In-situ detection and self-calibration method and device for water quality-ecological index of river and lake multi-source complementation

Abstract

The present application relates to river and lake water environment in-situ monitoring and analysis detection technical field, especially to a kind of river and lake multi-source complementary water quality-ecological index in-situ detection and self-calibration method and device, by switching between water sample, blank and standard liquid and standard addition to the same water sample, generate at least two standard addition gradient points of standard addition increment concentration and online update zero point and sensitivity;Convert optical / electrochemical and other multi-mechanism channels into index values, based on uncertainty, drift / mutation characterization and physical constraints, consistency fusion is carried out based on adjacent station consistency, and the data packet containing fusion value, reliability and status code is output.The present application can significantly reduce the influence of long-term drift, biological fouling and matrix change on measurement results, improve the stability and consistency of in-situ continuous monitoring data, so it is suitable for long-term unattended deployment, improve the availability and traceability of continuous monitoring data.

Description

Technical Field

[0001] This invention relates to the field of in-situ monitoring and analysis of river and lake water environment, and in particular to a method and device for in-situ detection and self-calibration of water quality-ecological indicators with multi-source complementarity in rivers and lakes, which is mainly used for online analysis, data fusion and reliability output of on-site water samples. Background Technology

[0002] River and lake water environment management places higher demands on continuous and refined understanding of water quality and ecological status. Compared to traditional manual sampling and laboratory analysis, in-situ / online monitoring can acquire dynamic changes in key indicators such as pH, dissolved oxygen, conductivity, turbidity, and chlorophyll at a higher temporal resolution, providing a data foundation for water process identification, pollution event tracking, and ecological risk assessment. However, the availability of continuous monitoring data is highly dependent on the stability and traceable calibration of sensors during long-term deployment. Monitoring equipment requires supporting cleaning, calibration, and data quality control processes; otherwise, high-frequency data may not accurately reflect changes in water bodies.

[0003] In practical engineering scenarios, one of the core challenges faced by in-situ sensors is biofouling and long-term drift. Biofilms, algae, and particulate matter in water can adhere to the surfaces of electrodes or optical windows, leading to signal attenuation, zero-point drift, and changes in sensitivity. Even with mechanical cleaning or antifouling measures, periodic maintenance and calibration are still often required. Multiple engineering guidelines and reviews have pointed out that biofouling is one of the important factors limiting long-term unattended continuous monitoring of water quality instruments, and data reliability must be ensured through maintenance / calibration and quality control mechanisms.

[0004] To address the aforementioned issues, existing technologies primarily develop along three paths: The first is the modular integration and online operation of multi-parameter probes / analyzers. For example, Chinese utility model publication CN206002529U discloses an online multi-parameter water quality analyzer that uses a secondary instrument and a multi-sensor interface board to connect multiple types of sensors, storing calibration and related parameters on the corresponding interface board to reduce recalibration and configuration costs. Another example is Chinese utility model publication CN206248652U, which discloses a real-time in-situ water quality monitoring device. While this approach improves the engineering deployability of multi-parameter monitoring, it typically still relies on sensor readings as the core output. Deviations caused by drift / contamination depend more on periodic manual calibration or post-accident correction, making it difficult to establish a traceable in-situ quantitative calibration chain at the device level. The reliability of the data often requires secondary judgment on the platform side.

[0005] Secondly, there are systematic solutions oriented towards quality control / verification. For example, Chinese invention publication CN111596022A proposes a remote quality control method and system for water quality probe-type sensors. This type of solution strengthens the closed loop of supervision and quality control, but its core is still verification / judgment. It relies heavily on external standard sample verification processes and platform-side rules, and may not have the ability to perform in-situ quantitative spiked calibration to form multi-point calibration and update the calibration parameters of each channel online. Furthermore, it has limited support for the mutual verification and fusion of multi-source heterogeneous sensors on the same indicator and the reliable quantitative output.

[0006] Thirdly, microfluidics and standard solutions are introduced into the sensor to attempt self-calibration. For example, Chinese invention patent CN106290515B discloses a self-calibrating marine multi-parameter chemical sensor with a microfluidic replaceable cavity structure. This scheme has significant reference value for probe-level self-calibration of "microfluidics + standard solution," but it still has limitations in the context of applications with high disturbance, strong pollution, and multi-index coupling in rivers and lakes: On the one hand, its self-calibration is more reflected in standard solution switching and cavity structure design, lacking a multi-mechanism mutual verification and fusion mechanism for the same index, addressing the matrix effect, cross-sensitivity, and multi-source inconsistency problems commonly found in river and lake scenarios; on the other hand, existing schemes usually output calibrated readings, lacking a mechanism to integrate drift rate, mutation rate, physical constraint consistency, and neighbor station consistency into a unified credibility and output it along with the data, thus making it difficult to directly provide a data foundation of "usable data + credibility" for the upper-level platform.

[0007] In addition, some patents construct flow path switching and closed-loop circulation structures from the perspective of in-situ analysis using reagent methods. For example, Chinese invention with publication number CN119688583B discloses an in-situ water quality analyzer and method. This type of scheme is suitable for reagent reaction analysis of specific parameters, but its system focuses on single-process analysis of "reagent-reaction-optical detection". Complementary measurement of multi-source sensor arrays, standard addition self-calibration, and unified reliability output are not its focus.

[0008] In summary, while existing technologies have made progress in areas such as multi-parameter integration, cleaning and maintenance, remote quality control, and microfluidic self-calibration, they still struggle to simultaneously meet the requirements of multi-source complementary measurement (mutual verification of the same indicator through multiple mechanisms) in long-term in-situ monitoring of rivers and lakes, microfluidic blank / standard switching and standard-added quantitative self-calibration, and quantifying data consistency and drift status into credibility and outputting it along with the results. Therefore, there is an urgent need for an in-situ detection and self-calibration technology for water quality-ecological indicators in river and lake scenarios to improve the usability and interpretability of monitoring data under long-term deployment conditions. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a method and apparatus for in-situ detection and self-calibration of water quality-ecological indicators from multiple sources in rivers and lakes. It addresses common problems in in-situ continuous monitoring of rivers and lakes, such as data distortion caused by long-term sensor drift and biofouling, inconsistent and difficult-to-verify measurement results of the same ecological indicator from multiple heterogeneous sensors, and the reliance of existing devices on manual calibration or post-construction platform rejection, lack of traceable in-situ quantitative self-calibration and data quality quantification output. This invention aims to solve the following key technical problems: How can we achieve in-situ, quantitative, and traceable self-calibration under unattended, long-term deployment conditions by using microfluidic switching between blank and standard solutions and standard addition (spiking gradient) to suppress the impact of zero-point drift and sensitivity decay on measurement results? Simultaneously, for detection channels that are not suitable for defining zero-point response with blank solutions or for standard addition-based quantitative calibration, how can we obtain the corresponding zero-point baseline, sensitivity verification value, drift status, or noise parameters through standard solutions, calibration solutions, or stable state verification? How to fuse measurement information of the same ecological indicator from multiple mechanisms such as optical and electrochemical methods with consistency constraints, so that stable and reliable ecological indicator values ​​can still be output when the sensor state changes; and how to identify abnormal states such as single channel drift, contamination, bubble interference or channel inconsistency by cross-consistency test between different detection mechanism channels of the same ecological indicator. How to achieve real-time quantification and classification of data quality at the device end, integrate information such as drift, mutation, physical and mechanistic constraint consistency, channel consistency and neighbor station consistency into a credibility parameter, and output it synchronously with the fusion result, so as to provide the upper platform with a data foundation of "usable data + credibility".

[0010] To address the aforementioned technical problems, this invention provides an in-situ detection and self-calibration method and apparatus for multi-source complementary water quality-ecological indicators in rivers and lakes. This method is executed in a long-term deployable in-situ flow-through detection structure, and an integrated microfluidic self-calibration module and data processing model collaboratively realize a detection and analysis process of "multi-mechanism mutual verification measurement—in-situ quantitative calibration—consistency fusion—reliable quantitative output." The method includes the following steps: S1, in-situ sampling and multi-channel detection.

