A boiler steam temperature prediction method and device

Abstract

The present application provides a kind of boiler steam temperature prediction method and device, including determining target operating parameter related to target variable steam temperature from the power generation boiler operating parameter collected from boiler generator unit;With target operating parameter as the input of steam temperature prediction model, target operating parameter is processed based on steam temperature prediction model, and the predicted steam temperature value is output, and steam temperature prediction model is obtained based on historical power generation boiler operating parameter set training.In the present application, the steam temperature prediction model based on historical power generation boiler operating parameter set is trained to predict target operating parameter, and the steam temperature value at future time can be predicted, so as to realize the prediction of steam temperature value.Through the above-mentioned mode, not only the service life of boiler system can be improved, but also the reliability and safety of boiler system operation can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power generation process, and particularly relates to a boiler steam temperature prediction method and device. BACKGROUND

[0002] At present, the main steam and the reheated steam are generated by heating in the superheater and reheater in the power generation boiler system to drive the steam turbine to work and carry out power generation operation. Since the main steam and the reheated steam temperature represents the working operation condition of the superheater and the reheater, the overall operation state of the power generation boiler system can be determined by the main steam and the reheated steam temperature.

[0003] The steam temperature reflecting the operation state of the boiler system in the power generation operation can be collected in real time by the temperature sensor in the system. Since the temperature sensor cannot predict the change of the steam temperature in the future period, and the superheater and the reheater structural members are mostly made of low-grade steel, when the boiler system works at the steam temperature exceeding the design temperature of the superheater and the reheater, the service life of the superheater and the reheater pipeline is reduced. When the steam temperature is too high or too low, the thermal strain of the steam turbine in the boiler system is increased, the creep speed is accelerated, and the fastener is loosened, thereby reducing the service life of the boiler system and easily reducing the reliability and safety of the operation of the boiler system. SUMMARY

[0004] Therefore, the embodiment of the present application provides a boiler steam temperature prediction method and device to solve the problems of reducing the service life of the boiler system and easily reducing the reliability and safety of the operation of the boiler system in the prior art.

[0005] To achieve the above object, the embodiment of the present application provides the following technical scheme:

[0006] The first aspect of the embodiment of the present application shows a boiler steam temperature prediction method, which comprises:

[0007] determining a target operation parameter related to the target variable steam temperature from the power generation boiler operation parameters collected by the boiler generator unit;

[0008] taking the target operation parameter as the input of the steam temperature prediction model, processing the target operation parameter based on the steam temperature prediction model, and outputting the predicted steam temperature value, wherein the steam temperature prediction model is obtained by training based on the historical power generation boiler operation parameter set.

[0009] Optionally, the determination of the target operation parameter related to the target variable steam temperature from the power generation boiler operation parameters collected by the boiler generator unit comprises:

[0010] Determine the correlation degree between the operation parameters collected by the boiler generator set and the steam temperature of the power generation boiler respectively.

[0011] According to the correlation degree, select the target operation parameter related to the target variable steam temperature.

[0012] Optionally, the process of obtaining the steam temperature prediction model based on the historical power generation boiler operation parameter set includes:

[0013] Obtain the historical power generation boiler operation parameter set collected by the boiler generator set;

[0014] Perform time series transformation on the historical power generation boiler operation parameter set to obtain a historical time series set;

[0015] Divide the historical time series set into a training set and a validation set based on a preset proportion;

[0016] Determine an initial LSTM regression model, and train the initial LSTM regression model based on the training set and the validation set to determine the LSTM regression model obtained by the current training as the steam temperature prediction model.

[0017] Optionally, the training of the initial LSTM regression model based on the training set and the validation set to determine the LSTM regression model obtained by the current training as the steam temperature prediction model includes:

[0018] Using cross-validation method, train the initial LSTM regression model on the training set to obtain a trained initial LSTM regression model;

[0019] Obtain optimal model parameters by fitting the trained initial LSTM regression model to the steam temperature of the power generation boiler on the validation set;

[0020] Perform performance evaluation on the initial steam temperature prediction model constructed based on the optimal model parameters;

[0021] When it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters meets the preset condition, determine that the initial steam temperature prediction model is the final steam temperature prediction model.

[0022] Optionally, it further includes:

[0023] When it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters does not meet the preset condition, continue to train the initial steam temperature prediction model based on the training set and the validation set.

[0024] Optionally, it further includes:

[0025] draw a prediction curve based on the predicted steam temperature value, and display.

[0026] The second aspect of the embodiment of the present application shows a boiler steam temperature prediction device, which comprises:

[0027] A determination unit is configured to determine target operation parameters related to the target variable steam temperature from the operation parameters of the power generation boiler collected by the boiler power generation unit.

