to obtain the fourth master control parameter; The loss function is determined based on the fourth and third master control parameters, and the prediction model is adjusted based on the loss function.

7. The method according to any one of claims 1-6, characterized in that, The operating data includes control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level.

8. A training method for a prediction model based on industrial processes, characterized in that, The method includes: Historical data of the industrial boiler is acquired, and historical operating data and historical main control parameters are obtained from the historical data, wherein the main control parameters include: main steam flow rate, main steam pressure and steam drum pressure; The first master control parameters and operating data at the same point in time are normalized respectively; The processed first master control parameter and the processed running data are concatenated to obtain the corresponding input data; Sample data is constructed based on the historical master control parameters corresponding to the specified time points of the input data. The input data is fed into the prediction model to predict the master control parameters at a specified time point; The predicted master control parameters are compared with the historical master control parameters at a specified time point to determine the corresponding loss function; The prediction model is adjusted according to the loss function to obtain the trained prediction model.

9. The method according to claim 8, characterized in that, The prediction model predicts the master control parameters for a specified time point through the following steps: The input data is reduced in dimensionality based on the kernel function matrix to obtain the reduced-dimensional input data. The weight coefficient vector is obtained by iterative processing of the input function based on dimensionality reduction. The main control parameters for prediction at a specified time point are determined based on the weight coefficient vector.

10. The method according to claim 9, characterized in that, The dimensionality reduction processing of the input data based on the kernel function matrix to obtain dimensionality-reduced input data includes: The kernel function matrix is ​​calculated using the input data; The dimension reduction matrix is ​​determined based on the confidence threshold dimension; The input data is dimensionality reduced based on the dimensionality reduction matrix and kernel function matrix to obtain dimensionality-reduced input data.

11. The method according to any one of claims 8-10, characterized in that, The operating data includes control data and measurement data. The control data includes at least one of the following: primary air fan frequency, secondary air fan frequency, and feeding grate cycle. The measurement data includes at least one of the following: furnace temperature, oxygen content after economizer, and steam drum water level.

12. An electronic device, comprising: processor; And a memory, on which executable code is stored, which, when executed by a processor, performs the method as described in any one of claims 1-11.

13. One or more machine-readable media having executable code stored thereon, which, when executed by a processor, performs the method as described in any one of claims 1-11.

Images