Optimized control method for grate cooler under multiple working conditions

Abstract

The application discloses a kind of grate cooler multi-condition optimization control methods, comprising the following steps: step 1, obtain the working condition data of four kinds of working conditions of grate cooler, and using KNN algorithm to establish working condition identification model to distinguish the working condition data of four kinds of working conditions;Step 2, establish the CARIMA model of grate cooler as the pressure prediction model of grate cooler, and using GPC algorithm as objective function;Step 3, establish the membership function of four kinds of working conditions as multi-condition parameter controller, and according to the membership function, select the control parameter of corresponding working condition;Then make objective function take minimum value, thereby obtain the optimal control amount of the control parameter selected by each kind of working condition corresponding;Step 4, based on the optimal control amount obtained in step 3, control the working of grate cooler.

Description

Technical Field

[0001] This invention relates to the field of grate cooler control methods, specifically a multi-condition optimized control method for grate coolers. Background Technology

[0002] The operational performance of the grate cooler is a crucial factor determining the energy efficiency of the entire cement production process, and accurately grasping the operating conditions is a prerequisite for ensuring optimal cooling performance. Grate coolers experience frequent fluctuations in operating conditions and exhibit nonlinearity during operation, causing them to constantly fluctuate under different conditions. The control requirements for the grate cooler also vary under these different conditions. Traditional operating condition classification and single control methods are insufficient to handle the complex and variable operating conditions of the grate cooler. To address the problems of frequent operating condition fluctuations and the difficulty in achieving control requirements under different conditions, a method of operating condition classification and multi-condition control is proposed. Summary of the Invention

[0003] This invention provides a multi-condition optimization control method for grate coolers based on an adaptive neighbor weight KNN operating condition identification model. The aim is to accurately classify different operating conditions of the grate cooler and adopt different control parameters according to different operating conditions to achieve better control results, thereby solving the problems of frequent operating condition fluctuations and difficulty in achieving control requirements under different operating conditions in existing grate cooler technologies.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] The multi-condition optimized control method for grate coolers includes the following steps:

[0006] Step 1: Obtain operating data for four operating conditions of the grate cooler: grate pressure drop, grate pressure stability, grate pressure rise, and grate pressure fluctuation. After preprocessing the operating data, establish an operating condition identification model based on the operating data using the KNN algorithm with adaptive neighbor weights. Output the operating condition label corresponding to the operating data for each operating condition through the operating condition identification model, thereby distinguishing the operating data of the four operating conditions.

[0007] Step 2: Establish the CARIMA model of the grate cooler as the pressure prediction model of the grate cooler, and use the GPC algorithm as the control algorithm to establish the objective function of the pressure prediction model of the grate cooler.

[0008] Step 3: Based on the grate pressure and grate pressure critical value working condition data corresponding to each working condition, establish the membership function of the four working conditions as a multi-working-condition parameter controller, and select the control parameters corresponding to the working conditions according to the membership function; then make the objective function established in Step 2 take the minimum value, thereby obtaining the optimal control quantity of the selected control parameters for each working condition.

[0009] Step 4: Control the operation of the grate cooler based on the optimal control quantity obtained in Step 3.

[0010] In the further step 1, preprocessing includes outlier identification and removal, filtering, and normalization.

[0011] Further step 1 uses The principle is to identify and remove abnormal data.

[0012] In the further step 1, a moving average filtering method is used for filtering.

[0013] In further step 1, step 4 also includes: inputting the optimal control quantity obtained in step 3 into the grate cooler pressure prediction model established in step 2, and using the difference between the actual value of the grate pressure output by the grate cooler during operation and the predicted value of the grate pressure output by the grate cooler pressure prediction model to adjust the reference trajectory of the desired output in the objective function in order to optimize the control effect.

[0014] Compared with the prior art, the advantages of the present invention are:

[0015] 1. Compared with conventional working condition classification algorithms, this invention introduces the KNN classification method with adaptive neighbor weights, which can accurately capture the relative importance between neighbor samples. It has significant effects on the accuracy and running time of working condition classification, and can quickly and accurately classify the complex and ever-changing working conditions of the grate cooler.

[0016] 2. This invention addresses the problem of frequent fluctuations in the operating conditions of grate coolers by proposing a multi-condition parameter controller with an adaptive operating condition strategy. This controller enables parameter switching between different operating conditions, thereby achieving the control objective of maintaining the grate pressure within a reasonable range under various operating conditions of the grate cooler, in conjunction with different control parameters of the GPC algorithm. Attached Figure Description

[0017] Figure 1 This is the membership curve in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the multi-condition controller in an embodiment of the present invention.

