Temperature monitoring method and system for vertical buried pipe ground source heat pump

Abstract

The invention provides a temperature monitoring method and system for a vertical buried pipe ground source heat pump, and belongs to the technical field of temperature measuration.The method and system comprise the steps that firstly, original temperature data are collected through a distributed optical fiber sensor, a temperature fluctuation matrix is constructed, and then time domain decomposition is conducted to obtain temperature fluctuation characteristic quantity; based on this, a temperature anomaly parameter is calculated and a temperature anomaly matrix is constructed. And inputting the temperature anomaly matrix into a pre-trained temperature error prediction model to obtain an error prediction value, and generating a temperature compensation matrix according to the error prediction value. And then a three-level compensation strategy is adopted, first-level compensation, second-level compensation and final correction of the temperature data are achieved in sequence, the pre-training model constructs training data through a conditional generative adversarial network embedded into a physical equation, and the reliability of the compensation effect is ensured. The technical problem that the temperature error compensation effect is poor due to the fact that temperature field distribution around the vertical buried pipe of the ground source heat pump is difficult to accurately measure in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of temperature measurement, and in particular, relates to a temperature monitoring method and system for a vertical buried pipe ground source heat pump. Background Art

[0002] The ground source heat pump system is an efficient and energy-saving technology that utilizes shallow underground geothermal energy. The vertical buried pipe heat exchange system is the most commonly used underground heat exchange method. In practical applications, accurately measuring and monitoring the temperature field distribution around the vertical buried pipe is of great significance for evaluating system performance, optimizing operating parameters, and predicting system life. Traditional temperature monitoring methods mainly rely on arranging discrete temperature sensors around the vertical buried pipes to obtain temperature distribution information through multi-point temperature measurement. With the development of distributed fiber optic sensing technology, it has the advantages of high spatial resolution, continuous measurement, and anti-electromagnetic interference, and has gradually become the main means of monitoring the temperature field of vertical buried pipes.

[0003] However, existing temperature monitoring technologies still face many challenges in practical applications. First, the underground environment is complex, and multiple factors such as soil properties, groundwater flow, and ambient temperature changes jointly affect the temperature field distribution. The coupling of these factors leads to large errors in temperature measurement. Secondly, due to the depth distribution characteristics of the underground temperature field, it is difficult for traditional temperature compensation methods to obtain sufficient calibration data, and the training data of the compensation model is seriously insufficient. Thirdly, existing temperature error compensation methods are mostly based on simplified models, which fail to fully consider the dynamic changes of the underground environment, and the compensation effect is limited. In addition, the uncertainty of groundwater flow and the nonlinear changes in soil thermophysical parameters pose severe challenges to the accuracy of temperature field measurement.

[0004] At present, the main technical means to solve the above problems include optimizing the sensor layout scheme, improving the signal processing algorithm, and establishing a temperature compensation model. However, these methods often only focus on the influence of a single factor and lack systematic consideration of the coupling of multiple factors in the underground environment. At the same time, since it is difficult to conduct large-scale actual measurements of the underground temperature field, the existing compensation model training data is seriously insufficient and the generalization ability of the model is limited. Therefore, how to achieve accurate measurement and effective compensation of the temperature field around the vertical buried pipe under the condition of data scarcity has become a key technical problem that needs to be solved urgently. In other words, there is a technical problem in the prior art that the temperature field distribution around the vertical buried pipe of the ground source heat pump is difficult to accurately measure, resulting in poor temperature error compensation effect. Summary of the invention

[0005] In view of this, the present invention provides a temperature monitoring method and system for a vertical buried pipe ground source heat pump, which can solve the technical problem in the prior art that the temperature field distribution around the vertical buried pipe of the ground source heat pump is difficult to accurately measure, resulting in poor temperature error compensation effect.

[0006] The present invention is implemented as follows: In a first aspect, the present invention provides a temperature monitoring method for a vertical buried pipe ground source heat pump, wherein a distributed optical fiber sensor is used to collect a sequence of original temperature data from multiple measuring points to construct a temperature fluctuation matrix; the temperature fluctuation matrix is ​​decomposed in the time domain to extract the fluctuation amplitude, fluctuation frequency and fluctuation phase of the temperature data to obtain a temperature fluctuation characteristic; the original temperature data is normalized based on the temperature fluctuation characteristic to generate a standardized temperature data sequence; temperature anomaly parameters are calculated based on the standardized temperature data sequence to construct a temperature anomaly matrix; the temperature anomaly matrix is ​​input into a pre-trained temperature error prediction model to obtain a temperature error prediction value; a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix is ​​established to generate a temperature compensation matrix; and the temperature compensation matrix is ​​used to normalize the temperature data collected in real time. Perform first-level compensation to obtain first-level corrected temperature data; calculate the steady-state deviation value between the first-level corrected temperature data and the historical calibration data; establish a temperature correction model according to the steady-state deviation value; use the temperature correction model to perform second-level compensation on the first-level corrected temperature data to obtain second-level corrected temperature data; calculate system dynamic response parameters; use the system dynamic response parameters to perform final correction on the second-level corrected temperature data, and output the corrected temperature data; it is characterized in that the pre-trained temperature error prediction model adopts a deep convolutional neural network structure, and the deep convolutional neural network structure consists of a feature extraction network layer, a feature fusion network layer and a prediction network layer; the temperature correction model includes a temperature offset compensation term and a reference point correction term; the system dynamic response parameters include temperature response time, temperature maximum value time and temperature adjustment time.

[0007] Among them, the feature extraction network layer adopts a four-layer convolution layer structure, the number of convolution kernels in each convolution layer is 32, 64, 128, and 256 respectively, the convolution kernel size is 3×3, the step size is 1, and each convolution layer is connected to a normalization network layer and an activation function layer. The feature fusion network layer adopts a fully connected layer structure with 1024 neurons. The prediction network layer adopts a linear output layer, and the output dimension is the number of measurement points.

[0008] Among them, the training data set of the pre-trained temperature error prediction model constructs virtual training data of underground temperature distribution through a conditional generative adversarial network. The conditional generative adversarial network includes a generative network and a discriminative network. The generative network has an embedded temperature generation equation group for simulating the distribution law of the underground temperature field.

[0009] Among them, the temperature generation equation group includes a heat conduction equation, a water flow influence equation and a thermal conductivity equation. The heat conduction equation is used to describe the heat conduction process in the soil around the vertical buried pipe, the water flow influence equation is used to simulate the influence of groundwater flow on the temperature field, and the thermal conductivity equation is used to calculate the thermal conductivity of the soil under different depths and different water contents.

[0010] Among them, in the process of performing primary compensation using the temperature compensation matrix, the compensation coefficient includes a correction weight and an offset, the correction weight ranges from 0 to 1, and the offset ranges from plus or minus 5 degrees Celsius; the steady-state deviation value between the primary corrected temperature data and the historical calibration data is calculated using a sliding time window method, the time window size is set to 1 hour, and the window sliding step is 10 minutes.

[0011] The temperature offset compensation term is calculated using the exponential smoothing method, the smoothing coefficient is 0.3, the reference point correction term is calculated using the piecewise linear interpolation method, and 5 reference points are evenly selected within the measurement range.

[0012] The temperature response time is defined as the time required for the temperature change to reach 63.2% of the final steady-state value, and the temperature adjustment time is defined as the time required for the temperature to enter the steady-state interval, and the range of the steady-state interval is plus or minus 2% of the final steady-state value.

[0013] Among them, the local anomaly factor algorithm is used to detect the degree of anomaly of temperature data. The size of the local neighborhood is taken as 10% of the total number of measuring points, and the measuring points with a local anomaly factor greater than 2 are marked as abnormal points.

[0014] The empirical mode decomposition method is used to decompose the temperature fluctuation matrix, and the obtained intrinsic mode function sequence is Hilbert transformed to calculate the instantaneous frequency and instantaneous amplitude, and extract the fluctuation amplitude, fluctuation frequency and fluctuation phase corresponding to the main frequency components.

[0015] The second aspect of the present invention provides a temperature monitoring system for a vertical buried pipe ground source heat pump, comprising a box body, a vertical plate fixedly installed on the top of the box body, a gear rotatably installed on one end of the vertical plate, tooth plates movably installed on both ends of the vertical plate, the tooth plates are meshingly connected with the outer surface of the gear, connecting rods are respectively fixedly installed on the tooth plates, a movable door is fixedly installed on one end of the connecting rod, and the movable door includes a left movable door and a right movable door, a roller and a rotating shaft are rotatably installed on the other end of the vertical plate, a belt is movably sleeved on the outer surfaces of the roller and the rotating shaft, one end of the roller passes through the vertical plate and is fixedly connected to one end of the gear, a circular shaft is fixedly installed on one end of the rotating shaft, a protective box is fixedly installed on one end of the vertical plate, a motor is fixedly installed on one end of the protective box, and the output end of the motor passes through the protective box and is fixedly connected to one end of the circular shaft.

[0016] Among them, a fixing rod is fixedly installed at one end of the box body, a strip groove is opened on the top of the fixing rod, a round rod is fixedly installed inside the strip groove, and moving blocks are movably sleeved on both sides of the surface of the round rod, and the top of the moving block is fixedly connected to the bottom of the movable door.

