Coal mill operation index determination method and device and storage medium
By training the BP neural network model to replace the XGB model and using counterfactual interpretation optimization, the accuracy and complexity of the nonlinear problem in coal mill operation prediction is solved, and more efficient coal mine processing production efficiency is achieved.
Patent Information
- Application Number
- CN202411064302.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-05
AI Technical Summary
The existing coal mill operation prediction model cannot effectively handle complex nonlinear relationships, resulting in low prediction accuracy and complex optimization processes, consuming a lot of computing resources, and making it difficult to maximize coal mine processing production efficiency.
By training the BP neural network model to replace the XGB model, the counterfactual interpretation is used to optimize the prediction model, simplify the optimization process, improve the prediction accuracy and interpretability of nonlinear problems, and adjust the coal mill process parameters to optimize the coal mine processing process.
It improves the prediction accuracy and interpretability of coal mill operating indicators, simplifies the model optimization process, saves computing resources, and improves the production efficiency of the mineral processing process.
Smart Images

Figure CN118892906A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of mineral processing, and in particular to a method for determining an operating index of a coal mill, a device for determining an operating index of a coal mill, and a computer-readable storage medium. Background Art
[0002] The mineral processing process is a highly complex dynamic system involving multi-process associations, multi-level couplings, and multi-time and space scale characteristics. In order to understand and control the status of these processing operations in real time, a large number of operating indicators need to be monitored. Although advances in sensor technology have enabled most process indicators to be monitored in real time, there are still some important indicators that cannot be obtained online and need to be obtained through model prediction. However, due to the problems of different raw material quality and frequent fluctuations in production conditions in the coal mine processing process, the existing prediction models cannot predict complex nonlinear relationships and are difficult to effectively match actual operating condition changes, resulting in the inability to maximize production efficiency. In addition, the existing prediction models are based on gradient optimization and structure, which makes their optimization process complex and consumes a lot of computing resources.
[0003] In order to overcome the above-mentioned defects of the prior art, the art urgently needs an improved method for determining coal mill operating indicators to improve the prediction accuracy of nonlinear problems and the interpretability of prediction results, and to simplify the optimization process of the model, so as to adjust the coal mill process parameters and optimize the coal mine processing operation process to improve the production efficiency of the mineral processing process. Summary of the invention
[0004] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or decisive elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.
[0005] In order to overcome the above-mentioned defects in the prior art, the present invention provides a method for determining the operating indicators of a coal mill, a device for determining the operating indicators of a coal mill and a computer-readable storage medium, which can replace the XGB model by training a BP neural network model to improve the prediction accuracy of nonlinear problems and the interpretability of prediction results, and optimize and train the prediction model based on counterfactual interpretation to simplify the optimization process of the model, thereby providing a reference for adjusting the process parameters of the coal mill and optimizing the coal mine processing operation process, so as to improve the production efficiency of the mineral processing process.
[0006] Specifically, the method for determining the above-mentioned coal mill operating index provided in accordance with the first aspect of the present invention comprises the following steps: obtaining a plurality of operating parameters of the coal mill. The plurality of operating parameters are selected from at least one of voltage, current, and temperature; and inputting the plurality of operating parameters into a pre-trained first BP neural network model to determine the operating index of the coal mill. The operating index at least includes the unit load of the coal mill. The first BP neural network model is obtained by training based on the training sample data, the predicted value data, and the counterfactual explanation of the expected direction of the operating index of the XGB model.
[0007] Further, in some embodiments of the present invention, the step of training the first BP neural network model includes: obtaining multiple groups of historical operating parameters X of the coal mill. Each group of the historical operating parameters X includes the multiple operating parameters; inputting each group of the historical operating parameters X into the pre-trained XGB model in sequence to obtain the corresponding first prediction value Y′; associating each group of the historical operating parameters X with the corresponding first prediction value Y′ to construct a first training sample data set; constructing an initial BP neural network model to be trained, and training the initial BP neural network model according to the first training sample data set to obtain a second BP neural network model;
[0008] Determine the expected value Y″ of the operating index:
[0009] Y″=Y′+ ε
[0010] Wherein, ε represents the expected direction of the operation index; each group of the historical operation parameters X in the first training sample data set is sequentially input into the second BP neural network model to obtain a second prediction value And using the Adam optimizer, according to the second predicted value The loss between the expected value Y″ The second BP neural network model is optimized to obtain the first BP neural network model.