[0019] Figure 3 This is a flowchart of a multi-condition controller based on a condition identification model in an embodiment of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] This embodiment discloses a multi-condition optimized control method for a grate cooler, including the following steps:

[0022] Step 1: Obtain operating data for the grate cooler under four operating conditions: pressure drop under the grate, stable pressure under the grate, pressure rise under the grate, and pressure fluctuation under the grate. Then, preprocess the obtained operating data as follows:

[0023] First, adopt The principle is to identify and remove outliers in the operating data.

[0024] Then, the moving average filtering method is used to filter the operating data after removing outliers, as shown in formula (1):

[0025] (1)

[0026] In formula (1): for Time-filtered operating condition data; for Operating data before filtering at any given time; Indicates the window size.

[0027] Finally, in order not to lose the characteristics of the original data, the filtered operating condition data is normalized, as shown in formula (2):

[0028] (2)

[0029] In formula (2): The filtered value. The data is after normalization. and They are respectively The maximum and minimum values ​​in the range.

[0030] After preprocessing, an adaptive neighbor weighted KNN algorithm is used to establish a working condition identification model based on the working condition data. The process is as follows:

[0031] Step S1: Select m sample points from the preprocessed operating condition data to form a sample set. The first sample in sample set D sample points for 3D feature vectors, and sample points Contains The eigenvalues ​​are denoted as follows: ;

[0032] Step S2: Calculate the Euclidean distance between each sample point and its neighboring sample points. The calculation formula is shown in formula (3):

[0033] (3)

[0034] Formula (3): For the first sample points The One eigenvalue; For the first sample points The One eigenvalue;

[0035] Step S3: Considering the existence of abnormal situations such as noise or disturbance in the working condition identification, based on the Euclidean distance obtained by formula (3), the Gaussian kernel function is used to calculate the weight of each sample point and its neighbors. The calculation formula is shown in formula (4):

[0036] (4)

[0037] In formula (4): Indicates the sample point and the first The weight of each neighbor; Indicates the sample point and the first Euclidean distance between neighbors; The variance of the Gaussian function is used to adjust the degree of weight decay.

[0038] Step S4: Based on the weights calculated by formula (4), calculate the probability of each sample point belonging to each working condition. The calculation formula is shown in formula (5):

[0039] (5),

[0040] In formula (5): K represents the number of neighbors; P represents the weight of the sample point relative to the l-th working condition; l Let l represent the probability that a sample point belongs to the l-th working condition. Four typical working conditions are selected for training the working condition recognition model, and l = 1, 2, 3, 4.

[0041] Let the working condition labels of four typical working conditions of the grate cooler be L1, L2, L3, and L4, respectively: grate pressure drop, grate pressure stability, grate pressure rise, and grate pressure fluctuation. Then, grate speed, kiln current, raw material feed rate, and grate pressure are selected as sample feature points. Based on the probability calculated by formula (5), the working condition with the highest probability is taken as the working condition to which the sample point belongs. The sample points are divided according to the working conditions to which they belong and the corresponding working condition labels are output. Thus, the working condition identification model is obtained.

[0042] Step S5: After inputting the working condition sample points into the working condition identification model, the working condition identification model outputs the working condition label corresponding to the working condition data of each working condition, thereby distinguishing the working condition data of the four working conditions. The working condition label is used for subsequent control parameter switching tasks.

[0043] In this embodiment, the operating condition labels are randomly shuffled and then divided into training and testing sets in a 7:3 ratio. For the operating condition identification scenario of the grate cooler, grate speed, kiln current, raw material feed rate, and grate pressure are selected as inputs to the operating condition identification model, and the operating condition label of grate pressure is the model output. The KNN algorithm of this embodiment and other different classification algorithms are selected. In this embodiment, the number of nearest neighbors K in the KNN algorithm is set to 50, while the other algorithms use default parameters. Finally, the performance of each algorithm is evaluated by accuracy and running time, and the results are shown in Table 1.

[0044] Table 1. Performance Comparison of Different Classification Algorithms for Identifying Working Conditions

[0045] Classification Algorithm accuracy time Bayes 94.7% 0.037s SVM 97.9% 2.806s KNN 98.6% 0.099s

[0046] As shown in Table 1, the accuracy and running time of the KNN algorithm in this embodiment are relatively balanced, which can meet the various requirements for grate cooler condition identification.

[0047] Step 2: Establish the CARIMA model of the grate cooler as the pressure prediction model, and use the GPC algorithm as the control algorithm to establish the objective function of the grate cooler pressure prediction model. The process is as follows:

[0048] Step (2.1): The CARIMA model is used as the pressure prediction model for the grate cooler. The pressure prediction model for the grate cooler is shown in formula (6):

[0049] (6)

[0050] In formula (6):

[0051] This is a delayed operation for the autoregressive part. ;

[0052] For the delayed operation of the moving average portion, ;

[0053] Delayed operation of the differential part, .

