Sampling optimization method of high-spatial-resolution OTDR (Optical Time Domain Reflectometry) optical fiber temperature measurement system

Abstract

The invention provides a sampling optimization method of a high-spatial-resolution OTDR (Optical Time Domain Reflectometry) optical fiber temperature measurement system, which is based on a variable-frequency sampling frequency selection optimization method of a multi-objective particle swarm optimization (MOPSO) algorithm, finds an optimal frequency combination under a given frequency range and spatial resolution, collects data under different sampling frequencies in the frequency combination, and utilizes a reconstruction algorithm to obtain a high-spatial-resolution OTDR optical fiber temperature measurement system. The sampling point closest to the ideal sampling point in space distance is found, and the temperature value of the point is obtained through calculation. According to the invention, the sampling frequency combination and the corresponding sampling point can be automatically selected, and higher-efficiency variable-frequency sampling is realized.

Description

Technical Field

[0001] This invention relates to a temperature measurement method for an OTDR fiber optic temperature measurement system. Background Technology

[0002] A distributed temperature monitoring system (DTS) is a real-time, online, and continuous temperature monitoring system. The system is designed based on the temperature-sensitive effect of Raman scattering and the principle of optical time-domain reflectometry (OTDR). It can use sensing optical fibers as sensing probes to achieve online monitoring of the temperature field along the entire length of the fiber. Due to its corrosion resistance, immunity to electromagnetic interference, intrinsic safety, and ability to continuously monitor long-distance, wide-range environmental temperatures, it is widely used in pipeline monitoring in many scenarios.

[0003] A typical OTDR fiber optic temperature measurement system includes a host computer, a data transmission line, a pulsed light source, a wavelength division multiplexing (WDM) device, a photodetector, a high-speed data acquisition card, and a sensing fiber optic cable. The pulsed light source emits nanosecond-level pulsed light with a wavelength of 1550nm, which is then transmitted into the sensing fiber optic cable after passing through the WDM device. Raman scattering occurs in the fiber, generating a backscattered signal containing temperature information. This backscattered signal is then filtered by the WDM device to separate the Stokes light and the single-strokes light. The APD converts the two optical signals into electrical signals, which are then acquired by the data acquisition card and transmitted to the host computer for processing.

[0004] Spatial resolution, a crucial indicator of distributed temperature measurement systems, represents the system's ability to characterize temperature abrupt changes within narrow units. It is defined as the distance between 10% and 90% of the temperature step change that the fiber optic temperature measurement system can measure when the ambient temperature of the sensing fiber undergoes a step change. Smaller spatial resolution indicates a stronger ability to characterize temperature abrupt changes within narrow units. Distributed temperature measurement systems based on optical time-domain reflectometry (OTDR) technology, due to the fixed speed of light in the fiber, can locate the signal position based on the time delay of the received optical signal, enabling distributed measurement. This indicator is limited by the hardware parameters of the temperature measurement system: the pulse width emitted by the pulsed light source, the bandwidth of the photoelectric converter, and the sampling rate of the data acquisition card. The first two determine the waveform shape acquired by the data acquisition card. As is known from OTDR technology, the sampling rate determines the interval between sampling points. The system's spatial resolution is determined by the lower limits of these three parameters.

[0005] Existing literature has disclosed several optimization schemes to address the problem of low spatial resolution caused by low sampling rates of data acquisition cards. Among them, variable frequency sampling technology utilizes the property in OTDR principles that the time interval between sampling points can be converted into spatial intervals. It designs a method to change the sampling frequency of the data acquisition card, thereby changing the density of the sampling point distribution, and then filtering and reconstructing a set of high-precision data. However, the proposed variable frequency sampling and reconstruction algorithms rely on subjective setting of frequency selection, resulting in partial sampling redundancy and low sampling point utilization. This not only increases the system response time but also consumes a lot of storage and computing resources. Therefore, it is necessary to design an optimization algorithm for frequency selection and an evaluation index for different frequency combinations to comprehensively consider measurement accuracy and response time. Summary of the Invention

[0006] The purpose of this invention is to provide a sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system, which can automatically select sampling frequency combinations and their corresponding sampling points to achieve more efficient frequency conversion sampling. To this end, this invention adopts the following technical solution:

[0007] A sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system is characterized by a frequency selection optimization method based on the multi-objective particle swarm optimization (MOPSO) algorithm. This method finds the optimal frequency combination within a given frequency range and spatial resolution, collects data at different sampling frequencies within the frequency combination, and uses a reconstruction algorithm to find the sampling point that is spatially closest to the ideal sampling point, and calculates the temperature value of that point.

