Seawater parameter calibration system

Abstract

The invention provides a seawater parameter calibration system, and belongs to the technical field of seawater parameter measurement. A thermohaline depth device is used for constructing a three-dimensional structure model to identify a spring layer position, a multi-parameter water quality sensing device is used for collecting data to establish a correlation model, and a dissolved oxygen analysis device is combined with thermohaline data to calculate a correction coefficient; and the spectral analysis device extracts feature recognition chemical components. The method comprises the key steps that a parameter interaction compensation function and an OceanParamNet model are adopted to construct a seawater comprehensive parameter calibration model, original data are corrected, and abnormal values are detected. The system realizes layered storage and wireless transmission of data, dynamically adjusts sampling frequency and power consumption, periodically executes a self-calibration program to compensate aging drift of a sensor, realizes high-precision self-adaptive calibration of seawater parameters, and solves the technical problems of low precision and poor environmental adaptability of seawater multi-parameter collaborative calibration in the prior art.

Description

Technical Field

[0001] The invention belongs to the technical field of seawater parameter measurement, and in particular relates to a seawater parameter calibration system. Background Art

[0002] Seawater parameter monitoring is an important means of marine scientific research and environmental supervision. Traditional seawater parameter measurement mainly relies on single sensors or parameter acquisition systems, such as CTD (conductivity-temperature-depth) meters, pH meters, dissolved oxygen meters, etc., which monitor the physical, chemical and biological characteristics of seawater respectively. These systems usually use independent calibration methods, that is, standard solutions or reference instruments are used for calibration of single parameters, and are applied in the fields of marine surveys, water quality monitoring and marine ecosystem research.

[0003] However, the traditional single-parameter calibration method ignores the complex interactions between seawater parameters, such as the impact of temperature and salinity on pH and dissolved oxygen measurements, and the correction of pressure on the responses of various sensors. In addition, the calibration accuracy of existing systems is significantly reduced in complex marine environments such as thermocline and halocline areas. Sensors are greatly affected by biofouling and long-term drift, and lack adaptive correction capabilities, resulting in reduced reliability of long-term deployed monitoring data.

[0004] The current technology is difficult to solve the comprehensive calibration problem under the interactive influence of multiple seawater parameters, especially under the joint action of multiple environmental factors. How to establish an interactive compensation model between seawater parameters to achieve high-precision and adaptive parameter calibration, while meeting the monitoring needs in different sea areas and different environmental conditions, has become a key technical bottleneck restricting the development of seawater parameter monitoring systems. In other words, the existing technology has the technical problems of low accuracy and poor environmental adaptability of seawater multi-parameter collaborative calibration. Summary of the invention

[0005] In view of this, the present invention provides a seawater parameter calibration system, which can solve the technical problems of low accuracy and poor environmental adaptability of seawater multi-parameter collaborative calibration in the prior art.

[0006] The present invention is implemented as follows: the present invention provides a seawater parameter calibration system with a control chip, the control chip is electrically connected to a multi-parameter water quality sensor device, a temperature-salinity-depth detection device, a dissolved oxygen analysis device, a spectrum analysis device, a data storage device, a wireless communication device, a power supply device, a positioning device and a clock synchronization device, and a system control module is arranged in the control chip, and the system control module performs the following steps: using the temperature-salinity-depth detection device to perform profile scanning along the water depth direction; using the multi-parameter water quality sensor device to collect seawater pH, redox potential and turbidity data; using the dissolved oxygen analysis device to collect dissolved oxygen concentration and oxygen saturation data in seawater; using the spectrum analysis device to collect seawater light absorption coefficient; using the parameter interaction compensation function to input all parameters into a multivariate regression model to construct a seawater comprehensive parameter calibration model; based on the OceanParamNet model and the calibration model, the newly collected raw data is corrected; abnormal value detection is performed on the calibrated data; the sampling frequency is adjusted according to the battery power state fed back by the power supply device; and the self-calibration program is periodically executed.

[0007] Among them, the multi-parameter water quality sensor device is used to collect seawater pH, redox potential and turbidity data; the temperature, salt and depth detection device is used to collect seawater temperature, salinity and depth data; the dissolved oxygen analysis device is used to collect dissolved oxygen concentration and oxygen saturation data in seawater; the spectral analysis device is used to collect seawater light absorption coefficient, fluorescence intensity and pigment concentration data; the data storage device is used to store all collected data, system operating parameters and calibration models; the wireless communication device is used for data transmission between systems and communication with the host computer; the power supply device is used to provide the required power for the system and monitor the battery status; the positioning device is used to determine the system deployment location and depth profile information; the clock synchronization device is used to ensure the time consistency of multi-point measurement data.

[0008] Among them, the multi-parameter water quality sensing device samples water quality parameters once per minute, the temperature, salinity and depth detection device samples water quality parameters once every 10 seconds, the dissolved oxygen analysis device samples water quality parameters once every 30 seconds, and the spectral analysis device samples water quality parameters once per minute.

[0009] Among them, a temperature-salinity-depth detection device is used to perform profile scanning along the water depth direction. Specifically, the seawater temperature, salinity and pressure data at different depths are collected to construct a three-dimensional structure model of temperature-salinity-depth and identify the locations of thermocline and halocline.

[0010] Among them, seawater pH, redox potential and turbidity data are collected through a multi-parameter water quality sensor device, and the three-point calibration method is used to compensate for the influence of temperature and pressure on the measurement, and a correlation model between pH and redox potential is established.

[0011] Among them, the spectral analysis device is used to collect seawater light absorption coefficient, fluorescence intensity and pigment concentration data in the wavelength range of 400 to 700 nanometers, and the principal component analysis method is used to extract the absorption peak at the characteristic wavelength to identify the chlorophyll concentration and the content of colored soluble organic matter.

[0012] The input parameters of the parameter interaction compensation function include a temperature parameter matrix, a salinity parameter matrix, a depth parameter matrix, an optical characteristic parameter matrix, and a chemical characteristic parameter matrix, and the output result of the parameter interaction compensation function is a seawater parameter correction coefficient matrix.

[0013] The parameter interaction compensation function adopts a piecewise nonlinear mapping algorithm. First, the first-order correlation coefficients between the parameters are calculated, then the second-order cross-term influence matrix is ​​constructed, and finally the optimal compensation coefficient is solved by the iterative least squares method.

[0014] The specific structure of the OceanParamNet model is a multi-layer temporal attention network architecture, which includes an encoder-decoder dual-path structure and a residual connection mechanism. The encoder consists of a three-layer bidirectional long short-term memory network, each layer contains 128 neurons, which are used to capture the time series characteristics of seawater parameters; the decoder uses a four-layer fully connected neural network with 256, 128, 64 and 32 neurons respectively, which are used to map the feature space to the parameter correction space.

[0015] The steps for establishing the training data set during the pre-training process of the OceanParamNet model include: first, collecting at least 12 months of continuous monitoring data of seawater parameters covering four seasons; then classifying the data according to geographical location and marine environmental characteristics; then quality control of the original data; then data enhancement; and finally constructing labeled data.

[0016] Compared with the prior art, the present invention provides a seawater parameter calibration system. The present invention proposes a seawater parameter calibration system based on the collaborative work of a multi-parameter water quality sensing device, a temperature, salinity and depth detection device, a dissolved oxygen analysis device and a spectral analysis device, and realizes high-precision calibration of seawater parameters through parameter interaction compensation function and OceanParamNet model.

[0017] The system successfully solves the limitations of traditional single-parameter calibration methods. By constructing interactive compensation functions between parameters, the multi-dimensional parameter matrices such as temperature, salinity, depth, optical properties and chemical properties are input into the calculation to generate a correction coefficient matrix, effectively eliminating the mutual interference between parameters. The OceanParamNet model used by the system combines time series analysis and multi-scale attention mechanisms, can adapt to different sea areas and seasonal changes, realize adaptive calibration, and significantly improve the measurement accuracy in complex environments such as thermoclines and haloclines.

[0018] Through a comprehensive and intelligent self-calibration program and anomaly detection algorithm, the present invention solves the technical problems of low accuracy and poor environmental adaptability of seawater multi-parameter collaborative calibration in the prior art, realizes dynamic and accurate calibration of seawater parameters, meets the requirements of high-precision, adaptability and long-term stability of seawater parameter monitoring in different sea environments, and solves the technical problems of low accuracy and poor environmental adaptability of seawater multi-parameter collaborative calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the steps executed by the system control module in the present invention.

[0020] Figure 2 This is a schematic diagram of the composition of the seawater parameter calibration system in Example 2. DETAILED DESCRIPTION

[0021] 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.