[0054] In the above formula: The control input at time t-1 Let t be the pressure under the grate of the grate cooler; Sampling time point; It is a lag operator.

[0055] The delayed operation in the autoregressive part Delayed operation of moving average portion Delayed operation of the sum and difference part middle, , and These are the parameters of the CARIMA model, which can all be identified by the least squares method according to the control requirements.

[0056] They are respectively , and The highest order; It is a difference operator; A noise sequence with zero mean.

[0057] Considering that the complexity of the model would increase the computational difficulty of the actual control process, this embodiment sets... Based on the above conditions, the pressure prediction model of the grate cooler in input-output form can be obtained from formula (6) as shown in formula (7):

[0058] (7).

[0059] Step (2.2): When facing systems with large lags and models with significant biases, the GPC algorithm can optimize using historical data over a period of time, minimizing the variance between the predicted value and the desired output trajectory, and exhibiting better robustness. Therefore, this embodiment uses the GPC algorithm as the control algorithm to establish the objective function of the grate cooler pressure prediction model. The GPC algorithm in... The objective function J(t) at time t is shown in equation (8):

[0060] (8)

[0061] In formula (8): and , yes The expected output reference trajectory at any given time. yes The expected output reference trajectory at time -1 Indicates the predicted length. As a softening factor and In actual control, a softening factor needs to be added to the desired output reference trajectory to balance the relationship between the system's speed and robustness.

[0062] The setpoint for the grate pressure at time t; and The number of steps to start and end the prediction; Weighting coefficients for control; Minimum number of predictive control steps and E{} indicates that the GPC algorithm in Expected value at any given moment.

[0063] The weighting coefficients in formula (8) Its main function is to reduce large fluctuations in the control quantity, thereby reducing frequent fluctuations in the grate pressure caused by large algorithm adjustments, or situations where it exceeds its reasonable range. If the weighting coefficients... Increasing the control increment decreases the control increment and reduces the output response speed, but improves system stability. Conversely, decreasing the control increment increases the control response speed but decreases system stability. Therefore, it is necessary to set the weighting coefficients according to the actual conditions of the grate cooler under different operating conditions. In this embodiment, four typical operating conditions are used to optimize the control of the grate cooler. Therefore, the weighting coefficient is set to... .

[0064] Step (2.3), as shown in formula (7), the algorithm contains multiple unknown values ​​from historical moments during its operation, leading to repeated calculations. To skip the calculation of unknown values ​​from historical moments, the algorithm... Only with This embodiment is based on Define two parameters based on the prediction length. Deterministic polynomial and The specific expressions are shown in formulas (9) and (10):

[0065] (9)

[0066] (10)

[0067] In formulas (9) and (10): and All by The relationship between them is shown in formula (11):

[0068] (11)

[0069] In formula (11): ,and Then, the Diophantine equation is introduced as shown in formula (12):

[0070] (12)

[0071] Combining formulas (6) and (12), the predicted length can be obtained as follows: Output at time t+j The prediction expression is shown in formula (13):

[0072] (13)

[0073] Step 3: Based on the grate pressure and critical grate pressure data corresponding to each working condition, establish membership functions for the four working conditions as a multi-condition parameter controller. Select the control parameters corresponding to the working condition based on the membership functions. Then, minimize the objective function established in Step 2 to obtain the optimal control quantity for the selected control parameters corresponding to each working condition, such as... Figure 2 As shown, the specific process is as follows:

[0074] Step (3.1): Based on the different characteristics of the four typical working conditions, set the control parameters corresponding to different working conditions, including: prediction step size. Control step size Weighting coefficients for different working conditions and softening factor Then, based on the selected four typical working conditions and their labels L1, L2, L3, and L4, membership functions are used to switch control parameters between different working conditions. The membership curves for each working condition are shown below. Figure 1 As shown, the membership functions for various working conditions are given by formulas (14), (15), (16), (17), and (18):

[0075] (14)

[0076] (15)

[0077] (16)

[0078] (17)

[0079] (18)

[0080] In formulas (14), (15), (16), (17), (18): The critical values ​​of grate pressure for L1 and L4 operating conditions; The critical pressure values ​​under the left grate for L2 and L4 operating conditions; The critical pressure values ​​under the right grate for L2 and L4 operating conditions; The critical values ​​of grate pressure for L4 and L3 operating conditions; It is the membership function of the L1 working condition; and These are the membership functions of the L4 conditions located to the left and right of the L2 condition, respectively; It is the membership function of the L2 working condition; It is the membership function for the L3 working condition; This indicates the pressure under the comb.

[0081] After the working condition data is input into the working condition identification model established in step 1, the working condition identification model outputs the corresponding working condition label. The working condition label will correspond to the membership function of the corresponding working condition. The control parameters of the corresponding working condition are selected through the membership function.