[0011] The present invention introduces a water sample to be tested into an in-situ flow-through detection chamber and arranges a multi-source sensor array in or connected to the in-situ flow-through detection chamber to simultaneously acquire at least two types of raw output signals with different mechanisms.

[0012] Furthermore, the in-situ flow-through detection cavity is a flow-through cavity with a constant optical path or constant volume, and is equipped with a flow stabilizing component, a bubble trapping component, and an anti-reflection inner wall to reduce the interference of bubbles, eddies, and scattering on optical and electrochemical measurements.

[0013] Furthermore, the multi-source sensor array includes at least an optical detection channel and an electrochemical detection channel, wherein the optical detection channel includes one or more of a UV / Vis absorption channel, a fluorescence channel, and a turbidity scattering channel, and the electrochemical detection channel includes one or more of pH, ORP, dissolved oxygen, and conductivity.

[0014] Furthermore, for ecological indicators of the same target, at least two detection channels with different mechanisms are set up for parallel measurement to form mutual verification redundancy. For example, dissolved oxygen is measured in parallel using optical DO and electrochemical DO, algae-related ecological indicators are measured in parallel using fluorescence and spectral absorption, and turbidity is measured in parallel using scattering and optical background transmittance.

[0015] Furthermore, the detection channels can be categorized according to their calibration methods into detection channels with standard additive quantitative detection conditions and state-type detection channels unsuitable for standard additive quantitative calibration conditions. The detection channel with standard additive quantitative detection is a concentration-type detection channel capable of generating a spiked concentration increment through a standard solution and establishing a spiked concentration-detection signal relationship. The state-type detection channel includes one or more of pH, ORP, dissolved oxygen, and conductivity, and its corresponding zero-point baseline, drift state, sensitivity verification value, or noise parameters can be obtained through standard solutions, calibration solutions, or stable state verification methods.

[0016] S2, Blank liquid zero-point calibration.

[0017] This invention uses a microfluidic self-calibration module to switch blank liquid into the in-situ flow-through detection chamber, performing zero-point calibration to obtain the zero-point baseline of each detection channel. With zero-point noise Among them, the zero-point baseline Used for zero-point calibration of the detection signals of each detection channel.

[0018] The blank solution is deionized water or a blank matrix solution prepared with the same ionic strength.

[0019] Furthermore, the zero-point calibration includes at least the zero-point baseline for each detection channel. With zero-point noise The optimal estimation method is the sliding window mean and robust variance estimation, i.e.: ; ; in, For the first The detection channel is in a blank liquid state at The raw output at time step; MAD represents the median absolute deviation; This indicates taking the median.

[0020] Furthermore, for detection channels that are not suitable for defining the zero-point response using blank solution, the corresponding zero-point baseline, sensitivity verification value, drift state, or noise parameters are obtained by using standard solution, calibration solution, or stable state verification.

[0021] Furthermore, the zero-point calibration can be linked with the cleaning action, which is microfluidic rinsing and / or mechanical brushing and / or ultrasonic micro-vibration to restore the optical window and electrode surface condition; the cleaning is only a measurement assurance process and still belongs to the maintenance link of the detection and analysis system.

[0022] S3, standard addition to generate scalar gradient points.

[0023] This invention involves re-entering the water sample into the in-situ flow-through detection chamber, and then performing standard-addition quantitative calibration on the same water sample using the microfluidic self-calibration module. This generates spiking increment concentrations corresponding to one unspecified reference point and at least two spiking gradient points. And obtain the detection channel with standard additive quantitative detection conditions corresponding to the spiked incremental concentration. The detection signal.

[0024] The detection signal is used in a subsequent regression step to update the calibration model parameters of the detection channel with standard additive quantitative detection conditions online. These calibration model parameters include channel sensitivity. and the current water sample without added labels response item Among them, channel sensitivity For the calibration model parameters that can be inherited and updated, the current water sample has an unspecified response term. This is only used for the conversion of current water sample concentration or index value; the current water sample is not spiked in the response item. In linear regression, this is represented by the intercept.

[0025] The microfluidic self-calibration module includes at least a blank solution storage unit, a standard solution storage unit, a quantitative transport unit, a microvalve assembly, and a quantitative mixing network. The quantitative transport unit is a micro-peristaltic pump, a piezoelectric micropump, or an injection micropump. The microvalve assembly is used to switch between water samples, blank solutions, and standard solutions. The quantitative mixing network is a split-flow-combination or series-parallel quantitative mixing structure used to generate mixtures with different volume ratios.

[0026] Preferably, the standard addition-based quantitative calibration employs two-point or multi-point spiking to add standard solution at least twice with different quantitative volumes to the same water sample. This includes: mixing the water sample, standard solution, and blank solution or diluent in a controllable volume ratio using a quantitative mixing network to generate an unspiked reference point and at least two mixtures with different spiking ratios. The mixtures are then sequentially introduced into the in-situ flow-through detection chamber as the incremental spiking concentration at the spiking gradient points. ; The spiking increment concentration The concentration increment is calculated based on the standard solution concentration, the volume of standard solution added, the volume of blank solution or diluent added, and the final volume of the mixture, and the spiking increment concentration is... Volume dilution correction for quantitative mixing process has been included.

[0027] Preferably, different spiking increments are achieved by maintaining a consistent water sample volume fraction at each regression point and varying the replacement ratio of standard solution to blank solution or diluent. .

[0028] Furthermore, in scenarios involving rapid sensitivity verification or low-frequency calibration verification, a two-point verification can be formed by using an unspecified point and a quantitatively specified point. This two-point verification is not used to replace the complete standard addition calibration test, but is used to estimate the sensitivity verification value and determine whether a complete calibration needs to be triggered.

[0029] When using two-point spiking, obtain the corrected detection signal from the unspiked water sample. The detection signal after single quantitative spiking correction ,in, , The first The original output signals collected by each detection channel under conditions of unspecified water sample and spiked mixture;

[0030] Furthermore, under the condition that the channel response is approximately linear and can be considered to pass through the origin after zero-point calibration, the channel sensitivity... It can be represented as: ; in, The zero-point baseline.

[0031] When using multi-point scaling, generate The spiking increment concentration at each spiking gradient point ( For each scalated gradient point, the first gradient is collected. The raw output signal of each detection channel and based on the zero-point baseline The detection signal of the corresponding channel response after zero-point calibration The expression is: ; in, After zero-point calibration, the first The detection channel corresponds to the first spiking increment concentration The detection signal.

[0032] S4, Based on detection signal Regression fitting is performed to obtain the channel sensitivity used for online updating of the calibration model parameters for the corresponding detection channel. And the unspecified response item of the current water sample used for converting the current water sample index values. And converted into channel estimates of the same ecological indicator. Simultaneously calculate the channel estimate. The corresponding drift consistency penalty and mutation penalty.

[0033] Based on detection signal Constructing regression data pairs The least squares or robust regression model is used to fit the linear model to obtain the channel sensitivity used for online updating of the calibration model parameters of the corresponding detection channel. And the unspecified response item of the current water sample used for converting the current water sample index values. .

[0034] Spiking increment concentration From the concentration of standard solution With quantitative mixing volume ratio Confirmed, the expression is: ; The expression for the least squares or robust regression fitted linear model is: ; The calibration model parameters can be calculated using the following formula: ; ; in, For the first The detection channel corresponds to the first spiking increment concentration The mean of the calibrated detection signal; This represents the average concentration of the spiked increment; This represents the zero-point correction response term for the current water sample under unspecified conditions, which is expressed as the intercept in linear regression. It is only used for the conversion of current water sample concentration or index value, and is not used as a long-term inherited channel calibration model parameter.

