CN110909508A - Heating furnace temperature field real-time prediction method based on convolution long and short term memory network - Google Patents
Heating furnace temperature field real-time prediction method based on convolution long and short term memory network Download PDFInfo
- Publication number
- CN110909508A CN110909508A CN201911036206.9A CN201911036206A CN110909508A CN 110909508 A CN110909508 A CN 110909508A CN 201911036206 A CN201911036206 A CN 201911036206A CN 110909508 A CN110909508 A CN 110909508A
- Authority
- CN
- China
- Prior art keywords
- temperature field
- data
- short term
- term memory
- heating furnace
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000010438 heat treatment Methods 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000006403 short-term memory Effects 0.000 title claims description 21
- 230000007787 long-term memory Effects 0.000 title claims description 20
- 230000015654 memory Effects 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 4
- 239000007789 gas Substances 0.000 claims description 4
- 239000001301 oxygen Substances 0.000 claims description 4
- 229910052760 oxygen Inorganic materials 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 239000012530 fluid Substances 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 description 11
- 238000005259 measurement Methods 0.000 description 11
- 238000004519 manufacturing process Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000002485 combustion reaction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008033 biological extinction Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000000571 coke Substances 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Feedback Control In General (AREA)
Abstract
A heating furnace temperature field real-time prediction method based on a convolution long-short term memory network comprises the following steps: building a temperature field real-time prediction network; training a real-time prediction network of a temperature field; respectively carrying out normalization processing on each data of the industrial heating furnace; carrying out three-dimensional expansion on each data of the industrial heating furnace; sending the expanded data into a trained real-time temperature field prediction network to obtain a three-dimensional temperature field prediction result; and restoring the obtained prediction result of the three-dimensional temperature field into the pipe wall temperature field through the reserved space coordinates of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field. Compared with the method for calculating the temperature field once by the fluid mechanics method, the method can obtain the predicted temperature field only by seconds, the efficiency is improved remarkably, and meanwhile, the prediction error is small.
Description
Technical Field
The invention relates to a method for predicting a temperature field of an industrial heating furnace. In particular to a heating furnace temperature field real-time prediction method based on a convolution long and short term memory network.
Background
An industrial heating furnace is an important device of an industrial refining device, and whether the industrial heating furnace is operated safely or not directly influences the service life, the production capacity and the economic benefit of the device. The combustion process of an industrial heating furnace is unstable, and local overtemperature can occur at random positions during operation. However, the furnace tube of the industrial heating furnace is mostly composed of a process medium which is easy to coke, if a certain local position of the heating furnace runs in an overtemperature state for a long time, the loss and the damage of the furnace tube can be caused, so that measures must be taken to optimize the combustion condition of the temperature field of the heating furnace. However, the heating furnace is huge in equipment and severe in environment, so that the related physical quantity parameters are difficult to measure on line, the combustion adjustment cannot be reliably based, and the combustion optimization operation is difficult to realize. Therefore, only soft-measurement solutions can be used to obtain the temperature field of the industrial furnace.
Soft measurement refers to the mathematical inference of variables that are difficult to measure and cannot be measured by using already obtained measurement quantities in industrial production in combination with computer application techniques, thereby obtaining values of important variables that cannot be directly measured. Soft measurements have the advantage of a fast dynamic response, which can continuously give the values of important variables in an industrial process.
The soft measurement Models can be divided into Model-drive Models (MDM for short) based on principle and Data-drive Models (DDM for short) [4], also called Data-Driven soft measurement Models.
The MDM is based on prior knowledge of an operation mechanism of an industrial process and relevant theoretical knowledge of complex physical chemistry and the like involved in the industrial process, and a principle model is generally high in numerical calculation accuracy and strong in interpretability. However, the traditional model based on the mechanism is often composed of a large number of algebraic equations, and has large calculation amount and slow convergence, so that the requirement of the soft measurement real-time property is difficult to meet.