[0028] A steam temperature prediction model is configured to take the target operation parameters as inputs of the steam temperature prediction model, process the target operation parameters based on the steam temperature prediction model, and output predicted steam temperature values, wherein the steam temperature prediction model is constructed by the construction unit.

[0029] Optionally, the determination unit is specifically configured to determine the correlation degrees of the operation parameters of the power generation boiler collected by the boiler power generation unit with the steam temperature of the power generation boiler respectively, and select the target operation parameters related to the target variable steam temperature according to the correlation degrees.

[0030] Optionally, the construction unit comprises:

[0031] An acquisition subunit is configured to acquire a set of historical operation parameters of the power generation boiler collected by the boiler power generation unit.

[0032] A transformation subunit is configured to perform time sequence transformation on the set of historical operation parameters of the power generation boiler to obtain a historical time sequence set.

[0033] A division subunit is configured to divide the historical time sequence set into a training set and a verification set based on a preset proportion.

[0034] A training subunit is configured to determine an initial LSTM network model, train the initial LSTM regression model based on the training set and the verification set, and determine the LSTM regression model obtained through the current training as the steam temperature prediction model.

[0035] Optionally, the device further comprises:

[0036] A drawing unit is configured to draw a prediction curve based on the predicted steam temperature value, and display.

[0037] Based on the above boiler steam temperature prediction method and device provided by the embodiment of the present application, the method comprises: determining the target operation parameter related to the target variable steam temperature from the power generation boiler operation parameters collected by the boiler generator set; taking the target operation parameter as the input of the steam temperature prediction model, processing the target operation parameter based on the steam temperature prediction model, and outputting the predicted steam temperature value, wherein the steam temperature prediction model is obtained by training based on the historical power generation boiler operation parameter set. In the embodiment of the present application, the steam temperature prediction model trained based on the historical power generation boiler operation parameter set is used to predict the target operation parameter, so that the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. Through the above-mentioned mode, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know the change of the system steam temperature in advance, so as to reasonably and timely deal with the possible steam temperature abnormality, so as to ensure that the boiler power generation operation is reliable and safe, that is, to improve the reliability and safety of the boiler system operation. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0039] Figure 1 The flowchart of the boiler steam temperature prediction method provided by the embodiment of the present application is shown in the figure;

[0040] Figure 2 The flowchart of training the steam temperature prediction model provided by the embodiment of the present application is shown in the figure;

[0041] Figure 3 The schematic diagram of constructing the time series data set provided by the embodiment of the present application is shown in the figure;

[0042] Figure 4 The schematic diagram of setting the cross-validation data set provided by the embodiment of the present application is shown in the figure;

[0043] Figure 5 The flowchart of another boiler steam temperature prediction method provided by the embodiment of the present application is shown in the figure;

[0044] Figure 6 The schematic diagram of drawing the predicted steam temperature and the real steam curve respectively provided by the embodiment of the present application is shown in the figure;

[0045] Figure 7A schematic diagram of an overlapping drawing prediction steam temperature and a real steam curve provided by an embodiment of the present application is shown in the figure;

[0046] Figure 8 A structural schematic diagram of a boiler steam temperature prediction device provided by an embodiment of the present application is shown in the figure;

[0047] Figure 9 A structural schematic diagram of another boiler steam temperature prediction device provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0049] In the present application, the terms "comprise", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the sentence "comprises a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0050] In the embodiments of the present application, the steam temperature prediction model trained based on the historical power generation boiler operation parameter set is used to predict the target operation parameter, so that the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. Through the above-mentioned manner, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know in advance the change of the system steam temperature in the power generation operation, so as to reasonably and timely dispose the possible steam temperature abnormality, so as to ensure that the power generation operation of the boiler is reliable and safe, that is, to improve the reliability and safety of the operation of the boiler system.

[0051] Referring to Figure 1 A flowchart of a boiler steam temperature prediction method provided by an embodiment of the present application is shown in the figure, and the method comprises:

[0052] Step S101: determining the target operation parameter related to the target variable steam temperature from the power generation boiler operation parameters collected from the boiler generator set.

[0053] Optionally, at each data collection time point, the values (pressure, temperature, etc.) of all sensors of the boiler generator set are read by the programmable logic controller (PLC) to form a data sample, i.e., a power generation boiler operating parameter.

[0054] In the embodiment of the present application, since there are many sensors arranged in the boiler generator set, and the collected unit operating state parameters are also extremely diverse. In the process of implementing step S101, the characteristic variables significantly associated with the boiler steam temperature change are screened out by using the distributed gradient boosting library (XGBoost) algorithm, so as to ensure the fitting effect of the regression prediction model and reduce the time consumption of the prediction model training.

[0055] It should be noted that not all sensor values are used, so data selection is required.

[0056] Step S102: Taking the target operating parameter as an input of the steam temperature prediction model, processing the target operating parameter based on the steam temperature prediction model, and outputting a predicted steam temperature value.