[0008] Based on the above-mentioned technical approach, the present invention can also adopt the following further technical solutions:

[0009] By utilizing temperature information from nearby sampling points, the temperature measurement results at the ideal sampling point location can be predicted using trend extrapolation or interpolation.

[0010] The frequency conversion sampling frequency selection optimization method based on the multi-objective particle swarm optimization (MOPSO) algorithm includes the following steps:

[0011] Step 1: Determine the sampling rate range supported by the hardware device, and select the available frequency range f. max ~f min The algorithm inputs the ideal spatial resolution Δx and randomly generates a set of particle swarm optimization variables x(t) = (x1, x2, ..., x) based on the multi-objective particle swarm optimization algorithm. i ,…,x m Each optimized particle x i (i = 1, 2, ..., m) corresponds to a frequency combination; algorithm parameters include: maximum number of iterations T max Number of particles m, maximum inertia factor ωmax Minimum inertia factor w min Acceleration constants c1 and c2, maximum velocity v max Let the current optimization algebra be t = 1, t ≤ T. max In a space of dimension range, m particles x1, x2, ..., x3 are randomly generated. i ,…,x m A population X(t) is formed, and the initial velocities of each particle are randomly generated as v1, v2, ..., v i ,…,v m The position of the i-th particle is x. i =(x i,1 ,x i,2 , ..., x i,j ), with a speed of v i =(v i,1 ,v i，2 , ..., v i,j ), where j = f max -f min Particle x i The value of each element in (i = 1, 2, ..., m) is either 0 or 1;

[0012] Step 2: Based on the optimized particle x i The frequency combinations are used to calculate the distribution of sampling points at each sampling frequency. Based on the reconstruction algorithm and the ideal spatial resolution Δx, two sets of sampling points are re-selected to obtain two sets of high-precision reconstructed data distributions p1, p2, ..., p... n and q1, q2, ..., q n ;

[0013] Step 3: Establish a mathematical model to obtain an objective function that measures the accuracy of the reconstructed data and the number of samplings:

[0014]

[0015] Value2(x i ) = count

[0016] Where Value1(x) i ) and Value2(x i Let Δd be the two objective functions of the optimization algorithm. i It is the deviation between the reconstructed data and the position of that point under ideal spatial resolution, Δd i =|i·Δx-q i |+|i·Δx-p i|, count is the number of samples. The objective function for each optimized particle is calculated according to this model to select the individual's historical best position Pbest and the population's global historical best position Gbest;

[0017] Step 4: Generate a new set of particle swarm optimization variables X(t+1) according to the transformation rules of the multi-objective particle swarm algorithm, and return to step 2 to continue the next optimization until the particle swarm algorithm reaches the maximum number of iterations. Then stop the calculation and output the frequency combination corresponding to the best particle.

[0018] Further, step 2 includes the following steps in sequence:

[0019] Step 201: Let the set of sampling points on the sensing fiber be {d} i |i=1,2,…,N},whered i It is an arithmetic sequence starting from 0 with intervals of Δx; based on particle x i The frequency combinations can be determined by first calculating the deviation between each ideal sampling point and its two closest actual sampling points at a single sampling frequency in the combination. Let the set of actual sampling points at each frequency form a dataset A.

[0020]

[0021] Where a f，n Let f be the position of the nth sampling point from the initial point;

[0022] Step 202: Based on the calculated sampling point distribution matrix, select the sampling points that are closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution.

[0023]

[0024] For any n∈[1, N], m∈{f1, f2, ..., f m All of them are satisfied. where i∈[1, N]

[0025] Step 203: Exclude the sampling points already selected in C1, and perform the selection again to select the sampling points that are the second closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution, thus obtaining C2;

[0026] Step 204: At different sampling frequencies, find the actual sampling points that are closest to each ideal sampling point according to C1 and C2, and obtain the final index matrix:

[0027]

[0028] For any m∈{f1, f2, ..., f m}, all satisfy

[0029] In the index matrix F above, the first row F 1,i This indicates the sampling frequency selected for the i-th point of the reconstructed data curve, and the second row F 2，i Frequency F 1，i Next F 2，i One sampling point;