[0011] Further, in some embodiments of the present invention, the Adam optimizer is used to calculate the second predicted value. The loss between the expected value Y″ The step of optimizing the second BP neural network model to obtain the first BP neural network model includes: The loss between the expected value Y″ Adjusting each group of the historical operating parameters X to obtain a counterfactual explanation X′ of each group of the historical operating parameters X; and optimizing the second BP neural network model according to the counterfactual explanation X′ to obtain the first BP neural network model.
[0012] Further, in some embodiments of the present invention, the step of training the initial BP neural network model according to the first training sample data set to obtain the second BP neural network model includes: inputting each group of the historical operating parameters X in the first training sample data set into the initial BP neural network model in sequence to obtain a third prediction value Y″′; determining a first model loss value of the initial BP neural network model according to the difference between the third prediction value Y″′ and the first prediction value Y′; in response to the first model loss value being greater than a preset first loss threshold, adjusting the parameters of the initial BP neural network model and re-training the initial BP neural network model; and in response to the first model loss value being less than or equal to the first loss threshold, determining that the training of the initial BP neural network model is completed, and obtaining the second BP neural network model.
[0013] Further, in some embodiments of the present invention, the step of training the XGB model includes: obtaining the true value Y of the operating indicator corresponding to each group of the historical operating parameters X, and associating each group of the historical operating parameters X with the corresponding true value Y of the operating indicator to construct a second training sample data set; constructing the XGB model to be trained; and training the XGB model according to the second training sample data set.
[0014] Furthermore, in some embodiments of the present invention, the step of constructing the XGB model to be trained includes: constructing an objective function of the XGB model:
[0015]
[0016] in, is the loss function of the first training sample of the ith order. k ) represents the complexity of the XGB model;
[0017] Perform a second-order Taylor expansion on the loss function to optimize the objective function:
[0018]
[0019] Among them, g i is the first-order derivative of the first training sample in the t-th learner. iis the second-order derivative of the i-th first training sample in the t-th learner; and uses a random grid search method to search for hyperparameters of the XGB model to minimize the objective function.
[0020] Further, in some embodiments of the present invention, the step of training the XGB model according to the second training sample data set includes: preprocessing the second training sample data set and dividing it into a training set and a test set; training the XGB model using the training set, and testing the trained XGB model using the test set to determine a second model loss value of the XGB model; in response to the second model loss value being greater than a preset second loss threshold, redefining the hyperparameters of the XGB model, and re-using the training set to train the XGB model according to the redefined hyperparameters; and in response to the second model loss value being less than or equal to the second loss threshold, determining that the training of the XGB model is completed.
[0021] Furthermore, in some embodiments of the present invention, the step of preprocessing the second training sample data set includes: screening the second training sample data set; and normalizing the second training sample data set:
[0022]
[0023] Wherein, n is the number of first training samples in the second training sample data set, and m is the number of variables in the first training sample.
[0024] Furthermore, in some embodiments of the present invention, the step of screening the second training sample data set includes: calculating the mutual information between each group of the operating parameters X and the corresponding operating indicator true value Y:
[0025]
[0026] Wherein, p(x) is the marginal probability density function of the operating parameter X. p(y) is the marginal probability density function of the operating indicator Y. p(x, y) is the joint probability density function of the operating parameter X and the operating indicator Y; and in response to any of the mutual information being less than a preset first threshold, the corresponding sample is deleted from the second training sample data set.
[0027] In addition, the above-mentioned coal mill operation index determination device provided according to the second aspect of the present invention includes a memory and a controller. The memory stores computer instructions. The controller is connected to the memory and is configured to execute the computer instructions stored thereon to implement the coal mill operation index determination method provided by the first aspect of the present invention.
[0028] In addition, the above-mentioned computer-readable storage medium provided according to the third aspect of the present invention stores computer instructions thereon, and is characterized in that when the computer instructions are executed by the processor, the method for determining the counterfactual interpretation of the coal mill parameters provided by the first aspect of the present invention is implemented. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.
[0030] Figure 1 A schematic diagram of a process for training a first BP neural network model according to some embodiments of the present invention is shown.
[0031] Figure 2 A line graph of historical operating data and counterfactual explanations provided according to some embodiments of the present invention is shown.