CFD is an abbreviation for Computational Fluid Dynamics (MDM), also a type of MDM. CFD solves problems related to fluid flow using methods of numerical analysis. The working condition data of the temperature, the flow and the like of the industrial heating furnace can be used as the basis for CFD calculation, so that the temperature calculation values of a series of points in the industrial heating furnace can be obtained. The CFD can calculate the three-dimensional temperature field data of the industrial heating furnace which is very detailed and abundant. However, CFD can not get rid of the large calculation amount of MDM, and a real-time temperature field is difficult to obtain.
A soft measurement modeling method based on Deep learning belongs to DDM and is applied to a Deep neural network (Deep neural network, DNN for short). Compared with the traditional neural network, the DNN network has higher layer depth and more advanced convergence method, thereby having stronger approximation and convergence capability. DNN shows very good performance for complex, highly nonlinear processes in industrial processes.
And is counted
A Recurrent Neural Network (RNN) and an improved version thereof, namely, a Long Short Term Memory (LSTM) Network, which have been proposed in the field of machine learning as early as the last century and are currently widely used. In an actual industrial process, the current system state and output are closely related to one or more previous states. RNNs are therefore well suited to address the time series problem in industrial processes. The LSTM contains memory and forgetting structures. Compared with the traditional RNN, the LSTM can solve the problems of gradient extinction and gradient explosion in the original RNN, is already applied to the fields of natural language processing, voice recognition and the like, is also applied to the field of soft measurement, and has excellent performance in processing the time sequence problem of the industrial process. After training is finished, the method is short in calculation time and can meet the requirement of real-time performance.
Disclosure of Invention
The invention aims to solve the technical problem of providing a heating furnace temperature field real-time prediction method based on a convolution long-short term memory network, which can realize the soft measurement of the temperature field of an industrial heating furnace.
The technical scheme adopted by the invention is as follows: a heating furnace temperature field real-time prediction method based on a convolution long and short term memory network is characterized by comprising the following steps:
1) building a temperature field real-time prediction network;
2) training a real-time prediction network of a temperature field;
3) respectively carrying out normalization processing on each data of the industrial heating furnace;
4) carrying out three-dimensional expansion on each data of the industrial heating furnace;
5) sending the expanded data into a trained real-time temperature field prediction network to obtain a three-dimensional temperature field prediction result;
6) and (3) restoring the obtained prediction result of the three-dimensional temperature field into the pipe wall temperature field through the space coordinates of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field reserved in the step 2).
The temperature field real-time prediction network in the step 1) has 4 layers of convolution long and short term memory networks (ConvLSTM), the number of input data channels is 18, the number of output data channels of a first layer of long and short term memory network (ConvLSTM1) is 24, the number of output data channels of a second layer of long and short term memory network (ConvLSTM2) is 48, the number of output data channels of a third layer of long and short term memory network (ConvLSTM3) is 24, output data of the first layer of long and short term memory network (ConvLSTM1) and output data of the third layer of long and short term memory network (ConvLSTM3) are spliced to obtain data with the number of channels being 48, the data are sent to a fourth layer of long and short term memory network (ConvLSTM4), and finally temperature field output (output) of 24 channels is obtained;
meanwhile, the lengths of the input and output time series of the first layer long-short term memory network (ConvLSTM1), the second layer long-short term memory network (ConvLSTM2) and the third layer long-short term memory network (ConvLSTM3) are both 10, the length of the input time series of the fourth layer long-short term memory network (ConvLSTM4) is both 10, and the length of the output time series is 3.
The step 2) comprises the following steps:
(1) respectively carrying out normalization processing on each data of the industrial heating furnace by adopting the following formula:
wherein, x is data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample;
(2) carrying out three-dimensional expansion on each data of the industrial heating furnace;
(3) converting a tube wall temperature field to a three-dimensional temperature field
(4) Training a temperature field real-time prediction network, wherein network parameters are not frozen during training, the initial learning rate during training is 0.01, the batch size during training is 1, an L1 loss function is selected as the loss function, and an Adam algorithm is adopted to optimize the loss function; the total iteration is 600 times, the learning rate is multiplied by 0.8 for attenuation every iteration of 40 times, and the network achieves the best effect in the 480 th iteration to obtain an optimal model; the average absolute error of the optimal model output was 28.3 kelvin.