[0017] Among them, a card slot is opened on one side of the left movable door, and a card plate is fixedly installed on one side of the right movable door, and the card slot is adapted to the card plate.

[0018] Among them, sensing optical fibers are fixedly installed at equal distances on the bottom of the box, the top of the sensing optical fibers penetrates the box and extends to the inner surface of the box, and the surface of the sensing optical fibers is movably sleeved with sleeve rods.

[0019] Among them, both ends of the bottom of the sleeve rod are provided with sliding grooves, the inside of the sliding grooves are fixedly installed with cylinders, the surface of the cylinders are movably sleeved with sliders, springs are fixedly installed between the sliders and the sliding grooves, the inner surface of the springs is movably connected to the outer surface of the cylinders, and a sealing rod is fixedly installed at the bottom of the sliders.

[0020] Wherein, both ends of the top of the sleeve rod are provided with limiting grooves, and moving rods are movably installed inside the limiting grooves. The bottom of the moving rod is fixedly connected to the top of the slider, and a clamping ring is fixedly installed on the top of the moving rod.

[0021] A temperature detection module and a data processing module are fixedly installed at one end of the interior of the box, and a data transmission module and a power supply module are fixedly installed at one end of the interior of the box, which are located on the right side of the temperature detection module and the data processing module.

[0022] Wherein, a mounting plate is fixedly installed on the top of the box body, and mounting holes are opened at both ends of the mounting plate.

[0023] Wherein, heat dissipation grooves are provided at equal distances on both sides of the box body, and the heat dissipation grooves have the same size.

[0024] Wherein, a display screen is fixedly installed at one end of the left movable door, and a PLC controller is fixedly installed at one end of the right movable door.

[0025] Among them, the data processing module adopts a single-chip microcomputer, and a storage medium is arranged in the data processing module. The storage medium stores program code. When the single-chip microcomputer executes the program code, it is used to execute the temperature monitoring method of the vertical buried pipe ground source heat pump.

[0026] During daily use of the system, the operator starts the motor, and the operation of the motor causes the round shaft to rotate, and then the round shaft drives the rotating shaft to rotate, and then the rotating shaft drives the belt to rotate, and then the belt drives the roller to rotate, and then the roller drives the gear to rotate, and then the gear drives the toothed plate to slide, and then the toothed plate drives the connecting rod to slide, and at the same time the connecting rod drives the movable door to move towards each other, and then the movable door moves to slide inside the strip groove and on the surface of the round rod until the card plate is placed inside the card slot, so that the box can be automatically closed, otherwise it can be opened, so as to facilitate the maintenance of multiple modules inside the box.

[0027] The system, during daily use, when the sensing optical fiber is connected to the vertical buried pipe, the sealing rod is pulled, and the sealing rod drives the slider to slide inside the slide groove and on the surface of the cylinder when sliding, and then the slider squeezes and compresses the spring, at which time the slider drives the moving rod to slide inside the limit groove, and then the moving rod drives the clamping ring to slide, and then the sleeve rod is slid until the sealing rod is placed at the connection between the sensing optical fiber and the vertical buried pipe, and at the same time the sealing rod is released, and then the clamping ring clamps the sensing optical fiber, and then the sealing rod can be used to auxiliary fix the connection between the sensing optical fiber and the vertical buried pipe to make the connection tighter.

[0028] Compared with the prior art, the temperature monitoring method and system of a vertical buried pipe ground source heat pump provided by the present invention realizes the generation of virtual data of underground temperature field distribution by constructing a conditional generative adversarial network based on a physical model, innovatively solving the problem of lack of training data. This method integrates physical processes such as the basic principles of heat conduction, the influence of groundwater flow, and changes in soil thermal conductivity into the generation network, ensuring the physical rationality of the generated data. At the same time, through the adversarial training of the discriminant network, the quality of the generated data is continuously optimized, and the reliability of the virtual training data is improved.

[0029] On this basis, the present invention designs a multi-level temperature error compensation strategy. First, the spatiotemporal variation characteristics of the temperature field are captured by the temperature fluctuation matrix, and a temperature anomaly detection mechanism is established. Secondly, the pre-trained deep convolutional neural network model is used to achieve accurate prediction of temperature errors. Thirdly, by establishing a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix, a dynamic temperature compensation matrix is ​​generated. Finally, a multi-level compensation strategy is adopted to sequentially realize steady-state error compensation, dynamic error correction, and system response characteristic compensation of temperature data, significantly improving the accuracy of temperature measurement.

[0030] The present invention effectively solves the technical problem in the prior art that the temperature field distribution around the vertical buried pipe of the ground source heat pump is difficult to accurately measure, resulting in poor temperature error compensation effect. By combining the physical model with the deep learning method, the physical interpretability of the model is guaranteed and the reliability of the compensation effect is improved. At the same time, the innovative use of conditional generative adversarial networks to construct training data overcomes the limitation of the lack of measured data in the underground environment and provides a new technical approach to improve the accuracy of temperature field measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a front view of the present invention; Figure 2 is a cross-sectional view of a vertical plate of the present invention; Figure 3 It is a front cross-sectional view of the present invention; Figure 4 is a side sectional view of the present invention; Figure 5 For the present invention Figure 4 A partial enlarged view of the middle part; Figure 6 For the present invention Figure 4 A partial enlarged view of point B in the middle; Figure 7 It is a schematic diagram of the internal structure of the box of the present invention; Figure 8 For the present invention Figure 7 A partial enlarged view of point C in the middle; Fig. 9 It is a schematic diagram of the top structure of the box body of the present invention; Fig.10 It is a schematic diagram of the bottom structure of the box body of the present invention.

[0032] Fig.11 is a flow chart of the method of the present invention.

[0033] In the attached drawings: 1. box body; 2. vertical plate; 3. gear; 4. tooth plate; 5. connecting rod; 61. left movable door; 62. right movable door; 7. roller; 8. rotating shaft; 9. belt; 10. round shaft; 11. protective box; 12. motor; 13. fixing rod; 14. strip groove; 15. round rod; 16. moving block; 17. card slot; 18. card plate; 19. sensing optical fiber; 20. sleeve rod; 21. slide groove; 22. cylinder; 23. slider; 24. spring; 25. sealing rod; 26. limit groove; 27. moving rod; 28. clamping ring; 29. ​​temperature detection module; 30. data processing module; 31. data transmission module; 32. power module; 33. mounting plate; 34. mounting hole; 35. heat dissipation slot; 36. display screen; 37. PLC controller. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0035] like Fig.11 FIG. 1 is a flow chart of a temperature monitoring method for a vertical buried pipe ground source heat pump provided by the present invention. The method comprises the following steps: S01. Collecting original temperature data sequences of multiple measuring points within a preset time interval through a distributed optical fiber sensor, and constructing a temperature fluctuation matrix according to the trend of temperature values ​​in the original temperature data sequence changing with time; S02, performing time domain decomposition on the temperature fluctuation matrix, extracting the fluctuation amplitude, fluctuation frequency and fluctuation phase of the temperature data, and obtaining the temperature fluctuation characteristic quantity; S03, performing data normalization processing on the original temperature data based on the temperature fluctuation characteristic quantity to generate a standardized temperature data sequence; S04, calculating temperature anomaly parameters according to the standardized temperature data sequence, and constructing a temperature anomaly matrix based on the temperature anomaly parameters according to the corresponding relationship between the time series and the spatial position; S05, inputting the temperature anomaly matrix into a pre-trained temperature error prediction model to obtain a temperature error prediction value; S06, establishing a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix according to the temperature error prediction value, and generating a temperature compensation matrix; S07, using the compensation coefficient of the corresponding position in the temperature compensation matrix to perform primary compensation on the temperature data collected in real time to obtain primary corrected temperature data; S08, calculating the steady-state deviation value between the primary corrected temperature data and the historical calibration data; S09, establishing a temperature correction model according to the steady-state deviation value, wherein the temperature correction model includes a temperature offset compensation term and a reference point correction term; S10, performing secondary compensation on the primary corrected temperature data using the temperature correction model to obtain secondary corrected temperature data; S11, calculating system dynamic response parameters according to the secondary corrected temperature data, including temperature response time, temperature maximum time and temperature adjustment time; S12, performing final correction on the secondary corrected temperature data using the system dynamic response parameter, and outputting the corrected temperature data; The pre-trained temperature error prediction model adopts a deep convolutional neural network structure, which is composed of a feature extraction network layer, a feature fusion network layer and a prediction network layer. The feature extraction network layer adopts a four-layer convolution layer structure, and the number of convolution kernels in each convolution layer is 32, 64, 128, and 256 respectively. The convolution kernel size is 3×3, the step size is 1, and each convolution layer is followed by a normalization network layer and an activation function layer. The feature fusion network layer adopts a fully connected layer structure with 1024 neurons. The prediction network layer adopts a linear output layer, and the output dimension is the number of measurement points. The training data set establishment step of the pre-trained temperature error prediction model constructs virtual training data of underground temperature distribution through a conditional generative adversarial network, wherein the conditional generative adversarial network includes a generative network and a discriminative network, and the generative network has an embedded temperature generation equation group for simulating the distribution law of underground temperature field; The temperature generation equation group includes: The heat conduction equation is used to describe the heat conduction process in the soil around the vertical buried pipe. The input includes soil thermal conductivity, soil mass density, soil specific heat at constant pressure, soil initial temperature distribution and soil boundary temperature. The output is the temperature field distribution value at different depths in the vertical direction. The soil thermal conductivity, soil mass density and soil specific heat at constant pressure are obtained through soil sample experiments. The soil initial temperature distribution and the soil boundary temperature are measured by temperature sensors. The water flow influence equation is used to simulate the influence of groundwater flow on the temperature field. The input includes groundwater flow velocity, groundwater temperature, soil permeability, soil porosity and water flow angle. The output is the temperature field correction value after considering the groundwater convection heat transfer. The groundwater flow velocity and groundwater temperature are obtained through groundwater monitoring equipment, the soil permeability and soil porosity are obtained through soil penetration experiments, and the water flow angle is obtained through a water flow detector. The thermal conductivity equation is used to calculate the thermal conductivity of soil at different depths and different water contents. The input includes soil apparent density, soil water content, soil mineral mass ratio, soil organic matter mass ratio and soil temperature value. The output is the soil thermal conductivity value at the corresponding position. The soil apparent density, soil water content, soil mineral mass ratio and soil organic matter mass ratio are obtained through soil composition experiments, and the soil temperature value is obtained by measuring with a temperature sensor. Among them, the temperature fluctuation matrix refers to a multidimensional data matrix formed by arranging the collected temperature data according to the corresponding relationship between time series and spatial position; The temperature anomaly matrix refers to the characteristic representation matrix obtained after anomaly detection of temperature data, which is used to indicate the degree of deviation of temperature data; The temperature compensation matrix refers to a calibration coefficient matrix used to correct temperature measurement errors, which includes temperature compensation coefficients and correction weight coefficients.