[0082] Step (3.2), Definition , can be obtained The expression is shown in formula (19):

[0083] (19)

[0084] In formula (19), coefficient can be and The common representation, and the specific relationship are shown in formula (20):

[0085] (20)

[0086] Step (3.3): When the objective function shown in formula (8) is minimized, the optimal control quantity corresponding to each typical working condition can be obtained. As shown in formula (21):

[0087] (twenty one)

[0088] According to formula (21), the incremental form of the optimal control quantity can be obtained as shown in formula (22):

[0089] (twenty two)

[0090] In formula (21) and formula (21): ,in express The expected output reference trajectory at any given time; ,in:

[0091] ;

[0092] ;

[0093] ;

[0094] and middle This corresponds to four different operating conditions of the grate cooler; Represents the identity matrix.

[0095] Step (3.4) After inputting the operating condition data at any given time into the operating condition identification model in step 1, the corresponding operating condition label is output. Then, according to the membership function given in step (3.1), the control parameter switching task between different operating conditions is completed. Finally, through steps (3.2) and (3.3), the operating condition of the grate cooler is obtained. The optimal control quantity at time t is shown in formula (23):

[0096] (twenty three)

[0097] In formula (23): The control quantity at the previous moment, and when Initial control quantity at time .

[0098] Step 4: Control the operation of the grate cooler based on the optimal control quantity obtained in Step 3, and input the optimal control quantity obtained in Step 3 into the grate cooler pressure prediction model established in Step 2. Based on the difference between the actual value of the grate pressure output by the grate cooler during operation and the predicted value of the grate pressure output by the grate cooler pressure prediction model, the reference trajectory of the desired output in the objective function is adjusted to optimize the control effect.

[0099] In this embodiment, the multi-condition controller for the grate cooler has a switch value as an enable / disable position. When the enable / disable position is 1, steps (3.3) and (3.4) are repeated; when the enable / disable position is 0, the optimized control of the grate cooler ends. Thus, the optimized control of the grate cooler is achieved using a multi-condition controller based on the KNN condition recognition model. The flowchart is as follows: Figure 3 As shown.

[0100] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. These embodiments are merely descriptions of preferred embodiments and are not intended to limit the scope or concept of the invention. The specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. Such combinations, as long as they do not violate the spirit of the present invention, should also be considered as part of this disclosure. To avoid unnecessary repetition, the present invention will not further describe the various possible combinations.

[0101] This invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this invention and without departing from the design idea of ​​this invention, all modifications and improvements made by those skilled in the art to the technical solutions of this invention should fall within the protection scope of this invention. The technical content for which protection is sought in this invention has been fully described in the claims.

Claims

1. A method for optimizing control of a grate cooler in multiple operating conditions, characterized in that, Includes the following steps: Step 1: Obtain operating data for four operating conditions of the grate cooler: grate pressure drop, grate pressure stability, grate pressure rise, and grate pressure fluctuation. After preprocessing the operating data, establish an operating condition identification model based on the operating data using the KNN algorithm with adaptive neighbor weights. Output the operating condition label corresponding to the operating data for each operating condition through the operating condition identification model, thereby distinguishing the operating data of the four operating conditions. Step 2: Establish the CARIMA model of the grate cooler as the pressure prediction model of the grate cooler, and use the GPC algorithm as the control algorithm to establish the objective function of the pressure prediction model of the grate cooler. Step 3: Based on the grate pressure and grate pressure critical value working condition data corresponding to each working condition, establish the membership function of the four working conditions as a multi-working-condition parameter controller, and select the control parameters corresponding to the working conditions according to the membership function; then make the objective function established in Step 2 take the minimum value, thereby obtaining the optimal control quantity of the selected control parameters for each working condition. Step 4: Control the operation of the grate cooler based on the optimal control quantity obtained in Step 3.

2. The multi-condition optimized control method of the grate cooler according to claim 1, characterized in that, In step 1, preprocessing includes outlier identification and removal, filtering, and normalization.

3. The multi-condition optimization control method for a grate cooler according to claim 2, characterized in that, In step 1, the 3σ principle is used to identify and remove abnormal data.

4. The multi-condition optimization control method for a grate cooler according to claim 2, characterized in that, In step 1, a moving average filtering method is used for filtering.

5. The multi-condition optimization control method for a grate cooler according to claim 1, characterized in that, Step 4 also includes: inputting the optimal control quantity obtained in step 3 into the grate cooler pressure prediction model established in step 2, and using the difference between the actual value of the grate pressure output by the grate cooler during operation and the predicted value of the grate pressure output by the grate cooler pressure prediction model to adjust the reference trajectory of the desired output in the objective function in order to optimize the control effect.

Images