[0035] The first result is obtained by conversion based on the calibration model parameters. Channel estimates for each detection channel The expression is: ; The calibration model parameters that are inherited and updated online include at least the zero baseline, channel sensitivity, temperature compensation parameters, and / or drift state parameters.

[0036] Furthermore, to mitigate the systemic impact of temperature on the electrochemical and optical detection channels, a temperature compensation term can be introduced into the model calibration parameters. For example, a linear or exponential temperature-compensated model can be used for sensitivity, expressed as: ; in, For the first Each detection channel is at temperature Channel sensitivity at lower levels; Reference temperature Channel sensitivity at lower levels; To detect the real-time temperature of the water sample inside the cavity; For the first Temperature coefficient of each detection channel.

[0037] For dissolved oxygen indices, it is preferable to introduce saturated dissolved oxygen based on temperature and air pressure. As a physical consistency constraint, it is converted into a constraint penalty in step S5 for credibility calculation.

[0038] Furthermore, the drift consistency penalty amount It can be characterized by a combination of the zero-point baseline change rate and the calibration parameter change rate, with the preferred form being: ; in, , This is the drift penalty weighting coefficient, used to adjust the drift consistency penalty between the zero-baseline drift term and the channel sensitivity drift term. The relative contribution in, and preferably satisfying , ; For time variables; zero-point baseline drift term With channel sensitivity drift term Estimation can be performed using the difference or linear fit slope of the discrete-time series within a sliding window, for example: ; ; in, For the first The timestamp is updated for the next zero-point calibration or verification.

[0039] Mutation penalty It can be obtained by short-time difference or robust change point detection, preferably by first-order difference with sliding window and threshold discrimination or robust residual discrimination based on Huber loss.

[0040] S5. Multiple channel estimates for the same ecological indicator The fusion value is obtained by performing consistency constraint fusion. And calculate the confidence level used for current data quality assessment and self-calibration trigger determination. At the same time, a status code is generated.

[0041] The conditions for triggering the self-calibration include: triggering according to a preset cycle and / or the reliability of the output of any ecological indicator in the most recent detection or verification cycle. Below the threshold The sensitivity change rate obtained by checking the calibration solution. Exceeding the threshold Or, inter-channel consistency penalty Exceeding the threshold This may be triggered by the expiration of the calibration validity period; under non-triggered conditions, maintain low-frequency zero-point calibration, blank flushing, or calibration solution verification. Among these, , These are the sensitivity check values ​​and their changes.

[0042] The consistency constraint fusion adopts a combination of weighted fusion and constraint penalties, namely: for the channel estimates of the same ecological indicator Set physical boundary constraints and cross constraints, and base them on the channel uncertainty of each channel. With channel-level credibility Determine the fusion weights The fusion weights Recommended: ; in, Channel-level reliability (obtained from drift, mutation, calibration residuals, etc.); For the first One detection channel, channel uncertainty The value is determined by zero-point noise, regression residuals, and state terms, and is preferably: ; in, Zero-point noise is obtained from robust variance estimation of the blank / steady-state sequence; The regression residuals are from the most recent calibrated regression. and These are the dimension transformation coefficients for the state term. Channel-level reliability. Preferred option: ; in, This is the amount of the drift consistency penalty. This is the mutation penalty amount; According to the fusion weight Channel estimates The fusion value is obtained by weighted fusion. The expression is: ; Subsequently, physical boundary constraints and cross-consistency checks were applied to the fusion results. The physical boundary constraints included: dissolved oxygen not exceeding the upper tolerance limit of the saturation concentration under the same temperature and pressure, pH within a reasonable range, and consistency between turbidity and transmittance / scattering signal. The cross constraints include intrinsic consistency constraints between homologous variables, such as the consistency between conductivity and temperature-compensated conductivity models, and the coupling consistency between chlorophyll fluorescence and turbidity / absorption background.

[0043] Furthermore, the consistency penalty for neighboring stations Robust consistency statistics on the same ecological indicator can be obtained from adjacent sites or multiple probes at the same cross section, including: Construct a sample set of adjacent sites with the same ecological indicators And using its median Interquartile range function Construct an acceptable range and calculate the deviation, expressed as: ; Among them, fusion value The output of the indicators after consistency constraint fusion; It is a set of samples of the same ecological indicators from adjacent sites or multiple probes at the same cross section within a preset time window; The interquartile range of the sample set is preferably defined as follows: , and These are the 75th and 25th quantiles, respectively; To prevent small positive numbers with a denominator of zero, it is preferable to take the value as 1% to 5% of the lower limit of the index range or the lower limit of the preset minimum interquartile range.

[0044] Furthermore, the neighbor station consistency penalty is only used as an auxiliary item for credibility calculation and status code generation, and does not directly replace the fusion value; when local pollution, stormwater inflow, algal bloom spatial gradient or hydrodynamic spatial difference is identified, a spatial difference status code is generated or the weight of the neighbor station consistency penalty is reduced.

[0045] Furthermore, the consistency penalty for physical and mechanistic constraints Used to characterize fusion value The degree of deviation from physical boundary / mechanism constraints and the cross-consistency between different detection mechanism channels of the same ecological indicator is preferably determined by the range boundary constraint penalty term. Physical boundary penalty term corresponding to ecological indicators and / or inter-channel consistency penalty It is composed and then normalized and aggregated.

[0046] The cross-consistency test includes a consistency test between channel estimates of the same ecological indicator obtained under two or more different detection mechanism channels.

[0047] The range boundary constraint penalty item It can be determined as follows: ; in, and These are the lower and upper limits of the reasonable range for ecological indicators; To prevent small positive numbers from being divided by zero.

[0048] For dissolved oxygen indices, it is preferable to introduce temperature-based indices. With air pressure saturated dissolved oxygen As a physical consistency constraint, a physical boundary penalty term is constructed. for: ; in, The dissolved oxygen concentration at the fusion output; The saturation tolerance coefficient is preferably 1.05 to 1.20. To prevent small positive numbers from being divided by zero.

[0049] Inter-channel consistency penalty This is used to characterize the degree of deviation between channel estimates obtained from two or more different detection mechanisms for the same ecological indicator. It is preferably constructed in a manner where the difference exceeds the allowable uncertainty range. For the case of two detection channels, it can be determined as follows: ; in, and These are channel estimates obtained for the same ecological indicator under two different detection mechanism channels; , These are respectively the channel estimates. , The corresponding channel uncertainty variance; To prevent small positive numbers with a denominator of zero.

[0050] Alternatively, consider the consistency penalty based on physical and mechanistic constraints. Physical boundary penalty item and / or inter-channel consistency penalty The weighted combination is then pruned to the [0,1] interval for subsequent credibility assessment. The calculation.

[0051] Based on the above-mentioned characteristics, the confidence level is... The output is calculated and normalized as follows: ; in, This is the drift consistency penalty amount obtained based on zero-point drift and / or changes in model calibration parameters; This is the mutation penalty amount obtained based on short-term mutation characteristics; This is the physical and mechanistic constraint consistency penalty amount obtained based on physical boundaries, mechanistic constraints, and inter-channel consistency. This refers to the neighbor station consistency penalty amount obtained based on the robust consistency of data from adjacent stations within the same water body. These are the weighting coefficients, and .

[0052] The status codes include at least the following types: bubble interference, biofouling, drift exceeding limits, mutation exceeding limits, calibration failure, flow path abnormality, channel inconsistency, calibration fluid verification abnormality, and spatial difference, which are used to characterize the current data quality and equipment status.

[0053] S6. Output includes the fusion value. Credibility The data packets containing status codes are recorded along with traceable metadata, serving as diagnostic and early warning data input for the upper-level platform.