[0036] The specific implementation of the above steps is described in detail below.

[0037] The specific implementation method of step S01 is to collect temperature data through distributed optical fiber sensors and construct a temperature fluctuation matrix. First, a distributed Brillouin scattering optical fiber temperature measurement system is selected as the data acquisition device, and the optical fiber sensor is spirally wound and installed along the surface of the vertical buried pipe. The measurement point spacing is set to 0.5 meters, the sampling frequency is set to 1 Hz, and the continuous collection time is not less than 24 hours. The collected original temperature data includes timestamps, spatial position coordinates, and corresponding temperature values. The original data is preprocessed, including outlier removal, data smoothing, and time synchronization. The preprocessing adopts the principle of three times the standard deviation to remove outliers, and uses the sliding average method for data smoothing. The time window size is set to 60 seconds. Subsequently, the temperature fluctuation matrix is ​​constructed according to the time series and depth position of the preprocessed temperature data. The rows of the matrix represent different moments, and the columns represent the measurement point positions at different depths. The purpose of this step is to obtain accurate and reliable original temperature data sequences to provide a data basis for subsequent analysis.

[0038] The specific implementation method of step S02 is to perform time domain decomposition on the temperature fluctuation matrix to extract feature quantities. The temperature fluctuation matrix is ​​decomposed by the empirical mode decomposition method. First, the local maximum and minimum points of the temperature data are determined, and the upper and lower envelopes are obtained by cubic spline interpolation, and the mean envelope is calculated. The first intrinsic mode function is obtained by subtracting the mean envelope from the original data, and the above process is repeated until the remaining signal becomes a monotonic function. The obtained intrinsic mode function sequence is Hilbert transformed, the instantaneous frequency and instantaneous amplitude are calculated, and the fluctuation amplitude, fluctuation frequency and fluctuation phase corresponding to the main frequency components are extracted. The fluctuation amplitude represents the severity of the temperature change, the fluctuation frequency reflects the speed of the temperature change, and the fluctuation phase reflects the leading or lagging relationship of the temperature change. The complex temperature fluctuation signal can be decomposed into several components with obvious characteristics through time domain decomposition, which is conducive to in-depth analysis of the temperature change law.

[0039] The specific implementation method of step S03 is to perform data normalization based on the temperature fluctuation characteristic quantity. First, the mean and standard deviation of the fluctuation amplitude are calculated, and the amplitude data is standardized to a zero mean unit variance distribution. The fluctuation frequency is normalized to the minimum and maximum values, and the frequency data is mapped to the interval of 0 to 1. The fluctuation phase is periodically normalized, and the phase angle is converted to the range of negative π to π. Then, the original temperature data is weighted and normalized according to the standardized characteristic quantity, and the weight coefficient is determined by the proportion of temperature fluctuation energy. The greater the fluctuation energy, the greater the weight of the characteristic quantity. Finally, a standardized temperature data sequence is obtained, which eliminates the scale differences between different measuring points and makes the data comparable. The role of this step is to convert temperature data of different dimensions into a unified scale space for subsequent analysis and modeling.

[0040] The specific implementation method of step S04 is to calculate the temperature anomaly parameters to construct a temperature anomaly matrix. The local anomaly factor algorithm is used to detect the degree of anomaly of the temperature data. First, the size of the local neighborhood is determined, and the value is 10% of the total number of measuring points. The distance between each measuring point and its neighborhood point is calculated, and the Euclidean distance is selected as the distance metric. The local reachable density of each measuring point is counted, and the local anomaly factor is obtained by comparing the local reachable density of the neighborhood point. The measuring points with a local anomaly factor greater than 2 are marked as abnormal points, and the degree of anomaly is represented by the value of the anomaly factor. Then, the temperature anomaly matrix is ​​constructed according to the correspondence between the anomaly factor and the spatial position. The larger the value in the matrix, the higher the degree of anomaly of the temperature data at that position. The purpose of this step is to identify abnormal patterns in temperature data and provide a basis for subsequent error prediction.

[0041] The specific implementation method of step S05 is to use a pre-trained deep convolutional neural network to predict the temperature error. The network consists of three functional network layers: feature extraction, feature fusion and prediction. The feature extraction network adopts a four-layer convolution structure, and the activation function uses a corrected linear unit function. In the feature extraction network, the first layer of convolution kernels is 32 3×3 in size, which extracts local temperature change features; the second layer of convolution kernels is 64 3×3 in size, which extracts temperature gradient features; the third layer of convolution kernels is 128 3×3 in size, which extracts temperature curvature features; the fourth layer of convolution kernels is 256 3×3 in size, which extracts high-order temperature features. The feature fusion network uses a fully connected layer of 1024 neurons to combine multi-scale features. The prediction network uses a linear output layer, and the output dimension is the same as the number of measurement points. This step predicts the temperature measurement error through a deep learning method to provide a benchmark for temperature data correction.

[0042] The specific implementation of step S06 is to establish a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix. The mapping model is constructed by using the multivariate regression analysis method, the independent variable is the temperature fluctuation feature quantity, and the dependent variable is the temperature anomaly parameter. First, feature selection is performed to calculate the correlation coefficient between the feature quantity and the anomaly parameter, and the feature quantity with an absolute value of the correlation coefficient greater than 0.3 is selected as the effective feature. Then a multivariate linear regression equation is established, and the regression coefficient is estimated by the least squares method. A significance test is performed on the regression equation. If the P value is less than 0.05, it means that the regression relationship is significant. Finally, a temperature compensation matrix is ​​generated according to the regression coefficient, and the numerical value in the matrix represents the temperature compensation coefficient of the corresponding position. The role of this step is to quantify the relationship between temperature fluctuation and temperature anomaly, and provide a basis for temperature data correction. The compensation matrix takes into account the corresponding relationship between the temperature fluctuation characteristics and the degree of anomaly, and can effectively correct the temperature measurement error.

[0043] The specific implementation method of step S07 is to use the temperature compensation matrix for primary compensation. When collecting temperature data in real time, first obtain the compensation coefficient of the corresponding position of the current measuring point in the temperature compensation matrix, and the compensation coefficient includes two parts: correction weight and offset. Linear correction is performed on the real-time temperature data, and the correction formula is the original temperature value plus the offset multiplied by the correction weight. The value range of the correction weight is 0 to 1, and the value range of the offset is plus or minus 5 degrees Celsius. Different compensation coefficients are used for measuring points of different depths. The correction weight of shallow measuring points is larger, and the correction weight of deep measuring points is smaller. The compensated temperature data is called primary corrected temperature data, which eliminates the systematic error in temperature measurement. The role of this step is to make a preliminary correction to the temperature data through the compensation matrix to improve the accuracy of temperature measurement.

[0044] The specific implementation method of step S08 is to calculate the steady-state deviation value between the first-level corrected temperature data and the historical calibration data. First, a historical calibration database is established, which contains standard temperature values ​​under different working conditions, and the standard temperature values ​​are obtained by calibrating high-precision temperature sensors. Match the first-level corrected temperature data with the historical calibration data, and select the data under the same working conditions for comparison. The sliding time window method is used to calculate the steady-state deviation value, the time window size is set to 1 hour, and the window sliding step is 10 minutes. In each time window, the root mean square error between the corrected temperature data and the calibration data is calculated as the steady-state deviation value. The purpose of this step is to evaluate the systematic deviation between the first-level corrected temperature data and the standard value, and provide a basis for the second-level compensation.

