Concrete temperature field reconstruction method and system, electronic equipment and storage medium

Abstract

The invention belongs to the field of data processing, and relates to a concrete temperature field reconstruction method and system, electronic equipment and a storage medium, and the method comprises the steps: S1, storing the obtained temperature of a temperature measurement point and the coordinates of the temperature measurement point in a temperature measurement point database; s2, dividing the temperature measurement point database into a training database and a test database; s3, obtaining an LSTM model output result through the LSTM model; s4, obtaining temperature measurement point interpolation temperatures corresponding to all temperature measurement points in the test database at the tj moment; s5, obtaining a temperature data training result of the temperature measurement point in the test database at the tj moment through a DNN model; s6, based on the temperature data training result of the temperature measuring point in the test database at the tj moment, judging whether an iteration stopping condition is met, and if the iteration stopping condition is not met, repeatedly executing S3 to S5; and if a stop condition is met, iteration is ended to complete temperature field reconstruction. The problem that the reconstruction precision of the concrete temperature field is low is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and specifically discloses a method, a system, an electronic device, and a storage medium for reconstructing a concrete temperature field. Background Art

[0002] With the development of sensor technology, the accuracy of sensors used for temperature monitoring in hydropower projects is getting higher and higher, and the quantity and variety are also increasing. Especially with the application of distributed optical fiber technology, traditional point-type temperature acquisition has been transformed into linear perception, greatly increasing the fineness and density of temperature data. On this basis, obtaining or reconstructing the true temperature field of the monitored object has become one of the research hotspots in the current hydraulic engineering discipline.

[0003] Currently, the construction of the true concrete temperature field based on monitored temperatures mainly uses spatial interpolation methods. As is well known, temperature is a physical quantity representing the degree of hotness or coldness of an object, and it has correlations in both spatial and temporal distributions. When performing global temperature field interpolation based on a limited number of temperature monitoring points, if only the spatial correlation of temperature is considered while ignoring the temporal correlation, the interpolation accuracy of the temperature field will be insufficient. Therefore, in order to improve the interpolation accuracy of the temperature field, it is essential to construct a spatio-temporal interpolation method that integrates the temporal and spatial information of temperature data. In recent years, some scholars have introduced the time dimension on the basis of spatial interpolation methods and proposed a series of spatio-temporal data interpolation methods, including spatio-temporal Kriging interpolation, spatio-temporal inverse distance weighted interpolation, interpolation considering spatio-temporal heterogeneity, etc. Among them, the spatio-temporal Kriging interpolation method is the most widely used, but its variogram model needs to consider both the time and space dimensions, with a high degree of complexity and subjective selection of the variogram model by manual intervention, resulting in a low accuracy of the obtained temperature field; while the interpolation accuracies of other spatio-temporal interpolation methods are relatively low, also making the obtained temperature field inaccurate. Therefore, conducting research on new methods for reconstructing the true spatio-temporal temperature field is of great significance for overcoming the current problem of insufficient interpolation accuracy of the temperature field, as well as obtaining the true spatio-temporal evolution law of the concrete temperature field and improving the temperature control quality. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, a system, an electronic device, and a storage medium for reconstructing a concrete temperature field, so as to solve the problem of low accuracy in reconstructing the concrete temperature field.

[0005] The specific solution of the present invention is as follows: A method for reconstructing a concrete temperature field includes the following steps: S1. Store the temperatures of temperature measurement points and the coordinates of temperature measurement points obtained into a temperature measurement point database; S2. Divide the temperature measurement point database into a training database and a test database; S3. According to t in the training database j-l ~tj The temperature of the temperature measurement points in the time period, and the output result of the LSTM model is obtained through the LSTM model, where l is the length of the time period; S4. Based on the temperature of the temperature measurement points at time t in the training database j and the temperature measurement point coordinates in the test database, through the OK spatial interpolation method, obtain the j interpolated temperature of the temperature measurement points corresponding to all temperature measurement points in the test database at time t; S5. According to the output result of the LSTM model and the interpolated temperature of the temperature measurement points, obtain the j training result of the temperature data of the temperature measurement points in the test database at time t through the DNN model; S6. Based on the j training result of the temperature data of the temperature measurement points in the test database at time t, determine whether the iteration stop condition is reached. If the iteration stop condition is not reached, repeat steps S3 to S5; if the stop condition is reached, end the iteration and complete the temperature field reconstruction.