[0054] The data packet also includes calibration-related timeliness and status parameters, including calibration timestamp and calibration validity period. Drift rate estimate, calibration residual statistic, channel uncertainty It may also include quality traceability fields such as trigger cause, calibration fluid verification result, inter-channel consistency penalty amount, and physical and mechanistic constraint consistency penalty amount; the data packet is uploaded to the upper platform through a wired or wireless communication module, and the platform can directly use the structured detection result of "fusion value + credibility + status code" for subsequent diagnosis and early warning calculation.

[0055] Corresponding to the method described above, the present invention also provides an in-situ detection and self-calibration device for water quality-ecological indicators, comprising an in-situ flow-through detection chamber, a multi-source sensor array, a microfluidic self-calibration module, a data processing module, and a communication output module.

[0056] The in-situ flow-through detection chamber is used to introduce the water sample to be tested and provide a stable detection space; A multi-source sensor array is disposed within or connected to the in-situ flow-through detection cavity to obtain at least two types of raw output signals with different mechanisms. The microfluidic self-calibration module includes a blank solution storage unit, a standard solution storage unit, a quantitative transport unit, a microvalve assembly, and a quantitative mixing network. It is used to switch between blank solution, water sample, and standard solution; perform zero-point calibration; and add standard solution to the same water sample, creating at least two spiking gradient concentration increments. ; The data processing module is electrically or communicatively connected to the multi-source sensor array and the microfluidic self-calibration module, and is used to execute the in-situ detection and self-calibration method for water quality-ecological indicators, and output fused values. Credibility With status codes; The communication output module is used to output the fused value. Credibility The status code is sent to the host platform.

[0057] The microfluidic self-calibration module includes two independent flow paths that converge at the in-situ flow-through detection chamber: one is a water sample flow path, and the other is a blank / standard solution flow path. The flow paths are equipped with check valve structures, flow monitoring structures, and bubble trapping structures to determine the flow path status and provide a basis for status code generation.

[0058] The data processing module consists of an embedded processor, analog-to-digital converter, and signal conditioning circuit. It has a built-in program model for zero-point calibration, standard regression, temperature compensation, consistency constraint fusion, and reliability calculation, which is used to perform detection analysis and quality quantification output according to the above methods.

[0059] By employing the above technical solution, the present invention provides a method and apparatus for in-situ detection and self-calibration of water quality-ecological indicators in rivers and lakes with multi-source complementarity, which has at least the following beneficial effects: 1. Achieve in-situ traceable self-calibration under unattended conditions: By switching between microfluidic blank / standard solutions and standard addition quantitative spiking, the zero point and sensitivity parameters are updated online, significantly reducing the impact of long-term drift, biofouling and matrix changes on measurement results, and improving the stability and consistency of in-situ continuous monitoring data.

[0060] 2. Enhance the credibility and anti-anomaly capability of multi-source monitoring results: By using multiple mechanisms such as optical and electrochemical methods to cross-validate the same ecological indicator and fusion the output value under consistency constraints, it can effectively suppress misjudgments caused by single-channel distortion, sudden interference, or local failures, and improve robustness in complex river and lake scenarios.

[0061] 3. A usable data base that directly outputs "fusion results + credibility + status codes": Quantifies quality information such as drift, mutation, physical constraints and neighbor station consistency at the device end, generates credibility and status codes and outputs them synchronously with the results, so that the upper platform can directly perform diagnosis and early warning modeling, reduce the cost of manual screening and repeated quality control after the fact, and improve data interpretability and engineering application efficiency. Attached Figure Description

[0062] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the multi-source complementary in-situ detection and self-calibration method for water quality-ecological indicators according to the present invention. Figure 2 This is a schematic diagram of the multi-source complementary in-situ detection and self-calibration device for water quality and ecological indicators according to the present invention. Figure 3 This is a structural block diagram of the microfluidic self-calibration module of the present invention; Figure 4 This is a cross-sectional view of the in-situ flow-through detection cavity of the present invention; Figure 5 This is a block diagram illustrating the principle of the data processing module and the communication output module of the present invention.

[0063] In the diagram: 1. Probe housing; 2. Inlet; 3. Outlet; 4. In-situ flow-through detection chamber; 5. Multi-source sensor array; 51. Optical detection channel; 52. Electrochemical detection channel; 6. Microfluidic self-calibration module; 61. Blank solution storage unit; 62. Standard solution storage unit; 63. Calibration solution storage unit; 64. Quantitative transport unit; 641. Water sampling pump; 642. Calibration solution pump; 65. Microvalve assembly; 66. Quantitative mixing network; 7. Bubble-blocking and accumulation chamber; 8. Data processing module; 9. Communication output module; 10. Cleaning components; 11. Power supply module. Detailed Implementation

[0064] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0065] Example 1: In-situ self-calibration and online updating of calibration parameters added to microfluidic standards.

[0066] This embodiment demonstrates, under controlled laboratory conditions, the in-situ self-calibration and online parameter update verification of a multi-source complementary water quality-ecological indicator in-situ detection and self-calibration device. It focuses on showcasing the process of microfluidic blank / standard solution switching, standard solution addition to form a spiking gradient, online regression updating of calibration parameters, and the generation of a "fusion result—confidence—status code" data packet. Figure 2 As shown, the device includes an in-situ flow-through detection chamber 4, a multi-source sensor array 5, a microfluidic self-calibration module 6, a data processing module 8, a communication output module 9, and a power supply module 11, all built into the probe housing 1. The probe housing 1 has an inlet 2 and an outlet 3 that communicate with the in-situ flow-through detection chamber 4. The multi-source sensor 5 includes an optical detection channel 51 and an electrochemical detection channel 52 that cooperate with the in-situ flow-through detection chamber 4.

[0067] Furthermore, the device also includes a cleaning action linked to zero-point calibration. The cleaning action is executed by a cleaning assembly 10, which includes microfluidic rinsing and / or mechanical brushing and / or ultrasonic micro-vibration, to restore the optical window and electrode surface condition. The cleaning action is only a measurement assurance process and still belongs to the maintenance of the detection and analysis system.

[0068] In this embodiment, the "equivalent absorbance" signal of optical detection channel 51 is selected as a representative ecological indicator channel for calculating nitrate nitrogen (as shown in the figure). (Calculation) Concentration. This optical detection channel is a concentration-type detection channel with standard additive quantitative detection conditions. It can determine the spiking increment concentration through the formation of nitrate standard solution and establish a linear relationship between the spiking increment concentration and the detection signal. For state-type detection channels such as pH, ORP, dissolved oxygen, and conductivity, which are not suitable for standard additive quantitative calibration conditions, the corresponding zero-point baseline, drift state, sensitivity check value, or noise parameters can be obtained by using standard solution, calibration solution, or stable state verification method.

[0069] The experimental platform is a constant-temperature stirred water tank (effective volume 50L), with the temperature controlled at 25.0±0.5℃. Magnetic stirring is used to maintain homogeneity within the tank. The device probe is connected to the water tank via a bypass circulation method, with the bypass flow rate stabilized at 0.20L / min. The in-situ flow-through detection chamber 4 adopts a flow-through structure with an optical path of 10mm, and a single detection sampling volume of approximately 8mL. Figure 4 As shown, the in-situ flow-through detection cavity 4 is a flow-through cavity with a constant optical path or constant volume. Inside it is a bubble-blocking and accumulation cavity 7 containing a flow stabilizing component, a bubble trapping component, and an anti-reflection inner wall, in order to reduce the interference of bubbles, eddies, and scattering on optical and electrochemical measurements.