[0045] The specific implementation method of step S09 is to establish a temperature correction model. The model includes two parts: a temperature offset compensation item and a reference point correction item. The temperature offset compensation item is used to correct the zero drift of the temperature measurement, and the reference point correction item is used to correct the linearity error of the temperature measurement. The temperature offset compensation item is calculated using the exponential smoothing method, and the smoothing coefficient is 0.3, taking into account the cumulative effect of temperature drift. The reference point correction item is calculated using the piecewise linear interpolation method, and 5 reference points are evenly selected within the measurement range. The reference point temperature value is determined by on-site calibration. For any measuring point, the reference point interval is determined according to its temperature value, and the correction coefficient is calculated using linear interpolation. The temperature correction model established in this step can effectively compensate for the nonlinear error in temperature measurement.

[0046] The specific implementation method of step S10 is to use the temperature correction model for secondary compensation. First, the offset of the primary corrected temperature data is calculated, and the temperature data is subtracted from the offset compensation value to obtain the zero-point correction result. Then the reference point interval where the temperature value is located is determined, and the linear correction coefficient within the interval is calculated. The zero-point correction result is multiplied by the linear correction coefficient to obtain the secondary corrected temperature data. The secondary correction takes into account the nonlinear characteristics of temperature measurement and can more accurately reflect the actual temperature value. The role of this step is to further improve the accuracy of temperature measurement and make the corrected temperature data closer to the true value.

[0047] The specific implementation of step S11 is to calculate the dynamic response parameters of the system. The temperature response time is defined as the time required for the temperature change to reach 63.2% of the final steady-state value, which is calculated by exponential fitting the secondary corrected temperature data. The temperature maximum time is the moment when the temperature reaches the peak value, which is determined by finding the local maximum value of the temperature data. The temperature adjustment time is defined as the time required for the temperature to enter the steady-state interval, and the range of the steady-state interval is plus or minus 2% of the final steady-state value, which is determined by judging the temperature fluctuation amplitude. These dynamic response parameters reflect the dynamic characteristics of the temperature measurement system and can be used to evaluate the control performance of the system.

[0048] The specific implementation method of step S12 is to make a final correction based on the dynamic response parameters of the system. First, a dynamic compensation model is established, and the model input is the temperature response time, the maximum temperature time and the temperature adjustment time, and the output is the dynamic correction coefficient. The fuzzy control method is used to design the dynamic compensation rules, and the correction coefficient is adjusted according to the size of the dynamic response parameters. When the response time is long, the correction coefficient is increased to speed up the response speed; when the adjustment time is long, the correction coefficient is reduced to suppress temperature fluctuations. The secondary corrected temperature data is multiplied by the dynamic correction coefficient to obtain the final correction result. The purpose of this step is to optimize the dynamic characteristics of temperature measurement and improve the control accuracy of the system. The final corrected temperature data has high accuracy and good dynamic characteristics, and reliably reflects the temperature distribution around the vertical buried pipe.

[0049] The calculation process involved in the present invention is described in detail as follows.

[0050] 1. The construction process of the temperature fluctuation matrix is ​​specifically expressed as follows: ; In the formula, is the temperature fluctuation matrix; Indicates Moment The temperature value of each measuring point; is the number of sampling time points; is the number of measuring points.

[0051] The calculation process of matrix elements: ; In the formula, is the reference temperature value, obtained through on-site calibration; is the temperature fluctuation, obtained by measurement; is the measurement error, which follows a normal distribution The steps of constructing the temperature fluctuation matrix are as follows: first, preprocess the temperature measurement data, including outlier removal and data smoothing; then arrange the processed data by time series and spatial position; finally, form dimensional matrix. This matrix can reflect the temporal and spatial distribution characteristics of the temperature field.

[0052] 2. The mathematical expression of the empirical mode decomposition process is as follows: ; In the formula, is the original temperature signal; For the Intrinsic mode functions; is the remaining signal; is the number of decomposition layers.

[0053] The extraction process of the intrinsic mode function is: ; ; In the formula, For the The signal obtained by the second screening; is the mean of the upper and lower envelopes. The screening termination condition is: ; In the formula, is the standard deviation; is the threshold value, which is 0.2. This method can adaptively decompose nonlinear and non-stationary signals.

[0054] 3. The calculation process of Hilbert transform is as follows: ; ; In the formula, is the Hilbert transform; To analyze the signal; is the instantaneous amplitude; is the instantaneous phase; Is an imaginary unit.

[0055] 4. The calculation formula for data normalization is as follows: ; In the formula, is the normalized data; is the original data; is the mean; is the standard deviation.

[0056] 5. The calculation formula of local anomaly factor is as follows: ; ; In the formula, For point The local reachable density of For point of Neighbor set; is the reachable distance; is the local abnormal factor.

[0057] 6. The expression of the heat conduction equation is as follows: ; In the formula, is the soil density, in kg / m³; is the specific heat capacity, in J / (kg·K); is the temperature, in K; is time, unit is s; is the thermal conductivity, unit is W / (m·K); is the heat source term, and its unit is W / m³.

[0058] 7. The expression of the water flow influence equation is as follows: ; In the formula, is the water velocity component, in m / s; is the thermal diffusion coefficient, in m² / s.

[0059] 8. The expression of thermal conductivity equation is as follows: ; In the formula, is the thermal conductivity of dry soil, in W / (m·K); is the moisture content; is the water density; is the specific heat of water; is the volumetric moisture content; is the saturated moisture content; is the empirical coefficient, and its value range is 0.5~0.8; is the thermal conductivity enhancement coefficient.

[0060] 9. The temperature compensation coefficient calculation formula is as follows: ; In the formula, is the compensation coefficient; is the benchmark compensation coefficient; is the temperature coefficient; is the temperature deviation.

[0061] 10. The formula for exponential smoothing is as follows: ; In the formula, for Smoothed value at time instant; for The observed value at time; is the smoothing coefficient, and its value is 0.3.

[0062] 11. The expression of the dynamic compensation model is as follows: ; In the formula, is the dynamic correction coefficient; is the temperature response time; is the time of maximum temperature; Adjust the time for temperature; is the corresponding reference time value; is the weight coefficient and satisfies .

[0063] 12. The expression of the temperature correction model is as follows: ; In the formula, is the corrected temperature value; To measure the temperature value; is the offset compensation coefficient; is the temperature offset; is the reference point correction factor; is the reference point temperature value.

[0064] Among them, the offset compensation term is: ; In the formula, is the initial compensation coefficient; is the attenuation coefficient. Reference point correction term: ; In the formula, is the correction coefficient, which is obtained by least square fitting. The model takes into account the nonlinear characteristics of temperature measurement and the time drift effect.

[0065] 13. The calculation formula of steady-state deviation value is as follows: ; In the formula, is the root mean square error; is the number of data points in the time window; is the corrected temperature value; is the calibration temperature value.

[0066] 14. The temperature anomaly matrix is ​​expressed as follows: ; In the formula, For the Moment The calculation formula for the abnormal factor value of each measuring point has been given above. The construction of the abnormal matrix takes into account the local density characteristics of the temperature data and can effectively identify abnormal measuring points.

[0067] When used as input to a deep convolutional neural network, the expression is: ; In the formula, is the output of the first layer of convolutional network; is the convolution kernel weight; Represents the convolution operation; is the bias term; is the activation function. This matrix is ​​used to characterize the abnormal characteristics of the temperature field and help predict temperature measurement errors.

[0068] 15. Dynamic compensation coefficient vector: ; Motion compensation coefficient vector During temperature correction use: ; In the formula, is the final corrected temperature value; It is the secondary correction temperature value; is a dynamic characteristic parameter vector, including temperature response time, temperature maximum time and temperature adjustment time. This vector is used to adjust the temperature dynamic response characteristics.

[0069] Dynamic compensation coefficient matrix: ; In the formula, is the dynamic compensation coefficient matrix element, indicating the The measurement point The influence coefficient of each measuring point. The calculation of dynamic compensation is: ; In the formula, is the temperature vector after compensation; is the measured temperature vector. This compensation method takes into account the mutual influence between the measuring points.

[0070] Dynamic compensation coefficient matrix In the final temperature correction use: ; In the formula, is the final corrected temperature vector; is the secondary correction temperature vector; is the corrected error vector. This matrix takes into account the spatial correlation between the measuring points and is used to optimize the overall distribution of the temperature field.

[0071] Specifically, the principle of the present invention is: the technical principle of the present invention is based on a deep learning framework driven by a physical model. First, the heat conduction equation, the water flow influence equation and the thermal conductivity equation are embedded in the conditional generative adversarial network. These three core equations respectively describe the heat conduction process, the influence of groundwater flow and the change of soil thermal conductivity characteristics. In this way, it is ensured that the generated data meets the constraints of physical laws. Among them, the heat conduction equation describes the diffusion characteristics of the temperature field, the water flow influence equation describes the influence of convective heat transfer, and the thermal conductivity equation reflects the dynamic changes of soil thermal physical parameters. The coupled solution of these three equations realizes the accurate simulation of the evolution of the temperature field in a complex underground environment.

[0072] On this basis, the present invention adopts a deep convolutional neural network to construct a temperature error prediction model. The network extracts the spatiotemporal characteristics of temperature data through a multi-layer convolution structure, uses a feature fusion network layer to achieve effective integration of multi-scale features, and finally outputs the temperature error prediction value through a prediction network layer. This network structure design fully considers the spatial correlation and temporal continuity of the temperature field distribution, and can effectively capture the changing law of the temperature error. At the same time, the present invention innovatively proposes a multi-level compensation strategy, which realizes comprehensive compensation from steady-state error to dynamic response characteristics through the synergistic effect of the temperature fluctuation matrix, the temperature anomaly matrix and the temperature compensation matrix.