[0022] The present invention provides a seawater parameter calibration system, comprising a control chip, a multi-parameter water quality sensor device, a temperature-salinity-depth detection device, a dissolved oxygen analysis device, a spectrum analysis device, a data storage device, a wireless communication device, a power supply device, a positioning device and a clock synchronization device, wherein the control chip is electrically connected to the multi-parameter water quality sensor device, the temperature-salinity-depth detection device, the dissolved oxygen analysis device, the spectrum analysis device, the data storage device, the wireless communication device, the power supply device, the positioning device and the clock synchronization device, respectively, and a system control module is arranged in the control chip, the multi-parameter water quality sensor device is used to collect seawater pH, redox potential and turbidity data, the temperature-salinity-depth detection device is used to collect seawater temperature, salinity and depth data, and the dissolved oxygen analysis device is used to collect The data of dissolved oxygen concentration and oxygen saturation in seawater, the spectral analysis device is used to collect seawater light absorption coefficient, fluorescence intensity and pigment concentration data, the data storage device is used to store all collected data, system operation parameters and calibration models, the wireless communication device is used for data transmission between systems and communication with the host computer, the power supply device is used to provide the power required by the system and monitor the battery status, the positioning device is used to determine the system deployment location and depth profile information, the clock synchronization device is used to ensure the time consistency of multi-point measurement data; the multi-parameter water quality sensor device samples water quality parameters once per minute, the temperature, salt and depth detection device samples once every 10 seconds, the dissolved oxygen analysis device samples once every 30 seconds, and the spectral analysis device samples once per minute.

[0023] like Figure 1As shown, the system control module is used to perform the following steps: S01, the system starts the self-checking stage, collects standard sample data through the multi-parameter water quality sensing device, the temperature-salinity-depth detection device, the dissolved oxygen analysis device and the spectrum analysis device, compares them with the standard values ​​pre-stored in the data storage device, calculates the initial deviation of each sensor, and generates a calibration parameter matrix; S02, obtaining the current geographical location information of the system based on the positioning device, retrieving the corresponding sea area historical parameter model from the data storage device, and initializing the seawater parameter prediction model in combination with the current season and tidal information; S03, using the temperature-salinity-depth detection device to perform profile scanning along the water depth direction, collect seawater temperature, salinity and pressure data at different depths, construct a temperature-salinity-depth three-dimensional structure model, and identify the positions of thermocline and halocline; S04, collecting seawater pH, redox potential and turbidity data through the multi-parameter water quality sensor device, using a three-point calibration method to compensate for the effects of temperature and pressure on the measurement, and establishing a pH and redox potential correlation model; S05, using the dissolved oxygen analysis device to collect dissolved oxygen concentration and oxygen saturation data in seawater, combining the temperature and salinity data obtained in S03, calculating the theoretical dissolved oxygen value according to the dissolved oxygen theoretical model, and comparing it with the measured value to obtain the dissolved oxygen correction coefficient; S06. Collecting seawater light absorption coefficient, fluorescence intensity and pigment concentration data within a wavelength range of 400 to 700 nanometers through the spectral analysis device, extracting absorption peaks at characteristic wavelengths using principal component analysis, and identifying chlorophyll concentration and colored soluble organic matter content; S07, using a parameter interaction compensation function to input all the parameters obtained from S03 to S06 into a multiple regression model, calculate the interaction between the parameters, and construct a seawater comprehensive parameter calibration model, wherein the input parameters of the parameter interaction compensation function include a temperature parameter matrix, a salinity parameter matrix, a depth parameter matrix, an optical property parameter matrix, and a chemical property parameter matrix, and the output result of the parameter interaction compensation function is a seawater parameter correction coefficient matrix; S08, based on the OceanParamNet model and the calibration model built in S07, corrects the newly collected raw data, including temperature compensation, salinity compensation, pressure compensation and mutual interference correction, to generate calibrated high-precision seawater parameter data; S09. Perform outlier detection on the calibrated data, implement outlier identification based on the moving median through the OceanParamNet model, identify data anomalies, and evaluate data validity in combination with the seawater parameter change rate threshold; S10, using the data storage device to record the original data, intermediate calculation results and calibrated data, using a hierarchical storage structure to ensure data integrity, and transmitting the data to the host computer system regularly through the wireless communication device; S11. According to the battery power status fed back by the power supply device, dynamically adjust the sampling frequency, data processing depth and wireless transmission power of each sensor device to achieve long-term stable operation of the system; S12, periodically execute the self-calibration procedure, use the inherent relationship between the physical and chemical parameters of seawater and the correlation constraints to fine-tune the calibration model parameters to compensate for the measurement errors caused by sensor aging and drift; The parameter interaction compensation function is used to optimize the multivariate regression model calculation process in step S07, and improve the accuracy and adaptability of the seawater comprehensive parameter calibration model. The temperature parameter matrix is ​​derived from the seawater temperature data collected by the temperature-salinity-depth detection device, the salinity parameter matrix is ​​derived from the seawater salinity data collected by the temperature-salinity-depth detection device, the depth parameter matrix is ​​derived from the seawater depth data collected by the temperature-salinity-depth detection device, the optical characteristic parameter matrix is ​​derived from the seawater light absorption coefficient, fluorescence intensity and pigment concentration data collected by the spectral analysis device, the chemical characteristic parameter matrix is ​​derived from the seawater pH, redox potential and turbidity data collected by the multi-parameter water quality sensor device and the dissolved oxygen concentration and oxygen saturation data in the seawater collected by the dissolved oxygen analysis device, and the seawater parameter correction coefficient matrix is ​​used to calibrate the original measurement data; the parameter interaction compensation function adopts a piecewise nonlinear mapping algorithm, firstly calculates the first-order correlation coefficient between the parameters, then constructs a second-order cross-term influence matrix, and finally solves the optimal compensation coefficient by iterative least squares method; The specific structure of the OceanParamNet model is a multi-layer temporal attention network architecture, including an encoder-decoder dual-path structure and a residual connection mechanism. The encoder is composed of a three-layer bidirectional long short-term memory network, each layer containing 128 neurons, which is used to capture the time series characteristics of seawater parameters; the decoder adopts a four-layer fully connected neural network, with the number of neurons being 256, 128, 64 and 32 respectively, which is used to map the feature space to the parameter correction space; the OceanParamNet model also introduces a four-head multi-scale attention mechanism, the number of heads of the attention mechanism is dynamically adjusted according to the sampling frequency of the temperature, salinity and depth detection device, the head width of the attention mechanism is proportional to the wavelength resolution of the spectral analysis device, and the depth of the attention mechanism is inversely proportional to the response time of the dissolved oxygen analysis device; the output layer of the OceanParamNet model uses a fully connected layer with a LeakyReLU activation function; The steps of establishing the training data set in the pre-training process of the OceanParamNet model include: first, collecting at least 12 months of continuous monitoring data of seawater parameters covering four seasons to ensure the integrity of the data in the time dimension; then classifying the data according to geographical location and marine environmental characteristics, and establishing a hierarchical data subset of different marine environmental types including nearshore waters, open ocean areas, upwelling areas, and thermocline areas; then quality control of the original data is performed, and an outlier detection algorithm based on physical constraints is used to remove erroneous data points; then data enhancement is performed to expand the diversity of training samples by adding Gaussian noise that conforms to the laws of marine physics, simulating sensor drift and parameter mutations; finally, label data is constructed, and the standard values ​​measured by high-precision research-type marine survey equipment are used as true labels to achieve the supervised learning training goals; The steps of pre-training the OceanParamNet model include: firstly, initializing the model, initializing the network weights using the Xavier method, and ensuring the effective propagation of the signal in the network at the initial stage of training; then executing a two-stage training strategy, in which a single parameter prediction task is used for pre-training in the first stage to learn the independent change law of each parameter, and a multi-parameter joint prediction task is used for fine-tuning in the second stage to learn the interaction relationship between the parameters; then implementing a cyclic decay learning rate strategy, with the initial learning rate set to 0.001, decaying to 80% of the original every 50 cycles, and a total training cycle of 300; then introducing regularization technology, including L2 weight regularization and Dropout, with a Dropout ratio of 0.2, to effectively prevent the model from overfitting; finally, adopting an early stopping strategy, stopping training when the validation set loss has not improved within 20 consecutive cycles, and selecting the model parameters with the best validation performance as the final pre-training model; Optionally, the multi-parameter water quality sensing device includes a conductivity electrode, a glass composite electrode, a redox electrode and a light scattering turbidity sensor, which is used to simultaneously measure the conductivity, pH, redox potential and turbidity of seawater. The conductivity electrode adopts a four-electrode structure to eliminate polarization effects, the glass composite electrode has a built-in temperature compensation circuit, the redox electrode adopts a combined structure of a platinum electrode and a silver-silver chloride reference electrode, and the light scattering turbidity sensor has an operating wavelength of 860 nanometers and a measurement angle of 90 degrees.

[0024] Optionally, the temperature, salt and depth detection device includes a high-precision platinum resistance temperature sensor, a conductivity sensor and a pressure sensor. The platinum resistance temperature sensor has an accuracy of ±0.002 degrees Celsius, the conductivity sensor adopts an inductive structure, and the measurement range is 0 to 70 millisiemens / cm. The pressure sensor adopts a silicon resonant structure, with a range of 0 to 2000 decibels and a resolution of 0.002 decibels.

[0025] Optionally, the dissolved oxygen analysis device includes a fluorescence quenching oxygen sensor and an electrochemical oxygen sensor. The fluorescence quenching oxygen sensor is based on the principle of kinetic fluorescence quenching, adopts blue light excitation and red light detection, and has a response time of less than 8 seconds. The electrochemical oxygen sensor adopts a combination structure of gold electrode and silver electrode, with a film thickness of 25 microns and a temperature compensation range of 0 to 40 degrees Celsius.