[0070] like Figure 3As shown, in the preferred configuration of this embodiment, the microfluidic self-calibration module 6 is equipped with a blank solution storage unit 61, a standard solution storage unit 62, and a calibration solution storage unit 63, which are used for zero-point calibration, standard addition calibration, and rapid drift verification, respectively. The quantitative transport unit 64 includes a water sample pump 641 and a calibration solution pump 642, combined with a two-position three-way microvalve assembly 65. The quantitative transport unit 64 adopts one of a micro peristaltic pump, a piezoelectric micropump, or an injection micropump. The microvalve assembly 65 is used to switch between water sample, blank solution, and standard solution. The quantitative mixing network 66 is a split-flow-combination or series-parallel quantitative mixing structure, used to generate mixtures with different volume ratios, and can output 4 fixed volume ratio mixing points. The standard solution is... A nitrate standard solution with a mass concentration of 50.00 mg / L; the calibration solution is... A calibration solution with a mass concentration of 10.00 mg / L was used; the blank solution was deionized water at the same temperature. The water sample to be tested in the water tank was prepared by weighing, and the concentration was controlled. The target concentration is 6.20 mg / L (converted based on preparation mass and volume), and no further additions will be made during the experiment. The device operates automatically and records data according to the following procedure: Figure 1 As shown.

[0071] S1. The water sample enters the in-situ flow-through detection chamber 4 and stabilizes for 180 seconds. The raw output of the optical detection channel is then acquired at a frequency of 1 Hz. The median of the last 60 seconds was taken as the original output signal of the water sample at that moment. .

[0072] S2. Switch the blank liquid into the in-situ flow-through detection chamber 4 and stabilize for 120 seconds. Perform blank liquid zero-point calibration on the optical detection channel applicable to the blank response definition. Acquire data under blank liquid conditions. raw output of time The zero-point baseline is obtained from the median. Zero-point noise get.

[0073] S3. Under the condition that the same water sample is re-entered into the in-situ flow-through detection chamber 4, the standard solution is quantitatively introduced through the quantitative mixing network 66 to form the spiking increment concentration corresponding to one unspiked reference point and three spiked gradient points. Among them, the spiking increment concentration corresponding to the unspecified baseline point. The three spiking gradient points corresponded to spiking increments of 1.00, 2.00, and 4.00 mg / L, respectively. .

[0074] The spiking increment concentration This is the concentration increment calculated based on the standard solution concentration, the volume of standard solution added, the volume of blank solution or diluent added, and the final volume of the mixture, and already includes volume dilution correction during the quantitative mixing process. In this embodiment, different spiking increment concentrations are formed by maintaining a consistent water sample volume fraction at each regression point and changing the replacement ratio of standard solution to blank solution or diluent. Each spiking gradient point is held for 90 seconds, and the median of the last 30 seconds is taken as the original output signal for that point. .

[0075] S4. Zero-point correction and regression data construction.

[0076] For the first spiking increment concentration ( ,in This indicates that the water sample was not labeled and ), collect the first The raw output signal of the detection channel Original output signal Select the median or average value within the stable sampling window at the end of the steady-state residence time at the spiking gradient point. Based on the zero-point baseline. Zero-point calibration is performed on the original output signals at each spiking gradient point to obtain the corrected detection signal: ; in, For the first Zero-point baseline of the detection channel; This is the detection signal after zero-point calibration.

[0077] Further construct regression data pairs As the regression input for step S5.

[0078] S5. Linear regression updates the calibration parameters and converts the water sample concentration / channel estimate.

[0079] Incremental concentration of spiking The independent variable is the detection signal after zero-point correction. As the dependent variable, for the regression data set The least squares or robust regression model is used to fit the linear model to obtain the first... Channel sensitivity of each detection channel With zero-point correction response term The preferred model is: ; in, The unit is (channel output unit)·(mg / L) -1 This is used to update the calibration model parameters of the corresponding detection channel online; This is the zero-point correction response term for the current water sample under unspecified conditions. In linear regression, it is represented by the intercept and is only used for the conversion of the current water sample concentration or index value. It is not used as a long-term inherited channel calibration model parameter.

[0080] In this embodiment, the ecological indicator is the concentration of the target component. Due to the increase in spiking concentration corresponding to the unspiked water sample. And at this time the detection signal and satisfy Therefore, the water sample concentration can be converted from the regression parameters as follows: ; Furthermore, the regression residuals are calculated. for: ; Based on this, residual statistics (such as residual mean square and root mean square error) and the fitting uncertainty are obtained, which are then used for subsequent confidence assessment. Calculations related to status code generation. Online inherited and updated calibration model parameters include at least the zero-point baseline. Channel sensitivity Temperature compensation parameters and / or drift state parameters.

[0081] S6, Output includes fusion value Credibility Structured data packets with status codes, and records calibration timestamps and zero-point baselines. Channel sensitivity Metadata such as residual variance, etc.

[0082] (1) Initial self-calibration (t=0h) Zero-point baseline measured with blank solution. The robust zero-point noise estimation of the blank sequence is as follows: According to the spiking increment concentration One unspecified baseline point and three spiked gradient points were formed, and the data are shown in Table 1 (unit: A).

[0083] Table 1. Record data of un-paired benchmark points and paired gradient points

[0084] By analyzing the regression data set Linear regression fitting yielded the unspecified response term and channel sensitivity for the current water sample: ; Then water sample concentration The calculation is as follows: ; The variance of the regression residuals is The standard uncertainty of concentration, obtained from parametric covariance propagation, is approximately [value missing]. (df=2). At the same time, a rapid check was performed on the calibration solution (10.00 mg / L) to obtain the calibration response. Based on this, the sensitivity check value is estimated. This aligns with the regression sensitivity. The rapid verification using this calibration solution does not generate multi-point spiking gradients; it is only used to obtain sensitivity verification values ​​and determine if sensitivity drift exists. Based on the credibility model set in this embodiment, under conditions of no significant abrupt changes and relatively small constraint penalties, the output credibility is... The status code is "CAL_OK".

[0085] (2) Continuous operation and drift accumulation (0–168h) The device operated continuously in the water tank for 7 days, during which a blank check (S2) was performed every 6 hours, and a rapid calibration solution check was performed every 6 hours (without generating multi-point gradients, only used to estimate sensitivity drift). The confidence level of the output for any detection or check cycle was... The sensitivity change rate obtained by checking the calibration solution. Exceeding the threshold At this time, a complete standard calibration is triggered (S2–S5). Under the conditions of this embodiment, the device completes a complete calibration at 72 hours and triggers a complete calibration at 168 hours due to decreased reliability. Typical zero-point baseline and sensitivity verification changes over time are shown in Table 2 (unit: A). ).

[0086] Table 2 Typical zero-baseline and sensitivity verification changes over time

[0087] The zero-point baseline drift rate can be calculated from data from 0–168 h. With sensitivity change rate ,for: ; ; in, For sensitivity verification values ​​obtained when using a known concentration of calibration solution for rapid verification, it is preferable to follow the formula... The calculation is used to characterize the drift trend of sensitivity over time and serves as the basis for determining whether to trigger the inclusion of a full standard in the calibration. This is the concentration of the calibration solution. Based on this, a drift consistency penalty is applied at the equipment end. It is used for credibility assessment and calibration triggering.

[0088] (3) Complete calibration results over 72 hours (t=72h) Perform a full calibration from S2 to S5 at 72 hours, and measure the zero-point baseline using a blank zero point. The spiking increment concentration was obtained. The data is shown in Table 3 (unit: A).

[0089] Table 3. Data on unspecified baseline points and spiked gradient points over 72 hours.

[0090] Linear regression yielded the following unspecified response term and channel sensitivity for the current water sample: ; water sample concentration The calculation is as follows: ; The standard uncertainty of concentration obtained from residual propagation is approximately (df=2). At the same time, a rapid check was performed on the calibration solution (10.00 mg / L) to obtain the calibration response. Corresponding sensitivity check value Consistent with the regression results, the output confidence level is... The status code is "CAL_OK".