[0030] Step 205: Exclude the sampling points already selected in F from C1 and C2, recalculate and find the actual sampling points that are closest to each ideal sampling point, and obtain the second index matrix F′:

[0031]

[0032] Step 205: For each ideal sampling point, obtain the locations and temperature data of the two actual sampling points closest to the ideal sampling point from the index matrix F and the second index matrix F′. Assume the locations are D1 and D2, and the temperature measurement results are T1 and T2. Based on the proximity of the locations, use the trend extrapolation method or the interpolation method to predict the temperature of the ideal sampling point. If the two points are on the same side of the ideal sampling point, the trend extrapolation method is used for prediction; if the two points are on opposite sides of the ideal sampling point, the interpolation method is used for prediction. Both methods use a linear regression model.

[0033] Furthermore, step 4 specifically includes the following steps:

[0034] Step 401: Use the objective function obtained in Step 3 as the fitness value to evaluate the quality of each particle. Embed the Pareto dominance relation into PSO to determine the best position pbest in the entire population and the global best position gbest, based on the objective function Value1(x) obtained in Step 3. i ) and Value2(x i According to the Pareto dominance principle, the local optimum pbest is updated. If the current value does not dominate the original pbest, it is updated randomly. The global optimum gbest is selected in the Archive set by calculating the crowding in the external archive set; and the velocity and position of the particles are updated to generate a new set of particle swarm optimization variables.

[0035] Step 402, according to the formula:

[0036] v i,j (i+1)=ωv i,j (t)+c1r1[P i，j -x i，j (t)]+c2r2[P g,i -x i,j (t)]

[0037] Update particle velocity and formula Update the particle positions to generate a new population X(t+1), v i，j Let be the current velocity of the i-th particle with the j-th parameter; c1 and c2 represent positive acceleration coefficients, and r1 and r2 are random numbers between 0 and 1; p i,j pbest represents the best position found so far for the i-th particle; p g,i gbest represents the best position found by the entire particle swarm; x i,j (t) represents the current position of the j-th parameter of the i-th particle;

[0038] Step 403: According to the formula ω = ω min +r3·(ω max -ω min Update the inertia factor of the optimization algorithm, where r3 is a random number between 0 and 1; update the Archive set (which stores the current non-dominated solutions). If the number of particles in the Archive set exceeds the specified size, a truncation operation is required.

[0039] Step 404: Determine if t equals T. max If the conditions are met, the frequency combination corresponding to the best particle is output, along with the high-precision data index matrix under that frequency combination; otherwise, t = t + 1, and the search returns to step 402.

[0040] The beneficial effects of this invention are as follows: Based on the structure of a conventional OTDR, this invention uses an RF signal source as the external clock input of the acquisition card to dynamically adjust the sampling frequency, effectively improving the problem of low spatial resolution caused by the sampling frequency limitation of the acquisition card, while minimizing hardware modification costs. Furthermore, it utilizes the MOPSO algorithm to optimize the frequency selection for variable frequency sampling, alleviating potential sampling redundancy. This frequency selection optimization method, using a selectable frequency range and ideal spatial resolution as input, can output the optimal frequency combination, avoiding insufficient utilization of sampling data due to random or artificial combinations of sampling frequencies. Considering the key concerns of measurement accuracy and response time in variable frequency sampling, the objective function of the multi-objective optimization algorithm is set as measurement deviation and the number of samplings. Because the particle evolution process uses Pareto dominance relations for sorting, the final output is not a single optimization result but a set of selectable optimization solutions. Operators can flexibly choose according to their preferences for the number of samplings and accuracy. As a random search algorithm, the multi-objective particle swarm optimization algorithm has excellent adaptability in the frequency selection optimization process, enabling the pre-calculation of optimization results and their deployment in the temperature measurement system, thus possessing certain engineering application value.

[0041] To avoid positional differences between actual and ideal measurement points in the reconstructed signal, this invention finds two nearest actual points as references during the sampling point indexing process. Temperature prediction is performed on the ideal measurement position based on the positional relationship, thus avoiding the impact of positional deviation on the accuracy of temperature measurement and further optimizing the spatial accuracy of the high-resolution measurement signal constructed by frequency conversion sampling. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the logical structure of an OTDR fiber optic temperature measurement system.

[0043] Figure 2 This is a flowchart of the variable frequency sampling data acquisition process based on the optimal frequency combination.