[0026] Optionally, the spectral analysis device includes a miniature spectrometer, a dual-optical path colorimetric system and a multi-wavelength fluorescence sensor. The miniature spectrometer has an operating wavelength range of 350 to 750 nanometers and a wavelength resolution of 2 nanometers. The dual-optical path colorimetric system uses a beam of reference light and a beam of measurement light to simultaneously detect the structure and compensate for light source drift. The multi-wavelength fluorescence sensor includes three groups of excitation light sources with excitation wavelengths of 435 nanometers, 470 nanometers and 530 nanometers, respectively.

[0027] Optionally, the data storage device adopts a layered storage architecture, including a fast cache layer, a primary storage layer and a backup storage layer, the fast cache layer is implemented using a static random access memory with a capacity of 512 kilobytes, the primary storage layer is implemented using a flash memory with a capacity of 32 megabytes, and the backup storage layer is implemented using a micro memory card with a capacity of 64 megabytes.

[0028] Optionally, the wireless communication device supports multiple communication protocols, including short-range Bluetooth communication, medium-range wireless LAN communication and long-range satellite communication. The short-range Bluetooth communication is used to connect to a portable terminal with a transmission rate of 3 megabits per second. The medium-range wireless LAN communication is used to connect to a shore-based site with a transmission rate of 150 megabits per second. The long-range satellite communication is used for ocean data transmission with a transmission rate of 9600 bits per second.

[0029] Optionally, the power supply device includes a main battery module, a solar charging module and a power management module. The main battery module uses a lithium thionyl chloride battery with a rated voltage of 3.6 volts and a capacity of 19 ampere hours. The solar charging module uses a monocrystalline silicon solar panel with a peak power of 2 watts. The power management module realizes battery charging and discharging protection, low voltage alarm and dynamic power consumption regulation.

[0030] Optionally, the positioning device integrates a global positioning system receiver, a Beidou positioning system receiver and an inertial measurement unit, the global positioning system receiver supports L1 and L2 frequency band signal reception, and the positioning accuracy is better than 2.5 meters, the Beidou positioning system receiver supports B1 and B2 frequency band signal reception, and the positioning accuracy is better than 2 meters, and the inertial measurement unit includes a three-axis accelerometer, a three-axis gyroscope and a three-axis electronic compass for auxiliary positioning and attitude measurement.

[0031] Optionally, the clock synchronization device uses a high-precision real-time clock with a temperature compensated crystal oscillator as the core, with a frequency stability better than one part per million and a temperature drift coefficient less than ±0.5 microseconds / degree Celsius. It is regularly synchronized with a reference clock source through a wireless time synchronization protocol to ensure that multi-point measurement data have a unified time standard.

[0032] Preferably, the control chip adopts a low-power 32-bit ARM Cortex-M4 microprocessor with an operating frequency of 120MHz, a hardware floating-point unit, a built-in 512KB flash memory and a 128KB SRAM. The control chip is connected to a multi-parameter water quality sensor and a temperature, salinity and depth detection device via an SPI bus, with a sampling rate controlled at 100Hz; communicates with a dissolved oxygen analysis device and a spectrum analysis device via an I²C bus, with a transmission rate of 400kbit / s; connects to a data storage device via a USB interface, with a transmission rate of 12Mbit / s; connects to a wireless communication device via a UART interface, with a baud rate of 115200bps; uses GPIO to directly control the power supply device, with a signal level of 3.3V; uses an RS-485 interface to communicate with a positioning device and a clock synchronization device, with a baud rate of 9600bps.

[0033] Preferably, the multi-parameter water quality sensor device supports an operating temperature of 0 to 45 degrees Celsius, and the measurement range includes: pH 2 to 12 pH units, accuracy ±0.01 pH; redox potential ±2000 mV, accuracy ±0.5 mV; turbidity 0 to 1000 NTU, accuracy ±2%. The device is equipped with a 13-bit ADC, a sampling frequency of 10 Hz, and a data output mode of RS-485.

[0034] Preferably, the working depth range of the temperature-salinity-depth detection device is 0 to 2000 meters, the temperature measurement range is -5 to 45 degrees Celsius, and the accuracy is ±0.002 degrees Celsius; the salinity measurement range is 0 to 70PSU, and the accuracy is ±0.003PSU; the pressure measurement range is 0 to 2000 decibels, and the accuracy is ±0.1%FS. The sensor adopts a titanium alloy shell, with a pressure resistance level of 2000 meters, and the signal output adopts a 24-bit ADC.

[0035] Preferably, the dissolved oxygen analysis device has a measurement range of 0 to 20 mg / L, an accuracy of ±0.1 mg / L, and a response time of <8 seconds; the oxygen saturation measurement range is 0 to 200%, with an accuracy of ±1%. The device adopts an optical isolation design, a power supply voltage of 5V±0.2V, and a power consumption of <120 mW.

[0036] Preferably, the spectrum analysis device has a wavelength range of 350 to 750 nanometers, a wavelength resolution of 2 nanometers, an optical path length of 10 millimeters, and a signal-to-noise ratio of >1000: 1. It is powered by 12V, has a peak power consumption of 0.8 watts, a standby power consumption of 0.1 watts, and a data output of a 16-bit resolution spectrum matrix.

[0037] Preferably, the data storage device uses an industrial-grade memory chip, has an operating temperature of -40 to 85 degrees Celsius, a read and write speed of >20MB / s, a total capacity of 96.5MB, supports FAT32 file system, and an interface type of USB2.0.

[0038] Preferably, the wireless communication device integrates multiple communication modules, the Bluetooth working frequency band is 2.4GHz, and the transmission distance is 30 meters; the wireless LAN supports 802.11b / g / n protocols, and the transmission distance is 200 meters; the satellite communication module supports short message transmission and has global coverage.

[0039] Preferably, the power supply device uses a lithium thionyl chloride battery pack with an operating voltage range of 3.0 to 3.9V and a rated capacity of 19Ah; the solar panel has a rated power of 2 watts, an open circuit voltage of 6.5V, and an operating current of 0.4A; and the power management module supports an operating temperature of -10 to 60 degrees Celsius.

[0040] Preferably, the positioning device and the clock synchronization device hardware are integrated into one module, supporting GPS / Beidou dual-mode positioning, with a positioning accuracy better than 2 meters and an update frequency of 1Hz; a built-in temperature-compensated crystal oscillator, a frequency stability better than 1ppm, and a timing accuracy better than 0.5 microseconds.

[0041] Preferably, the entire system is encapsulated in a waterproof casing with a protection level of IP68, a maximum pressure-resistant depth of 2000 meters, an operating temperature range of -5 to 45 degrees Celsius, a relative humidity of 0 to 100%, a total system weight of no more than 2.5 kilograms, and an overall dimension of 120 mm in diameter and 380 mm in length.

[0042] The specific implementation of the above steps is described in detail below.

[0043] The specific implementation method of step S01 is that after the system is started, the self-test program is first executed, and the initialization command is sent to each sensor device through the control chip, and the standard sample data is collected and compared with the pre-stored standard value. The control chip first reads the calibration parameters in the data storage device, sends a sampling instruction to the multi-parameter water quality sensor device, and collects the pH, redox potential and turbidity data of the standard sample; then instructs the temperature, salt and depth detection device to measure the temperature, salinity and pressure values ​​of the standard solution; then starts the dissolved oxygen analysis device to obtain the dissolved oxygen concentration and oxygen saturation of the standard oxygen solution; finally controls the spectral analysis device to scan the standard spectral solution and record the light absorption coefficient, fluorescence intensity and pigment concentration information. All collected data are compared with the preset standard values ​​in the data storage device, and the least squares method is used to calculate the deviation coefficient of each parameter to construct the initial calibration parameter matrix. The initial calibration parameter matrix adopts a diagonal matrix structure, the main diagonal elements represent the sensitivity correction coefficients of each sensor, and the non-diagonal elements represent the cross-interference coefficients between sensors. The deviation thresholds are set as: pH ± 0.02 pH units, redox potential ± 5 mV, turbidity ± 0.5 NTU, temperature ± 0.01 degrees Celsius, salinity ± 0.02 PSU, dissolved oxygen ± 0.05 mg / L, light absorption coefficient ± 0.01 / m. This step ensures that the sensor is in good condition before the system starts working, and establishes an accurate baseline calibration for subsequent measurements.

[0044] The specific implementation method of step S02 is to initialize the seawater parameter prediction model based on the current geographical location information of the system. The control chip first controls the positioning device to collect geographical coordinates, time and current environmental factors, and determines the precise longitude and latitude coordinates through the GPS and Beidou dual systems with an accuracy better than 2 meters; then retrieves the matching historical parameter library from the data storage device, uses the spatial proximity algorithm to find the nearest historical monitoring point data, and constructs the initial environmental parameter lookup table; then matches the current date information with the tidal database, determines the current tidal phase, and calculates the tidal coefficient; finally, the acquired geographical location, seasonal characteristics and tidal information are input into the environmental parameter prediction model, and the prior distribution of each parameter is initialized through the Bayesian probability network to generate a parameter prediction baseline for the current environment. The Bayesian network structure includes three layers, the input layer is the environmental feature node, the middle layer is the tidal and seasonal influencing factor node, and the output layer is the parameter prediction probability distribution node. The environmental feature similarity threshold is set to 0.85, the tidal influence coefficient is set to ±0.2, and the seasonal adjustment coefficient is adjusted between 0.6 and 1.4 according to the latitude change. The main purpose of this step is to combine historical data and current environmental factors to establish an initial prediction model to provide prior knowledge support for subsequent parameter calibration.