[0091] (4) Comparison before and after calibration triggered at 168h (t=168h) At 168 hours, perform routine water sample measurement according to S1, and simultaneously execute blank verification and calibration solution rapid verification. Then, determine whether to trigger complete calibration based on the reliability of the output from the most recent testing or verification cycle and the sensitivity verification results of the calibration solution. At this time, the blank zero-point verification obtains the zero-point baseline. The calibration fluid was used to verify the calibration response. That is, sensitivity check value The device was calibrated using the parameters from the most recent full model calibration 72 hours ago. , The water sample is converted in real time, and the original output signal of the water sample is obtained. Then the instant correction value Instantaneous concentration The estimate is: ; Simultaneously, a drift consistency penalty is constructed based on the drift check. In this embodiment, We can obtain: ; Combined with mutation penalty Consistency penalty for physical and mechanistic constraints And calculate the confidence level according to the following formula. for: ; Calculated credibility If the value falls below the threshold of 0.80, the status code is set to "CAL_REQ," triggering a complete standard addition calibration (S2–S5). The spiking increment concentration is obtained after calibration. The data is shown in Table 4 (unit: A).

[0092] Table 4. Data of unspecified reference points and spiked gradient points after calibration.

[0093] The regression yielded the following unspecified response terms and channel sensitivities for the current water sample: ; water sample concentration Updated to: ; The concentration standard uncertainty obtained from residual propagation is approximately (df=2). After calibration, the device updates the status code to "CAL_OK" and outputs the final confidence level. Simultaneously, the calibration timestamp and zero baseline are recorded in the data packet. Channel sensitivity Metadata such as residual statistics.

[0094] This embodiment presents the complete data and calculation process of the device performing zero-point verification, rapid calibration fluid verification, and standard addition-based complete self-calibration under continuous operation conditions. Among these, channel sensitivity... As an inheritable and updatable calibration model parameter, the current unspecified response term of the water sample. This is only used for current water sample concentration conversion; the calibration solution rapid verification is used to estimate sensitivity drift and assist in determining whether a complete calibration has been triggered. The data package is output in the form of "concentration result - confidence level - status code" and records metadata such as calibration timestamp, zero baseline, channel sensitivity, residual statistics, calibration solution verification results and trigger reasons, providing a traceable detection data foundation for multi-channel consistency fusion and platform-side diagnostic early warning calls in subsequent embodiments.

[0095] Example 2: Consistency Constraint Fusion and Credibility of Multi-Mechanism Mutual Verification calculate.

[0096] This embodiment, based on a multi-source complementary in-situ detection and self-calibration device for water quality and ecological indicators, focuses on verifying "mutual verification of multiple mechanisms for the same ecological indicator—fusion of consistency constraints—reliability". The data processing flow is described as "quantitative output". The in-situ flow-through detection chamber 4 of the device integrates a multi-source sensor array 5, preferably including a dissolved oxygen (DO) optical detection channel and an electrochemical detection channel, a chlorophyll a (Chl-a) fluorescence channel and an absorption channel. Temperature and pressure sensors are also configured for physical constraint modeling. Adjacent monitoring nodes are arranged within the same water body as a reference for neighboring station consistency. The microfluidic self-calibration module 6 can update the zero point and sensitivity of concentration-type detection channels with standard addition quantitative conditions as described in Example 1. For state-type detection channels such as pH, ORP, dissolved oxygen, and conductivity, which are not suitable for standard addition quantitative calibration, the corresponding zero-point baseline, drift state, sensitivity verification value, or noise parameters are obtained through standard solutions, calibration solutions, or stable state verification methods. In this embodiment, each detection channel has completed its most recent effective calibration or verification, with a focus on demonstrating the calculation process of fusion and mass quantification output.

[0097] The experiment was conducted in a constant-temperature circulating water bath with an effective volume of 120 L, a water conductivity of approximately 550 μS / cm, and a temperature controlled at [temperature value missing]. Atmospheric pressure The water body circulates slowly and continuously to ensure cross-sectional uniformity. Monitoring node A (for fusion output) and neighboring node B (for consistency reference) are set up, with sampling locations 3m apart and at the same height. Node A operates on a 10-minute evaluation cycle: within each cycle, each channel collects data at 1Hz for 60 seconds, and the median is used as the channel estimate. Simultaneously, the abrupt change penalty is calculated based on the short window difference and the change in calibration parameters. Consistency penalty for drift and output the fusion value. Credibility With status codes.

[0098] The channel uncertainty model and weighting model used in this embodiment are as follows. For the first... One detection channel, channel uncertainty The value is determined by zero-point noise, regression residuals, and state terms, and is preferably: ; in, Zero-point noise is obtained from robust variance estimation of the blank / steady-state sequence; This refers to the fitting residual, verification residual, or equivalent uncertainty obtained from the most recent standard addition regression, standard solution calibration, or calibration solution verification; for detection channels equipped with standard addition-based quantitative detection conditions, The most recent standard can be used to incorporate the regression residuals; for state-based detection channels, The equivalent error can be obtained by checking the standard solution, calibration solution, or stable state. and These are the state dimension conversion coefficients. Channel-level reliability. Preferred option: ; And based on this, the fusion weight is obtained. for: ; The combined value of the same ecological indicator The weighted fusion is as follows: ; Drift consistency penalty at the fusion layer Take the weighted average, and calculate the mutation penalty. Take the maximum value of the channel, that is: ; ; in, For the first Drift consistency penalty for each detection channel; For the first Mutation penalty for each detection channel; Consistency penalty for physical and mechanistic constraints The preferred option includes range boundary constraint penalty terms and physical boundary penalty terms. Consistency penalty between channels Taking DO as an example, saturated dissolved oxygen is calculated based on temperature and air pressure. In this embodiment, the temperature is 25°C. The maximum allowed is Physical boundary penalty terms are acceptable: ; in, The dissolved oxygen concentration at the fusion output; The saturation tolerance coefficient is preferably 1.05 to 1.20. To prevent small positive numbers from being divided by zero; Inter-channel consistency penalty To characterize the deviation between channel estimates of the same ecological indicator obtained under different detection mechanism channels, it is preferably constructed in a manner that "the difference exceeds the allowable range of uncertainty". For the case of two detection channels, it can be determined as follows: ; in, and These are channel estimates obtained for the same ecological indicator under two different detection mechanism channels; , These are respectively the channel estimates. , The corresponding channel uncertainty variance; To prevent small positive numbers with a denominator of zero.

[0099] In this embodiment, the consistency penalty for physical and mechanistic constraints is taken as follows: ; For other indicators involving reasonable upper and lower limits, the range boundary constraint penalty term can also be incorporated into the physical and mechanistic constraint consistency penalty. of Operations or weighted combinations.

[0100] Neighbor site consistency penalty The median of the same ecological indicator sequence of neighboring node B in the last 30 minutes Interquartile range The construction, expressed as: ; in, To prevent interquartile range Small positive numbers that are too small or zero, resulting in a denominator of zero, have dimensions consistent with the aforementioned ecological indicators; it should be noted that the neighbor station consistency penalty amount... It is only used to assist in generating credibility and status codes, and does not directly replace the fusion value of this node. When local pollution, stormwater inflow, spatial gradient of algal blooms, or spatial differences in hydrodynamics are identified, the system can generate a spatial difference status code or reduce the weight of neighbor station consistency penalty in the credibility calculation.

[0101] Preferred, Take 1% to 5% of the lower limit of the range of the index or the lower limit of the preset minimum interquartile range. For example, for the dissolved oxygen index, one could take... .

[0102] Final credibility The output is obtained by linear normalization and pruning: ; in, This is the amount of the drift consistency penalty. This is the mutation penalty amount; This is the physical and mechanistic constraint consistency penalty amount obtained based on physical boundaries, mechanistic constraints, and inter-channel consistency. This is the consistency penalty for neighboring stations.

[0103] The status code is generated by the logic of various penalty amounts and thresholds. When the penalty amount changes abruptly... When the value is large, the output is "bubble / transient interference"; when the drift consistency penalty is large... When the value is large, the output is "drift exceeds limit" and the channel consistency penalty is applied. Or physical and mechanistic constraints consistency penalty amount When the value is large, the system outputs "channel inconsistency / constraint inconsistency" and synchronously outputs various penalty amounts and channel weights in the data packet for platform-side traceability.