[0044] Figure 3 This is a schematic diagram of the measurement execution process of an OTDR fiber optic temperature measurement system based on MOPSO.

[0045] Figure 4 This is a flowchart of the MOPSO algorithm. Detailed Implementation

[0046] This invention provides a sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system, which can be applied to a distributed temperature measurement system with adjustable sampling frequency. The distributed temperature measurement system includes a host computer, a data transmission line, a pulsed light source, a wavelength division multiplexing (WDM) device, a photodetector, a high-speed data acquisition card, and a sensing temperature fiber. The OTDR fiber optic temperature measurement system is a Raman scattering-based distributed temperature measurement system and requires the following structure: the hardware includes a pulsed light source, a WDM, an APD photoelectric converter, a high-speed data acquisition card, an RF signal source, a temperature-sensing fiber, and a host computer; the RF signal source communicates with the host computer via a serial port, dynamically adjusting the output signal frequency according to instructions; the RF signal source output is connected externally to the data acquisition card. Clock input terminal; the acquisition card communicates with the host computer via USB 3.0 and simultaneously receives the clock signal emitted by the RF signal source for acquisition; during system operation: first, the host computer sends a set of control signals to control the frequency of the pulse signal output by the RF signal source, thus completing one frequency setting; at the current frequency, the high-speed acquisition card completes data acquisition and uploads it to the host computer, which then saves the acquired data; then, the host computer sends the next set of control signals to control the RF signal source to adjust its frequency output, thereby changing the sampling frequency of the high-speed acquisition card for the next set of data acquisition and data saving; through multiple frequency settings, the host computer completes the acquisition of optical signal data at multiple different sampling frequencies.

[0047] Using the principle of optical time-domain reflectometry, a sequence of sampling points arranged at certain time intervals can be converted into a sequence of sampling points arranged at certain spatial intervals. Variations in the sampling interval at different sampling frequencies result in a dense distribution of sampling points.

[0048] This invention utilizes a frequency selection optimization method based on the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm to find the optimal frequency combination within a given frequency range and spatial resolution. It collects data at different sampling frequencies within this frequency combination, and uses a reconstruction algorithm to find the sampling point spatially closest to the ideal sampling point, calculating the temperature value at that point. This invention simultaneously considers the utilization rate of the sampling frequency and the accuracy of the reconstructed sampling point location, providing a set of optimal solutions for operators to choose from flexibly.

[0049] The reconstructed sampling points may not perfectly coincide with the ideal sampling points in location. By utilizing the temperature information of nearby sampling points, we can use trend extrapolation or interpolation to predict the temperature measurement results at the location of the ideal sampling points.

[0050] Furthermore, the pulsed light source periodically emits pulsed light with a wavelength on the order of nanoseconds, which is then injected into the sensing temperature fiber through a wavelength division multiplexer. The sensing temperature fiber generates fiber Raman backscattered light signals along the line, which return to the wavelength division multiplexer. The wavelength division multiplexer splits the fiber Raman backscattered light signals into two paths: Stokes light and anti-Stokes light. The two scattered light signals are converted into electrical signals by a photoelectric converter and then uploaded to the host computer by a high-speed acquisition card.

[0051] Furthermore, the frequency conversion sampling can obtain multiple groups of fiber Raman backscattered light signal data at different sampling frequencies. Based on the principle of optical time-domain reflection, different sampling intervals are converted into different spatial intervals. Since the sensing temperature fiber is fixed to the object to be measured, the initial position of the sampling signal is fixed, and the sampling point sequence can be rearranged according to the positional relationship. When the set sampling frequency does not change much, the sampling point sequence exhibits a dense distribution within a specific interval. Based on the distribution of the sampling points, a set of sampling points at specific locations can be selected to form an ideal sampling point set, which can avoid the limitation of spatial resolution by the sampling frequency of the acquisition card.