[0045] The specific implementation method of step S03 is to perform water depth profile scanning using a temperature-salinity-depth detection device. The control chip first controls the device to perform depth calibration to determine the water surface reference position; then collects temperature, salinity and pressure data at different depths in sequence according to the preset depth intervals, and the conventional setting is 0.5 meters from the surface to 10 meters, 1 meter from 10 meters to 100 meters, and 5 meters below 100 meters; then the collected data is synchronized in time and aligned in depth to eliminate sampling delays and sensor position differences; then the discrete depth point data is processed continuously using a cubic spline interpolation algorithm to generate a smooth temperature, salinity and depth relationship curve; finally, the thermocline and halocline positions are identified by gradient analysis, and the temperature gradient change rate threshold is set to 0.05 degrees Celsius / meter, and the salinity gradient change rate threshold is set to 0.01PSU / meter. The three-dimensional structural model obtained by calculation is stored as a three-dimensional matrix, including depth, temperature and salinity dimensions, and the matrix elements represent the parameter values ​​of the corresponding state points. This step aims to obtain vertical layered structural information of seawater and provide key data for understanding the characteristics of seawater at different depths.

[0046] The specific implementation method of step S04 is to collect and calibrate the chemical parameters of seawater through a multi-parameter water quality sensor. The control chip first controls the glass composite electrode and the redox electrode to be cleaned and pre-stabilized before measurement, and the stabilization time is 30 seconds; then the sampling program is started, and the pH and redox potential of the seawater are measured by the electrode, and the light scattering turbidity sensor measures the turbidity value; then the three-point calibration method is used to compensate the original data for temperature and pressure, and the standard buffer solution with pH values ​​of 4.01, 7.00 and 10.01 is selected to establish a calibration curve, the redox potential is calibrated with a +220 millivolt standard solution, and the turbidity is calibrated with 0, 20 and 100 NTU standard suspensions; then the temperature data collected by the temperature-salinity-depth detection device is used to calculate the influence coefficient of temperature on the electrode potential through the Nernst equation, and the temperature compensation coefficient is -0.003 pH / degree Celsius; finally, the pH and redox potential correlation model is established based on the measurement data, and the exponential relationship is used to fit the nonlinear relationship between the two. The sensor sampling frequency of this step is set to once per minute, the data averaging window is 5 samples, and the data validity judgment threshold is a continuous change rate of less than ±0.05pH / minute or ±10mV / minute. The purpose of this step is to obtain accurate seawater chemical characteristic parameters and provide basic data for comprehensive assessment of seawater quality.

[0047] The specific implementation of step S05 is to use a dissolved oxygen analysis device to measure the dissolved oxygen parameters of seawater and perform theoretical correction. The control chip first starts the fluorescence quenching oxygen sensor and the electrochemical oxygen sensor, and the preheating time is 60 seconds; then the two sensors are controlled to simultaneously collect dissolved oxygen concentration and oxygen saturation data, and the sampling frequency is once every 30 seconds; then the temperature and salinity data obtained in step S03 are substituted into the dissolved oxygen saturation calculation formula to calculate the theoretical dissolved oxygen value; then the measured dissolved oxygen value is compared with the theoretical calculated value, and the correction coefficient matrix is ​​calculated. The correction coefficient matrix includes temperature influence items, salinity influence items and pressure influence items; finally, the correction coefficient is applied to correct the real-time dissolved oxygen data to generate calibrated dissolved oxygen concentration data. In the dissolved oxygen theoretical model, the temperature influence coefficient is set to -0.28 mg / L·Celsius, the salinity influence coefficient is set to -0.05 mg / L·PSU, and the pressure influence coefficient is set to 0.01 mg / L·dB. The effective range of dissolved oxygen measurement is 0 to 20 mg / L, and the accuracy requirement is ±0.1 mg / L or ±2% of the reading, whichever is greater. This step ensures the accuracy of dissolved oxygen data under different temperature and salinity conditions, providing reliable data for marine biogeochemical research.

[0048] The specific implementation method of step S06 is to collect and analyze the optical characteristic parameters of seawater through a spectral analysis device. The control chip first controls the micro-spectrometer to perform baseline calibration and record the pure water background spectrum; then, within the wavelength range of 400 to 700 nanometers, the seawater sample is scanned in full spectrum with a step length of 2 nanometers to obtain the light absorption coefficient spectrum; then the multi-wavelength fluorescence sensor is started, and the corresponding fluorescence emission intensity is measured at 435 nanometers, 470 nanometers and 530 nanometers respectively; then the collected spectral data is processed by principal component analysis to extract the absorption peak at the characteristic wavelength, and the wavelength selection is based on the characteristic absorption peak of chlorophyll a at 438 nanometers and 678 nanometers, and the absorption characteristics of colored soluble organic matter at 350 to 400 nanometers; finally, the chlorophyll concentration and the content of colored soluble organic matter are calculated by the established spectral feature and concentration relationship model. The principal component analysis adopts the singular value decomposition algorithm, the characteristic value threshold is set to 90% of the total variance contribution rate, the fluorescence intensity detection limit is 0.05 relative fluorescence unit, and the chlorophyll a concentration detection limit is 0.05 micrograms / liter. The purpose of this step is to non-destructively detect biological and chemical components in seawater by optical methods, so as to provide a scientific basis for the assessment of marine primary productivity and water quality.

[0049] The specific implementation method of step S07 is to construct a seawater comprehensive parameter calibration model using a parameter interaction compensation function. The control chip first obtains the temperature parameter matrix, salinity parameter matrix, depth parameter matrix, optical characteristic parameter matrix and chemical characteristic parameter matrix from each sensor device S03 to S06; then the parameter interaction compensation function is applied to process each matrix, which first calculates the Pearson correlation coefficient between each parameter and constructs a parameter correlation matrix; then identifies strongly correlated parameter pairs based on the correlation matrix, and the correlation coefficient threshold is set to ±0.7; then constructs a second-order cross-term influence matrix, which includes the interaction influence coefficients between parameters; finally, the optimal compensation coefficient is solved by the iterative least squares method, and the iterative termination condition is that the parameter change rate is less than 0.1% or the maximum number of iterations is 500 times. The parameter interaction compensation function outputs a seawater parameter correction coefficient matrix, which is applied to the original measurement data for calibration. The function uses a piecewise nonlinear mapping algorithm to process the nonlinear relationship between parameters, and uses a cubic polynomial fitting for the temperature-salinity related area, and a logarithmic linear fitting for the optical-chemical parameter related area. The main purpose of the parameter interaction compensation function is to accurately characterize the complex correlation between seawater parameters and improve the accuracy and adaptability of the comprehensive calibration model.

[0050] The specific implementation method of step S08 is to correct the raw data based on the OceanParamNet model and the calibration model. The control chip first loads the pre-trained OceanParamNet model parameters and initializes the model calculation environment; then the newly collected raw data is preprocessed and aligned according to the sensor type and sampling time; then the calibration model constructed by S07 is used to calculate the basic correction parameters, including temperature compensation coefficient, salinity compensation coefficient, pressure compensation coefficient and mutual interference correction coefficient; then the raw data and basic correction parameters are used as the input of the OceanParamNet model, and the optimized correction parameters are calculated through the forward propagation of the model; finally, the optimized correction parameters are applied to calibrate the raw data to generate high-precision seawater parameter data. The temperature compensation range is -2 to 40 degrees Celsius, the salinity compensation range is 0 to 42PSU, and the pressure compensation range is 0 to 2000 decibels. The model reasoning process uses floating point 16-bit precision calculation to balance the calculation accuracy and efficiency. This step significantly improves the accuracy of seawater parameter measurement by combining traditional physical models and deep learning methods, especially in areas with rapid environmental changes.

[0051] The specific implementation method of step S09 is to use the OceanParamNet model to detect outliers on the calibrated data. The control chip first divides the calibrated parameter data into time series segments, with the segment length set to 10 minutes; then calculates statistical features for each segment of data, including the median, interquartile range, and moving average; then applies the outlier identification algorithm based on the moving median to calculate the deviation of each data point from the moving median, and if the deviation exceeds 3 times the interquartile range, it is marked as a potential outlier; then the potential outlier point is input into the anomaly detection module of the OceanParamNet model, and its rationality is evaluated in combination with the parameter historical trend and physical constraints; finally, the validity of the data point is determined based on the anomaly probability output by the model and the threshold of the seawater parameter change rate. The anomaly probability threshold is set to 0.85, and the parameter change rate threshold is: temperature 0.1 degrees Celsius / minute, salinity 0.05PSU / minute, pH 0.03pH unit / minute, dissolved oxygen 0.2 mg / L / minute. The detected outliers are processed in three ways: minor outliers are smoothed, moderate outliers are replaced by model interpolation, and severe outliers are marked as invalid data. This step ensures the reliability of the system output data and avoids the impact of outliers on subsequent analysis and application.