[0057] In step S102, the steam temperature prediction model is obtained by training based on a historical power generation boiler operating parameter set.

[0058] It should be noted that the process of obtaining the steam temperature prediction model based on the historical power generation boiler operating parameter set includes the following steps, as shown in Figure 2

[0059] Step S11: Obtaining a historical power generation boiler operating parameter set collected by a boiler generator set.

[0060] In the process of implementing step S11, the historical power generation boiler operating parameter set collected by each boiler sensor of the boiler generator set in a preset historical time period is obtained.

[0061] It should be noted that the characteristic variables of the historical power generation boiler operating parameter set correspond to the steam temperature value at the current sampling time point.

[0062] The preset historical time period is set according to multiple experiences, for example, it can be set to the past year.

[0063] Step S12: Time sequence transformation is performed on the historical power generation boiler operating parameter set to obtain a transformed historical power generation boiler operating parameter set.

[0064] The specific content of S12 is as follows: first, the historical target operating parameter significantly associated with the steam temperature change of the power generation boiler is screened out by using the XGBoost algorithm; the vector of the first target operating parameter in the historical target operating parameter is taken as the input of the XGBoost algorithm, and the vector of the second target operating parameter in the historical target operating parameter is taken as the output of the XGBoost algorithm. ​The system replicates and expands n times along the horizontal direction, with time steps of t0, t1, ..., t2. n The feature sequence; the second to the last vector, shift all its feature attribute columns up by 1 to n rows; add the steam temperature column (Label) of the power generation boiler, and shift this column up by m (m>n) rows; delete the last m rows of the constructed time series set to obtain the transformed historical time series set.

[0065] It should be noted that since the shift operation caused blank markers to appear in the last m rows of the time series set, the last m rows of the entire time series set need to be deleted.

[0066] For example: Figure 3 As shown, the original feature vector is arranged according to the sample indices 0, 1, 2, ..., N-2, N-1, N, corresponding to the features a, e, i, ..., m, q, u. The vector of the first target operating parameter in the original feature vector is then used. The system replicates and expands n times along the horizontal direction, with time steps of t0, t1, ..., t2. n The feature sequence; the second to the last vector, shift all its feature attribute columns up by 1 to n rows; add the steam temperature column (Label) of the power generation boiler, and shift this column up by m (m>n) rows; at this time, the blank nan tags that appear in rows N-1 and N need to be deleted. The last two rows containing nan in the constructed time series set are then deleted to obtain the transformed historical time series set.

[0067] Step S13: Divide the historical time series set into a training set and a validation set based on a preset ratio.

[0068] It should be noted that the preset ratio can be set to 4:1, and this can be set according to the actual situation. This application does not impose any restrictions on this.

[0069] The training set is used to train the regression model, and the validation set is used to optimize the model's fit.

[0070] In this embodiment of the invention, to avoid overfitting in the trained regression model, cross-validation is used to divide the historical time series set, i.e., the historical validation set, into N parts. One part of the historical time series set is used as validation data, and the other N-1 parts are used as training data for model validation.

[0071] For example, when N is equal to 5, the validation set can be divided into 5 parts, including A part, B part, C part, D part and E part. E part is used as the validation data, and B part, C part, D part and A part are used as the training data for model validation, so as to ensure that the ratio of the training set to the validation set is 4:1. From the first round to the fifth round of training, the validation set data of E, D, C, B and A parts are sequentially used as the validation data Validation Set, as shown in the following table. Figure 4

[0072] Step S14: determining an initial LSTM regression model, and training the initial LSTM regression model based on the training set and the validation set, and determining the LSTM regression model obtained by the current training as a steam temperature prediction model.

[0073] Optionally, based on the pre-construction process of the prediction model shown above, in the process of performing step S14 to train the initial LSTM regression model based on the training set and the validation set, and determining the LSTM regression model obtained by the current training as a steam temperature prediction model, the following steps are included:

[0074] Step S21: using the cross-validation method, training the initial LSTM regression model by using the training set, and obtaining the trained initial LSTM regression model.

[0075] It should be noted that the LSTM loop structure is composed of an input gate, a forget gate and an output gate.

[0076] The initial LSTM regression model is composed of two consecutive loop structures (LSTM cells), and each loop structure will pass the state of the current unit to the next loop structure.

[0077] Among them, the input gate determines how much time series data corresponding to the current loop unit is stored as the state of the current unit; the forget gate determines how much state corresponding to the previous loop unit is introduced into the current loop unit; and the output gate controls how much state of the current loop unit is transmitted to the next loop unit.