[0052] Furthermore, the frequency conversion sampling frequency selection optimization method based on the multi-objective particle swarm optimization (MOPSO) algorithm includes the following steps in sequence:

[0053] Step 1: Determine the sampling rate range supported by the hardware device, and select the available frequency range f. max ~f min The algorithm inputs the ideal spatial resolution Δx and randomly generates a set of particle swarm optimization variables x(t) = (x1, x2, ... x) based on the multi-objective particle swarm optimization algorithm. i ,…,x m Each optimized particle x i (i = 1, 2, ..., m) corresponds to a frequency combination. Algorithm parameters include: maximum number of iterations T. maxNumber of particles m, maximum inertia factor ω max Minimum inertia factor w min Acceleration constants c1 and c2, maximum velocity v max Let the current optimization algebra be t = 1, t ≤ T. max In a space of dimension range, m particles x1, x2, ..., x3 are randomly generated. i ,…,x m A population X(t) is formed, and the initial velocities of each particle are randomly generated as v1, v2, ..., v i ,…,v m The position of the i-th particle is x. i =(x i，1 ,x i,2 ,…,x i，j ), with a speed of v i =(v i，1 v i，2 , ..., v i，j ), where j = f max -f min Particle x i The value of each element in (i = 1, 2, ..., m) is either 0 or 1;

[0054] Step 2: Based on the optimized particle x i The frequency combinations are used to calculate the distribution of sampling points at each sampling frequency. Based on the reconstruction algorithm and the ideal spatial resolution Δx, two sets of sampling points are re-selected to obtain two sets of high-precision reconstructed data distributions p1, p2, ..., p... n and q1, q2, ..., q n ;

[0055] Step 3: Establish a mathematical model to obtain an objective function that measures the accuracy of the reconstructed data and the number of samplings:

[0056]

[0057] Value2(x i ) = count

[0058] Where Value1(x) i ) and Value2(x i Let Δd be the two objective functions of the optimization algorithm. i It is the deviation between the reconstructed data and the position of that point under ideal spatial resolution, Δd i =|i·Δx-q i |+|i·Δx-p i|, count is the number of samples. The objective function for each optimized particle is calculated according to this model to select the individual's historical best position Pbest and the population's global historical best position Gbest;

[0059] Step 4: Generate a new set of particle swarm optimization variables X(t+1) according to the transformation rules of the multi-objective particle swarm optimization algorithm, and return to Step 2 to continue the next optimization until the particle swarm optimization algorithm reaches the maximum number of iterations. Then stop the calculation and output the frequency combination corresponding to the best particles, specifically including:

[0060] Step 401: Use the objective function obtained in Step 3 as the fitness value to evaluate the quality of each particle. Embed the Pareto dominance relation into PSO to determine the best position pbest in the entire population and the global best position gbest, based on the objective function Value1(x) obtained in Step 3. i ) and Value2(x i According to the Pareto dominance principle, the local optimum pbest is updated. If the current value does not dominate the original pbest, it is updated randomly. The global optimum gbest is selected from the Archive set by calculating the crowding in the external archive set; and the velocity and position of the particles are updated to generate a new set of particle swarm optimization variables.

[0061] Step 402, according to the formula:

[0062] v i,j (t+1)=ωv i,j (t)+c1r1[P i,j -x i,j (t)]+c2r2[P g，j -x i，j (t)]

[0063] Update particle velocity and formula Update the particle positions to generate a new population X(t+1), v i,j Let be the current velocity of the i-th particle with the j-th parameter; c1 and i2 represent positive acceleration coefficients, and r1 and r2 are random numbers between 0 and 1; p i，j pbest represents the best position found so far for the i-th particle; p g,j gbest represents the best position found by the entire particle swarm; x i，j (t) represents the current position of the j-th parameter of the i-th particle;

[0064] Step 403: According to the formula ω = ω min +r3·(ω max -ω minUpdate the inertia factor of the optimization algorithm, where r3 is a random number between 0 and 1; update the Archive set (which stores the current non-dominated solutions). If the number of particles in the Archive set exceeds the specified size, a truncation operation is required.

[0065] Step 404: Determine if t equals T. max If the conditions are met, the frequency combination corresponding to the best particle is output, along with the high-precision data index matrix under that frequency combination; otherwise, t = t + 1, and the process returns to step 402 to continue the search.

[0066] Furthermore, step 2 is based on optimizing particle x i The frequency combinations are used to calculate the distribution of sampling points at each sampling frequency. Based on the reconstruction algorithm, sampling points are re-selected according to the ideal spatial resolution Δx, and a set of high-precision temperature measurement data d1, d2, ..., d... is designed. n The steps are as follows:

[0067] Step 201: Let the set of sampling points on the sensing fiber be {d} i |i=1,2,…,N},whered i It is an arithmetic sequence starting from 0 with intervals of Δx; based on particle x i The frequency combinations can be determined by first calculating the deviation between each ideal sampling point and its two closest actual sampling points at a single sampling frequency in the combination. Let the set of actual sampling points at each frequency form a dataset A.