[0032] Figure 3 A schematic flow chart of a method for determining coal mill operating indicators provided according to some embodiments of the present invention is shown.
[0033] Figure 4 A result diagram of predicting operating indicators using an XGB model provided according to some embodiments of the present invention is shown.
[0034] Figure 5 A result diagram of predicting operating indicators using a BP neural network model provided according to some embodiments of the present invention is shown. DETAILED DESCRIPTION
[0035] The following specific embodiments illustrate the implementation of the present invention, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this implementation. On the contrary, the purpose of introducing the invention in conjunction with the implementation is to cover other options or modifications that may be extended based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will include many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description.
[0036] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0037] In addition, the terms "upper", "lower", "left", "right", "top", "bottom", "horizontal" and "vertical" used in the following description should be understood as the directions shown in the paragraph and the related drawings. Such relative terms are only used for the convenience of description and do not mean that the device described therein must be manufactured or operated in a specific direction, and therefore should not be understood as limiting the present invention.
[0038] It is understood that although the terms "first", "second", "third", etc. may be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first component, region, layer and / or part discussed below may be referred to as a second component, region, layer and / or part without departing from some embodiments of the present invention.
[0039] As mentioned above, due to the problems of different raw material quality and frequent fluctuations in production conditions in the coal mining process, the existing prediction model cannot predict complex nonlinear relationships and is difficult to effectively match actual working condition changes, resulting in the inability to maximize production efficiency. In addition, the existing prediction model is based on gradient optimization and structure, which makes its optimization process complex and consumes a lot of computing resources.
[0040] In order to overcome the above-mentioned defects in the prior art, the present invention provides a method for determining the operating indicators of a coal mill, a device for determining the operating indicators of a coal mill and a computer-readable storage medium, which can replace the XGB model by training a BP neural network model to improve the prediction accuracy of nonlinear problems and the interpretability of prediction results, and optimize and train the prediction model based on counterfactual interpretation to simplify the optimization process of the model, thereby providing a reference for adjusting the process parameters of the coal mill and optimizing the coal mine processing operation process, so as to improve the production efficiency of the mineral processing process.
[0041] In some non-limiting embodiments, the method for determining the coal mill operating index provided in the first aspect of the present invention can be implemented based on the device for determining the coal mill operating index provided in the second aspect of the present invention. Specifically, the device for determining the coal mill operating index can be configured with a memory and a processor. The memory includes but is not limited to the computer-readable storage medium provided in the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the method for determining the coal mill operating index provided in the first aspect of the present invention.
[0042] Please refer to the following for details Figure 1 . Figure 1 A schematic diagram of a process for training a first BP neural network model according to some embodiments of the present invention is shown.
[0043] like Figure 1 As shown, in the process of training the first BP neural network model, the processor can obtain multiple groups of historical operating parameters X of the coal mill, and input each group of historical operating parameters X into the pre-trained XGB model in sequence to obtain the corresponding first prediction value Y'. Here, each group of historical operating parameters X includes multiple operating parameters.
[0044] Specifically, during the training of the XGB model, the processor may obtain the true value Y of the operating indicator corresponding to each set of historical operating parameters X, and associate each set of historical operating parameters X with the corresponding true value Y of the operating indicator to construct a second training sample data set D={X, Y}.
[0045] After that, the processor can build the XGB model to be trained. Specifically, the processor can first build the objective function of the XGB model:
[0046]
[0047] in, is the loss function of the first training sample of the ith order. k ) represents the complexity of the XGB model.
[0048] Afterwards, the processor can perform a second-order Taylor expansion on the above loss function to optimize the objective function:
[0049]
[0050] Among them, g i is the first-order derivative of the i-th first training sample in the t-th learner. i is the second-order derivative of the i-th first training sample in the t-th learner.
[0051] Please refer to Table 1, which shows a parameter result table of a random grid search method provided according to some embodiments of the present invention.
[0052] Table 1 Parameter results of random grid search method
[0053]
[0054] As shown in Table 1, the processor can use the random grid search method to search for the hyperparameters of the XGB model to minimize the objective function. Since the XGB model contains a large number of parameters, the workload of manual confirmation is huge and prone to errors. The processor can use the random grid search method for the four parameters of the number of weak classifiers "n_estimators" in the XGB model, the sampling ratio "subsample", the strength of L1 regularization in the objective function "reg_alpha", and the strength of L2 regularization in the objective function "reg_lambda", randomly extract parameter subspaces in the parameter space, and search the model in the subspace for the optimal hyperparameters of the XGB model.