[0078] In the process of implementing step S21, the number of training sets is used as the input node number of the input gate of the initial LSTM regression model. The training set input from the input layer at each time will first pass through the input gate, and the switch of the input gate will determine whether information will be input to the storage unit at this time. The data in the storage unit at each time will undergo a process of whether to be forgotten. If it is forgotten, the value in the storage unit will be cleared, that is, forgotten; finally, the output gate determines whether data is output from the storage unit at each time.

[0079] In the embodiment of the application, when training the training set, first, the output h of the previous loop unit is input to the input gate of the current loop unit, and the output of the input gate is the input of the forget gate. The output of the forget gate is the input of the output gate. The output of the output gate is the output of the current loop unit.​t-1 and the timing data of the current unit, i.e., the training set input x t , the input gate (Sigmoid function, σ) obtains i t , the hyperbolic tangent function tanh obtains the temporary state of the current unit

[0080] Specifically, the output h t-1 of the previous cycle unit and the timing data of the current unit, i.e., the training set input x t , are substituted into formula (1) to obtain the input gate output i t ; the output h t-1 of the previous cycle unit and the timing data of the current unit input x t are substituted into formula (2) to obtain the temporary state of the current unit

[0081] Formula (1):

[0082] i t = σ (ω i · [h t-1 , x t ] + b i ) (1)

[0083] wherein h t-1 is the output of the previous cycle unit, x t is the timing data of the current unit, i.e., the training set input, σ is the constant value corresponding to the input gate Sigmoid function, ω i is the weight coefficient of the input gate, and b i is the bias of the input gate.

[0084] It should be noted that ω i and b i are coefficients set according to actual conditions.

[0085] Formula (2)

[0086]

[0087] wherein h t-1 is the output of the previous cycle unit, x t is the timing data of the current unit, i.e., the training set input, ω c is the weight coefficient used when calculating the temporary state, and b C is the bias used when calculating the temporary state.

[0088] Then, the output h t-1 of the previous cycle unit and the timing data of the current unit input x t obtain the forgetting gate output ft .

[0089] Specifically, the output h t-1 of the previous cycle unit and the time series data input x t of the current unit are input into formula (3) to obtain the forget gate output f t .

[0090] Formula (3):

[0091] f t = σ (ω f · [h t-1 , x t ] + b f ) (3)

[0092] wherein h t-1 is the output of the previous cycle unit, x t is the time series data of the current unit, i.e., the input of the training set, ω f is the weight coefficient of the forget gate, and b f is the bias of the forget gate.

[0093] Then, the forget gate output f t , the input gate output i t and the temporary state C of the current unit obtained by calculating the above formulas (1), (2) and (3), and the state C t-1 of the previous cycle unit are substituted into formula (4) to obtain the device C t of the current unit.

[0094]

[0095] wherein f t is the forget gate output, i t is the input gate output, C t-1 is the state of the previous cycle unit, and C t-1 is the temporary state of the current unit.

[0096] Continuing, the output h t of the previous cycle unit and the time series data input x t of the current unit are input into formula (5) to obtain the output gate output o t , and the final output h t of the current unit is obtained by combining the state C t-1 of the current unit.

[0097] Specifically, the output h t of the previous cycle unit and the time series data input x t of the current unit are substituted into formula (5) to determine the output gate output oand the current cell state C t The input formula (6) is calculated to obtain the current cell output h t .

[0098] Formula (5):

[0099] o t = σ (ω o · [h t-1 , x t ] + b o ) (5)

[0100] where ω o and b o are the weight coefficient and bias of the output gate, h t-1 is the output of the previous cycle cell, and x t is the time series data of the current cell.

[0101] Formula (6):

[0102] h t = o t *tanh (C t ) (6)

[0103] where o t is the output of the output gate, and tanh () is the hyperbolic tangent function.

[0104] Finally, the initial LSTM regression model can be constructed according to the current cell output h t .

[0105] Optionally, L2 square error is used as the loss function of the LSTM regression model.

[0106] In implementation, the loss function D L2 can be calculated by substituting the estimated value of the data sample by the LSTM regression model and the true value corresponding to the data sample into formula (7).

[0107] Formula (7):

[0108]

[0109] where, is the estimated value of a data sample by the LSTM regression model, y i is the true value corresponding to the data sample, and D L2 is the loss function that calculates the sum of the square deviations of the predicted values and the true values of all p data samples by the LSTM regression model.

[0110] It should be noted that when the weight coefficient and bias in the above loop structure are valued to make D L2The optimal LSTM neural network model is determined when the loss function takes the minimum value.

[0111] Step S22: Using cross-validation method, the fitting degree of the trained initial LSTM regression model to the steam temperature of the power boiler is determined on the validation set to obtain the optimal model parameters.

[0112] In the process of implementing step S22, the LSTM regression model is trained based on the training data in the validation set, the validation data is predicted using the trained LSTM regression model, the prediction structure is determined, and step S21 is returned for execution until N cross-validation is performed. After each validation set validation, the model parameters of the trained LSTM regression model need to be verified, and then the optimized optimal model parameters are determined.