[0006] In some embodiments, the step of repeating steps S3 to S5 if the iteration stop condition is not reached includes: If the iteration stop condition is not reached, repeat steps S3 to S5 through the Adam optimization method and the gradient descent method.

[0007] In some embodiments, the gradient descent method includes updating the LSTM model parameters based on the first-order moment estimation and the second-order moment estimation of the stochastic gradient. The formulas for the first-order moment estimation and the second-order moment estimation are respectively: , where P k is the first-order moment estimation of the k-th iteration; P k-1 is the first-order moment estimation of the (k - 1)-th iteration; v k is the second-order moment estimation of the k-th iteration; v k-1 is the second-order moment estimation of the (k - 1)-th iteration; θ k is the LSTM model parameter of the k-th iteration; θ k-1 is the LSTM model parameter of the (k - 1)-th iteration; is the gradient of the loss function of the k-th iteration; ω1 is the exponential decay rate of the first-order moment estimation; ω2 is the exponential decay rate of the second-order moment estimation; α is the learning rate; ε is a constant.

[0008] In some embodiments, the iteration stop condition is that the error value obtained by comparing and calculating the j training result of the temperature data of the temperature measurement points in the test database at time t with the j temperature measurement point data information at time t in the test database is less than the set threshold.

[0009] In some embodiments, the calculation of the error value includes: using the mean square error (MSE) as the loss function L(W) to calculate the error value, and the formula of the loss function L(W) is: , where W is the weight coefficient matrix; n is the number of samples; y i is the i-th measured value; f(x i ) is the i-th predicted value.

[0010] In some embodiments, the iteration stop condition is the maximum value of the preset number of iterations.

[0011] In some embodiments, it further includes a preprocessing step, and the preprocessing step includes: Normalizing the temperatures and coordinates of abnormal and missing temperature measurement points of the temperature measurement points.

[0012] The present invention also relates to a concrete temperature field reconstruction system for the above-mentioned concrete temperature field reconstruction method, including: An acquisition module for acquiring the temperatures and coordinates of the temperature measurement points and storing them in the temperature measurement point database; A preprocessing module for normalizing the temperatures and coordinates of abnormal and missing temperature measurement points of the temperature measurement points; A database segmentation module for segmenting the temperature measurement point database into a training database and a test database; An LSTM-DNN network training module for obtaining the output result of the LSTM model based on the temperatures of the temperature measurement points in the training database during the time period from t j-l to t j , obtaining the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at time t j through the OK spatial interpolation method based on the temperatures of the temperature measurement points at time t j in the training database and the coordinates of the temperature measurement points in the test database, and obtaining the training result of the temperature data of the temperature measurement points in the test database at time t j through the DNN model based on the output result of the LSTM model and the interpolated temperatures of the temperature measurement points, determining whether the LSTM-DNN network reaches the iteration stop condition, if not reaching the iteration stop condition, continuing the iterative training, and if reaching the stop condition, ending the iterative training to complete the temperature field reconstruction.

[0013] The present invention also relates to an electronic device, including: a processor and a memory; the memory is used to store the executable instructions of the processor, and the processor is configured to execute the above-mentioned concrete temperature field reconstruction method by executing the executable instructions.

[0014] The present invention also relates to a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned method for reconstructing the concrete temperature field.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: Firstly, according to the temperature of the temperature measurement points in the training database during the time period from t j-l to t j , the time variation characteristic law of the temperature is trained by the LSTM model. Secondly, based on the temperature of the temperature measurement points at the moment t j in the training database and the coordinates of the temperature measurement points in the test database, the interpolated temperature of the temperature measurement points corresponding to all the temperature measurement points in the test database at the moment t j is obtained through the OK spatial interpolation method to make up for the missing spatial variation characteristics of the temperature in the LSTM model. The time variation characteristic law of the temperature trained by the LSTM model and the spatial variation characteristics of the temperature obtained through the spatial interpolation method are used to simultaneously train the time variation characteristic law and the spatial variation characteristic law of the temperature through the DNN model. Finally, the accurate temperature of the temperature measurement points in the test database is reconstructed. By fully integrating the time variation characteristics and the spatial variation characteristics of the temperature, the problem of insufficient interpolation accuracy of the temperature field is solved, and the accuracy of the temperature field reconstruction is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of a method for reconstructing a concrete temperature field in Embodiment 1 of the present invention.

[0017] Figure 2 It is a flowchart of the LSTM-DNN network training in Embodiment 1 of the present invention.