[0055] After that, the processor can first screen the second training sample data set. The processor can identify the operating status of the coal mill based on the current index of the coal mill to delete samples in abnormal operating status. In normal operating state, the current range of the coal mill is about 40A to 50A. When the current of the coal mill is lower than zero or obviously abnormal, the processor can determine that the coal mill is in an abnormal operating state. After the above screening operation, the processor can delete 58155 data in the original second training sample data set to 25587 data.
[0056] In addition, in some preferred embodiments, during the screening of the second training sample data set, the processor may further calculate the mutual information between each group of operating parameters X and its corresponding operating indicator true value Y:
[0057]
[0058] Wherein, p(x) is the marginal probability density function of the operating parameter X, p(y) is the marginal probability density function of the operating index Y, and p(x, y) is the joint probability density function of the operating parameter X and the operating index Y. In response to any mutual information being less than a preset first threshold, the processor may delete the corresponding sample from the second training sample data set.
[0059] Afterwards, the processor may normalize the second training sample data set:
[0060]
[0061] Wherein, n is the number of second training samples in the second training sample data set, and m is the number of variables in the second training sample. For example, the second training sample data set after the above screening has 25587 sample data and 30 feature variables, then n=25587, m=30.
[0062] Afterwards, the processor may divide the second training sample data set into a training set and a test set. The processor may use the training set to train the XGB model, and use the test set to test the trained XGB model to determine the second model loss value of the XGB model.
[0063] Thereafter, in response to the second model loss value being greater than a preset second loss threshold, the processor may redefine the hyperparameters of the XGB model, and re-use the training set to train the XGB model according to the redefined hyperparameters.
[0064] Thereafter, in response to the second model loss value being less than or equal to the second loss threshold, the processor may determine that the training of the XGB model is completed.
[0065] Afterwards, the processor may associate each set of historical operating parameters X with the corresponding first predicted value Y′ output by the XGB model to construct a first training sample data set D′={X, Y′}.
[0066] Afterwards, the processor can construct an initial BP neural network model to be trained, and train the initial BP neural network model according to the first training sample data set to obtain a second BP neural network model. Here, the initial BP neural network model consists of an input layer, a hidden layer and an output layer.
[0067] Specifically, in the process of training the initial BP neural network model, the processor can input each group of historical operating parameters X in the first training sample data set D′={X, Y′} into the initial BP neural network model in sequence to obtain a third prediction value Y″′, and then determine the first model loss value of the initial BP neural network model based on the difference between the third prediction value Y″′ and the first prediction value Y′.
[0068] Afterwards, in response to the first model loss value being greater than a preset first loss threshold, the processor can return the error signal along the original path of the output layer, hidden layer, and input layer, adjust the weights, bias and other parameters of the initial BP neural network model, and re-train the initial BP neural network model.
[0069] Thereafter, in response to the first model loss value being less than or equal to the first loss threshold, the processor may determine that the training of the initial BP neural network model is completed, and obtain a second BP neural network model.
[0070] After obtaining the second BP neural network model, the processor can determine the expected value Y″ of the operating indicator according to the expected direction of the operating indicator:
[0071] Y″=Y′+ε
[0072] Among them, ε represents the expected direction of the operating indicator.
[0073] Furthermore, in some embodiments, when the expected direction of the operating indicator is rising, ε>0. When the expected direction of the operating indicator is falling, ε<0.
[0074] Afterwards, the processor may sequentially input each set of historical operating parameters X in the first training sample data set D′={X, Y′} into the second BP neural network model to obtain a second prediction value
[0075] Please refer to Figure 2 . Figure 2 A line graph of historical operating data and counterfactual explanations provided according to some embodiments of the present invention is shown.
[0076] After that, if Figure 2 As shown, the processor can use the Adam optimizer to calculate the second prediction value according to The loss between the expected value Y″ Each group of historical operating parameters X is adjusted to obtain a counterfactual explanation X′ of each group of historical operating parameters X. Afterwards, the processor may optimize the second BP neural network model according to the counterfactual explanation X′ to obtain the first BP neural network model.