The data in the step 2) comprises the following steps: measuring the values of the temperature, the pressure, the oxygen content and the gas flow of the distributed control system in the industrial heating furnace according to set time;
step 2), the expansion method of the step (2) is the same as that of the step 3), 1 piece of working condition data is expanded into a matrix of 32 x 652, and each element in the matrix is filled with the working condition data; and for the temperature data of 8 thermocouples on the pipe wall, only 50 elements near the spatial position corresponding to the thermocouple are filled with the temperature data measured by the thermocouple during expansion, and the rest elements are set to be zero, and finally, input data of 18 x 32 x 652 is obtained.
Step 2) the step (3) is that the pipeline is divided into 24 pipelines according to the z axis, then the pipeline wall is cut along the axis and unfolded into a rectangle, the unfolding method is that 32 points are sequentially selected according to the axis direction of the pipeline, and the 32 points are arranged according to the z axis direction to obtain a first row of data; and sequentially selecting 32 points in the axial direction of the pipeline to be continuously arranged, repeating for 652 times to obtain a 24 x 32 x 652 matrix, simultaneously dividing the temperature value of the temperature field by 1000 to enable the value range to be between 0 and 1, and keeping the space coordinate of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field.
The normalization processing in the step 3) is carried out by adopting the following formula:
wherein, x is the data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample.
The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network is completely driven by data, and uses ConvLSTM to dynamically extract the internal relation of a time sequence to obtain a predicted temperature field, so that the soft measurement of the industrial heating furnace temperature field is realized, and the average absolute error MAE of the predicted temperature field is 31.7K. Compared with the computational fluid dynamics method which needs several days for calculating the temperature field once, the method can obtain the predicted temperature field only in several seconds, the efficiency is obviously improved, and meanwhile, the prediction error is small.
Drawings
FIG. 1 is a schematic view of the temperature field of the pipe of the present invention;
FIG. 2 is a schematic diagram of a temperature field real-time prediction network according to the present invention;
FIG. 3 is a thermocouple trend chart for number 1;
FIG. 4 is a thermocouple trend chart for number 2;
FIG. 5 is a thermocouple trend chart for number 3;
FIG. 6 is a thermocouple trend chart for number 4;
FIG. 7 is a thermocouple trend chart for number 5;
FIG. 8 is a thermocouple trend chart for number 6;
FIG. 9 is a thermocouple trend graph for number 7;
FIG. 10 is a thermocouple trend graph for number 8;
FIG. 11 is a calculated temperature field modeling diagram;
FIG. 12 is a predicted temperature field modeling diagram.
Detailed Description
The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network is described in detail below with reference to the embodiment and the accompanying drawings.
The invention discloses a heating furnace temperature field real-time prediction method based on a convolution long and short term memory network, which comprises the following steps:
1) building a temperature field real-time prediction network;
as shown in fig. 2, the temperature field real-time prediction network has 4 layers of convolutional long-short term memory networks (ConvLSTM), the number of input data channels is 18, the number of output data channels of the first layer of long-short term memory network (ConvLSTM1) is 24, the number of output data channels of the second layer of long-short term memory network (ConvLSTM2) is 48, the number of output data channels of the third layer of long-short term memory network (ConvLSTM3) is 24, output data of the first layer of long-short term memory network (ConvLSTM1) and output data of the third layer of long-short term memory network (ConvLSTM3) are spliced to obtain data with the number of channels being 48, and the data are sent to the fourth layer of long-short term memory network (ConvLSTM4), and finally a temperature field output (output) of 24 channels is obtained;
meanwhile, the lengths of the input and output time series of the first layer long-short term memory network (ConvLSTM1), the second layer long-short term memory network (ConvLSTM2) and the third layer long-short term memory network (ConvLSTM3) are both 10, the length of the input time series of the fourth layer long-short term memory network (ConvLSTM4) is both 10, and the length of the output time series is 3.