[0104] (1) The fusion of two mechanisms of dissolved oxygen (DO) and mutual verification Output.

[0105] The DO ecological index of node A is measured simultaneously using both the DO optical detection channel (denoted as O) and the DO electrochemical detection channel (denoted as E). To verify the accuracy of the fusion results, a benchtop optical DO meter is used as a reference in this embodiment (denoted as reference value). The sampling was synchronized with node A and recorded as a 60-second average. By applying different disturbances to the operation process (high reading due to electrochemical membrane fouling, low reading due to optical window contamination, and spikes due to transient bubbles), the data and calculation results for four typical cycles are shown in the table below (unit: mg / L). The uncertainty calculated according to the above model is: The error is... ).

[0106] Table 5. Data and Calculation Results for Typical Cycles

[0107] To facilitate verification, the key calculation process is explained using the period "B: High Electrochemical Drift" as an example. Within this period, the characterization measurements of the detection channel are taken as follows: , and take The state dimension conversion coefficient is taken as .in, and The zero-point noise (preferably the robust standard deviation of the signal within the stable window) obtained by optical detection channel O and electrochemical detection channel E during the blank liquid zero-point verification stage are respectively. ); and These are the regression residuals obtained in the most recent standard-added regression calibration of the channel, preferably the root mean square or standard deviation of the regression residuals, used to reflect the contribution of the calibration model fitting error to the channel uncertainty.

[0108] Channel uncertainty for: ; ; Channel-level reliability for: ; ; Fusion weights for: ; ; The combined value of the same ecological indicator for: ; Drift consistency penalty at the fusion layer for: ; Channel difference Combined uncertainty Take the consistency threshold Then the inter-channel consistency penalty item for: ; Fusion value for this period Not exceeding Therefore, the physical boundary penalty term The consistency penalty for physical and mechanistic constraints is obtained. .

[0109] The median DO concentration at neighboring station B in the most recent 30 minutes was 8.23 ​​mg / L, and the interquartile range was 0.20 mg / L. [The value is taken as...] mg / L, then the neighbor station consistency penalty for: ; Ultimate credibility for: ; Because the drift penalty of electrochemical channels is large, and the consistency penalty term between channels is also large. When a high level is reached, the device outputs a status code of "DRIFT_E (electrochemical drift) + INCONS (channel inconsistency)" and records the weight of the detection channel for this cycle in the data packet. The characterization fusion results are primarily contributed by the optical channel, thus achieving suppression and traceable expression of single-channel anomalies. Similarly, the periodicity of the "D: bubble interference spike" is penalized by mutation. Furthermore, when the channel inconsistency penalty reaches its limit, the reliability is significantly reduced and "BUBBLE_E (electrochemical transient bubble interference)" is output.

[0110] (2) Chlorophyll a (Chl-a) dual-channel fusion and Output.

[0111] To reflect the multi-source complementarity of ecological indicators, this embodiment uses parallel estimation of Chl-a via both the fluorescence channel (denoted as F) and the absorption channel (denoted as A). The absorption channel outputs an equivalent Chl-a after online correction of the turbidity background, while the fluorescence channel outputs a fluorescence equivalent Chl-a. Reference value Data were obtained by laboratory extraction spectrophotometry and recorded synchronously with sampling. Data and calculation results for three typical periods were obtained by adding kaolin to the water to cause turbidity interference and adding humic acid to cause fluorescence quenching, and are shown in the table below (unit: µg / L). The uncertainty calculated according to the above model is: The error is... ).

[0112] Table 6. Data and Calculation Results for Typical Cycles

[0113] Among them, F: "Turbidity interference absorption is higher" periodically appears in the channel estimate value. The fusion weights are significantly higher and the uncertainty is significantly increased. Automatically tilted towards the fluorescence channel to adjust the fusion value. Still in line with the reference value To maintain consistency, while also imposing consistency penalties based on physical and mechanistic constraints. With credibility This clearly reflects the decline in data quality during this period; specifically, for the Chl-a metric, the consistency penalty for physical and mechanistic constraints in this embodiment... The main component is the inter-channel consistency penalty term between the fluorescence and absorption channels, rather than the dissolved oxygen saturation upper limit constraint. In contrast, the "low fluorescence quenching" period (G: low fluorescence quenching) shows that the fluorescence channels are downweighted, and the fusion output is mainly contributed by the absorption channels, with higher reliability. Medium to high to reflect the data status of "available but needing to be labeled".

[0114] As can be seen from the parallel verification of DO and Chl-a above, this embodiment fully demonstrates the device-side mutual verification and fusion of multi-mechanism results for the same ecological indicator, the quantitative expression of drift, mutation, consistency of physical and mechanistic constraints, consistency between channels and deviations from neighboring stations, and the "fusion result". + Credibility The structured output method of "+status code" provides a usable data foundation that can be directly accessed for subsequent platform-side diagnostic and early warning models.

[0115] Example 3: Long-term field deployment and comparative verification.

[0116] This embodiment verifies the long-term operational capability and data availability of the in-situ detection and self-calibration method and device for water quality and ecological indicators under real-world river and lake conditions. The implementation site is a typical cross-section of a shallow urban lake (average water depth approximately 1.2–1.8 m, gentle bank slope, significantly affected by seasonal fluctuations in rainfall and algae growth). Three in-situ monitoring nodes A, B, and C are deployed at the same cross-section, with a node spacing of approximately 200–300 m for neighbor station consistency constraints. Node A outputs a "fusion value" according to the present invention. + Credibility The data packet with the "+status code" indicates that nodes B and C serve as neighboring references and also possess the detection capabilities of this invention. The neighboring station consistency penalty is only used as auxiliary information for confidence calculation and status code generation, and does not directly replace the fusion output of node A. When local pollution, stormwater inflow, algal bloom spatial gradient, or hydrodynamic spatial differences occur, the system can generate a spatial difference status code or reduce the weight of the neighboring station consistency penalty. For ease of comparison, a traditional multi-parameter probe (which only outputs single-channel readings and does not have the microfluidic standard to add self-calibration and consistency fusion / confidence output) is simultaneously deployed next to node A, and "manual sampling - laboratory analysis / portable calibrator" is used as the reference value source.

[0117] The multi-source sensor array 5 at node A includes at least: dissolved oxygen (DO) optical channel O and electrochemical channel E, chlorophyll a (Chl-a) fluorescence channel F and absorption channel A, turbidity, temperature, and air pressure; and is also configured with a UV absorption channel for... Estimate. Among them, the UV absorption channel is a concentration-type detection channel with standard addition quantitative conditions. It can form a definite spiking increment concentration using nitrate standard solution and complete standard addition regression. For multi-mechanism channels such as DO and Chl-a, the channel estimate, uncertainty, and state parameters can be obtained by combining the most recent valid calibration, standard solution or calibration solution verification, stable state verification, and inter-channel consistency testing. The microfluidic self-calibration module 6 is configured with blank solution, Standard solution (50.00 mg / L) and The calibration solution (10.00 mg / L) was used, and the spiking increment concentration was generated through a quantitative mixing network 66. Zero-point calibration and standard addition regression update were performed according to Example 1. The device operation cycle was set to 10 min: stable sampling for 60 s within each cycle, and the median was taken to form the channel estimate. Then, following Example 2, consistency constraint fusion is performed to obtain the fused value. And calculate the credibility. Output the data packet after the status code.