[0052] The specific implementation method of step S10 is to record data using a data storage device and transmit it through a wireless communication device. The control chip first organizes the original data, intermediate calculation results and calibrated data in a predetermined format; then writes the data storage device using a hierarchical storage structure, stores the high-frequency access real-time data in the fast cache layer, stores the processed standard data in the main storage layer, and backs up the original data to the backup storage layer; then adds metadata tags to the stored data, including timestamps, geographic locations, sensor status and processing levels; then performs timed data compression and integrity checks, uses lossless compression algorithms to reduce data volume, and ensures data integrity through cyclic redundancy checks; finally, according to the preset transmission strategy, the data is sent to the host computer system through a wireless communication device, and the nearshore area uses wireless LAN communication, and the offshore area switches to satellite communication. The data transmission frequency is set to once every 10 minutes near the coast and once every 1 hour in the offshore area. The transmitted data contains the latest calibration results and key original data. This step ensures reliable storage and timely transmission of data, providing support for remote monitoring and data analysis.

[0053] The specific implementation method of step S11 is to dynamically adjust the system working mode according to the state of the power supply device. The control chip first reads the battery power status, solar charging efficiency and system power consumption data fed back by the power supply device; then divides the working mode according to the remaining battery power, when the power is higher than 80%, it is the full-function mode, between 40% and 80% is the standard mode, between 20% and 40% is the energy-saving mode, and below 20% is the emergency mode; then dynamically adjust the sampling frequency of each sensor device according to different working modes, the full-function mode maintains the design frequency, the standard mode is reduced by 25%, the energy-saving mode is reduced by 50%, and the emergency mode is reduced by 75%; then optimize the data processing depth, reduce the function mode to reduce secondary calculations and complex algorithm execution; finally, adjust the wireless transmission power and transmission frequency according to communication requirements and battery status, and give priority to key data transmission when the power is low. The battery voltage monitoring threshold is set to 3.2 volts, and the low power consumption mode is entered below this value; the night work strategy is started when the solar charging efficiency is lower than 0.5 watts; the total power consumption of the system is controlled within 0.8 watts at full load and 0.3 watts in energy-saving mode. This step extends the system life through intelligent power management to ensure long-term stable operation.

[0054] The specific implementation of step S12 is to periodically execute the self-calibration procedure to compensate for sensor drift. The control chip first sets the self-calibration cycle according to the system operation schedule, and is generally set to execute once every 168 hours of operation; then collects the data of various seawater parameters during the calibration period, focusing on recording multiple sets of repeated measurements under a stable environment; then uses the inherent relationship between the physical and chemical parameters of seawater to construct constraints, such as the theoretical relationship between dissolved oxygen concentration and temperature and salinity, and the relationship between pH and carbonate system equilibrium; then cross-validates the data of each sensor based on these constraints, calculates the deviation trend and drift rate; finally, compensates for the detected drift by fine-tuning the calibration model parameters, and updates the calibration coefficient matrix. The sensor aging drift judgment threshold is: temperature sensor ±0.005 degrees Celsius / month, salinity sensor ±0.01PSU / month, pH electrode ±0.03pH unit / month, dissolved oxygen sensor ±0.1 mg / L / month. The self-calibration process uses the recursive least squares method to gradually adjust the parameters so that the model dynamically adapts to changes in sensor performance. This step compensates for hardware performance degradation through software, greatly extending the system calibration cycle and service life.

[0055] The mathematical model or calculation process involved in the present invention is described in detail below.

[0056] The calculation of the initial calibration parameter matrix in step S01 is specifically expressed as follows: ; In the formula, is the initial calibration parameter matrix; For the The sensitivity correction factor of each sensor; For the The sensor pairs The cross-interference coefficient of each sensor; is the total number of sensors in the system.

[0057] Sensitivity correction factor The calculation method is: ; In the formula, For the The standard value of the parameter; For the Measured values ​​of parameters.

[0058] Cross-interference coefficient The calculation method is: ; In the formula, Indicates The change of the parameter The cross-interference coefficient is obtained by fitting the measurement results of multiple sets of standard samples using the least squares method. This matrix structure can fully characterize the mutual influence between sensors and improve calibration accuracy.

[0059] The temperature-salinity-depth three-dimensional structure model in step S03 adopts a cubic spline interpolation algorithm, and its mathematical expression is: ; In the formula, For Depth The parameter value (temperature or salinity) at For the sampling depth points; For interval The spline coefficients on .

[0060] The spline coefficients are determined by: (function value is continuous); (function value is continuous); (first-order derivative is continuous); (Second-order derivative is continuous); In the formula, For the The measured parameter values ​​of the sampling points; and denote the first and second order derivatives respectively.

[0061] Thermocline location The judgment conditions are: ; In the formula, For Depth The temperature value at is the temperature gradient threshold, set to 0.05 degrees Celsius / meter.

[0062] Halocline location The judgment conditions are: ; In the formula, For Depth Salinity value at is the salinity gradient threshold, set to 0.01 PSU / m.

[0063] The expression of the relationship between pH and temperature compensation in step S04 is: ; In the formula, is the pH value after temperature compensation; is the original pH value measured; is the temperature compensation coefficient, the value is -0.003pH / degree Celsius; is the actual water temperature; is the reference temperature, usually 25 degrees Celsius.

[0064] The correlation model between pH and redox potential: ; In the formula, is the redox potential (mV); is the standard redox potential, related to the reference electrode; is the gas constant, 8.314 J / (mol·K); is the absolute temperature (K); is the Faraday constant, 96485C / mol; is the pH value; is the organic matter influence coefficient, ranging from 30 to 50 mV; is the attenuation coefficient, ranging from 0.2 to 0.5; is the concentration of dissolved organic carbon (mg / L). This model takes into account the comprehensive effects of water pH, temperature and organic matter content on redox potential.

[0065] The expression of the dissolved oxygen theoretical model in step S05 is: ; In the formula, is the saturated dissolved oxygen concentration (mg / L); is the absolute temperature (K); is salinity (PSU); to and to are empirical coefficients, and their values ​​are: , , , , , , .

[0066] The expression of the effect of pressure on dissolved oxygen is: ; In the formula, is the dissolved oxygen concentration after pressure compensation; is the measured original dissolved oxygen concentration; is the pressure influence coefficient, and its value is 0.00004 / dB; is the water pressure (dB).

[0067] Dissolved oxygen correction factor matrix: ; In the formula, are the first-order correction coefficients for temperature, salinity and pressure, respectively; are the second-order correction coefficients for temperature, salinity, and pressure, respectively. Each coefficient is obtained by comparing the theoretical dissolved oxygen value with the measured value using the multiple regression method.

[0068] In step S06, principal component analysis is applied to spectral data processing: ; In the formula, is the original spectral data matrix, where rows represent samples and columns represent wavelengths; is the left singular vector matrix; is a diagonal matrix of singular values; is the transpose of the right singular vector matrix.

[0069] Characteristic wavelength selection conditions: ; In the formula, is the selected characteristic wavelength set; For the wavelength; is the right singular vector matrix Line Column elements; Select the threshold value for the characteristic wavelength and set it to 0.2; For the number of principal components to be selected, the first few principal components with a contribution rate of 90% are usually selected.

[0070] Chlorophyll concentration calculation formula: ; In the formula, is the chlorophyll a concentration (μg / L); , and The absorbances at 438 nm, 678 nm, and 750 nm, respectively; to is the calibration coefficient, obtained by regression of standard samples.

[0071] The key calculation process of the parameter interaction compensation function in step S07 is: Parameter correlation matrix: ; In the formula, For the Parameters and The Pearson correlation coefficient of the parameters is calculated as follows: ; In the formula, For the In the sample The value of the parameter; For the The average value of the parameters; is the sample size.

[0072] Second-order cross-term influence matrix: ; In the formula, Representation parameters and parameters The cross-effect coefficient.

[0073] Comprehensive calibration model of seawater parameters: ; In the formula, For parameters Correction value of For parameters The measured value of For parameters Reference value of is the first-order correction coefficient; is the second-order cross correction coefficient.

[0074] The correction formula of the original data by the OceanParamNet model in step S08 is: ; In the formula, For parameters The final revised value of For parameters The original measurement value of is the temperature compensation coefficient; is the salinity compensation coefficient; is the pressure compensation coefficient; is the mutual interference correction factor.

[0075] The calculation method of each compensation coefficient is: ; ; ; ; In the formula, , , are the actual temperature, salinity and pressure values ​​respectively; , , are reference temperature, reference salinity and reference pressure respectively; For parameters The actual value of For parameters Reference value of , , For parameters Temperature, salinity and pressure compensation coefficients; For parameters Parameters The interference coefficient.

[0076] The outlier identification algorithm based on the moving median in step S09: ; ; ; In the formula, For the The deviation of a data point from the moving median; For the The parameter value of each data point; is the median window width, usually set to 5 to 10; is the interquartile range; and are the first and third quartiles, respectively; Abnormal flag, 1 indicates abnormality, 0 indicates normality.