2) Training a real-time prediction network of a temperature field; the method comprises the following steps:
(1) respectively carrying out normalization processing on each data of the industrial heating furnace by adopting the following formula:
wherein, x is data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample;
(2) the actual working condition is only a 1-dimensional vector and is not matched with the 3-dimensional temperature field output by the target network, so that the 1-dimensional working condition is expanded into 3-dimensional working condition. The three-dimensional expansion of the data of the industrial heating furnace is needed, wherein the data comprises the following data: measuring the values of the temperature, the pressure, the oxygen content and the gas flow of the distributed control system in the industrial heating furnace according to set time; the expansion method is the same, namely 1 piece of working condition data is expanded into a matrix of 32 x 652, and each element in the matrix is filled with the working condition data; and for the temperature data of 8 thermocouples on the pipe wall, only 50 elements near the spatial position corresponding to the thermocouple are filled with the temperature data measured by the thermocouple during expansion, and the rest elements are set to be zero, and finally, input data of 18 x 32 x 652 is obtained.
(3) The pipe wall temperature field is converted into the three-dimensional temperature field, and the 3-dimensional temperature field also contains the structural information of the pipe wall, but the temperature field cannot be output as a network, so the calculation points on the pipe wall temperature field are rearranged into the three-dimensional temperature field.
Dividing a pipeline into 24 pipelines according to a z axis, cutting and unfolding the pipeline wall into a rectangle along the axis, wherein the unfolding method comprises the steps of sequentially selecting 32 points according to the axis direction of the pipeline, and arranging the 32 points according to the z axis direction to obtain a first row of data; and sequentially selecting 32 points in the axial direction of the pipeline to be continuously arranged, repeating for 652 times to obtain a 24 x 32 x 652 matrix, simultaneously dividing the temperature value of the temperature field by 1000 to enable the value range to be between 0 and 1, and keeping the space coordinate of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field.
(4) Training a temperature field real-time prediction network, wherein network parameters are not frozen during training, the initial learning rate during training is 0.01, the batch size during training is 1, an L1 loss function is selected as the loss function, and an Adam algorithm is adopted to optimize the loss function; the total iteration is 600 times, the learning rate is multiplied by 0.8 for attenuation every iteration of 40 times, and the network achieves the best effect in the 480 th iteration to obtain an optimal model; the average absolute error of the optimal model output was 28.3 kelvin.
3) Respectively carrying out normalization processing on each data of the industrial heating furnace; the data described herein include: measuring the values of the temperature, the pressure, the oxygen content and the gas flow of the distributed control system in the industrial heating furnace according to set time; the expansion method is the same, namely 1 piece of working condition data is expanded into a matrix of 32 x 652, and each element in the matrix is filled with the working condition data; and for the temperature data of 8 thermocouples on the pipe wall, only 50 elements near the spatial position corresponding to the thermocouple are filled with the temperature data measured by the thermocouple during expansion, and the rest elements are set to be zero, and finally, input data of 18 x 32 x 652 is obtained. The normalization process described here is performed using the following equation:
wherein, x is data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample;
4) carrying out three-dimensional expansion on each data of the industrial heating furnace; the expansion method is that 1 piece of working condition data is expanded into a matrix of 32 x 652, and each element in the matrix is filled with the working condition data; and for the temperature data of 8 thermocouples on the pipe wall, only 50 elements near the spatial position corresponding to the thermocouple are filled with the temperature data measured by the thermocouple during expansion, and the rest elements are set to be zero, and finally, input data of 18 x 32 x 652 is obtained.
5) Sending the expanded data into a trained real-time temperature field prediction network to obtain a three-dimensional temperature field prediction result;
6) and (3) restoring the obtained prediction result of the three-dimensional temperature field into the pipe wall temperature field through the space coordinates of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field reserved in the step 2). The temperature field output by the temperature field real-time prediction network is only a three-dimensional matrix and does not have the shape of a pipe wall. Therefore, it is necessary to reduce the predicted result to the tube wall temperature field.
Examples are given below
Because the temperature field obtained through CFD calculation only has data at one moment, and the network training needs a large amount of data to train the network, in order to enable the embodiment to be smoothly carried out, a series of simulated temperature fields are constructed according to actual conditions to be used as the data of the network training.