[0113] Optionally, when training and verifying the LSTM regression model, the number of neuron nodes in each LSTM loop structure is set to 50, and the LSTM neural network parameters are continuously iteratively optimized in the way of back propagation BP gradient adjustment. The initial learning rate (Learning Rate) is set to 0.02, and the Adam function is used as the optimization algorithm for training the LSTM regression model to adjust different network parameters to change the amplitude during training iteration. After the first LSTM loop structure, a Dropout layer is added to regularize the neural network structure, and the dropout ratio of the LSTM node is set to 0.02; before the second LSTM loop structure, batch normalization (BatchNormalization) is introduced to stabilize the model training gradient and accelerate convergence.

[0114] Step S23: Performance evaluation is performed on the initial steam temperature prediction model constructed based on the optimal model parameters, and when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters meets the preset condition, step S24 is performed, and when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters does not meet the preset condition, step S25 is performed.

[0115] Step S24: The initial steam temperature prediction model is determined as the final steam temperature prediction model.

[0116] In the process of implementing steps S23 to S25, the performance evaluation effect of the initial steam temperature prediction model can be determined according to the root mean square error (RMSE) and the determination coefficient (R2Score).

[0117] Specifically, first, the root mean square error (RMSE) is determined, and the root mean square error (RMSE) can be calculated as follows: is the estimated value of the LSTM regression model for a data sample, y iThe formula (8) is calculated by substituting the real value corresponding to the data sample.

[0118] Formula (8):

[0119]

[0120] wherein q is the number of times of steam temperature value prediction of the power generation boiler by the regression model, is the estimated value of the LSTM regression model for a data sample, y i is the real value corresponding to the data sample.

[0121] Then, since there is a numerical deviation in each prediction, the numerical deviation of each prediction can be determined, i.e., the R2 score, by calculating formula (9).

[0122] Formula (9):

[0123]

[0124] wherein the LSTM regression model predicts r data samples (feature sequences) in the training set, y i represents the real steam temperature value corresponding to the i-th data sample, is the predicted steam temperature value of the i-th data sample, is the average real steam temperature of the r data samples. The R2 score reflects the proportion of the change of the future period steam temperature of the power generation boiler that can be explained by the feature variables through the LSTM regression model.

[0125] It should be noted that the R2 score is in percentage form, and the maximum is 100%.

[0126] In the embodiment of the present application, the higher the R2 score, the closer the prediction result of the LSTM regression model to the real situation. The prediction accuracy of the steam temperature value reaches the order of magnitude. For example: RMSE, 5 (degrees Celsius, ℃); R2 score, more than 90%.

[0127] Finally, when the value of the R2 score is greater than or equal to the value in the preset condition, the initial steam temperature prediction model is determined as the final steam temperature prediction model. When the value of the R2 score is less than the value in the preset condition, the initial steam temperature prediction model can be continuously trained based on the training set and the validation set.

[0128] Optionally, after the final steam temperature prediction model is constructed, the obtained LSTM regression model can be saved in the form of an h5 file under a Python language environment.

[0129] Specifically, the saved final steam temperature prediction model is loaded and a regressor is obtained.

[0130] In the process of implementing step S102, the operating parameters of the power generation boiler collected by the boiler power generation unit can be processed by repeatedly calling the regressor, that is, predic_value = regressor.predict(sampling_data), to predict the steam temperature value in the future period.

[0131] sampling_data is a feature vector formed by using the actually collected data and performing feature extraction (the method determined in step one), and predic_value is the predicted steam temperature value output by the LSTM regression model.

[0132] It should be noted that the future period can be 10 seconds, which can be set according to actual conditions, and the embodiments of the present application do not impose any restrictions.

[0133] In the embodiments of the present application, the target operating parameters related to the target variable steam temperature are determined from the operating parameters of the power generation boiler collected by the boiler power generation unit; the target operating parameters are used as the input of the steam temperature prediction model, the target operating parameters are processed based on the steam temperature prediction model, and the predicted steam temperature value is output. The steam temperature prediction model is obtained by training based on the historical power generation boiler operating parameter set. As can be seen, the steam temperature prediction model trained based on the historical power generation boiler operating parameter set can predict the target operating parameters, and the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. Through the above-mentioned manner, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know the change of the system steam temperature in advance, and the possible steam temperature abnormality can be reasonably and timely disposed, so as to ensure that the power generation operation of the boiler is reliable and safe, that is, to improve the reliability and safety of the operation of the boiler system.

[0134] Based on the above-mentioned boiler steam temperature prediction method according to the embodiments of the present application, in the process of determining the target operating parameters related to the target variable steam temperature from the operating parameters of the power generation boiler collected by the boiler power generation unit in step S101, the following steps are included:

[0135] Step S31: determining the correlation degree of the operating parameters of the power generation boiler collected by the boiler power generation unit with the steam temperature of the power generation boiler.