[0077] The principles for constructing these equations are mainly based on the physical and chemical principles of marine environmental monitoring, taking into account the complex interactions between seawater parameters. For example, the theoretical model of dissolved oxygen is based on the physical relationship between gas solubility and temperature, salinity and pressure; the parameter interaction compensation function takes into account the combined effects of cross-interference between sensors and environmental factors. The use of power relationships and exponential relationships reflects the characteristics of nonlinear changes in many marine parameters, while the matrix structure effectively expresses the complex correlations in multi-parameter systems. Compared with existing technologies, these equations take into account the mutual influence of various parameters more comprehensively, can adapt to the complex and changeable marine environment, and provide more accurate parameter calibration results.

[0078] Specifically, the principle of the present invention is: the technical principle of the seawater parameter calibration system of the present invention is based on the comprehensive application of parameter correlation analysis and deep learning model. First, the system collects the physical, chemical and optical property data of seawater through multiple sensors to establish a multi-dimensional parameter space. In this space, there are complex nonlinear relationships between the parameters, such as temperature affects dissolved oxygen saturation, salinity affects conductivity and pH measurement, and pressure affects sensor response characteristics.

[0079] The core innovation of the system lies in the design of the parameter interaction compensation function, which uses a piecewise nonlinear mapping algorithm to first calculate the first-order correlation coefficient between parameters, construct a second-order cross-term influence matrix, and then solve the optimal compensation coefficient through iterative least squares method. This method can accurately capture the complex interaction between parameters, such as the effect of temperature on the response of pH electrodes and the interference of salinity on dissolved oxygen measurement, thereby achieving multi-parameter coordinated correction.

[0080] The OceanParamNet model is another key component of the system. It is based on a multi-layer temporal attention network architecture, including an encoder-decoder dual-path structure and a residual connection mechanism. The model learns the spatiotemporal variation patterns and intrinsic correlations of seawater parameters by pre-training on a large amount of historical data covering different sea areas and seasons. The four-head multi-scale attention mechanism in the model can dynamically adjust according to the sampling frequency and characteristics of different parameters, effectively extracting the key features of parameter changes.

[0081] The system also designs a periodic self-calibration procedure and an outlier detection algorithm. Through the inherent relationship between the physical and chemical parameters of seawater, constraints are established to fine-tune the calibration model and compensate for the errors caused by sensor aging and drift. Outlier identification based on the moving median combined with the threshold assessment of the rate of change of seawater parameters ensures data quality.

[0082] The organic combination of these technical principles enables the system to achieve high-precision, adaptive calibration of seawater parameters in complex and changeable marine environments, solving the problem of multi-parameter interaction that is difficult to overcome with traditional methods, thereby improving the accuracy and reliability of seawater parameter measurements.

[0083] A specific embodiment 1 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.

[0084] The specific implementation of step S01 is that after the system is started, the self-test program is first executed, and then parameter collection is performed. After the collected data is compared with the stored standard value, the initial calibration parameter matrix is ​​calculated, and its expression is: ; In the formula, is the initial calibration parameter matrix; For the The sensitivity correction factor of each sensor is calculated as ; For the The sensor pairs The cross-interference coefficient of the sensors is calculated as ; For the The standard value of the parameter; For the The measured values ​​of the parameters; is the total number of sensors in the system. The deviation thresholds are set as: pH ± 0.02 pH units, redox potential ± 5 mV, turbidity ± 0.5 NTU, temperature ± 0.01 degrees Celsius, salinity ± 0.02 PSU, dissolved oxygen ± 0.05 mg / L, light absorption coefficient ± 0.01 / m. The cross-interference coefficient is obtained by fitting the measurement results of multiple sets of standard samples using the least squares method. This step ensures that the sensors are in good condition before the system starts working, and establishes an accurate baseline calibration for subsequent measurements.

[0085] The specific implementation of step S02 is the same as above and will not be repeated here.

[0086] The specific implementation of step S03 is to use the temperature, salt and depth detection device to perform water depth profile scanning. The control chip first controls the device to perform depth calibration to determine the water surface reference position; then collects temperature, salinity and pressure data at different depths in sequence according to the preset depth intervals, which are generally set to 0.5 meters from the surface to 10 meters, 1 meter from 10 meters to 100 meters, and 5 meters below 100 meters; then the collected data is time synchronized and depth aligned to eliminate sampling delays and sensor position differences; then the discrete depth point data is processed continuously using the cubic spline interpolation algorithm, and its mathematical expression is: ; In the formula, For Depth The parameter value (temperature or salinity) at For the sampling depth points; For interval The spline coefficients on . The spline coefficients are determined by the following conditions: (function value is continuous); (function value is continuous); (first-order derivative is continuous); (Second-order derivative is continuous); In the formula, For the The measured parameter values ​​of the sampling points; and Represent the first and second order derivatives respectively. Finally, the thermocline and halocline positions are identified by gradient analysis. The judgment conditions are:

[0087] ; In the formula, For Depth The temperature value at is the temperature gradient threshold, set to 0.05 degrees Celsius / meter. The judgment conditions are: ; In the formula, For Depth Salinity value at is the salinity gradient threshold, which is set to 0.01 PSU / m. This step aims to obtain information on the vertical stratification structure of seawater and provide key data for understanding the characteristics of seawater at different depths.

[0088] The specific implementation of step S04 is to collect and calibrate the chemical parameters of seawater through a multi-parameter water quality sensor. The control chip first controls the glass composite electrode and the redox electrode to be cleaned and pre-stabilized before measurement, and the stabilization time is 30 seconds; then the sampling program is started, and the pH and redox potential of the seawater are measured by the electrode, and the light scattering turbidity sensor measures the turbidity value; then the three-point calibration method is used to compensate the original data for temperature and pressure, where the expression of the pH and temperature compensation relationship is: ; In the formula, is the pH value after temperature compensation; is the original pH value measured; is the temperature compensation coefficient, the value is -0.003pH / degree Celsius; is the actual water temperature; is the reference temperature, usually 25 degrees Celsius. Finally, a correlation model between pH and redox potential is established based on the measured data: ; In the formula, is the redox potential (mV); is the standard redox potential, related to the reference electrode; is the gas constant, 8.314 J / (mol·K); is the absolute temperature (K); is the Faraday constant, 96485C / mol; is the pH value; is the organic matter influence coefficient, ranging from 30 to 50 mV; is the attenuation coefficient, ranging from 0.2 to 0.5; is the concentration of dissolved organic carbon (mg / L). The sensor sampling frequency of this step is set to 1 time per minute, the data averaging window is 5 samplings, and the data validity judgment threshold is a continuous change rate of less than ±0.05pH / minute or ±10mV / minute. The purpose of this step is to obtain accurate seawater chemical characteristic parameters and provide basic data for comprehensive assessment of seawater quality.

[0089] The specific implementation of step S05 is to use a dissolved oxygen analysis device to measure the dissolved oxygen parameters of seawater and perform theoretical correction. The control chip first starts the fluorescence quenching oxygen sensor and the electrochemical oxygen sensor, and the preheating time is 60 seconds; then the two sensors are controlled to simultaneously collect dissolved oxygen concentration and oxygen saturation data, and the sampling frequency is once every 30 seconds; then the temperature and salinity data obtained in step S03 are substituted into the dissolved oxygen saturation calculation formula: ; In the formula, is the saturated dissolved oxygen concentration (mg / L); is the absolute temperature (K); is salinity (PSU); to and to are empirical coefficients, and their values ​​are: , , , , , , . Then consider the effect of pressure on dissolved oxygen: ; In the formula, is the dissolved oxygen concentration after pressure compensation; is the measured original dissolved oxygen concentration; is the pressure influence coefficient, and its value is 0.00004 / dB; is the water pressure (dB). Finally, compare the measured dissolved oxygen value with the theoretical calculated value and calculate the dissolved oxygen correction coefficient matrix: ; In the formula, are the first-order correction coefficients for temperature, salinity and pressure, respectively; are the second-order correction coefficients for temperature, salinity, and pressure, respectively. Each coefficient is obtained by comparing the theoretical dissolved oxygen value with the measured value using a multiple regression method. In the dissolved oxygen theoretical model, the temperature influence coefficient is set to -0.28 mg / L·°C, the salinity influence coefficient is set to -0.05 mg / L·PSU, and the pressure influence coefficient is set to 0.01 mg / L·dB. This step ensures the accuracy of dissolved oxygen data under different temperature and salinity conditions, providing reliable data for marine biogeochemical research.

[0090] The specific implementation of step S06 is to collect and analyze the optical characteristic parameters of seawater through a spectral analysis device. The control chip first controls the micro-spectrometer to perform baseline calibration and record the pure water background spectrum; then, in the wavelength range of 400 to 700 nanometers, the seawater sample is scanned in full spectrum with a step length of 2 nanometers to obtain the light absorption coefficient spectrum; then the multi-wavelength fluorescence sensor is started, and the corresponding fluorescence emission intensity is measured at 435 nanometers, 470 nanometers and 530 nanometers respectively; then the collected spectral data is processed by principal component analysis: ; In the formula, is the original spectral data matrix, where rows represent samples and columns represent wavelengths; is the left singular vector matrix; is a diagonal matrix of singular values; is the transpose of the right singular vector matrix. The characteristic wavelength selection condition is: ; In the formula, is the selected characteristic wavelength set; For the wavelength; is the right singular vector matrix Line Column elements; Select the threshold value for the characteristic wavelength and set it to 0.2; For the number of principal components selected, the first few principal components with a contribution rate of 90% are usually selected. Finally, the chlorophyll concentration is calculated by establishing a spectral feature and concentration relationship model: ; In the formula, is the chlorophyll a concentration (μg / L); , and The absorbances at 438 nm, 678 nm, and 750 nm, respectively; to is the calibration coefficient, which is obtained by regression of standard samples. The purpose of this step is to non-destructively detect biological and chemical components in seawater by optical methods, so as to provide a scientific basis for the assessment of marine primary productivity and water quality.