[0077] In this way, the processor can use the second prediction value The loss between the expected value Y″ The counterfactual explanations of each group of historical operating parameters are calculated to make the operating indicators change in the expected direction. Therefore, the model is optimized based on the counterfactual explanations without calculating the gradient, so as to simplify the optimization process of the first BP neural network model and save computing resources.
[0078] Please refer to Figure 3 . Figure 3 A schematic flow chart of a method for determining coal mill operating indicators provided according to some embodiments of the present invention is shown.
[0079] like Figure 3As shown, after completing the training and optimization of the first BP neural network model, the processor can obtain multiple operating parameters of the coal mill. Here, the multiple operating parameters are selected from at least one of voltage, current, and temperature. Afterwards, the processor can input the multiple operating parameters into the above-mentioned pre-trained first BP neural network model to determine the operating index of the coal mill. Here, the operating index at least includes the unit load of the coal mill.
[0080] Furthermore, in order to verify the performance of the optimized first BP neural network model and the XGB model, the present invention also performs statistical analysis on their prediction values respectively.
[0081] Please refer to Figure 4 , Figure 5 , Table 2 and Table 3. Figure 4 A result diagram of predicting operating indicators using an XGB model provided according to some embodiments of the present invention is shown. Figure 5 The result diagram of the first BP neural network model prediction operation index provided according to some embodiments of the present invention is shown. Table 2 shows the XGB model performance index table provided according to some embodiments of the present invention. Table 3 shows the first BP neural network model performance index table provided according to some embodiments of the present invention.
[0082] Table 2 XGB model performance index table
[0083] RMSE <![CDATA[R 2 ]]> 0.139 0.993
[0084] Table 3 Performance of the first BP neural network model
[0085] RMSE <![CDATA[R 2 ]]> 0.499 0.906
[0086] like Figure 4 , Figure 5 As shown in Table 2 and Table 3, the XGB model has a high accuracy in predicting the operating indicators, with a root mean square error (RMSE) of 0.139 and a determination coefficient R2 of 0.993. The root mean square error of the first BP neural network model is 0.139, and the determination coefficient R2 is 0.906. It can be seen that the first BP neural network model can improve the prediction accuracy of nonlinear problems and the interpretability of prediction results, and simplify the optimization process of the model while achieving prediction accuracy similar to that of the XGB model, so it is more suitable for data processing and result prediction of nonlinear problems.
[0087] In summary, the method for determining the operating indicators of the coal mill, the device for determining the operating indicators of the coal mill and the computer-readable storage medium provided by the present invention can replace the XGB model by training the BP neural network model to improve the prediction accuracy of nonlinear problems and the interpretability of the prediction results, and optimize and train the prediction model based on counterfactual interpretation to simplify the optimization process of the model, thereby providing a reference for adjusting the process parameters of the coal mill and optimizing the coal mine processing operation process, so as to improve the production efficiency of the mineral processing process.
[0088] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.
[0089] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for determining coal mill operating indicators, characterized in that: The following steps are involved: Acquiring a plurality of operating parameters of the coal mill, wherein the plurality of operating parameters are selected from at least one of voltage, current, and temperature; and The multiple operating parameters are input into a pre-trained first BP neural network model to determine the operating indicators of the coal mill, wherein the operating indicators at least include the unit load of the coal mill, and the first BP neural network model is obtained by training based on the training sample data, predicted value data and counterfactual explanation data of the XGB model corresponding to the expected direction of the operating indicators.
2. The determination method according to claim 1, characterized in that: The steps of training the first BP neural network model include: Acquire multiple groups of historical operating parameters X of the coal mill, wherein each group of the historical operating parameters X includes the multiple operating parameters; Input each group of historical operating parameters X into the pre-trained XGB model in sequence to obtain the corresponding first prediction value Y ′ ; Associating each group of the historical operating parameters X with the corresponding first predicted value Y ′ , to construct a first training sample data set; Constructing an initial BP neural network model to be trained, and training the initial BP neural network model according to the first training sample data set to obtain a second BP neural network model; Determine the expected value Y″ of the operating index: And″=And ′ +ε Wherein, ε represents the expected direction of the operating indicator; Input each group of the historical operating parameters X in the first training sample data set into the second BP neural network model in turn to obtain a second prediction value as well as Using the Adam optimizer, according to the second predicted value The loss between the expected value Y″ The second BP neural network model is optimized to obtain the first BP neural network model.