[0073] A specific embodiment 1 of the method of the present invention is provided below, and the specific implementation method of each step in this embodiment 1 is described in detail as follows.

[0074] The specific implementation method of step S01 is to collect temperature data and construct a temperature fluctuation matrix through a distributed Brillouin scattering optical fiber temperature measurement system. The system uses a pulsed light source to emit a laser with a wavelength of 1550 nanometers, and couples the optical signal into the sensing optical fiber through a fiber coupler. The optical signal interacts with the medium during transmission to produce Brillouin scattering. A single-mode optical fiber is selected as the sensing optical fiber and installed along the surface of the vertical buried pipe by spiral winding. The winding spacing is 0.1 meters, the measurement point spacing is set to 0.5 meters, the sampling frequency is 1 Hz, and the continuous collection time is not less than 24 hours. The collected raw temperature data includes timestamps, spatial position coordinates, and corresponding temperature values. Data preprocessing is first performed, including outlier removal, data smoothing, and time synchronization. The outlier removal adopts the principle of three times the standard deviation to calculate the mean of the data sequence. and standard deviation , will deviate from the mean by more than The data are marked as outliers and removed. The sliding average method is used for data smoothing. The time window size is set to 60 seconds. The arithmetic mean of the data in each window is calculated as the smoothing result. The linear interpolation method is used for time synchronization to unify the sampling time of different measuring points. Then the temperature fluctuation matrix is ​​constructed according to the time series and depth position of the preprocessed temperature data. . Matrix elements The calculation formula is ,in is the reference temperature value, obtained through on-site calibration. is the temperature fluctuation, obtained by measurement, is the measurement error, which follows a normal distribution The purpose of this step is to obtain accurate and reliable original temperature data series to provide a data basis for subsequent analysis.

[0075] The specific implementation of step S02 is to extract the characteristic quantity by time domain decomposition of the temperature fluctuation matrix. The temperature fluctuation matrix is ​​decomposed by the empirical mode decomposition method, which can adaptively decompose nonlinear and non-stationary signals. The basic principle of empirical mode decomposition is to decompose a complex signal into a finite number of intrinsic mode functions and a residual term. The decomposition formula is: ,in is the original temperature signal, For the The intrinsic mode functions, is the residual signal, is the number of decomposition layers. The extraction process of the intrinsic mode function adopts the iterative screening method. First, the local maximum and minimum points of the signal are calculated, and the upper and lower envelopes are obtained by cubic spline interpolation, and the mean envelope is calculated. The original data is subtracted from the mean envelope to obtain the candidate intrinsic mode function, and the above process is repeated until the termination condition is met. The mathematical expression of the iterative process is , ,in For the The signal obtained by the screening is is the mean of the upper and lower envelopes. The screening termination condition is the standard deviation Less than threshold , the calculation formula is ,in The value is 0.2. The Hilbert transform of the obtained intrinsic mode function sequence is calculated as follows: , ,in is the Hilbert transform, To analyze the signal, is the instantaneous amplitude, is the instantaneous phase, is an imaginary unit. The fluctuation amplitude, fluctuation frequency and fluctuation phase corresponding to the main frequency components are extracted through Hilbert transform. The fluctuation amplitude represents the severity of temperature change, the fluctuation frequency reflects the speed of temperature change, and the fluctuation phase reflects the leading or lagging relationship of temperature change.

[0076] The specific implementation of step S03 is to perform data normalization based on the temperature fluctuation characteristic quantity. First, the mean and standard deviation of the fluctuation amplitude are calculated, and the amplitude data is converted to a zero mean unit variance distribution using a standardized method. The calculation formula is: ,in is the normalized data, is the original data, is the mean, is the standard deviation. The fluctuation frequency is normalized to the minimum and maximum values, and the frequency data is mapped to the range of 0 to 1. The fluctuation phase is periodically normalized, and the phase angle is converted to the range of negative π to π. Then the original temperature data is weighted and normalized according to the standardized feature quantity. The weight coefficient is determined by the proportion of temperature fluctuation energy. The larger the fluctuation energy, the greater the weight of the feature quantity. Finally, a standardized temperature data sequence is obtained, which eliminates the scale differences between different measuring points and makes the data comparable. The purpose of this step is to convert temperature data of different dimensions into a unified scale space for subsequent analysis and modeling.

[0077] The specific implementation of step S04 is to calculate the temperature anomaly parameters and construct a temperature anomaly matrix. The local anomaly factor algorithm is used to detect the degree of anomaly of the temperature data. First, the size of the local neighborhood is determined, and the value is 10% of the total number of measurement points. The distance between each measurement point and its neighborhood point is calculated, and the Euclidean distance is used as the distance metric. The local reachable density of each measurement point is calculated, and the calculation formula is: ,in For point The local reachable density of For point of Neighbor set, is the reachable distance. Then calculate the local anomaly factor, the formula is ,in is the local anomaly factor. The measuring points with a local anomaly factor greater than 2 are marked as abnormal points, and the degree of abnormality is represented by the value of the anomaly factor. The temperature anomaly matrix is ​​constructed based on the calculated anomaly factors. , the matrix elements Indicates Moment The purpose of this step is to identify abnormal patterns in temperature data and provide a basis for subsequent error prediction.

[0078] The specific implementation method of step S05 is to use a pre-trained deep convolutional neural network to predict temperature errors. The network consists of three functional network layers: feature extraction, feature fusion, and prediction, and each layer adopts a specific structure and parameter settings. The feature extraction network adopts a four-layer convolution structure, and each convolution layer is followed by a normalization network layer and an activation function layer. The first layer of convolution kernels is 32 3×3 in size, which extracts local temperature change features; the second layer of convolution kernels is 64 3×3 in size, which extracts temperature gradient features; the third layer of convolution kernels is 128 3×3 in size, which extracts temperature curvature features; the fourth layer of convolution kernels is 256 3×3 in size, which extracts high-order temperature features. The feature fusion network uses a fully connected layer of 1024 neurons to combine multi-scale features. The prediction network uses a linear output layer, and the output dimension is the same as the number of measuring points. When the temperature anomaly matrix is ​​used as the network input, the output expression of the first layer of convolution network is ,in is the output of the first layer of convolutional network, is the convolution kernel weight, represents the convolution operation, is the bias term, As the activation function, the corrected linear unit function is selected. The output of this layer is processed by the subsequent network layer to finally obtain the temperature error prediction value. The network training adopts the back propagation algorithm, the loss function uses the mean square error, the optimizer uses the adaptive moment estimation algorithm, the initial value of the learning rate is set to 0.001, and the number of training rounds is 1000. This step predicts the temperature measurement error through the deep learning method to provide a benchmark for temperature data correction.

[0079] The specific implementation method of step S06 is to establish a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix. The mapping model is constructed by using the multivariate regression analysis method. The independent variable is the temperature fluctuation characteristic quantity, and the dependent variable is the temperature anomaly parameter. First, feature selection is performed to calculate the correlation coefficient between the characteristic quantity and the abnormal parameter, and the characteristic quantity with an absolute value of the correlation coefficient greater than 0.3 is selected as the effective feature. Then a multivariate linear regression equation is established, and the regression coefficient is estimated by the least squares method. A significance test is performed on the regression equation. If the P value is less than 0.05, it means that the regression relationship is significant. A temperature compensation matrix is ​​generated based on the regression coefficient, and the numerical value in the matrix represents the temperature compensation coefficient at the corresponding position. The calculation formula for the compensation coefficient is: ,in is the compensation coefficient, is the benchmark compensation coefficient, is the temperature coefficient, is the temperature deviation. This compensation matrix takes into account the corresponding relationship between the temperature fluctuation characteristics and the degree of abnormality, and can effectively correct the temperature measurement error. The role of this step is to quantify the relationship between temperature fluctuation and temperature anomaly, and provide a basis for temperature data correction.

[0080] The specific implementation method of step S07 is to perform primary compensation using a temperature compensation matrix. When collecting temperature data in real time, first obtain the compensation coefficient of the current measuring point at the corresponding position in the temperature compensation matrix. The compensation coefficient includes two parts: a correction weight and an offset. Perform a linear correction on the real-time temperature data, and the correction formula is the original temperature value plus the offset multiplied by the correction weight. The correction weight ranges from 0 to 1, and the offset ranges from plus or minus 5 degrees Celsius. Different compensation coefficients are used for measuring points at different depths. The correction weights for shallow measuring points are larger, and the correction weights for deep measuring points are smaller. The compensated temperature data is called primary corrected temperature data, which eliminates the systematic error in temperature measurement. The calculation process of the primary correction can be expressed as ,in is the temperature vector after compensation, is the dynamic compensation coefficient matrix, The purpose of this step is to make a preliminary correction to the temperature data through the compensation matrix to improve the accuracy of temperature measurement.