[0091] The specific implementation method of step S07 is to construct a seawater comprehensive parameter calibration model using a parameter interaction compensation function. The control chip first obtains the temperature parameter matrix, salinity parameter matrix, depth parameter matrix, optical characteristic parameter matrix and chemical characteristic parameter matrix from each sensor device S03 to S06; then the parameter interaction compensation function is applied to process each matrix. The function first calculates the Pearson correlation coefficient between each parameter and constructs a parameter correlation matrix: ; In the formula, For the Parameters and The Pearson correlation coefficient of the parameters is calculated as follows: ; In the formula, For the In the sample The value of the parameter; For the The average value of the parameters; is the number of samples. Then, based on the correlation matrix, strongly correlated parameter pairs are identified, and the correlation coefficient threshold is set to ±0.7; then the second-order cross-term influence matrix is ​​constructed: ; In the formula, Representation parameters and parameters Finally, the optimal compensation coefficient is solved by iterative least squares method to build a comprehensive calibration model of seawater parameters: ; In the formula, For parameters Correction value of For parameters The measured value of For parameters Reference value of is the first-order correction coefficient; is the second-order cross correction coefficient. The function uses a piecewise nonlinear mapping algorithm to deal with the nonlinear relationship between parameters. A cubic polynomial fitting is used for the temperature-salinity related region, and a logarithmic linear fitting is used for the optical-chemical parameter related region. The main purpose of the parameter interaction compensation function is to accurately characterize the complex correlation between seawater parameters and improve the accuracy and adaptability of the comprehensive calibration model.

[0092] The specific implementation method of step S08 is to correct the original data based on the OceanParamNet model and the calibration model. The control chip first loads the pre-trained OceanParamNet model parameters and initializes the model calculation environment; then the newly collected original data is pre-processed and aligned according to the sensor type and sampling time; then the calibration model constructed in S07 is used to calculate the basic correction parameters, including temperature compensation coefficient, salinity compensation coefficient, pressure compensation coefficient and mutual interference correction coefficient; then the original data and basic correction parameters are used as the input of the OceanParamNet model, and the optimized correction parameters are calculated through the forward propagation of the model; finally, the optimized correction parameters are applied to calibrate the original data, and the expression is:

[0093] ; In the formula, For parameters The final revised value of For parameters The original measurement value of is the temperature compensation coefficient; is the salinity compensation coefficient; is the pressure compensation coefficient; is the mutual interference correction coefficient. The calculation method of each compensation coefficient is: ; ; ; ; In the formula, , , are the actual temperature, salinity and pressure values ​​respectively; , , are reference temperature, reference salinity and reference pressure respectively; For parameters The actual value of For parameters Reference value of , , For parameters Temperature, salinity and pressure compensation coefficients; For parameters Parameters The interference coefficient is 0.04. The temperature compensation range is -2 to 40 degrees Celsius, the salinity compensation range is 0 to 42 PSU, and the pressure compensation range is 0 to 2000 decibels. The model reasoning process uses floating point 16-bit precision calculation to balance calculation accuracy and efficiency. This step significantly improves the accuracy of seawater parameter measurement by combining traditional physical models and deep learning methods, especially in areas with rapid environmental changes.

[0094] The specific implementation of step S09 is to use the OceanParamNet model to detect outliers on the calibrated data. The control chip first divides the calibrated parameter data into time series segments, with the segment length set to 10 minutes; then calculates statistical features for each segment of data, including the median, interquartile range, and moving average; then applies the outlier identification algorithm based on the moving median: ; ; ; In the formula, For the The deviation of a data point from the moving median; For the The parameter value of each data point; is the median window width, usually set to 5 to 10; is the interquartile range; and are the first and third quartiles, respectively; is an abnormal mark, 1 indicates abnormality and 0 indicates normality. Then the potential abnormal points are input into the anomaly detection module of the OceanParamNet model, and their rationality is evaluated in combination with the historical trend of parameters and physical constraints; finally, the validity of the data points is determined based on the abnormal probability output by the model and the threshold of the change rate of seawater parameters. The abnormal probability threshold is set to 0.85, and the parameter change rate threshold is: temperature 0.1 degrees Celsius / minute, salinity 0.05PSU / minute, pH 0.03pH unit / minute, dissolved oxygen 0.2mg / L / minute. This step ensures the reliability of the system output data and avoids the impact of abnormal values ​​on subsequent analysis and application.

[0095] The specific implementation of steps S10-S12 is the same as above and will not be described in detail here.

[0096] In order to better understand and implement the present invention, the following provides Example 2 of a specific application scenario of the present invention: When building an ocean observation network in a deep-sea area of ​​a certain sea area, researchers need to calibrate seawater parameters with high precision to obtain accurate and reliable marine environmental data. This example describes in detail a practical application scheme of a seawater parameter calibration system in a deep-sea area of ​​a certain sea area.

[0097] The seawater parameter calibration system is mainly composed of two parts: hardware system and software control system. The hardware system includes control chips, various sensors and auxiliary equipment; the software control system is built into the control chip and is responsible for system operation control, data processing and calibration.

[0098] The hardware system configuration is as follows: The control chip uses an industrial-grade ARM Cortex-M7 processor with a main frequency of 240MHz, 512KB SRAM, 2MB flash memory, and a floating-point unit to support real-time data processing. The multi-parameter water quality sensing device uses composite electrode technology, integrated pH, ORP and turbidity sensors, with measurement ranges of 0-14pH, ±2000mV, 0-1000NTU, and accuracies of ±0.01pH, ±1mV, and ±0.5NTU, respectively. The temperature, salt and depth detection device uses a high-precision platinum resistance thermometer (accuracy ±0.002℃), a conductivity sensor (accuracy ±0.003mS / cm) and a pressure sensor (accuracy ±0.01%FS), which can obtain accurate temperature, salinity and depth data. The dissolved oxygen analysis device uses fluorescence quenching technology, with a measurement range of 0-20mg / L, an accuracy of ±0.1mg / L, and a response time of <5 seconds. The spectral analysis device uses a 7-channel spectral sensor, covering the 400-700nm wavelength range and is equipped with an anti-biofouling optical window.

[0099] The data storage device uses 128GB industrial-grade flash memory, which supports cyclic storage and partition management. The wireless communication device integrates an underwater acoustic communication module (effective distance 1000m) and a satellite communication module (used when floating). The power supply device uses a 12V / 40Ah lithium battery pack, which supports continuous operation for 30 days at a depth of 1000m. The positioning device integrates a pressure sensor, an underwater acoustic positioning receiver and a GPS module (used when floating). The clock synchronization device uses a temperature-compensated crystal oscillator with a drift rate of <1ppm, and realizes time synchronization between multiple devices through underwater acoustic signals.

[0100] Before the system was deployed, the researchers calibrated each sensor with a standard solution in a laboratory environment. As shown in Table 1: Table 1 Initial calibration data of sensors in seawater parameter calibration system

[0101] The system is deployed in the northern part of a sea area (21°45'N, 118°12'E), at a water depth of about 1200 meters. After the system is started, the calibration process is automatically performed according to the preset steps in the software control system: When the system starts the self-check phase, the system imports the calibration coefficients in Table 1 above into the initial calibration parameter matrix. By analyzing the test results of the standard solution, the cross-interference coefficients between the sensors are calculated to form a complete initial calibration parameter matrix as shown in Table 2: Table 2 Sensor initial cross-interference coefficient matrix

[0102] Based on the geographic location information, the system retrieves the summer seawater parameter model of the northern part of a certain sea area from the data storage device, and initializes the initial parameters of the seawater parameter prediction model of the northern part of a certain sea area by considering the tidal conditions during actual deployment (high tide period, tidal range of 1.2 meters).

[0103] The system uses the temperature, salinity and depth detection device to perform profile scanning along the water depth direction, with a sampling interval of 10 meters, and collects temperature, salinity and depth data from the surface to a depth of 1200 meters. The temperature, salinity and depth three-dimensional structure model is constructed using the cubic spline interpolation algorithm. The profile scanning results are shown in Table 3: Table 3 Temperature, salinity and depth profile data of the northern part of a certain sea area (partial)

[0104] Based on the data in Table 3, the system identified that the thermocline is located between 105 and 175 meters in depth, with a maximum temperature gradient of 0.14°C / m; the halocline is located between 80 and 140 meters in depth, with a maximum salinity gradient of 0.012 PSU / m.