[0018] Figure 3 It is a block diagram of a concrete temperature field reconstruction system in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0020] Embodiment 1 A method for reconstructing a concrete temperature field, as Figure 1 shown, includes the following steps: S1. Obtain the data information of the temperature measurement points and the boundary coordinates of the temperature field reconstruction area; Monitor the data information of the temperature measurement points of the concrete and the boundary coordinates of the temperature field reconstruction area. The data information of the temperature measurement points includes the temperature of the temperature measurement points and the coordinates of the temperature measurement points.

[0021] S2. After preprocessing the obtained data information of the temperature measurement points and the boundary coordinates of the temperature field reconstruction area, store the data information of the temperature measurement points in the temperature measurement point database.

[0022] After normalizing the abnormal temperature of the temperature measurement points and the missing temperature of the temperature measurement points, save them to the temperature database; After normalizing the coordinates of the temperature measurement points and the boundary coordinates of the temperature field reconstruction area, save them to the coordinate database; After one-to-one correspondence between the temperature of the temperature measurement points in the temperature database and the coordinates of the temperature measurement points in the coordinate database, store them in the temperature measurement point database T(x,y,z,t). The expression of the temperature measurement point database is: , where, represents the data information of the temperature measurement points at the mth point (x,y,z,t) at time t n moment.

[0023] S3. Divide the temperature measurement point database into a training database and a test database.

[0024] For example, divide the temperature measurement point database according to a ratio of 4:1, that is, the training database T tr accounts for 80% of the temperature measurement point database, and the test database T te accounts for 20% of the measurement point database.

[0025] S4. Train the LSTM-DNN network based on the training database and the test database to complete the temperature field reconstruction.

[0026] The LSTM-DNN network includes a long short-term memory recurrent neural network LSTM and a deep neural network DNN connected in series.

[0027] Training of the LSTM-DNN network, as Figure 2 shown, includes: S41. According to the temperature of the temperature measurement points in the training database during the time period from t j-l to t j , obtain the output result of the LSTM model through the LSTM model; Input the temperature of the temperature measurement points in the training database during the time period from t j-l to t j into the LSTM model. Among them, l is the length of the time period. After training the time variation characteristic law of the temperature field by the LSTM model, output the output result of the LSTM model.

[0028] S42. Based on the temperature measurement points' temperatures at time t in the training database and the temperature measurement points' coordinates in the test database, through the OK spatial interpolation method, obtain the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at time t; j Based on the temperature measurement points' temperatures at time t in the training database, use the OK spatial interpolation method for temperature prediction on the temperature measurement points' coordinates in the test database to obtain the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at time t. j Based on the temperature measurement points' temperatures at time t in the training database, j use the OK spatial interpolation method for temperature prediction on the temperature measurement points' coordinates in the test database to obtain the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at time t. j

[0029] S43. According to the output result of the LSTM model and the interpolated temperatures of the temperature measurement points, obtain the training result of the temperature data of the temperature measurement points in the test database at time t through the DNN model; j Input the output result of the LSTM model and the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database into the DNN model. Through the DNN model, train the time-varying characteristics and spatial-varying characteristics of the temperature measurement points' temperatures simultaneously, and fully fuse the time-varying characteristics and spatial-varying characteristics of the temperature.

[0030] S44. Based on the training result of the temperature data of the temperature measurement points in the test database at time t, judge whether the iteration stop condition is reached. If the iteration stop condition is not reached, repeat the execution of S41 to S43; if the stop condition is reached, end the iteration and complete the temperature field reconstruction. j Compare and calculate the error value between the training result of the temperature data of the temperature measurement points in the test database at time t and the temperature measurement points' data information at time t in the test database. If the error value is greater than or equal to the set threshold, through the Adam optimization method and the gradient descent method, continuously update the weight and bias matrices, and repeat the execution of S41 to S43; if the error value is less than the set threshold, then end the iteration and reconstruct the accurate temperatures of the temperature measurement points in the test database to complete the temperature field reconstruction.

[0031] Compare the training result of the temperature data of the temperature measurement points in the test database at time t with the temperature measurement points' data information at time t in the test database to calculate the error value. j If the error value is greater than or equal to the set threshold, through the Adam optimization method and the gradient descent method, continuously update the weight and bias matrices, and repeat the execution of S41 to S43; if the error value is less than the set threshold, then end the iteration and reconstruct the accurate temperatures of the temperature measurement points in the test database to complete the temperature field reconstruction. j

[0032] Use the mean square error MSE as the loss function L(W) to calculate the error value. The formula of the loss function L(W) is: , where W is the weight coefficient matrix; n is the number of samples; y i is the i-th measured value; f(x i ) is the i-th predicted value.

[0033] ​​​​If the maximum number of preset iterations is not reached, the steps S41 to S43 are repeatedly executed by using the Adam optimization method and the gradient descent method; if the maximum number of preset iterations is reached, the iteration ends, and the accurate temperature of the temperature measurement points in the test database is reconstructed to complete the temperature field reconstruction.

[0034] The gradient descent method includes updating the LSTM model parameters based on the first-order moment estimation and the second-order moment estimation of the stochastic gradient. The formulas for the first-order moment estimation and the second-order moment estimation are respectively: , where, P k is the first-order moment estimation of the k-th iteration; P k-1 is the first-order moment estimation of the (k - 1)-th iteration; v k is the second-order moment estimation of the k-th iteration; v k-1 is the second-order moment estimation of the (k - 1)-th iteration; θ k is the LSTM model parameter of the k-th iteration; θ k-1 is the LSTM model parameter of the (k - 1)-th iteration; is the gradient of the loss function of the k-th iteration; ω1 is the exponential decay rate of the first-order moment estimation; ω2 is the exponential decay rate of the second-order moment estimation; α is the learning rate; ε is a small constant.

[0035] First, the present invention trains the temperature time variation characteristic law through the LSTM model according to the temperature of the temperature measurement points in the training database during the time period from t j-l to t j . Secondly, based on the temperature of the temperature measurement points at the moment t j in the training database and the coordinates of the temperature measurement points in the test database, the interpolation temperature of the temperature measurement points corresponding to all the temperature measurement points in the test database at the moment t j is obtained through the OK spatial interpolation method to make up for the missing spatial variation characteristics of the temperature in the LSTM model. The time variation characteristic law of the temperature trained by the LSTM model and the spatial variation characteristics of the temperature obtained through the spatial interpolation method are used to simultaneously train the time variation characteristic law and the spatial variation characteristic law of the temperature through the DNN model. Finally, the accurate temperature of the temperature measurement points in the test database is reconstructed. By fully integrating the time variation characteristics and the spatial variation characteristics of the temperature, the problem of insufficient interpolation accuracy of the temperature field is solved, and the accuracy of the temperature field reconstruction is improved.

[0036] The present invention also relates to a concrete temperature field reconstruction system for the above-mentioned concrete temperature field reconstruction method, as Figure 3 shown, including: A collection module, configured to acquire the temperature of the temperature measurement points and the coordinates of the temperature measurement points, and store them in the temperature measurement point database; A preprocessing module for normalizing the temperatures and temperature measurement point coordinates of abnormal and missing temperature measurement points. A database segmentation module for segmenting the temperature measurement point database into a training database and a test database. An LSTM-DNN network training module for obtaining the output result of the LSTM model based on the temperatures of the temperature measurement points in the t j-l ~t j time period in the training database through the LSTM model, and obtaining the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at the t j moment through the OK spatial interpolation method based on the temperatures of the temperature measurement points at the t j moment in the training database and the temperature measurement point coordinates in the test database, and obtaining the training result of the temperature data of the temperature measurement points in the test database at the t j moment through the DNN model based on the output result of the LSTM model and the interpolated temperatures of the temperature measurement points, and determining whether the LSTM-DNN network reaches the iteration stop condition. If the iteration stop condition is not reached, continue the iterative training. If the stop condition is reached, end the iterative training to complete the temperature field reconstruction.

[0037] The present invention also relates to an electronic device, including: a processor and a memory; the memory is used to store the executable instructions of the processor, and the processor is configured to execute the above-mentioned method for reconstructing the concrete temperature field by executing the executable instructions.

[0038] The present invention also relates to a computer storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for reconstructing the concrete temperature field is implemented.

[0039] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for reconstructing the temperature field of concrete, characterized in that, Including the following steps: S1. Store the temperature of the temperature measurement point and the coordinates of the temperature measurement point obtained into the temperature measurement point database; S2. Divide the temperature measurement point database into a training database and a test database; S3, according to the training database t j-l ~t j The temperature of the temperature measurement point in the time period is obtained through the LSTM model output result, where l is the length of the time period; S4. Based on the temperature measurement points' temperatures at time t in the training database and the temperature measurement points' coordinates in the test database, through the OK spatial interpolation method, obtain the interpolated temperatures of the temperature measurement points corresponding to all the temperature measurement points in the test database at time t; j At time t j ; S5. Based on the output result of the LSTM model and the interpolated temperature of the temperature measurement points, the training result of the temperature data of the temperature measurement points in the test database at time t is obtained through the DNN model; j ​ S6. Based on t j Test the training result of the temperature data of the temperature measurement points in the database at the moment, and determine whether the iteration stop condition is reached. If the iteration stop condition is not reached, repeat steps S3 to S5; If the stop condition is reached, end the iteration and complete the temperature field reconstruction.

2. The concrete temperature field reconstruction method according to claim 1, characterized in that, If the iteration stop condition is not reached, repeat the execution of S3 to S5, including: If the iteration stop condition is not reached, repeat the execution of S3 to S5 through the Adam optimization method and the gradient descent method.

3. A method for reconstructing the concrete temperature field according to claim 2, characterized in that, The gradient descent method includes updating the LSTM model parameters based on the first-order moment estimation and the second-order moment estimation of the stochastic gradient. The formulas for the first-order moment estimation and the second-order moment estimation are respectively: ， Among them, P k is the first moment estimate of the k-th iteration; P k-1 is the first moment estimate of the (k - 1)-th iteration; v k is the second moment estimate of the k-th iteration; v k-1 is the second moment estimate of the (k - 1)-th iteration; θ k is the LSTM model parameter of the k-th iteration; θ k-1 is the LSTM model parameter of the (k - 1)-th iteration; is the gradient of the loss function of the k-th iteration; ω1 is the exponential decay rate of the first moment estimate; ω2 is the exponential decay rate of the second moment estimate; α is the learning rate; ε is a constant.

4. A method for reconstructing the concrete temperature field according to claim 1, characterized in that: The iteration stop condition is that the error value obtained by comparing and calculating the training result of the temperature data of the temperature measurement points in the test database at time t j with the temperature measurement point data information at time t j in the test database is less than the set threshold value.

5. A method for reconstructing the concrete temperature field according to claim 4, characterized in that, The calculation of the error value includes: using the mean square error MSE as the loss function L(W) to calculate the error value. The formula for the loss function L(W) is: ， Among them, W is the weight coefficient matrix; n is the number of samples; y i is the i-th measured value; f(x i ) is the i-th predicted value.

6. A method for reconstructing the concrete temperature field according to claim 1, characterized in that: The iteration stop condition is the maximum value of the preset number of iterations.

7. A method for reconstructing the temperature field of concrete according to claim 1, characterized in that, It further includes a preprocessing step, and the preprocessing step includes: Normalize the temperature of the abnormal and missing temperature measurement points and the coordinates of the temperature measurement points.

8. A concrete temperature field reconstruction system, characterized in that, A concrete temperature field reconstruction method according to any one of claims 1-7, including: An acquisition module, configured to acquire the temperature of the temperature measurement point and the coordinates of the temperature measurement point, and store them into the temperature measurement point database; A preprocessing module, configured to normalize the temperature of the abnormal and missing temperature measurement points and the coordinates of the temperature measurement points; A database division module, configured to divide the temperature measurement point database into a training database and a test database; The LSTM-DNN network training module is used to obtain the output results of the LSTM model based on the temperatures of the temperature measurement points in the training database during the time period from t j-l to t j . Based on the temperatures of the temperature measurement points at time t j in the training database and the coordinates of the temperature measurement points in the test database, the interpolated temperatures of all the temperature measurement points corresponding to the test database at time t j are obtained through the OK spatial interpolation method. According to the output results of the LSTM model and the interpolated temperatures of the temperature measurement points, the training results of the temperature data of the temperature measurement points in the test database at time t j are obtained through the DNN model. It is judged whether the LSTM-DNN network reaches the iteration stop condition. If the iteration stop condition is not reached, the iteration training continues. If the stop condition is reached, the iteration training ends and the temperature field reconstruction is completed.

9. An electronic device, characterized in that, Including: A processor and a memory; The memory is used to store the executable instructions of the processor, and the processor is configured to execute a concrete temperature field reconstruction method according to any one of claims 1-7 by executing the executable instructions.

10. A computer storage medium, characterized in that: A computer program is stored on the computer storage medium, and when the computer program is executed by the processor, a concrete temperature field reconstruction method according to any one of claims 1-7 is implemented.

Images