3. The determination method according to claim 2, characterized in that: The Adam optimizer is used according to the second predicted value The loss between the expected value Y″ The step of optimizing the second BP neural network model to obtain the first BP neural network model includes: According to the second prediction value The loss between the expected value Y″ Adjust each group of historical operating parameters X to obtain the counterfactual explanation X of each group of historical operating parameters X ′ ;as well as Explain X according to the counterfactual ′ , optimize the second BP neural network model to obtain the first BP neural network model.
4. The determination method according to claim 2, characterized in that: The step of training the initial BP neural network model according to the first training sample data set to obtain a second BP neural network model comprises: Inputting each group of the historical operating parameters X in the first training sample data set into the initial BP neural network model in sequence to obtain a third prediction value Y″′; According to the third prediction value Y″′ and the first prediction value Y ′ The difference between the values of and determines the first model loss value of the initial BP neural network model; In response to the first model loss value being greater than a preset first loss threshold, adjusting the parameters of the initial BP neural network model, and retraining the initial BP neural network model; and In response to the first model loss value being less than or equal to the first loss threshold, it is determined that the training of the initial BP neural network model is completed, and the second BP neural network model is obtained.
5. The determination method according to claim 2, characterized in that: The steps of training the XGB model include: Obtaining the true value Y of the operating index corresponding to each group of the historical operating parameters X, and associating each group of the historical operating parameters X with the true value Y of the operating index corresponding thereto, so as to construct a second training sample data set; Build the XGB model to be trained; and The XGB model is trained according to the second training sample data set.
6. The determination method according to claim 5, characterized in that: The steps of constructing the XGB model to be trained include: Construct the objective function of the XGB model: in, is the loss function of the first training sample of the ith order, Ω(f k ) represents the complexity of the XGB model; Perform a second-order Taylor expansion on the loss function to optimize the objective function: Among them, g i is the first-order derivative of the first training sample in the t-th learner, h i is the second-order derivative of the i-th first training sample in the t-th learner; and A random grid search method is used to search for the hyperparameters of the XGB model to minimize the objective function.
7. The classification method according to claim 6, characterized in that: The step of training the XGB model according to the second training sample data set includes: Preprocessing the second training sample data set and dividing it into a training set and a test set; Using the training set to train the XGB model, and using the test set to test the trained XGB model to determine a second model loss value of the XGB model; In response to the second model loss value being greater than a preset second loss threshold, redefine the hyperparameters of the XGB model, and re-use the training set to train the XGB model according to the redefined hyperparameters; and In response to the second model loss value being less than or equal to the second loss threshold, it is determined that the training of the XGB model is completed.
8. The determination method according to claim 7, characterized in that: The step of preprocessing the second training sample data set includes: screening the second training sample data set; and Normalize the second training sample data set: Wherein, n is the number of second training samples in the second training sample data set, and m is the number of variables in the second training samples.
9. The determination method according to claim 8, characterized in that: The step of screening the second training sample data set includes: Calculate the mutual information between each group of operating parameters X and their corresponding operating index true value Y: Wherein, p(x) is the marginal probability density function of the operating parameter X, p(y) is the marginal probability density function of the operating index Y, and p(x, y) is the joint probability density function of the operating parameter X and the operating index Y; and In response to any of the mutual information being less than a preset first threshold, the corresponding sample is deleted from the second training sample data set.
10. A device for determining a coal mill operating index, characterized in that: include: a memory having computer instructions stored thereon; and A controller is connected to the memory and configured to execute computer instructions stored thereon to implement the method for determining the coal mill operation index according to any one of claims 1 to 9.
11. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by the processor, the method for determining the coal mill operating index according to any one of claims 1 to 9 is implemented.
Citation Information
Patent Citations
PMU primary frequency modulation load forecasting method based on LSTM and associative full-connected neural network
CN109242236A
Method, device and system for controlling coal mill, and storage medium
CN110673478A
Device and method for generating a counterfactual data sample for a neural network
CN112541520A
Coal mill startup and shutdown intelligent operation optimization method based on artificial intelligence
CN113843039A
Coal mill system operation monitoring method and device
CN113971363A