[0105] The pH, redox potential and turbidity data of seawater were collected at different depths by using a multi-parameter water quality sensor. The three-point calibration method was used to compensate for the effects of temperature and pressure, and a correlation model between pH and redox potential was established. The compensated data is shown in Table 4: Table 4 Water quality parameter data at different depths (partial)

[0106] The system collects dissolved oxygen concentration and oxygen saturation data of seawater through a dissolved oxygen analysis device. Combined with the temperature and salinity data obtained previously, the theoretical dissolved oxygen value is calculated according to the dissolved oxygen theoretical model, and the dissolved oxygen correction coefficient is obtained by comparing it with the measured value. The dissolved oxygen data at each depth are shown in Table 5: Table 5 Dissolved oxygen data at different depths

[0107] The system collects data on seawater light absorption coefficient, fluorescence intensity and pigment concentration in the wavelength range of 400 to 700 nanometers through a spectral analysis device. The absorption peaks at characteristic wavelengths of 438nm, 678nm and 750nm are extracted using the principal component analysis method to identify the chlorophyll concentration and colored soluble organic matter content at different depths, as shown in Table 6: Table 6 Optical properties data at different depths

[0108] The system uses the parameter interaction compensation function to input all the above parameters into the multivariate regression model and calculate the interaction between the parameters. By considering the interaction between temperature, salinity, depth, and optical and chemical properties, a complete seawater comprehensive parameter calibration model is constructed. The interaction compensation coefficients of the main parameters are shown in Table 7: Table 7 Interaction compensation coefficients of main parameters

[0109] Based on the pre-trained OceanParamNet model and the calibration model constructed above, the system corrects the newly collected raw data, including temperature compensation, salinity compensation, pressure compensation and mutual interference correction, to generate the final high-precision seawater parameter data. The comparison of data before and after calibration is shown in Table 8: Table 8 Comparison of seawater parameters before and after calibration (depth 500m)

[0110] Outlier detection was performed on the calibrated data, and outlier identification based on the moving median was realized through the OceanParamNet model. The system set the outlier judgment threshold to 3 times the interquartile range, identified a small number of data anomalies, and combined the seawater parameter change rate threshold to evaluate the data validity. In a complete profile measurement, the system identified a total of 15 abnormal data points, with an abnormal rate of 2.3%, which was lower than the preset 5% threshold.

[0111] The data storage device records the original data, intermediate calculation results and calibrated data, and adopts a hierarchical storage structure to ensure data integrity. The system transmits data to the host computer system through a wireless communication device every 2 hours, with a transmission success rate of 98.5%.

[0112] According to the battery power status feedback from the power supply device, the system dynamically adjusts the sampling frequency of each sensor device. When the battery power is less than 70%, the system automatically adjusts the sampling interval of the temperature, salt and depth detection device from 10 seconds to 30 seconds, and the sampling frequency of the spectrum analysis device from once per minute to once every 3 minutes, while reducing the wireless transmission power to achieve long-term stable operation of the system.

[0113] The system automatically performs a self-calibration procedure every 24 hours, using the inherent relationship between the physical and chemical parameters of seawater and the correlation constraints to fine-tune the calibration model parameters and compensate for the measurement errors caused by sensor aging and drift. Through 30 consecutive days of observation, the system obtained high-quality seawater parameter data at all depths, with the drift rate controlled within 0.5%, meeting the needs of scientific research.

[0114] Traditional seawater parameter measurement methods usually use a single sensor for independent measurement, lacking a compensation mechanism for the mutual influence between parameters. The measurement accuracy is easily affected by environmental factors such as temperature, pressure, and biofouling, and sensor drift is difficult to correct in real time. This leads to unstable quality of long-term observation data, especially in deep-sea areas with complex environmental conditions. In traditional methods, calibration usually requires manual periodic recovery of equipment for laboratory calibration, or on-site calibration using standard solutions, which is complicated to operate and has a long calibration cycle.

[0115] The seawater parameter calibration system in this embodiment integrates data from multiple sensors and establishes a model of interaction between parameters, thereby achieving high-precision automatic calibration of seawater parameters. The system takes into account the influence of environmental factors such as temperature, salinity, and pressure on the measurement of each parameter, and effectively eliminates measurement errors through parameter interaction compensation functions. At the same time, automatic outlier detection is achieved through the OceanParamNet model, which improves data quality. The system's self-calibration function can compensate for sensor drift and achieve long-term stable observation. Compared with traditional methods, this system improves measurement accuracy by about 25-40%, reduces maintenance costs by about 50%, and extends the calibration cycle by 3-5 times. It is particularly suitable for the construction of deep-sea long-term observation networks and provides reliable data support for marine scientific research.

[0116] It should be noted that the variables involved in the present invention are explained in detail as shown in Tables 9 and 10 below.

[0117] Table 9 Variable explanation table (Part I)

[0118] Table 10 Variable explanation table (Part II)

[0119] 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 seawater parameter calibration system, characterized in that: A control chip is provided, and the control chip is electrically connected to a multi-parameter water quality sensor device, a temperature-salinity-depth detection device, a dissolved oxygen analysis device, a spectrum analysis device, a data storage device, a wireless communication device, a power supply device, a positioning device, and a clock synchronization device. A system control module is provided in the control chip, and the system control module performs the following steps: using the temperature-salinity-depth detection device to perform a profile scan along the water depth direction; using the multi-parameter water quality sensor device to collect seawater pH, redox potential, and turbidity data; using the dissolved oxygen analysis device to collect dissolved oxygen concentration and oxygen saturation data in seawater; and using the spectrum analysis device to collect seawater light absorption coefficient; All parameters are input into the multivariate regression model using the parameter interaction compensation function to construct a seawater comprehensive parameter calibration model; the newly collected raw data are corrected based on the OceanParamNet model and the calibration model; outlier detection is performed on the calibrated data; the sampling frequency is adjusted according to the battery power status fed back by the power supply device; and the self-calibration procedure is executed periodically.

2. The seawater parameter calibration system according to claim 1, characterized in that: The multi-parameter water quality sensor device is used to collect seawater pH, redox potential and turbidity data. The temperature, salt and depth detection device is used to collect seawater temperature, salinity and depth data. The dissolved oxygen analysis device is used to collect dissolved oxygen concentration and oxygen saturation data in seawater. The spectral analysis device is used to collect seawater light absorption coefficient, fluorescence intensity and pigment concentration data. The data storage device is used to store all collected data, system operating parameters and calibration models. The wireless communication device is used for data transmission between systems and communication with the host computer. The power supply device is used to provide the required power for the system and monitor the battery status. The positioning device is used to determine the system deployment location and depth profile information. The clock synchronization device is used to ensure the time consistency of multi-point measurement data.

3. The seawater parameter calibration system according to claim 2, characterized in that: The sampling frequency of water quality parameters by the multi-parameter water quality sensing device is once per minute, the sampling frequency of the temperature, salinity and depth detection device is once every 10 seconds, the sampling frequency of the dissolved oxygen analysis device is once every 30 seconds, and the sampling frequency of the spectral analysis device is once per minute.

4. The seawater parameter calibration system according to claim 3, characterized in that: A temperature-salinity-depth detection device is used to perform profile scanning along the water depth direction. Specifically, seawater temperature, salinity and pressure data at different depths are collected to construct a three-dimensional structure model of temperature-salinity-depth and identify the locations of thermocline and halocline.

5. The seawater parameter calibration system according to claim 4, characterized in that: The seawater pH, redox potential and turbidity data were collected through a multi-parameter water quality sensor device. The three-point calibration method was used to compensate for the effects of temperature and pressure on the measurement, and a correlation model between pH and redox potential was established.

6. The seawater parameter calibration system according to claim 5, characterized in that: The light absorption coefficient, fluorescence intensity and pigment concentration data of seawater in the wavelength range of 400 to 700 nanometers were collected through a spectral analysis device. The absorption peak at the characteristic wavelength was extracted using the principal component analysis method to identify the chlorophyll concentration and the content of colored soluble organic matter.

7. The seawater parameter calibration system according to claim 6, characterized in that: The input parameters of the parameter interaction compensation function include temperature parameter matrix, salinity parameter matrix, depth parameter matrix, optical characteristic parameter matrix and chemical characteristic parameter matrix, and the output result of the parameter interaction compensation function is seawater parameter correction coefficient matrix.

8. The seawater parameter calibration system according to claim 7, characterized in that: The parameter interaction compensation function adopts a piecewise nonlinear mapping algorithm. First, the first-order correlation coefficients between the parameters are calculated, then the second-order cross-term influence matrix is ​​constructed, and finally the optimal compensation coefficient is solved by the iterative least squares method.

9. The seawater parameter calibration system according to claim 8, characterized in that: The specific structure of the OceanParamNet model is a multi-layer temporal attention network architecture, which includes an encoder-decoder dual-path structure and a residual connection mechanism. The encoder consists of a three-layer bidirectional long short-term memory network, each layer contains 128 neurons, which is used to capture the time series characteristics of seawater parameters. The decoder uses a four-layer fully connected neural network with 256, 128, 64 and 32 neurons respectively, which is used to map the feature space to the parameter correction space.

10. The seawater parameter calibration system according to claim 9, characterized in that: The steps for establishing the training dataset during the pre-training process of the OceanParamNet model include: first, collecting at least 12 months of continuous monitoring data on seawater parameters covering four seasons; then classifying the data according to geographical location and marine environmental characteristics; then quality control of the raw data; then data enhancement; and finally constructing labeled data.

Images