[0081] The specific implementation method of step S08 is to calculate the steady-state deviation value between the first-level corrected temperature data and the historical calibration data. First, a historical calibration database is established, which contains standard temperature values ​​under different working conditions. The standard temperature values ​​are obtained by calibrating high-precision temperature sensors. Match the first-level corrected temperature data with the historical calibration data, and select data under the same working conditions for comparison. The sliding time window method is used to calculate the steady-state deviation value. The time window size is set to 1 hour, and the window sliding step is 10 minutes. In each time window, the root mean square error between the corrected temperature data and the calibration data is calculated. The calculation formula is: ,in is the root mean square error, is the number of data points in the time window, To correct the temperature value, To calibrate the temperature value. The steady-state deviation value reflects the systematic error of the temperature correction and provides a basis for the secondary compensation. The purpose of this step is to evaluate the systematic deviation between the temperature data after the primary correction and the standard value to ensure the reliability of the temperature correction.

[0082] The specific implementation of step S09 is to establish a temperature correction model. The model includes two parts: temperature offset compensation term and reference point correction term. The temperature offset compensation term is used to correct the zero drift of temperature measurement, and the reference point correction term is used to correct the linearity error of temperature measurement. The expression of the temperature correction model is: ,in is the corrected temperature value, To measure the temperature, is the offset compensation coefficient, is the temperature offset, is the reference point correction coefficient, is the reference point temperature value. The temperature offset compensation term is calculated using the exponential smoothing method, with a smoothing coefficient of 0.3. The calculation formula is: ,in for The smoothed value at time, for The observed value at time, is the smoothing coefficient. The reference point correction term is calculated using piecewise linear interpolation. Five reference points are evenly selected within the measurement range, and the reference point temperature value is determined by on-site calibration. For any measurement point, the reference point interval is determined according to its temperature value, and the correction coefficient is calculated using linear interpolation. The temperature correction model established in this step can effectively compensate for the nonlinear error in temperature measurement and improve the accuracy of temperature measurement.

[0083] The specific implementation of step S10 is to use the temperature correction model for secondary compensation. First, the offset of the primary correction temperature data is calculated, and the temperature data is subtracted from the offset compensation value to obtain the zero point correction result. The offset compensation value is calculated using the exponential smoothing method, taking into account the cumulative effect of temperature drift. Then, the reference point interval where the temperature value is located is determined, and the linear correction coefficient within the interval is calculated. The linear correction coefficient is obtained by interpolation calculation, and the calculation formula is: ,in is the correction coefficient, which is obtained by least squares fitting. The zero point correction result is multiplied by the linear correction coefficient to obtain the secondary correction temperature data. The secondary correction takes into account the nonlinear characteristics of temperature measurement and can more accurately reflect the actual temperature value. The role of this step is to further improve the accuracy of temperature measurement and make the corrected temperature data closer to the true value.

[0084] The specific implementation of step S11 is to calculate the dynamic response parameters of the system. The temperature response time is defined as the time required for the temperature change to reach 63.2% of the final steady-state value, which is calculated by exponential fitting the secondary corrected temperature data. The maximum temperature time is the moment when the temperature reaches the peak value, which is determined by finding the local maximum value of the temperature data. The temperature adjustment time is defined as the time required for the temperature to enter the steady-state interval. The range of the steady-state interval is plus or minus 2% of the final steady-state value, which is determined by judging the temperature fluctuation amplitude. The expression of the dynamic compensation model is ,in is the dynamic correction coefficient, is the temperature response time, is the time of maximum temperature, is the temperature adjustment time, is the corresponding reference time value, is the weight coefficient and satisfies These dynamic response parameters reflect the dynamic characteristics of the temperature measurement system and can be used to evaluate the control performance of the system.

[0085] The specific implementation of step S12 is to make a final correction based on the system dynamic response parameters. First, a dynamic compensation model is established. The model input is the temperature response time, the maximum temperature time and the temperature adjustment time, and the output is the dynamic correction coefficient. The calculation expression for the final temperature correction is: ,in is the final corrected temperature vector, is the secondary corrected temperature vector, is the corrected error vector. Dynamic compensation coefficient matrix Elements Indicates The measurement point The influence coefficient of each measuring point. The fuzzy control method is used to design dynamic compensation rules, and the correction coefficient is adjusted according to the size of the dynamic response parameter. When the response time is long, the correction coefficient is increased to speed up the response; when the adjustment time is long, the correction coefficient is reduced to suppress temperature fluctuations. The final correction result comprehensively considers the static and dynamic characteristics of temperature measurement and can accurately reflect the temperature distribution around the vertical buried pipe. The purpose of this step is to optimize the dynamic characteristics of temperature measurement and improve the control accuracy of the system.

[0086] In order to better understand and implement the present invention, Example 2 of a specific application scenario of the method of the present invention is provided below: A research team applied the temperature monitoring method of the present invention in an actual engineering project of a ground source heat pump system. The system includes 10 vertical buried pipes with a burial depth of 100 meters. First, a distributed optical fiber temperature sensing system is installed on one of the typical buried pipes. A single-mode optical fiber is used as the sensing element. It is installed along the pipe wall in a spiral winding manner. The winding spacing is 0.1 meters, the measuring point spacing is 0.5 meters, and there are a total of 200 measuring points. The sampling frequency is set to 1 Hz, and the temperature data is continuously collected for 48 hours. The initial collected temperature data is shown in Table 1: Table 1 Example of raw temperature data (partial)

[0087] The original data was preprocessed, and the triple standard deviation principle was used to eliminate outliers, and a total of 15 abnormal data points were identified. The sliding average method with a 60-second time window was used for data smoothing. The processed temperature data was constructed into a temperature fluctuation matrix with a matrix dimension of 172800×200.

[0088] The temperature fluctuation matrix is ​​decomposed by the empirical mode decomposition method, the standard deviation threshold is set to 0.2, and four intrinsic mode functions are obtained by iterative calculation. The first intrinsic mode function reflects the short-term temperature fluctuation characteristics with a period of about 1 hour; the second intrinsic mode function reflects the medium-term temperature change characteristics with a period of about 6 hours; the third and fourth intrinsic mode functions reflect the long-term temperature change trend. The energy distribution of each intrinsic mode function is shown in Table 2: Table 2 Energy distribution of eigenmode function

[0089] The decomposed intrinsic mode function is subjected to Hilbert transform to extract the fluctuation characteristic quantity. Then the data is normalized and the standardized temperature data of each measuring point is calculated. The local anomaly factor algorithm is used to detect temperature anomalies. The local neighborhood size is set to 20 measuring points. The calculation results show that there are 32 measuring points with abnormal conditions, mainly concentrated near the surface and the bottom of the pipe. The anomaly detection results are shown in Table 3: Table 3 Temperature anomaly detection results (partial)

[0090] A deep convolutional neural network is established for temperature error prediction. The network structure includes 4 convolutional layers, 1 fully connected layer and an output layer. The field calibration data is used as the training set. After 1000 rounds of training, the prediction accuracy of the network on the test set reaches 0.15℃. The comparison between the prediction results and the actual errors is shown in Table 4: Table 4 Comparison of temperature error prediction results (partial)

[0091] A temperature compensation matrix is ​​established based on the prediction results, and the real-time temperature data is compensated for the first level. During the compensation process, the correction weight of the shallow measuring points (0-20 meters) is set to 0.8, the correction weight of the middle measuring points (20-80 meters) is set to 0.6, and the correction weight of the deep measuring points (80-100 meters) is set to 0.4. The root mean square error between the temperature data and the calibration data after the first level compensation is reduced to 0.12℃.

[0092] The steady-state deviation value is calculated by the sliding time window method, and the primary correction data is compensated for the secondary compensation. In the compensation model, the temperature offset compensation term is calculated by the exponential smoothing method, and the smoothing coefficient is 0.3; the reference point correction term evenly selects 5 reference points within the measurement range for correction. The temperature measurement accuracy is further improved after the secondary compensation, and the root mean square error is reduced to 0.08℃.

[0093] Finally, the system dynamic response parameters are calculated, including temperature response time, temperature maximum time and temperature adjustment time. The test results show that the average temperature response time of the system is 185 seconds, the temperature maximum time is 486 seconds, and the temperature adjustment time is 823 seconds. A dynamic compensation model is established based on these parameters to achieve the final temperature correction. The final corrected system performance parameters are shown in Table 5: Table 5 System performance parameters

[0094] The traditional buried pipe temperature monitoring method has the following main problems: First, it adopts a single-point temperature measurement method with low spatial resolution and cannot accurately reflect the temperature field distribution; second, it does not take into account the nonlinear characteristics of the measurement error and has a poor compensation effect; third, it lacks dynamic characteristic analysis, which affects the system control effect. The present invention adopts distributed optical fiber temperature measurement technology to achieve high spatial resolution temperature field monitoring; predicts temperature errors through deep learning methods and establishes a multi-level temperature compensation mechanism; introduces dynamic response parameter analysis to optimize system control performance. The implementation results show that compared with traditional methods, the present invention improves the measurement accuracy by 80%, the system response speed by 23.2%, and the operation stability by 10.7%, which significantly improves the operation effect of the ground source heat pump system.