The embodiment of the invention selects 30 working conditions of the accessory corresponding to the working conditions of each calculated temperature field to construct a simulated temperature field. And adjusting the calculated temperature field to obtain a simulated temperature field by taking the temperature of a single pipe wall thermocouple in a working condition as a standard. For example, for the calculated temperature field for condition number 15, if one wants to obtain a simulated temperature field for condition number 10, one should first calculate the percentage increase of the 8 thermocouple data for condition number 10 relative to the 8 thermocouple data for condition number 15. Then, the temperature of the pipeline corresponding to the temperature field is calculated according to 8 growth percentages and the No. 15 working condition, and for the pipeline without the thermocouple in the middle, the gradual change treatment is adopted according to the growth percentages of the two nearest thermocouples.
The temperature field prediction of this example is actually a regression task, and for this problem, this example will use the mean absolute error MAE to evaluate the algorithm performance, the calculation formula of MAE is as in equation 1.
Wherein m is the number of points in the three-dimensional temperature field, yiThe actual temperature value of the ith point of the three-dimensional temperature field,
the predicted temperature value for the ith point of the three-dimensional temperature field,
the MAE for the best results of training was 0.0283, corresponding to a test result of 0.0317, converted to kelvin temperatures, i.e., 28.3K and 31.7K. The error is 3.6% and 4.1% relative to the average temperature of the temperature field 773.6K.
It can be seen that the error of the prediction results as a whole is within an acceptable range. But also measures local errors to better reflect algorithm performance. As shown in table 1, the predicted temperature field for the 4805 th operating mode and the temperature differences corresponding to 8 thermocouples can be seen, and the network prediction result can well conform to the real data in terms of local details.
TABLE 1 temperature comparison of LSTM predictions at eight sites
Besides the requirement of predicting the temperature field to be consistent with the real working condition, the prediction result is expected to be capable of being consistent with the variation trend of the temperature. The actual and predicted values for 8 thermocouples at 4805 operating mode and 20 hours later are shown in fig. 3-10.
During the data preprocessing phase, the present example rearranges the temperature field into a 24 x 32 x 652 matrix for network input and output. In the last step, the matrix is required to be reduced into a pipe wall temperature field, so that the experimental result can be sensed more intuitively. As shown in FIGS. 11 and 12, the three-dimensional scatter modeling results at matlab for the calculated temperature field and the predicted temperature field for condition No. 4805 are shown. The gaps of the pipelines are caused by the fact that the corresponding calculation points are lacked in the reduction process due to the uneven distribution of the calculation points.
Claims (7)
1. A heating furnace temperature field real-time prediction method based on a convolution long and short term memory network is characterized by comprising the following steps:
1) building a temperature field real-time prediction network;
2) training a real-time prediction network of a temperature field;
3) respectively carrying out normalization processing on each data of the industrial heating furnace;
4) carrying out three-dimensional expansion on each data of the industrial heating furnace;
5) sending the expanded data into a trained real-time temperature field prediction network to obtain a three-dimensional temperature field prediction result;
6) and (3) restoring the obtained prediction result of the three-dimensional temperature field into the pipe wall temperature field through the space coordinates of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field reserved in the step 2).
2. The heating furnace temperature field real-time prediction method based on the convolution long-short term memory network as claimed in claim 1, wherein the temperature field real-time prediction network in step 1) has 4 layers of convolution long-short term memory networks (ConvLSTM), the number of input data channels is 18, the number of output data channels of the first layer of long-short term memory network (ConvLSTM1) is 24, the number of output data channels of the second layer of long-short term memory network (ConvLSTM2) is 48, the number of output data channels of the third layer of long-short term memory network (ConvLSTM3) is 24, the output data of the first layer of long-short term memory network (ConvLSTM1) and the output data of the third layer of long-short term memory network (ConvLSTM3) are spliced to obtain data with the number of channels of 48, and the data are sent to the fourth layer of long-short term memory network (ConvLSTM4), and finally the temperature field output (output) of 24 channels is obtained;
meanwhile, the lengths of the input and output time series of the first layer long-short term memory network (ConvLSTM1), the second layer long-short term memory network (ConvLSTM2) and the third layer long-short term memory network (ConvLSTM3) are both 10, the length of the input time series of the fourth layer long-short term memory network (ConvLSTM4) is both 10, and the length of the output time series is 3.