[0136] In the process of implementing step S31, the error of the predicted steam temperature value of the power generation boiler operating parameter and the actual value (Ground Truth) is compared, and then the correlation degree of different power generation boiler operating parameters and the power generation boiler steam temperature is judged according to the error size.

[0137] It should be noted that the smaller the error, the higher the correlation degree of the power generation boiler operating parameter and the power generation boiler steam temperature, and then the target operating parameter can be selected according to the correlation degree.

[0138] Step S32: selecting a target operating parameter related to the target variable steam temperature according to the correlation degree.

[0139] In the process of implementing step S32, the power generation boiler operating parameter with low error value is preferentially selected as the target operating parameter, that is, the power generation boiler operating parameter with high correlation degree is selected as the target operating parameter, that is, the feature data.

[0140] In the embodiment of the present application, the correlation degree of the power generation boiler operating parameter collected by the boiler generator set and the power generation boiler steam temperature is determined. According to the correlation degree, a target operating parameter related to the target variable steam temperature is selected. The steam temperature prediction model trained based on the historical power generation boiler operating parameter set is used to predict the target operating parameter, and the steam temperature value at the future time can be predicted, so as to realize the prediction of the steam temperature value. Through the above-mentioned way, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know the change of the system steam temperature in advance, and then the possible steam temperature abnormality can be reasonably and timely disposed, so as to ensure that the boiler power generation operation is reliable and safe, that is, to improve the reliability and safety of the boiler system operation.

[0141] Based on the above-mentioned boiler steam temperature prediction method shown in the embodiment of the present application, combined with Figure 1 , see Figure 5 , further comprising:

[0142] Step S103: after obtaining the predicted steam temperature value by executing step S102, drawing a prediction curve based on the predicted steam temperature value and displaying.

[0143] In the process of implementing step S103, the result visualization of the boiler steam temperature prediction is realized by using the drawing library matplotlib in Python. Specifically, the real situation and the predicted situation of the steam temperature are drawn respectively to present the scene of the batch prediction of the temperature value by the LSTM regression model, and to compare with the real situation. Or, the real and predicted situations are presented according to the overlapping drawing mode to quickly determine the deviation between the predicted situation and the real situation.

[0144] Based on the boiler steam temperature prediction method described above, the results of boiler steam temperature prediction are presented using the matplotlib plotting library in Python, and visualized as follows: Figure 6 As shown.

[0145] Figure 6 The actual and predicted steam temperatures of power generation boilers are represented by drawing them separately, presenting a scene diagram comparing the batch predicted temperature values ​​of the LSTM regression model with the actual situation.

[0146] In this embodiment of the invention, not only can it be achieved through... Figure 6 The predictions can be displayed separately, or the actual and predicted steam temperatures can be overlaid, as shown below. Figure 7 As shown.

[0147] Figure 7 By using an overlay plotting method, the actual and predicted steam temperatures of power generation boilers can be expressed, clearly showing the deviation between the predicted and actual conditions.

[0148] In this embodiment of the invention, a steam temperature prediction model trained based on historical power boiler operating parameter sets is used to predict target operating parameters, thereby predicting future steam temperature values. The actual and predicted steam temperatures are then plotted separately to present a scenario where the LSTM regression model batch-predicts temperature values ​​and compares them with the actual values. Alternatively, the actual and predicted values ​​can be presented using an overlay plotting method to quickly determine the deviation between the predicted and actual values.

[0149] Based on the boiler steam temperature prediction device shown in the above embodiments of the present invention, the present invention also discloses a structural schematic diagram of a boiler steam temperature prediction device, as follows: Figure 8 As shown, the device includes:

[0150] The determining unit 801 is used to determine the target operating parameters related to the target variable steam temperature from the operating parameters of the power generation boiler collected from the boiler generator set.

[0151] The steam temperature prediction model 802 is used to take the target operating parameters as input to the steam temperature prediction model, process the target operating parameters based on the steam temperature prediction model, and output the predicted steam temperature value. The steam temperature prediction model is constructed based on the construction unit 803.

[0152] It should be noted that the specific principles and execution processes of each unit in the boiler steam temperature prediction device disclosed in the embodiments of the present application are the same as the boiler steam temperature prediction method disclosed in the embodiments of the present application, and can be referred to the corresponding part of the boiler steam temperature prediction method disclosed in the embodiments of the present application, which will not be repeated here.