[0095] A specific embodiment 3 of the system of the present invention is provided below: Figure 1-Figure 10A vertical buried pipe ground source heat pump temperature monitoring system based on distributed optical fiber comprises a box body 1, a vertical plate 2 is fixedly installed on the top of the box body 1, a gear 3 is rotatably installed on one end of the vertical plate 2, tooth plates 4 are movably installed on both ends of the vertical plate 2, the number of tooth plates 4 is two, and the size of the two tooth plates 4 is the box body, the tooth plates 4 are meshed and connected with the outer surface of the gear 3, the gear 3 will drive the tooth plates 4 to move toward or away from each other when rotating, connecting rods 5 are respectively fixedly installed on the tooth plates 4, the tooth plates 4 will drive the connecting rods 5 to slide when sliding, one end of the connecting rod 5 is fixedly installed with a movable door, including a left movable door 61 and a right movable door 62, the connecting rod 5 will drive the movable door to slide when sliding, the two movable doors have the same size, and the other end of the vertical plate 2 is rotatably installed with a roller 7 and The outer surfaces of the rotating shaft 8, roller 7 and rotating shaft 8 are movably sleeved with a belt 9. When the rotating shaft 8 rotates, the belt 9 will drive the roller 7 to rotate. One end of the roller 7 passes through the vertical plate 2 and is fixedly connected to one end of the gear 3. When the roller 7 rotates, the gear 3 will be driven to rotate. A round shaft 10 is fixedly installed at one end of the rotating shaft 8. When the round shaft 10 rotates, it will drive the rotating shaft 8 to rotate. A protective box 11 is fixedly installed at one end of the vertical plate 2. The protective box 11 can protect the roller 7, the rotating shaft 8, the belt 9 and the round shaft 10 from collision. A motor 12 is fixedly installed at one end of the protective box 11, and the output end of the motor 12 passes through the protective box 11 and is fixedly connected to one end of the round shaft 10. When the motor 12 is running, the round shaft 10 will rotate.

[0096] In the present invention, when the box body 1 needs to be opened and closed, the operation of the motor 12 will cause the round shaft 10 to rotate, and then the round shaft 10 will drive the rotating shaft 8 to rotate, and then the rotating shaft 8 will drive the belt 9 to rotate, and then the belt 9 will drive the roller 7 to rotate, at this time the roller 7 will drive the gear 3 to rotate, and then the gear 3 will drive the tooth plate 4 to slide, and then the tooth plate 4 will drive the connecting rod 5 to slide, and at the same time the connecting rod 5 will drive the movable door to move towards each other, so that the box body 1 can be automatically closed, and vice versa, it can be opened, so as to facilitate the maintenance of multiple modules inside the box body 1.

[0097] A specific embodiment 4 of the system of the present invention is provided below: Figure 1-Figure 10, a vertical buried pipe ground source heat pump temperature monitoring system based on distributed optical fiber, a fixed rod 13 is fixedly installed at one end of the box body 1, a strip groove 14 is opened on the top of the fixed rod 13, a round rod 15 is fixedly installed inside the strip groove 14, and a moving block 16 is movably sleeved on both sides of the surface of the round rod 15. The inner surface of the moving block 16 and the outer surface of the round rod 15 are both smooth, which can make the moving block 16 slide more smoothly on the surface of the round rod 15 and reduce the occurrence of jamming. The top of the moving block 16 is fixedly connected to the bottom of the movable door, and the movable door is When sliding, the moving block 16 will be driven to slide inside the strip groove 14 and on the surface of the round rod 15. Due to the design of the round rod 15 and the moving block 16, the sliding of the movable door can be more stable. A card slot 17 is provided on one side of the left movable door 61, and a card plate 18 is fixedly installed on one side of the right movable door 62, and the card slot 17 is adapted to the card plate 18. The inner surface of the card slot 17 and the outer surface of the card plate 18 are both rough. When the card plate 18 is placed inside the card slot 17, the connection between the two can be made tighter, thereby improving the stability of the movable door when closed.

[0098] In the present invention, when the movable door slides, it will drive the moving block 16 to slide inside the strip groove 14 and on the surface of the round rod 15. Due to the design of the round rod 15 and the moving block 16, the sliding of the movable door can be more stable. When the card plate 18 is placed inside the card groove 17, the two movable doors will close the box body 1. When the card plate 18 slides out of the card groove 17, the movable doors will move in opposite directions, so that the box body 1 is in an open state.

[0099] A specific embodiment 5 of the system of the present invention is provided below: Figure 1-Figure 10, a vertical buried pipe ground source heat pump temperature monitoring system based on distributed optical fiber, a sensing optical fiber 19 is fixedly installed at an equal distance at the bottom of the box 1, and the sensing optical fiber 19 adopts a specially designed fast heat-conducting optical cable, which has the characteristics of high tensile strength, bending resistance, pressure resistance, waterproofness, corrosion resistance, etc., and can effectively protect the optical fiber and quickly transfer external heat. The top of the sensing optical fiber 19 penetrates the box 1 and extends to the inner surface of the box 1. The size of the sensing optical fiber 19 is the same. The sensing optical fiber 19 can be distributedly connected with the vertical buried pipe, and the surface of the sensing optical fiber 19 is movably sleeved with a sleeve rod 20 The inner surface of the sleeve rod 20 and the outer surface of the sensing optical fiber 19 are both smooth, and the position of the sealing rod 25 can be slid through the sleeve rod 20 to meet the auxiliary fixation of the connection between the sensing optical fiber 19 and the vertical buried pipe. The bottom ends of the sleeve rod 20 are provided with a slide groove 21, and a cylinder 22 is fixedly installed inside the slide groove 21. The surface of the cylinder 22 is movably sleeved with a slider 23. The inner surface of the slider 23 and the outer surface of the cylinder 22 are both smooth, which can make the slider 23 slide more smoothly on the surface of the cylinder 22, reducing the occurrence of jamming. The slider 23 and the slide groove 21 are connected to each other. A spring 24 is fixedly installed between the two ends of the sleeve rod 20, and the inner surface of the spring 24 is movably connected to the outer surface of the cylinder 22. When the slider 23 slides inside the slide groove 21 and on the surface of the cylinder 22, the spring 24 is squeezed and compressed, and a sealing rod 25 is fixedly installed at the bottom of the slider 23. When the sealing rod 25 slides, it drives the slider 23 to slide. At the same time, the connection between the sensing optical fiber 19 and the vertical buried pipe can be auxiliary fixed by the sealing rod 25. The top ends of the sleeve rod 20 are provided with a limit groove 26, and the inside of the limit groove 26 is movably installed with a moving rod 27, the inner surface of the limit groove 26 and the outer surface of the moving rod 27 are both smooth, which can make the moving rod 27 slide more smoothly inside the limit groove 26 and reduce the occurrence of jamming. The bottom of the moving rod 27 is fixedly connected to the top of the slider 23. When the slider 23 slides, it will drive the moving rod 27 to slide inside the limit groove 26, and the top of the moving rod 27 is fixedly installed with a clamping ring 28. When the moving rod 27 slides, it will drive the clamping ring 28 to slide. The clamping ring 28 can fix the surface of the sensing optical fiber 19, and the clamping ring 28 can be made of soft material.

[0100] In the present invention, when the sensing optical fiber 19 is connected to the vertical buried pipe, the sealing rod 25 is pulled accordingly, and the sealing rod 25 will drive the slider 23 to slide inside the slide groove 21 and on the surface of the cylinder 22 when sliding, and then the slider 23 will squeeze and compress the spring 24, at this time the slider 23 will drive the moving rod 27 to slide inside the limit groove 26, and then the moving rod 27 will drive the clamping ring 28 to slide, and then the sliding sleeve rod 20 will be slid until the sealing rod 25 is placed at the connection between the sensing optical fiber 19 and the vertical buried pipe, and at the same time the sealing rod 25 is released, and then the clamping ring 28 will clamp and fix the sensing optical fiber 19, and then the sealing rod 25 can be used to auxiliary fix the connection between the sensing optical fiber 19 and the vertical buried pipe to make the connection tighter.

[0101] A specific embodiment 6 of the system of the present invention is provided below: Figure 1-Figure 10 , a vertical buried pipe ground source heat pump temperature monitoring system based on distributed optical fiber, a temperature detection module 29 and a data processing module 30 are fixedly installed at one end of the interior of the box 1, the temperature detection module 29 can monitor the temperature and process the monitored data through the data processing module 30, and a data transmission module 31 and a power supply module 32 are fixedly installed at one end of the interior of the box 1, which are located on the right side of the temperature detection module 29 and the data processing module 30. The data transmission module 31 can transmit the processed data and provide power support to the temperature detection module 29, the data processing module 30 and the data transmission module 31 through the power supply module 32, so that they can work for a longer time, a mounting plate 33 is fixedly installed on the top of the box 1, and mounting holes 34 are opened at both ends of the mounting plate 33. Due to the design of the mounting plate 33 and the mounting holes 34, the box 1 can be easily installed.

[0102] In the present invention, the operator connects the temperature detection module 29, the data processing module 30, the data transmission module 31 and the power module 32 to the card board 18 through wires or connecting lines. At the same time, the temperature detection module 29 can monitor the temperature, the data processing module 30 can process the monitored data, and then transmit the processed data through the data transmission module 31. Then, the power module 32 can provide power support to the temperature detection module 29, the data processing module 30 and the data transmission module 31, so that they can work for a longer time.

[0103] A specific embodiment 7 of the system of the present invention is provided below: Figure 1-Figure 10, a vertical buried pipe ground source heat pump temperature monitoring system based on distributed optical fiber, heat dissipation grooves 35 are equidistantly provided on both sides of the box 1, and the heat dissipation grooves 35 have the same size. The temperature detection module 29, the data processing module 30, the data transmission module 31 and the power supply module 32 will generate heat during operation. If the heat is not dissipated in time, it may affect the normal monitoring work. Then the heat can be dissipated through the heat dissipation groove 35. A display screen 36 is fixedly installed at one end of the left movable door 61, and a PLC controller 37 is fixedly installed at one end of the right movable door 62. The PLC controller 37 is a digital operation electronic system specially designed for use in an industrial environment. It uses a programmable memory to store and execute logical operations, sequential control, timing, Instructions for operations such as counting and arithmetic operations control various types of mechanical equipment or production processes through digital or analog input and output, and adopts modern large-scale integrated circuit technology. PLC controller 37 has taken many measures in both hardware and software to improve its reliability. In terms of hardware, PLC controller 37 shields major components such as power transformer, CPU, programmer, etc. to prevent external interference, and uses a filtering network to eliminate high-frequency interference for the power supply system and input lines. The power supply required by the CPU is adjusted by multi-stage filtering and regulators to adapt to power grid fluctuations. In terms of software, PLC controller 37 regularly detects the external environment through a monitoring program, handles faults in a timely manner, and sets a monitoring timer to prevent the program from entering an infinite loop, thereby improving the degree of intelligence.

[0104] In the present invention, the operator connects the display screen 36 and the PLC controller 37 to the motor 12, the sensing optical fiber 19, the temperature detection module 29, the data processing module 30, the data transmission module 31, and the power module 32 through wires. The heat dissipation groove 35 can dissipate the heat generated by the temperature detection module 29, the data processing module 30, the data transmission module 31 and the power module 32 to prevent overheating and affect normal monitoring work. At the same time, the display screen 36 can display the monitored data, making it convenient for the operator to observe, and the degree of automation of the device can also be achieved through the PLC controller 37.

[0105] It should be noted that the variables involved in the present invention are explained in detail as shown in Table 6 below.

[0106] Table 6 Variable explanation table

[0107] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A temperature monitoring method for a vertical buried pipe ground source heat pump, comprising the following steps: The original temperature data sequence of multiple measuring points is collected by distributed optical fiber sensors to construct a temperature fluctuation matrix; Decomposing the temperature fluctuation matrix in the time domain, extracting the fluctuation amplitude, fluctuation frequency and fluctuation phase of the temperature data, and obtaining the temperature fluctuation characteristic quantity; Based on the temperature fluctuation characteristic quantity, the original temperature data is normalized to generate a standardized temperature data sequence; temperature anomaly parameters are calculated according to the standardized temperature data sequence to construct a temperature anomaly matrix; Inputting the temperature anomaly matrix into a pre-trained temperature error prediction model to obtain a temperature error prediction value; Establishing a mapping relationship between the temperature fluctuation matrix and the temperature anomaly matrix to generate a temperature compensation matrix; using the temperature compensation matrix to perform primary compensation on the temperature data collected in real time to obtain primary corrected temperature data; Calculating a steady-state deviation value between the primary corrected temperature data and the historical calibration data; Establishing a temperature correction model according to the steady-state deviation value; performing secondary compensation on the primary corrected temperature data using the temperature correction model to obtain secondary corrected temperature data; Calculate system dynamic response parameters; The secondary corrected temperature data is finally corrected by using the system dynamic response parameters, and the corrected temperature data is output; characterized in that the pre-trained temperature error prediction model adopts a deep convolutional neural network structure, and the deep convolutional neural network structure is composed of a feature extraction network layer, a feature fusion network layer and a prediction network layer; the temperature correction model includes a temperature offset compensation term and a reference point correction term; The system dynamic response parameters include temperature response time, temperature maximum value time and temperature adjustment time.

2. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 1, characterized in that: The feature extraction network layer adopts a four-layer convolution layer structure. The number of convolution kernels in each convolution layer is 32, 64, 128, and 256 respectively. The convolution kernel size is 3×3, and the step size is 1. Each convolution layer is followed by a normalization network layer and an activation function layer. The feature fusion network layer adopts a fully connected layer structure with 1024 neurons. The prediction network layer adopts a linear output layer with an output dimension of the number of measurement points.

3. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 2, characterized in that: The training data set of the pre-trained temperature error prediction model constructs virtual training data of underground temperature distribution through a conditional generative adversarial network. The conditional generative adversarial network includes a generative network and a discriminative network. The generative network has an embedded temperature generation equation group for simulating the distribution law of the underground temperature field.

4. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 3, characterized in that: The temperature generation equation group includes a heat conduction equation, a water flow influence equation and a thermal conductivity equation. The heat conduction equation is used to describe the heat conduction process in the soil around the vertical buried pipe. The water flow influence equation is used to simulate the influence of groundwater flow on the temperature field. The thermal conductivity equation is used to calculate the thermal conductivity of the soil under different depths and different water contents.

5. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 4, characterized in that: In the process of performing primary compensation using the temperature compensation matrix, the compensation coefficient includes a correction weight and an offset, the correction weight ranges from 0 to 1, and the offset ranges from plus or minus 5 degrees Celsius; The steady-state deviation value between the first-level corrected temperature data and the historical calibration data is calculated using a sliding time window method, with the time window size set to 1 hour and the window sliding step size set to 10 minutes.

6. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 5, characterized in that: The temperature offset compensation term is calculated using an exponential smoothing method, with a smoothing coefficient of 0.3; the reference point correction term is calculated using a piecewise linear interpolation method, and 5 reference points are evenly selected within the measurement range.

7. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 6, characterized in that: The temperature response time is defined as the time required for the temperature change to reach 63.2% of the final steady-state value, and the temperature adjustment time is defined as the time required for the temperature to enter the steady-state interval, and the range of the steady-state interval is plus or minus 2% of the final steady-state value.

8. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 7, characterized in that: The local anomaly factor algorithm is used to detect the degree of anomaly of temperature data. The size of the local neighborhood is taken as 10% of the total number of measuring points, and the measuring points with a local anomaly factor greater than 2 are marked as abnormal points.

9. The temperature monitoring method of a vertical buried pipe ground source heat pump according to claim 8, characterized in that: The empirical mode decomposition method is used to decompose the temperature fluctuation matrix, and the obtained intrinsic mode function sequence is subjected to Hilbert transform to calculate the instantaneous frequency and instantaneous amplitude, and extract the fluctuation amplitude, fluctuation frequency and fluctuation phase corresponding to the main frequency components.

10. A temperature monitoring system for a vertical buried pipe ground source heat pump, comprising a box, characterized in that: A vertical plate is fixedly installed on the top of the box body, a gear is rotatably installed on one end of the vertical plate, and toothed plates are movably installed on both ends of the vertical plate, the toothed plates are meshed and connected with the outer surface of the gear, and connecting rods are fixedly installed on the toothed plates respectively, and a movable door is fixedly installed on one end of the connecting rods, and a roller and a rotating shaft are rotatably installed on the other end of the vertical plate, and a belt is movably sleeved on the outer surfaces of the roller and the rotating shaft, one end of the roller passes through the vertical plate and is fixedly connected to one end of the gear, one end of the rotating shaft is fixedly installed with a round shaft, and one end of the vertical plate is fixedly installed with a protective box, and the protective box A motor is fixedly installed at one end, and the output end of the motor passes through the protective box and is fixedly connected to one end of the circular shaft; a fixed rod is fixedly installed at one end of the box body, a strip groove is provided on the top of the fixed rod, a round rod is fixedly installed inside the strip groove, and moving blocks are movably sleeved on both sides of the surface of the round rod, and the top of the moving block is fixedly connected to the bottom of the movable door; the movable door includes a left movable door and a right movable door, a card slot is provided on one side of the left movable door, and a card plate is fixedly installed on one side of the right movable door, and the card slot is adapted to the card plate; the bottom of the box body is fixedly installed at equal distances A sensing optical fiber is fixedly installed, the top of the sensing optical fiber passes through the box body and extends to the inner surface of the box body, and the surface of the sensing optical fiber is movably sleeved with a sleeve rod; both ends of the bottom of the sleeve rod are provided with a slide groove, the inside of the slide groove is fixedly installed with a cylinder, the surface of the cylinder is movably sleeved with a slider, a spring is fixedly installed between the slider and the slide groove, the inner surface of the spring is movably connected to the outer surface of the cylinder, and a sealing rod is fixedly installed at the bottom of the slider; both ends of the top of the sleeve rod are provided with a limit groove, the inside of the limit groove is movably installed with a moving rod, and the bottom of the moving rod The box body is fixedly connected to the top of the slider, and the top of the moving rod is fixedly installed with a clamping ring; a temperature detection module and a data processing module are fixedly installed at one end of the interior of the box body, and a data transmission module and a power supply module located on the right side of the temperature detection module and the data processing module are fixedly installed at one end of the interior of the box body; the data processing module adopts a single-chip microcomputer, and a storage medium is provided in the data processing module, and a program code is stored in the storage medium. When the single-chip microcomputer executes the program code, it is used to execute the temperature monitoring method of the vertical buried pipe ground source heat pump described in any one of claims 1-9.

Images