3. The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network as claimed in claim 1, wherein step 2) comprises:
(1) respectively carrying out normalization processing on each data of the industrial heating furnace by adopting the following formula:
wherein, x is data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample;
(2) carrying out three-dimensional expansion on each data of the industrial heating furnace;
(3) converting a tube wall temperature field to a three-dimensional temperature field
(4) Training a temperature field real-time prediction network, wherein network parameters are not frozen during training, the initial learning rate during training is 0.01, the batch size during training is 1, an L1 loss function is selected as the loss function, and an Adam algorithm is adopted to optimize the loss function; the total iteration is 600 times, the learning rate is multiplied by 0.8 for attenuation every iteration of 40 times, and the network achieves the best effect in the 480 th iteration to obtain an optimal model; the average absolute error of the optimal model output was 28.3 kelvin.
4. The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network as claimed in claim 1, wherein the data of step 2) comprises: and measuring the values of the temperature, the pressure, the oxygen content and the gas flow of the distributed control system in the industrial heating furnace according to set time.
5. The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network as claimed in claim 1, wherein the expansion method of step 2), step 2 and step 3) is the same, and 1 piece of working condition data is expanded into a matrix of 32 x 652, and each element in the matrix is filled with the working condition data; and for the temperature data of 8 thermocouples on the pipe wall, only 50 elements near the spatial position corresponding to the thermocouple are filled with the temperature data measured by the thermocouple during expansion, and the rest elements are set to be zero, and finally, input data of 18 x 32 x 652 is obtained.
6. The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network as claimed in claim 1, wherein step 2) step (3) is that the pipeline is divided into 24 pipelines according to the z axis, then the pipeline wall is cut along the axis and unfolded into a rectangle, the unfolding method is that 32 points are sequentially selected according to the axis direction of the pipeline, and the 32 points are arranged according to the z axis direction to obtain a first row of data; and sequentially selecting 32 points in the axial direction of the pipeline to be continuously arranged, repeating for 652 times to obtain a 24 x 32 x 652 matrix, simultaneously dividing the temperature value of the temperature field by 1000 to enable the value range to be between 0 and 1, and keeping the space coordinate of the pipe wall temperature field corresponding to each point of the three-dimensional temperature field.
7. The heating furnace temperature field real-time prediction method based on the convolution long and short term memory network as claimed in claim 1, wherein the normalization process in step 3) is performed by using the following formula:
wherein, x is the data to be normalized, mean, max and min are respectively the mean, the maximum and the minimum in the range of the working condition data training sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911036206.9A CN110909508A (en) | 2019-10-29 | 2019-10-29 | Heating furnace temperature field real-time prediction method based on convolution long and short term memory network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911036206.9A CN110909508A (en) | 2019-10-29 | 2019-10-29 | Heating furnace temperature field real-time prediction method based on convolution long and short term memory network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110909508A true CN110909508A (en) | 2020-03-24 |
Family
ID=69814769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911036206.9A Pending CN110909508A (en) | 2019-10-29 | 2019-10-29 | Heating furnace temperature field real-time prediction method based on convolution long and short term memory network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110909508A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111859746A (en) * | 2020-07-10 | 2020-10-30 | 西安交通大学 | Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction |
| CN112528569A (en) * | 2020-10-17 | 2021-03-19 | 中国石油化工股份有限公司 | Industrial heating furnace temperature field prediction method based on digital twinning |
| CN113012767A (en) * | 2021-02-24 | 2021-06-22 | 大唐环境产业集团股份有限公司 | Desulfurization system slurry pH value online prediction method and device based on time sequence |
| CN113806981A (en) * | 2021-09-16 | 2021-12-17 | 浙江衡玖医疗器械有限责任公司 | Water temperature prediction method and device for hemispherical ultrasonic imaging system and electronic equipment |
| CN113962452A (en) * | 2021-10-15 | 2022-01-21 | 佛山科学技术学院 | Thermal runaway spread prediction method and prediction system |
| CN114036861A (en) * | 2021-10-25 | 2022-02-11 | 中国石油化工股份有限公司 | Three-dimensional temperature field prediction method based on infoGAN |
| WO2022110115A1 (en) * | 2020-11-30 | 2022-06-02 | 西安交通大学 | Industrial process intelligent control method and system |
| CN114798763A (en) * | 2021-01-29 | 2022-07-29 | 宝山钢铁股份有限公司 | Method and system for predicting rough rolling outlet temperature of tapping plate blank of heating furnace |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2010140576A (en) * | 2010-10-04 | 2012-04-10 | Государственное образовательное учреждение высшего профессионального образования "Адыгейский государственный университет" (ГОУ ВПО " | METHOD FOR PREDICTING TIME OF OFFENSION AND LEVEL OF FLOOD |
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| CN108764460A (en) * | 2018-05-16 | 2018-11-06 | 华中科技大学 | A kind of Time Series Forecasting Methods based on time convolution sum LSTM |
| CN109492839A (en) * | 2019-01-17 | 2019-03-19 | 东华大学 | A kind of mineral hot furnace operating condition prediction technique based on RNN-LSTM network |
| CN109816008A (en) * | 2019-01-20 | 2019-05-28 | 北京工业大学 | A kind of astronomical big data light curve predicting abnormality method based on shot and long term memory network |
| CN109978279A (en) * | 2019-04-10 | 2019-07-05 | 青岛农业大学 | Marine surface temperature regional prediction method |
| CN110059694A (en) * | 2019-04-19 | 2019-07-26 | 山东大学 | The intelligent identification Method of lteral data under power industry complex scene |
| CN110197307A (en) * | 2019-06-03 | 2019-09-03 | 上海海洋大学 | A kind of domain type sea-surface temperature prediction technique of combination attention mechanism |
| CN110348059A (en) * | 2019-06-12 | 2019-10-18 | 西安交通大学 | A kind of channel flow field reconstructing method based on structured grid |
-
2019
- 2019-10-29 CN CN201911036206.9A patent/CN110909508A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2010140576A (en) * | 2010-10-04 | 2012-04-10 | Государственное образовательное учреждение высшего профессионального образования "Адыгейский государственный университет" (ГОУ ВПО " | METHOD FOR PREDICTING TIME OF OFFENSION AND LEVEL OF FLOOD |
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| CN108764460A (en) * | 2018-05-16 | 2018-11-06 | 华中科技大学 | A kind of Time Series Forecasting Methods based on time convolution sum LSTM |
| CN109492839A (en) * | 2019-01-17 | 2019-03-19 | 东华大学 | A kind of mineral hot furnace operating condition prediction technique based on RNN-LSTM network |
| CN109816008A (en) * | 2019-01-20 | 2019-05-28 | 北京工业大学 | A kind of astronomical big data light curve predicting abnormality method based on shot and long term memory network |
| CN109978279A (en) * | 2019-04-10 | 2019-07-05 | 青岛农业大学 | Marine surface temperature regional prediction method |
| CN110059694A (en) * | 2019-04-19 | 2019-07-26 | 山东大学 | The intelligent identification Method of lteral data under power industry complex scene |
| CN110197307A (en) * | 2019-06-03 | 2019-09-03 | 上海海洋大学 | A kind of domain type sea-surface temperature prediction technique of combination attention mechanism |
| CN110348059A (en) * | 2019-06-12 | 2019-10-18 | 西安交通大学 | A kind of channel flow field reconstructing method based on structured grid |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111859746A (en) * | 2020-07-10 | 2020-10-30 | 西安交通大学 | Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction |
| CN112528569A (en) * | 2020-10-17 | 2021-03-19 | 中国石油化工股份有限公司 | Industrial heating furnace temperature field prediction method based on digital twinning |
| WO2022110115A1 (en) * | 2020-11-30 | 2022-06-02 | 西安交通大学 | Industrial process intelligent control method and system |
| CN114798763A (en) * | 2021-01-29 | 2022-07-29 | 宝山钢铁股份有限公司 | Method and system for predicting rough rolling outlet temperature of tapping plate blank of heating furnace |