[0068]

[0069] Where a f，n Let f be the position of the nth sampling point from the initial point;

[0070] Step 202: Based on the calculated sampling point distribution matrix, select the sampling points that are closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution.

[0071]

[0072] For any n∈[1, N], m∈{f1, f2, ..., f m All of them are satisfied. where i∈[1, N]

[0073] Step 203: Exclude the sampling points already selected in C1, and perform the selection again to select the sampling points that are the second closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution, thus obtaining C2;

[0074] Step 204: At different sampling frequencies, find the actual sampling points that are closest to each ideal sampling point according to C1 and C2, and obtain the final index matrix:

[0075]

[0076] For any m∈{f1, f2, ..., f m}, all satisfy i = 1, 2, ..., N

[0077] In the index matrix F above, the first row F 1，i This indicates the sampling frequency selected for the i-th point of the reconstructed data curve, and the second row F 2，i Frequency F 1，i Next F 2，i One sampling point.

[0078] Step 205: Exclude the sampling points already selected in F from C1 and C2, recalculate and find the actual sampling points that are closest to each ideal sampling point, and obtain the second index matrix F′:

[0079]

[0080] Step 205: For each ideal sampling point, the locations and temperature data of the two actual sampling points closest to the ideal sampling point can be obtained from the index matrix F and the second index matrix F′. Assume the locations are D1 and D2, and the temperature measurements are T1 and T2. Based on the proximity of the locations, the temperature of the ideal sampling point is predicted using either trend extrapolation or interpolation. If the two points are on the same side of the ideal sampling point, trend extrapolation is used for prediction; if the two points are on opposite sides of the ideal sampling point, interpolation is used for prediction. Both methods use a linear regression model.

[0081] Finally, it should be noted that the above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system, characterized in that, The frequency selection optimization method based on the multi-objective particle swarm optimization (MOPSO) algorithm finds the optimal frequency combination within a given frequency range and spatial resolution. It collects data at different sampling frequencies within the frequency combination, uses a reconstruction algorithm to find the sampling point that is spatially closest to the ideal sampling point, and calculates the temperature value of that point.

2. The sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system as described in claim 1, characterized in that, By utilizing temperature information from nearby sampling points, the temperature measurement results at the ideal sampling point location can be predicted using trend extrapolation or interpolation.

3. The sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system as described in claim 1, characterized in that, The frequency conversion sampling frequency selection optimization method based on the multi-objective particle swarm optimization (MOPSO) algorithm includes the following steps: Step 1: Determine the sampling rate range supported by the hardware device, and select the available frequency range f. max ~f min The algorithm inputs the ideal spatial resolution Δx and randomly generates a set of particle swarm optimization variables x(t) = (x1, x2, ... x) based on the multi-objective particle swarm optimization algorithm. i ,…,x m Each optimized particle x i (i = 1, 2, ..., m) corresponds to a frequency combination; algorithm parameters include: maximum number of iterations T max Number of particles m, maximum inertia factor ω max Minimum inertia factor w min Acceleration constants c1 and c2, maximum velocity v max Set the current optimization algebra to t = 1, t ≤ T. max In a space of dimension range, m particles x1, x2, ..., x are randomly generated. i ,…,x m A population X(t) is formed, and the initial velocities of each particle are randomly generated as v1, v2, ..., v. i ,…,v m The position of the i-th particle is x. i =(x i，1 x i，2 , ..., x i，j ), with a speed of v i =(v i，1 v i，2 , ..., v i，j ), where j = f max -f min Particle x i The value of each element in (1,2,…,m) is either 0 or 1; Step 2: Based on the optimized particle x i The frequency combinations are used to calculate the distribution of sampling points at each sampling frequency. Based on the reconstruction algorithm and the ideal spatial resolution Δx, two sets of sampling points are re-selected to obtain two sets of high-precision reconstructed data distributions p1, p2, ..., p. n and q1, q2...,q n ; Step 3: Establish a mathematical model to obtain an objective function that measures the accuracy of the reconstructed data and the number of samplings: Value2(x i )=count Where Value1(x) i ) and Value2(x i Let Δd be the two objective functions of the optimization algorithm. i It is the deviation between the reconstructed data and the location of that point under ideal spatial resolution. i =|i·Δx-q i |+|i·Δx-p i |, count is the number of samples. The objective function for each optimized particle is calculated according to this model to select the individual's historical best position Pbest and the population's global historical best position Gbest; Step 4: Generate a new set of particle swarm optimization variables X(t+1) according to the transformation rules of the multi-objective particle swarm algorithm, and return to step 2 to continue the next optimization until the particle swarm algorithm reaches the maximum number of iterations. Then stop the calculation and output the frequency combination corresponding to the best particle.

4. The sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system as described in claim 1, characterized in that, Step 2 includes the following steps in sequence: Step 201: Let the set of sampling points on the sensing fiber be {d} i |i=1,2,…,N},whered i It is an arithmetic sequence starting from 0 with intervals of Δx; based on particle x i The frequency combinations can be determined by first calculating the deviation between each ideal sampling point and its two closest actual sampling points at each individual sampling frequency in the combination. Let the set of actual sampling points at each frequency form a dataset A. Where a f，n Let f be the position of the nth sampling point from the initial point; Step 202: Based on the calculated sampling point distribution matrix, select the sampling points that are closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution. For any n∈[1,N], m∈{f1,f2,…,f m All of them are satisfied. where i∈[1,N] Step 203: Exclude the sampling points already selected in C1, and perform the selection again to select the sampling points that are the second closest to the actual sampling points at each frequency and the sampling points at the ideal spatial resolution, thus obtaining C2; Step 204: At different sampling frequencies, find the actual sampling points that are closest to each ideal sampling point according to C1 and C2, and obtain the final index matrix: For any m∈{f1, f2, ..., f m }, all satisfy i = 1, 2, ..., N In the index matrix F above, the first row F 1,i This indicates the sampling frequency selected for the i-th point of the reconstructed data curve, and the second row F 2,i Frequency F 1,i Next F 2,i One sampling point; Step 205: Exclude the sampling points already selected in F from C1 and C2, recalculate and find the actual sampling points that are closest to each ideal sampling point, and obtain the second index matrix F′: Step 205: For each ideal sampling point, obtain the locations and temperature data of the two actual sampling points closest to the ideal sampling point from the index matrix F and the second index matrix F′. Assume the locations are D1 and D2, and the temperature measurement results are T1 and T2. Based on the proximity of the locations, use the trend extrapolation method or the interpolation method to predict the temperature of the ideal sampling point. If the two points are on the same side of the ideal sampling point, the trend extrapolation method is used for prediction; if the two points are on opposite sides of the ideal sampling point, the interpolation method is used for prediction. Both methods use a linear regression model.

5. The sampling optimization method for a high spatial resolution OTDR fiber optic temperature measurement system as described in claim 1, characterized in that, Step 4 specifically includes the following steps: Step 401: Use the objective function obtained in Step 3 as the fitness value to evaluate the quality of each particle. Embed the Pareto dominance relation into PSO to determine the best position pbest in the entire population and the global best position gbest, based on the objective function Value1(x) obtained in Step 3. i ) and Value2(x i According to the Pareto dominance principle, the local optimum pbest is updated. If the current value does not dominate the original pbest, it is updated randomly. The global optimum gbest is selected in the Archive set by calculating the crowding in the external archive set; and the velocity and position of the particles are updated to generate a new set of particle swarm optimization variables. Step 402, according to the formula: v i,j (t+1)=ωv i,j (t)+c1r1[P i，j -x i，j (t)]+c2r2[P g,j -x i，j (t)] Update particle velocity and formula Update the particle positions to generate a new population X(t+1), v i,j Let be the current velocity of the i-th particle with the j-th parameter; c1 and c2 represent positive acceleration coefficients, and r1 and r2 are random numbers between 0 and 1; p i,j pbest represents the best position found so far for the i-th particle; p g,j gbest represents the best position found by the entire particle swarm; x i，j (t) represents the current position of the j-th parameter of the i-th particle; Step 403: According to the formula ω = ω min +r3·(ω max -ω min Update the inertia factor of the optimization algorithm, where r3 is a random number between 0 and 1; update the Archive set (which stores the current non-dominated solutions). If the number of particles in the Archive set exceeds the specified size, a truncation operation is required. Step 404: Determine if t equals T. max If the conditions are met, the frequency combination corresponding to the best particle is output, along with the high-precision data index matrix under that frequency combination; otherwise, t = t + 1, and the search returns to step 402.

Images