[0153] In the embodiments of the present application, the target operation parameter related to the target variable steam temperature is determined from the power generation boiler operation parameters collected by the boiler generator set; the target operation parameter is taken as the input of the steam temperature prediction model, the target operation parameter is processed based on the steam temperature prediction model, and the predicted steam temperature value is output. The steam temperature prediction model is obtained by training based on the historical power generation boiler operation parameter set. As can be seen, the steam temperature prediction model trained based on the historical power generation boiler operation parameter set can predict the target operation parameter, and the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. Through the above-mentioned manner, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know the change of the system steam temperature in advance, and then the possible steam temperature abnormality can be reasonably and timely disposed, so as to ensure that the boiler power generation operation is reliably and safely performed, that is, the reliability and safety of the boiler system operation are improved.

[0154] Optionally, based on the boiler steam temperature prediction device shown in the embodiments of the present application, the determination unit 801 is specifically configured to: determine the correlation degrees of the power generation boiler operation parameters collected by the boiler generator set with the power generation boiler steam temperature respectively; and select the target operation parameter related to the target variable steam temperature according to the correlation degrees.

[0155] In the embodiments of the present application, the correlation degrees of the power generation boiler operation parameters collected by the boiler generator set with the power generation boiler steam temperature are determined. The target operation parameter related to the target variable steam temperature is selected according to the correlation degrees. The steam temperature prediction model trained based on the historical power generation boiler operation parameter set is used to predict the target operation parameter, and the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. Through the above-mentioned manner, not only the service life of the boiler system can be improved, but also the monitoring personnel of the power generation boiler operation can know the change of the system steam temperature in advance, and then the possible steam temperature abnormality can be reasonably and timely disposed, so as to ensure that the boiler power generation operation is reliably and safely performed, that is, the reliability and safety of the boiler system operation are improved.

[0156] Optionally, based on the boiler steam temperature prediction device shown in the embodiments of the present application, the construction unit 803 includes:

[0157] The acquisition subunit is configured to acquire a historical power generation boiler operation parameter set collected by a boiler power generation unit.

[0158] The transformation subunit is configured to perform time series transformation on the historical power generation boiler operation parameter set to obtain a historical time series set.

[0159] The division subunit is configured to divide the historical time series set into a training set and a verification set based on a preset ratio.

[0160] The training subunit is configured to determine an initial LSTM network model, and train the initial LSTM regression model based on the training set and the verification set, and determine an LSTM regression model obtained through the training as a steam temperature prediction model.

[0161] Optionally, the training subunit is configured to: train the initial LSTM regression model on the training set by using a cross-validation method to obtain a trained initial LSTM regression model; determine a fitting degree of the trained initial LSTM regression model on the steam temperature of the power generation boiler on the verification set to obtain optimal model parameters; perform performance evaluation on an initial steam temperature prediction model constructed based on the optimal model parameters; and determine the initial steam temperature prediction model as a final steam temperature prediction model when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters meets a preset condition.

[0162] Optionally, the training subunit is further configured to: continue to train the initial steam temperature prediction model based on the training set and the verification set when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters does not meet the preset condition.

[0163] In the embodiment of the present application, the historical power generation boiler operation parameter set collected by the boiler power generation unit is acquired, and time series transformation is performed to obtain a historical time series set. Then, the initial LSTM regression model is trained on the training set by using a cross-validation method to obtain a trained initial LSTM regression model, and the fitting degree of the trained initial LSTM regression model on the steam temperature of the power generation boiler is determined on the verification set to obtain optimal model parameters. Then, the evaluated steam temperature prediction model is determined. This can predict the steam temperature value at a future time based on the trained steam temperature prediction model, so as to realize prediction of the steam temperature value, improve the service life of the boiler system, and improve the reliability and safety of the operation of the boiler system.

[0164] Optionally, based on the boiler steam temperature prediction device shown in the embodiment of the present application, Figure 8 , see Figure 9 , the boiler steam temperature prediction device further has a drawing unit 804.

[0165] The drawing unit 804 is configured to draw a prediction curve based on the predicted steam temperature value, and display.

[0166] In the embodiments of the present application, the steam temperature prediction model trained based on the historical power generation boiler operation parameter set is used to predict the target operation parameter, so that the steam temperature value at the future time can be predicted, thereby realizing the prediction of the steam temperature value. The real situation and the prediction situation of the steam temperature are drawn respectively to present the batch prediction temperature value of the LSTM regression model, and the deviation between the prediction situation and the real situation is quickly determined according to the overlapping drawing mode.

[0167] Each of the embodiments in the specification is described in a progressive manner, and the same and similar parts of each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments. In particular, for the system or system embodiments, since it is basically similar to the method embodiments, it is described more simply, and the related parts can be referred to the part of the method embodiments. The above-described system and system embodiments are only illustrative, and the units described as separate components can be or can not be physically separated, and the components displayed as units can be or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to the actual needs. Those skilled in the art can understand and implement without creative labor.

[0168] The skilled person can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical scheme. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0169] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of boiler steam temperature prediction, characterized by, The method comprises: determining a target operation parameter related to a target variable steam temperature from power generation boiler operation parameters collected by a boiler power generating unit by using an extreme gradient rising algorithm; selecting a vector of target operation parameters of every n+1 time unit as an input of a steam temperature prediction model, processing the vector of target operation parameters of every n+1 time unit based on the steam temperature prediction model, and outputting a predicted steam temperature value of an m+1 time unit, wherein m and n are variables, and m is greater than n; the steam temperature prediction model is obtained by training based on a historical power generation boiler operation parameter set; the process of obtaining the steam temperature prediction model by training based on the historical power generation boiler operation parameter set comprises: obtaining a historical power generation boiler operation parameter set collected by a boiler power generating unit by using an extreme gradient rising algorithm; extending a vector of a first operation parameter in the historical power generation boiler operation parameter set n times in a horizontal direction to construct a feature sequence with time steps t0, t1, …, tn; sequentially moving all feature attribute columns of a second vector to an n-th vector by 1 to n rows; adding a power generation boiler steam temperature column and moving the power generation boiler steam temperature column by m rows; and deleting the last m rows of the constructed time sequence set to obtain a transformed historical time sequence set; dividing the historical time sequence set into a training set and a validation set based on a preset proportion; determining an initial LSTM regression model, training the initial LSTM regression model based on the training set and the validation set, and determining a current LSTM regression model obtained by training as a steam temperature prediction model.

2. The method of claim 1, wherein, The process of determining a target operation parameter related to a target variable steam temperature from power generation boiler operation parameters collected by a boiler power generating unit comprises: determining the correlation degrees of power generation boiler operation parameters collected by a boiler power generating unit with a power generation boiler steam temperature respectively; selecting a target operation parameter related to a target variable steam temperature according to the correlation degrees.

3. The method of claim 1, wherein, The process of training the initial LSTM regression model based on the training set and the validation set and determining a current LSTM regression model obtained by training as a steam temperature prediction model comprises: training the initial LSTM regression model on the training set by using a cross-validation method to obtain a trained initial LSTM regression model; determining the fitting degree of the trained initial LSTM regression model to a power generation boiler steam temperature on the validation set to obtain optimal model parameters; performing performance evaluation on an initial steam temperature prediction model constructed based on the optimal model parameters; when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters meets a preset condition, determining that the initial steam temperature prediction model is a final steam temperature prediction model.

4. The method of claim 3, wherein, The method further comprises: when it is determined that the initial steam temperature prediction model constructed based on the optimal model parameters does not meet the preset condition, continuing to train the initial steam temperature prediction model based on the training set and the validation set.

5. The method of claim 1, wherein, The method further comprises: drawing a prediction curve based on the predicted steam temperature value and displaying the prediction curve.

6. A boiler steam temperature prediction device characterized by comprising: The device comprises: The determining unit is configured to determine a target operation parameter related to a target variable steam temperature from power generation boiler operation parameters collected by a boiler power generating unit by using an extreme gradient ascent algorithm; The steam temperature prediction model is configured to select a vector of target operation parameters of every n+1 time unit as an input of the steam temperature prediction model, process the vector of target operation parameters of every n+1 time unit based on the steam temperature prediction model, and output a predicted steam temperature value of an m+1 time unit, where m and n are variables, and m is greater than n; the steam temperature prediction model is constructed by the constructing unit; The constructing unit comprises an obtaining subunit, a transforming subunit, a dividing subunit, and a training subunit; The obtaining subunit is configured to obtain a set of historical power generation boiler operation parameters collected by a boiler power generating unit by using an extreme gradient ascent algorithm; The transforming subunit is configured to copy and expand a vector of a first operation parameter in the set of historical power generation boiler operation parameters n times along a horizontal direction to construct a feature sequence with time steps of t0, t1, …, and tn; sequentially move all feature attribute columns of a second vector to an nth vector by 1 to n rows; add a power generation boiler steam temperature column and move the power generation boiler steam temperature column by m rows; and delete the last m rows of the constructed time sequence set to obtain a transformed historical time sequence set; The dividing subunit is configured to divide the historical time sequence set into a training set and a validation set based on a preset proportion. The training subunit is configured to determine an initial LSTM regression model, train the initial LSTM regression model based on the training set and the validation set, and determine an LSTM regression model obtained by the current training as the steam temperature prediction model.

7. The apparatus of claim 6, wherein, The determining unit is specifically configured to determine the correlation degrees of power generation boiler operation parameters collected by a boiler power generating unit with a power generation boiler steam temperature, and select a target operation parameter related to a target variable steam temperature according to the correlation degrees.

8. The apparatus of claim 6, wherein, Further comprising: The drawing unit is configured to draw a prediction curve based on the predicted steam temperature value and display the prediction curve.

Images