| CN114798763B (en) * | 2021-01-29 | 2024-07-12 | 宝山钢铁股份有限公司 | Method and system for predicting rough rolling outlet temperature of heating furnace tapping plate blank |
| CN113012767A (en) * | 2021-02-24 | 2021-06-22 | 大唐环境产业集团股份有限公司 | Desulfurization system slurry pH value online prediction method and device based on time sequence |
| CN113012767B (en) * | 2021-02-24 | 2024-08-20 | 大唐环境产业集团股份有限公司 | Online prediction method and device for pH value of desulfurization system slurry based on time sequence |
| CN113806981A (en) * | 2021-09-16 | 2021-12-17 | 浙江衡玖医疗器械有限责任公司 | Water temperature prediction method and device for hemispherical ultrasonic imaging system and electronic equipment |
| CN113806981B (en) * | 2021-09-16 | 2024-07-05 | 浙江衡玖医疗器械有限责任公司 | Water temperature prediction method and device for hemispherical ultrasonic imaging system and electronic equipment |
| CN113962452A (en) * | 2021-10-15 | 2022-01-21 | 佛山科学技术学院 | Thermal runaway spread prediction method and prediction system |
| CN114036861A (en) * | 2021-10-25 | 2022-02-11 | 中国石油化工股份有限公司 | Three-dimensional temperature field prediction method based on infoGAN |
| CN114036861B (en) * | 2021-10-25 | 2022-09-27 | 中国石油化工股份有限公司 | Three-dimensional temperature field prediction method based on infoGAN |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110909508A (en) | Heating furnace temperature field real-time prediction method based on convolution long and short term memory network | |
| CN110441065B (en) | Gas turbine on-line detection method and device based on LSTM | |
| CN113051831B (en) | Modeling method and thermal error control method for thermal error self-learning prediction model of machine tool | |
| CN109755683B (en) | Battery pack internal temperature real-time monitoring method based on compressed sensing theory | |
| CN113657660B (en) | Heat source load prediction method based on substation load and heat supply network hysteresis model | |
| CN111898237B (en) | Parallel simulated annealing rapid optimization method for inversion measurement of multiple thermophysical parameters of material | |
| CN112733443B (en) | Water supply network model parameter optimization checking method based on virtual monitoring points | |
| CN110442974B (en) | Method and device for optimizing performance of regenerative chamber of horseshoe flame glass kiln | |
| CN111158237A (en) | Industrial furnace temperature multi-step prediction control method based on neural network | |
| CN111460548B (en) | Normal-state roller compacted concrete gravity dam combined damming safety assessment method | |
| CN114462336B (en) | Method for calculating average temperature of coolant of main pipeline of nuclear reactor | |
| CN114444790B (en) | Method for predicting time sequence of various measuring points on gas turbine based on steady-state characteristic composition | |
| CN117705456A (en) | Digital test run method of aeroengine based on multi-model deep learning | |
| CN116861256A (en) | Furnace temperature prediction method, system, equipment and medium for solid waste incineration process | |
| JP2013137774A (en) | Method for predicting state of process and system for predicting state of process | |
| CN112016241A (en) | Soft measurement modeling method based on mLASSO-MLP model | |
| CN116401755A (en) | CNN-LSTM-Attention network structure-based solid rocket engine total thrust prediction method | |
| CN111210877A (en) | Method and device for deducing physical property parameters | |
| CN117634278A (en) | Design optimization method and design optimization device | |
| CN115828688A (en) | Digital twin cable temperature prediction method based on Gaussian process regression | |
| CN115481554A (en) | Thermal diffusion digital twin model, temperature field real-time optimization control model and method in explosive fusion casting and curing process | |
| CN115221791A (en) | Supercritical boiler wall temperature online prediction method and system | |
| CN113900379A (en) | Neural network-based predictive control algorithm | |
| Liu et al. | High precision temperature measurement for cryogenic temperature sensors based on deep learning technology | |
| Petrov et al. | Visualization of physical and mathematical models of thermodynamic processes of single-phase flows using MATLAB to diagnose the condition of external heat supply networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |