An Optical Dissolved Oxygen Sensor Calibration Method Based on Nonlinear Temperature Compensation

Abstract

The present invention discloses a calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation, belonging to the field of sensor calibration, which includes the following steps: S1, building a calibration platform; S2, data acquisition: at multiple temperature gradients, generating a dissolved oxygen saturated water sample through a nitrogen-oxygen mixed gas, and synchronously collecting the fluorescence lifetime and influencing factors; S3, establishing a non-linear compensation model between the fluorescence lifetime and temperature, and optimizing the model parameters through the Levenberg-Marquardt algorithm; S4, dynamically allocating weight coefficients; S5, predicting the drift amount of the optical dissolved oxygen sensor to be calibrated to achieve closed-loop self-maintenance calibration. By adopting the above-mentioned calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation, not only the influence of temperature on the optical dissolved oxygen sensor is considered, but also multiple environmental factors such as salinity and pressure are incorporated. Through multi-parameter collaborative correction, the accuracy and comprehensiveness of calibration are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor calibration, and particularly to a calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation. Background Art

[0002] An optical dissolved oxygen sensor is a precision device for measuring the dissolved oxygen concentration in water. It detects the oxygen content in water through a photochemical reaction. This type of sensor typically includes a driving light source (LED) and a photodetector. When blue light of a specific wavelength irradiates the fluorescent substance at the front end of the sensor, the substance absorbs energy and emits red light. If oxygen molecules exist in the surrounding environment, these molecules will collide with the excited fluorescent substance, resulting in a decrease in its luminescence intensity or the cessation of luminescence. The resulting signal change can be used to calculate the dissolved oxygen concentration.

[0003] The traditional calibration method for optical dissolved oxygen sensors has the following technical defects:

[0004] 1. The temperature compensation model is overly simplified: Conventional methods mostly use linear or second-order polynomial fitting to describe the relationship between temperature and fluorescence lifetime. However, it is difficult to describe the non-linear thermodynamic process (such as the change in fluorescence quenching activation energy) in a wide temperature range (0 - 40°C), resulting in an error accumulation of more than ±2.5% in the high-temperature region. For example, the existing technology obtains the dissolved oxygen reference value through the iodometric method. Although it can eliminate part of the temperature interference, it requires manual preparation of chemical reagents and cannot achieve in-situ calibration, with a high operation complexity (time-consuming ≥ 30 minutes per time), and the accuracy drops by 35% when the salinity > 20 PSU.

[0005] 2. The lack of collaborative correction of multiple environmental parameters: Interference factors such as salinity and pressure are not included in the calibration system. The existing technology mostly adopts a step-by-step correction strategy (such as temperature compensation first and then pressure correction), resulting in an error superposition effect.

[0006] 3. Insufficient suppression of dynamic drift: The traditional light source control relies on a constant current drive module and does not establish a closed-loop feedback mechanism, resulting in the inability to correct in real-time the LED light intensity drift (±3%) and the baseline drift caused by optical window contamination (> 0.8 mg / L / week). Some methods use static calibration coefficients to compensate for the temperature effect, but do not integrate dynamic algorithms (such as the LSTM prediction model), resulting in poor long-term stability (the drift amount in 30 days > 1.2 mg / L). Summary of the Invention

[0007] The purpose of the present invention is to provide a calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation to solve the above technical problems.

[0008] To achieve the above object, the present invention provides a calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation, comprising the following steps:

[0009] S1. Build a calibration platform: Fill water into a multi-temperature reaction chamber, and immerse the optical dissolved oxygen sensor to be calibrated, a temperature sensor, a salinity sensor, and a pressure sensor in the water;

[0010] S2. Data acquisition: Generate dissolved oxygen saturated water samples through a nitrogen-oxygen mixed gas at multiple temperature gradients, and synchronously collect fluorescence lifetimes and influencing factors;

[0011] S3. Model construction: Based on the combined fitting of the Arrhenius equation and a fourth-order polynomial, establish a non-linear compensation model between fluorescence lifetime and temperature, and optimize the model parameters through the Levenberg-Marquardt algorithm;

[0012] S4. Calculate the theoretical dissolved oxygen saturation concentration using the improved Weiss equation, and dynamically allocate weight coefficients in combination with the measurement standard deviations of the salinity sensor and the pressure sensor to achieve multi-parameter collaborative correction;

[0013] S5. Adjust the intensity fluctuation of the driving light source of the optical dissolved oxygen sensor to be calibrated through a PID controller, and use an LSTM network to predict the drift amount of the optical dissolved oxygen sensor to be calibrated to achieve closed-loop self-maintenance calibration.

[0014] Preferably, the multiple temperature gradients described in step S2 include 0 °C, 10 °C, 20 °C, 30 °C, 40 °C; the influencing factors include dissolved oxygen concentration , salinity and pressure .

[0015] Preferably, step S3 specifically includes the following steps:

[0016] S31. Use the Arrhenius equation to describe the effect of temperature on the activation energy of fluorescence quenching:

[0017] (1);

[0018] In the formula, represents the time constant of the fluorescence quenching process at the water temperature ; represents the activation energy of the fluorescence quenching process, with the unit of eV; represents the theoretical fluorescence lifetime at absolute zero, with the unit of μs; is the Boltzmann constant, and ;

[0019] S32. Use a fourth-order polynomial to compensate for the non-linear optical drift caused by temperature:

[0020] (2);

[0021] Wherein, represents the predicted fluorescence lifetime; , , , and are the baseline drift compensation constant term, the linear drift compensation constant term, the thermal expansion effect compensation constant term, the non-linear optical response constant term, and the high-order thermodynamic effect constant term respectively;

[0022] S33. Obtain the non-linear compensation model of fluorescence lifetime and temperature by combining formula (1) and formula (2):

[0023] (3);

[0024] Wherein, represents the water temperature and the final fluorescence lifetime;

[0025] S34. Optimize the model parameters by the Levenberg-Marquardt algorithm;

[0026] S341. Initialize the parameters: , , the initial damping factor , the initial step size scaling factor , and estimate by the least squares method;

[0027] S342. Iterative optimization:

[0028] S3421. Calculate the sum of squared residuals :

[0029] (4);

[0030] Wherein, represents the amount of data collected at each temperature gradient; represents the water temperature and the measured fluorescence lifetime;

[0031] S3422. Adjust the damping factor , and update the parameter vector ;

[0032] S3423. Iterate steps S3421 - S3422 until the set number of iterations is reached, and output the optimized non-linear compensation model of fluorescence lifetime and temperature;

[0033] S343. Verify the optimized non - linear compensation model of fluorescence lifetime and temperature using goodness - of - fit, residual control, and temperature sensitivity, and output the optimized non - linear compensation model of fluorescence lifetime and temperature after passing the verification.

[0034] Preferably, step S4 specifically includes the following steps:

[0035] S41. Construct the Weiss standard equation:

[0036] (5);

[0037] In the formula, represents the saturation vapor pressure at water temperature ; , and are all empirical coefficients, and ; represents the standard atmospheric pressure;

[0038] S42. Construct the comprehensive saturation concentration formula:

[0039] (6);

[0040] In the formula, represents the comprehensive saturation concentration; represents the theoretical saturation concentration considering only pressure; represents the theoretical saturation concentration considering only salinity;

[0041] Among them,

[0042] (7);

[0043] (8);

[0044] In the formula, represents the salinity, with the unit of PSU; , and are all salinity correction coefficients, and ; represents the real - time pressure, with the unit of kPa; represents the pressure non - linear correction coefficient, and ;

[0045] S43. Dynamically allocate weight coefficients:

[0046] (9);

[0047] In the formula, represents the corrected dissolved oxygen value; Represents the salinity sensor weight coefficient; Represents the original dissolved oxygen measurement value, with the unit of mg / L; Represents the pressure sensor weight coefficient;

[0048] Among them,

[0049] (10);

[0050] In the formula, Represents the measurement standard deviation of the salinity sensor; Represents the measurement standard deviation of the pressure sensor.

[0051] Preferably, step S5 specifically includes the following steps:

[0052] S51. Data acquisition: Synchronously acquire the driving light source current value every 1 second , the original dissolved oxygen measurement value , water temperature and salinity and perform preprocessing;

[0053] S52. PID regulates the light source intensity;

[0054] S521. Real-time collect the driving light source current fluctuation signal , and generate the error signal at the moment with the set value as the reference;

[0055] S522. Use the incremental PID algorithm to calculate the control quantity of the driving light source intensity :

[0056] (11);

[0057] In the formula, , and represent the proportional gain coefficient, integral gain coefficient, and differential gain coefficient; Represents the initial moment; Represents the sampling period; Represents the error signal at the moment;

[0058] S523. Output the control voltage through PWM modulation to suppress the driving light source current fluctuation within the set threshold range;

[0059] S53. Use the LSTM network to predict the drift amount of the optical dissolved oxygen sensor to be calibrated;

[0060] S531. Construct an LSTM network structure, which includes an input layer, two hidden layers, and an output layer arranged in sequence. The input variable of the input layer is the water temperature , salinity , and the cumulative working time of the driving light source ; The output layer outputs the predicted value of the moment drift , denotes the Sigmoid function, denotes the weight matrix, denotes the state of the hidden layer, denotes the bias term;

[0061] S532. Set dynamic compensation when the predicted value of the drift is greater than the set value:

[0062] (12);

[0063] In the formula, denotes the corrected dissolved oxygen value at time denotes the original dissolved oxygen value; denotes the drift change rate suppression factor;

[0064] S533. Parameter self-update:

[0065] (13);

[0066] In the formula, denotes the updated PID controller parameters; denotes the PID parameter learning rate; denotes the control performance index.‌

[0067] Preferably, the preprocessing described in step S51 includes removing abnormal current values with a fluctuation amplitude > ±5% of the set value, and performing sliding window mean filtering on the original dissolved oxygen measurement value .

[0068] Preferably, the set threshold range in step S523 is ±0.3 mA.

[0069] Therefore, the present invention adopts the above-mentioned optical dissolved oxygen sensor calibration method based on non-linear temperature compensation, and has the following beneficial effects:

[0070] 1. Improve the calibration accuracy: Through the non-linear temperature compensation model and multi-parameter collaborative correction, the calibration accuracy of the optical dissolved oxygen sensor is significantly improved, especially the accuracy under wide temperature range and high salinity conditions;

[0071] 2. Enhanced environmental adaptability: Considering the impacts of multiple environmental factors such as temperature, salinity, and pressure on the sensor comprehensively, the adaptability and stability of the sensor in different environments are enhanced;

[0072] 3. Achieved closed-loop self-maintenance calibration: Through the combination of a PID controller and an LSTM network, real-time monitoring and dynamic compensation of the light source intensity fluctuation and the sensor drift amount are achieved, reducing the maintenance cost and improving the long-term stability of the sensor;

[0073] 4. Simplified operation process: There is no need to manually prepare chemical reagents, and in-situ calibration can be achieved, simplifying the operation process and improving the calibration efficiency;

[0074] 5. Improved sensor performance: By optimizing the model parameters and dynamic correction strategies, the baseline drift and light intensity drift of the sensor are effectively suppressed, improving the measurement accuracy and stability of the sensor.

[0075] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0076] Figure 1 It is a flowchart of a calibration method for an optical dissolved oxygen sensor based on nonlinear temperature compensation according to the present invention. Detailed Embodiments

[0077] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of this application. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions from beginning to end.

[0078] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0079] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0080] As Figure 1As shown in the figure, a calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation includes the following steps:

[0081] S1. Set up a calibration platform: Fill water into a multi-temperature zone reaction chamber, and immerse the optical dissolved oxygen sensor to be calibrated, temperature sensor, salinity sensor, and pressure sensor in the water;

[0082] S2. Data acquisition: Generate dissolved oxygen saturated water samples through a nitrogen-oxygen mixed gas at multiple temperature gradients, and synchronously collect the fluorescence lifetime and influencing factors;

[0083] The multiple temperature gradients described in step S2 include 0 °C, 10 °C, 20 °C, 30 °C, and 40 °C; the influencing factors include dissolved oxygen concentration , salinity and pressure .

[0084] S3. Model construction: Based on the combined fitting of the Arrhenius equation and the fourth-order polynomial, establish a non-linear compensation model for the fluorescence lifetime and temperature, and optimize the model parameters through the Levenberg-Marquardt algorithm;

[0085] Step S3 specifically includes the following steps:

[0086] S31. Use the Arrhenius equation to describe the influence of temperature on the activation energy of fluorescence quenching:

[0087] (1);

[0088] In the formula, represents the time constant of the fluorescence quenching process at the water temperature ; represents the activation energy of the fluorescence quenching process, with the unit of eV; represents the theoretical fluorescence lifetime at absolute zero, with the unit of μs; is the Boltzmann constant, and ;

[0089] S32. Use the fourth-order polynomial to compensate for the non-linear optical drift caused by temperature:

[0090] (2);

[0091] In the formula, represents the predicted fluorescence lifetime; , , , and are the baseline drift compensation constant term, linear drift compensation constant term, thermal expansion effect compensation constant term, non-linear optical response constant term, and high-order thermodynamic effect constant term respectively;

[0092] S33. Obtain the non - linear compensation model of fluorescence lifetime and temperature by combining formula (1) and formula (2):

[0093] (3);

[0094] Wherein, represents the water temperature and the final fluorescence lifetime under this condition;

[0095] S34. Optimize the model parameters by Levenberg - Marquardt algorithm;

[0096] S341. Initialize the parameters: , , the initial damping factor , the initial step - size scaling factor , and estimate by the least - squares method;

[0097] S342. Iterative optimization:

[0098] S3421. Calculate the sum of squared residuals :

[0099] (4);

[0100] Wherein, represents the amount of data collected at each temperature gradient; represents the water temperature and the measured fluorescence lifetime under this condition;

[0101] S3422. Adjust the damping factor , and update the parameter vector ;

[0102] S3423. Iterate steps S3421 - S3422 until the set number of iterations is reached, and output the optimized non - linear compensation model of fluorescence lifetime and temperature;

[0103] S343. Verify the optimized non - linear compensation model of fluorescence lifetime and temperature by using the goodness of fit, residual control and temperature sensitivity, and output the optimized non - linear compensation model of fluorescence lifetime and temperature after passing the verification.

[0104] S4. Calculate the theoretical dissolved oxygen saturation concentration by using the improved Weiss equation, and dynamically allocate the weight coefficients in combination with the measurement standard deviations of the salinity sensor and the pressure sensor to achieve multi - parameter collaborative calibration;

[0105] Step S4 specifically includes the following steps:

[0106] S41. Construct the Weiss standard equation:

[0107] (5);

[0108] In the formula, represents the saturated vapor pressure at water temperature ; , and are all empirical coefficients, and ; represents the standard atmospheric pressure;

[0109] S42. Construct the comprehensive saturation concentration formula:

[0110] (6);

[0111] In the formula, represents the comprehensive saturation concentration; represents the theoretical saturation concentration considering only pressure; represents the theoretical saturation concentration considering only salinity;

[0112] Among them,

[0113] (7);

[0114] (8);

[0115] In the formula, represents the salinity, with the unit of PSU; , and are all salinity correction coefficients, and ; represents the real-time pressure, with the unit of kPa; represents the pressure nonlinear correction coefficient, and ;

[0116] S43. Dynamically allocate the weight coefficient:

[0117] (9);

[0118] In the formula, represents the corrected dissolved oxygen value; represents the salinity sensor weight coefficient; represents the original dissolved oxygen measurement value, with the unit of mg / L; represents the pressure sensor weight coefficient;

[0119] Among them,

[0120] (10);

[0121] In the formula, represents the measurement standard deviation of the salinity sensor; represents the measurement standard deviation of the pressure sensor.

[0122] S5. Adjust the driving light source intensity fluctuation of the optical dissolved oxygen sensor to be calibrated through a PID controller, and use an LSTM network (Long Short-Term Memory network) to predict the drift amount of the optical dissolved oxygen sensor to be calibrated, so as to achieve closed-loop self-maintenance calibration.

[0123] Step S5 specifically includes the following steps:

[0124] S51. Data acquisition: Synchronously acquire the driving light source current value , the original dissolved oxygen measurement value , the water temperature and the salinity per second and perform preprocessing;

[0125] Preferably, the preprocessing described in step S51 includes eliminating abnormal current values with a fluctuation amplitude > ±5% of the set value, and performing moving window mean filtering on the original dissolved oxygen measurement value .

[0126] S52. PID adjusts the light source intensity;

[0127] S521. Real-time collect the driving light source current fluctuation signal , and generate an error signal at time with the set value ;

[0128] S522. Calculate the control amount of the driving light source intensity using the incremental PID algorithm:

[0129] (11);

[0130] In the formula, , and represent the proportional gain coefficient, integral gain coefficient, and differential gain coefficient; represents the initial moment; represents the sampling period; represents the error signal at time

[0131] S523. Output a control voltage through PWM modulation (pulse width modulation) to suppress the driving light source current fluctuation within the set threshold range;

[0132] Preferably, the set threshold range described in step S523 is ±0.3 mA.

[0133] S53. Use the LSTM network to predict the drift amount of the optical dissolved oxygen sensor to be calibrated;

[0134] S531. Construct an LSTM network structure, which includes an input layer, two hidden layers, and an output layer arranged in sequence. The input variables of the input layer are water temperature , salinity , cumulative working time of the driving light source ; The output layer outputs the predicted value of the drift amount at time , represents the Sigmoid function, represents the weight matrix, represents the state of the hidden layer, represents the bias term;

[0135] S532. Set dynamic compensation when the predicted drift amount is greater than the set value:

[0136] (12);

[0137] In the formula, represents the corrected dissolved oxygen value at time represents the original dissolved oxygen value; represents the drift change rate suppression factor;

[0138] S533. Parameter self-update:

[0139] (13);

[0140] In the formula, represents the updated PID (Proportional-Integral-Derivative) controller parameters; represents the PID parameter learning rate; represents the control performance index.

[0141] Simulation experiment:

[0142] 1. Experimental preparation: Build a calibration platform: Fill water into the multi-temperature reaction chamber, and immerse the optical dissolved oxygen sensor, temperature sensor, salinity sensor, and pressure sensor to be calibrated in the water. Set the temperature gradient: Select 0°C, 10°C, 20°C, 30°C, 40°C as the experimental temperature gradient. Configure the dissolved oxygen saturated water sample: Generate the dissolved oxygen saturated water sample through a nitrogen-oxygen mixed gas.

[0143] 2. Data acquisition: At each temperature gradient, fluorescence lifetime, dissolved oxygen concentration, salinity, and pressure data are collected synchronously.

[0144] 3. Data preprocessing: Abnormal current values with a fluctuation amplitude > ±5% are removed, and a moving window mean filter is applied to the original dissolved oxygen measurement values.

[0145] 4. Model construction and optimization: Nonlinear compensation model: Based on the joint fitting of the Arrhenius equation and a fourth-order polynomial, a nonlinear compensation model of fluorescence lifetime and temperature is established.

[0146] Parameter optimization: The Levenberg-Marquardt algorithm is used to optimize the model parameters. By iteratively calculating the sum of squared residuals, adjusting the damping factor, and updating the parameter vector until the set number of iterations is reached.

[0147] 5. Multi-parameter collaborative calibration: The improved Weiss equation is used to calculate the theoretical dissolved oxygen saturation concentration. Combining the measurement standard deviations of the salinity sensor and the pressure sensor, the weight coefficients are dynamically allocated to achieve multi-parameter collaborative calibration.

[0148] 6. Closed-loop self-maintenance calibration: PID regulates the light source intensity: The current fluctuation signal of the driving light source is collected in real time. The control quantity is calculated through the incremental PID algorithm, and the control voltage is output through PWM modulation to suppress the light source current fluctuation within the range of ±0.3 mA.

[0149] LSTM network predicts the drift amount: An LSTM network structure is constructed to predict the sensor drift amount. When the predicted value is greater than the set value, dynamic compensation is performed, and the PID controller parameters are self-updated.

[0150] The simulation experiment results are as follows:

[0151] In the wide temperature range (0 - 40 °C), through the nonlinear temperature compensation model, the calibration accuracy of the optical dissolved oxygen sensor is significantly improved. Especially in the high-temperature range (40 °C), the cumulative error is reduced to within ±1.5%, which is more than 30% higher than the traditional method.

[0152] Under the conditions of salinity fluctuation of ±10‰ and pressure change of ±50 kPa, through the dynamic weight allocation strategy, multi-parameter collaborative calibration is achieved, and the comprehensive measurement error ≤ 2%, which is significantly better than the step-by-step correction strategy.

[0153] The fluctuation of the light source intensity is effectively suppressed, the light intensity drift rate is reduced to within ±1.5%, and the baseline drift amount is reduced to 1 / 8 of the original value, improving the long-term stability of the sensor. And the prediction accuracy of the LSTM network for the sensor drift amount exceeds 90%. After dynamic compensation, the long-term drift amount of the sensor is controlled within ±0.5 mg / L, thus proving the effectiveness of the present invention.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent substitutions, and these modifications or equivalent substitutions cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An optical dissolved oxygen sensor calibration method based on non-linear temperature compensation, characterized in that: It includes the following steps: S1. Build a calibration platform: Fill water into the multi-temperature reaction chamber, and immerse the optical dissolved oxygen sensor, temperature sensor, salinity sensor, and pressure sensor to be calibrated in the water; S2. Data acquisition: Generate dissolved oxygen saturated water samples through a nitrogen-oxygen mixed gas at multiple temperature gradients, and synchronously collect fluorescence lifetimes and influencing factors; S3. Model construction: Based on the joint fitting of the Arrhenius equation and the fourth-order polynomial, establish a non-linear compensation model of fluorescence lifetime and temperature, and optimize the model parameters through the Levenberg-Marquardt algorithm; S4. Calculate the theoretical dissolved oxygen saturation concentration using the improved Weiss equation, and dynamically allocate weight coefficients by combining the measurement standard deviations of the salinity sensor and the pressure sensor to achieve multi-parameter collaborative calibration; Step S4 specifically includes the following steps: S41. Construct the Weiss standard equation: （5）； In the formula, represents the saturated vapor pressure at water temperature; , and are all empirical coefficients, and ; represents the standard atmospheric pressure; S42. Construct the comprehensive saturation concentration formula: （6）； In the formula, represents the comprehensive saturation concentration; represents the theoretical saturation concentration considering only pressure; represents the theoretical saturation concentration considering only salinity; Where, （7）； （8）； wherein, represents the salinity, with the unit of PSU; , and are all salinity correction coefficients, and ; represents the real-time pressure, with the unit of kPa; represents the pressure non-linearity correction coefficient, and ; S43. Dynamically allocate weight coefficients: （9）； Wherein, represents the corrected dissolved oxygen value; represents the weight coefficient of the salinity sensor; represents the original dissolved oxygen measurement value, with the unit of mg / L; represents the weight coefficient of the pressure sensor; Where, （10）； In the formula, represents the measurement standard deviation of the salinity sensor; represents the measurement standard deviation of the pressure sensor; S5. Adjust the driving light source intensity fluctuation of the optical dissolved oxygen sensor to be calibrated through a PID controller, and use an LSTM network to predict the drift amount of the optical dissolved oxygen sensor to be calibrated to achieve closed-loop self-maintenance calibration.

2. A calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation according to claim 1, characterized in that: The multiple temperature gradients described in step S2 include 0°C, 10°C, 20°C, 30°C, and 40°C; the influencing factors include dissolved oxygen concentration , salinity , and pressure .

3. A calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation according to claim 2, characterized in that: Step S3 specifically includes the following steps: S31. Use the Arrhenius equation to describe the influence of temperature on the fluorescence quenching activation energy: （1）； In the formula, represents the time constant of the fluorescence quenching process at the water temperature; represents the activation energy of the fluorescence quenching process, with the unit of eV; represents the theoretical fluorescence lifetime at absolute zero, with the unit of μs; is the Boltzmann constant, and ; S32. Use the fourth-order polynomial term to compensate for the non-linear optical drift caused by temperature: （2）； In the formula, represents the predicted fluorescence lifetime; and are respectively the baseline drift compensation constant term, the linear drift compensation constant term, the thermal expansion effect compensation constant term, the nonlinear optical response constant term, and the high-order thermodynamic effect constant term; S33. Combine formula (1) and formula (2) to obtain a non-linear compensation model of fluorescence lifetime and temperature: （3）； Wherein, represents the water temperature and the final fluorescence lifetime thereunder; S34. Optimize the model parameters through the Levenberg-Marquardt algorithm; S341. Initialize parameters: , , the initial damping factor , the initial step size scaling factor , and estimate by the least squares method ; S342. Iterative optimization: S3421. Calculate the sum of squared residuals : （4）； In the formula, represents the amount of data collected for each temperature gradient; represents the water temperature at which the fluorescence lifetime is measured; S3422. Adjust the damping factor , and update the parameter vector ; S3423. Iterate steps S3421 - S3422 until the set number of iterations is reached, and output the optimized non-linear compensation model of fluorescence lifetime and temperature; S343. Verify the optimized non-linear compensation model of fluorescence lifetime and temperature using goodness of fit, residual control, and temperature sensitivity, and output the optimized non-linear compensation model of fluorescence lifetime and temperature after passing the verification.

4. A calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation according to claim 3, characterized in that: Step S5 specifically includes the following steps: S51. Data acquisition: Synchronously acquire the driving light source current value every 1 second , the original dissolved oxygen measurement value , water temperature and salinity and perform preprocessing; S52. PID regulate the light source intensity; S521. Real-time collect the current fluctuation signal of the driving light source , and generate an error signal at a certain moment with a set value as a reference; S522. Calculate the control quantity for driving the light source intensity using the incremental PID algorithm : （11）； Wherein, , and represent the proportional gain coefficient, the integral gain coefficient, and the derivative gain coefficient; represents the initial time; represents the sampling period; represents the error signal at time S523. Output a control voltage through PWM modulation to suppress the driving light source current fluctuation within the set threshold range; S53. Use an LSTM network to predict the drift amount of the optical dissolved oxygen sensor to be calibrated; S531. Construct an LSTM network structure, which includes an input layer, two hidden layers, and an output layer arranged in sequence. The input variable of the input layer is the water temperature , salinity , and the cumulative working time of the driving light source ; the output layer outputs the predicted value of the moment drift , denotes the Sigmoid function, denotes the weight matrix, denotes the state of the hidden layer, denotes the bias term; S532. Set to perform dynamic compensation when the predicted drift amount is greater than the set value: （12）； In the formula, represents the dissolved oxygen value after time correction; represents the original dissolved oxygen value; represents the drift change rate suppression factor; S533. Parameter self-update: （13）； In the formula, represents the updated PID controller parameters; represents the PID parameter learning rate; represents the control performance index.

5. A calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation according to claim 4, characterized in that: The preprocessing described in step S51 includes eliminating abnormal current values with a fluctuation amplitude > ±5% of the set value, and performing moving window mean filtering on the original dissolved oxygen measurement values and performing moving window mean filtering on the original dissolved oxygen measurement values 6. A calibration method for an optical dissolved oxygen sensor based on non-linear temperature compensation according to claim 4, characterized in that: The set threshold range described in step S523 is ±0.3 mA.

Images