[0118] The self-calibration strategy is set to perform a blank zero-point check every 6 hours to update the zero-point baseline. With zero-point noise Perform a rapid verification of the calibration solution every 24 hours to estimate the detection channels. Sensitivity check value Change; when the reliability of the output of any ecological indicator in the most recent detection or verification period meets the requirements. Furthermore, a complete self-calibration process is triggered when the sensitivity change rate obtained from calibration solution verification exceeds 5% for three consecutive cycles, or when the calibration validity period expires. For UV channels with standard additive quantitative detection conditions, the complete self-calibration process includes one unspecified baseline point and three spiked gradient points, corresponding to incremental spiking concentrations. mg / L; for other state-based or multi-mechanism channels, perform corresponding standard solution verification, calibration solution verification, stable state verification, or inter-channel consistency checks. The triggering reason is written to the status code and metadata fields.

[0119] To conduct comparative verification, manual sampling (12 times) was carried out simultaneously at node A during 21 days of continuous operation. DO reference values ​​were compared on-site using a benchtop optical DO analyzer, while Chl-a reference values ​​were obtained using laboratory extraction spectrophotometry. The reference values ​​were obtained using laboratory spectrophotometric / ion analysis methods; each sampling time was aligned with the device output cycle, and the fused values ​​from the same cycle data packets were used. Credibility The status code was used as comparison data, and the corresponding readings of the traditional probe were also recorded. Table 7 shows the key data from 12 comparative samplings (unit: DO / ). (mg / L, Chl-a is µg / L).

[0120] Table 7 Comparison of fused output and reference value of node A (n=12)

[0121] Based on the 12 sets of comparative data in Table 7, the fusion output of node A and the readings of the traditional probe are quantitatively evaluated using the following statistical measures: ; ; in, The number of data groups participating in the comparative evaluation (in this embodiment) ); For the first The fused output value of node A in the group data (or the fused value output by the device of the present invention); In order to be with the first The reference values ​​corresponding to the data sets were obtained through manual sampling laboratory analysis or measurement by a comparison instrument. For reference value sequence The arithmetic mean, i.e. .

[0122] Taking DO as an example, the sum of squared errors for 12 comparisons is ,but Regarding DO, Chl-a and The statistical results are summarized in Table 8.

[0123] Table 8. Comparison statistics between fused output and traditional probes (n=12)

[0124] In addition to accuracy, this embodiment further statistically analyzes the high-reliability available data rate and quality grading distribution during a 21-day long-term operation. Let the total number of output cycles be... For the output of this invention, a confidence level is defined. Define for highly reliable and available data Define for tagging data For low-reliability data, the number of retained data points for traditional probes was calculated based on commonly used field quality control rules (physical range + mutation rate threshold + missing data removal). The statistical results are shown in Table 9, where the high-reliability usable data rate for DO was calculated as follows: .

[0125] Table 9. Data quality distribution and high-reliability available data rate during the 21-day operation period (N=3024)

[0126] During the 21-day run, the low confidence level of node A was mainly related to three types of field disturbances: firstly, transient spikes in electrochemical DO caused by short-term bubble / flow disturbances (preferably marked as BUBBLE_E or FLOW_TRANS status codes); secondly, high turbidity background caused by rainstorm processes leading to the interaction between absorbing Chl-a and UV types. Physical and mechanistic constraints consistency penalty The third factor is the increase in inconsistency between channels caused by biofouling and long-term drift, triggering self-calibration (the status codes are preferably marked as FOUL / DRIFT and CAL_REQ). This embodiment records that node A triggered four complete standard addition calibrations during the operating period (all based on confidence level). (Continuous triggering or exceeding the sensitivity limit triggering), where the most important is... Taking the UV channel as an example, its sensitivity verification value The temperature dropped after rainwater and fouling accumulation, and returned to a reasonable range after calibration; meanwhile, the device records the calibration timestamp and zero baseline after each complete calibration. Channel sensitivity The residual statistics and calibration validity period are written into the data packet metadata field for easy traceability on the platform side. Based on consumable consumption, according to the example setting of one blank check every 6 hours (5 mL of blank solution per check) and 4 complete calibrations (approximately 12 mL of blank solution and 1.5 mL of standard solution per check), the total blank solution consumption within 21 days is approximately... The total consumption of standard solution is approximately If the device is preferably configured with a calibration solution for rapid verification, then based on one micro-calibration verification per day and 0.5 mL of calibration solution per verification, the calibration solution consumption is: mL, still meets the requirements of miniaturization and low maintenance for long-term deployment.

[0127] Through long-term on-site deployment and comparative verification, this embodiment fully presents the operational data, reference comparison data, accuracy statistics, and high-reliability usability data rate calculation process of the present invention under real river and lake conditions, and uses "fusion value" as the basis for this invention. + Credibility The output format of "+status code+traceable metadata" synchronously records quality information such as drift, mutation, physical and mechanistic constraint consistency, inter-channel consistency, neighbor station consistency, and calibration fluid verification results, providing a data foundation that can be directly called for subsequent platform-based evaluation, diagnosis, and early warning models.

[0128] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0130] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for in-situ detection and self-calibration of water quality-ecological indicators from multiple complementary sources in rivers and lakes, characterized in that, The method includes the following steps: S1. Introduce the water sample to be tested into the in-situ flow-through detection chamber, and arrange a multi-source sensor array in the in-situ flow-through detection chamber or at a location connected to it to synchronously collect at least two types of original output signals with different mechanisms. S2. The blank liquid is switched into the in-situ flow-through detection chamber through the microfluidic self-calibration module, and zero-point calibration is performed on the detection channels applicable to the blank response definition to obtain the zero-point baseline and zero-point noise of each detection channel; S3. The water sample is cut back into the in-situ flow-through detection chamber. The microfluidic self-calibration module performs standard additive quantitative calibration on the same water sample to form the spiking increment concentration corresponding to one unspecified reference point and at least two spiking gradient points. The detection signal corresponding to the spiking increment concentration is obtained from the detection channel with standard additive quantitative detection conditions. S4. Based on the detection signal, perform regression fitting to obtain the channel sensitivity used for online updating of the corresponding detection channel calibration model parameters, and the current water sample unspecified response term used for converting the current water sample index value, and convert it into the channel estimate value of the same ecological index. At the same time, calculate the drift consistency penalty and mutation penalty value corresponding to the channel estimate value. S5. Perform consistency constraint fusion on multiple channel estimates of the same ecological indicator to obtain a fused value, calculate the credibility for current data quality evaluation and self-calibration trigger judgment, and generate a status code. S6. Output a data packet containing the fusion value, credibility and status code and record traceable metadata as input for the diagnostic and early warning data of the upper platform.

2. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, The in-situ flow-through detection cavity is a flow-through cavity with a constant optical path or a constant volume; The multi-source sensor array includes at least an optical detection channel and an electrochemical detection channel, wherein the optical detection channel includes one or more of UV / Vis absorption channels, fluorescence channels, and turbidity scattering channels, and the electrochemical detection channel includes one or more of pH, ORP, dissolved oxygen, and conductivity. The detection channels are classified into detection channels with standard additive quantitative detection conditions and state-type detection channels that are not suitable for standard additive quantitative calibration conditions, according to the calibration method. The detection channel equipped with standard addition-type quantitative detection is a concentration-type detection channel that can generate a spike concentration increment through a standard solution and establish a relationship between the spike increment concentration and the detection signal. The state-based detection channel includes one or more of pH, ORP, dissolved oxygen, and conductivity, and obtains the corresponding zero-point baseline, drift state, sensitivity verification value, or noise parameter through standard solution, calibration solution, or stable state verification method.

3. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, The blank solution is deionized water or a blank matrix solution prepared with the same ionic strength; The zero-point calibration includes at least: The zero-point baseline of each detection channel is determined using a sliding window mean and robust variance. With zero-point noise To make an estimate, that is: ； ； in, For the first The detection channel is in a blank liquid state at The raw output at time step; MAD represents the median absolute deviation; This indicates taking the median.

4. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, The standard addition-based quantitative calibration employs two-point or multi-point spiking to add standard solution at least twice with different quantitative volumes to the same water sample, forming a spiking increment concentration corresponding to an unspiked baseline point and at least two spiking gradient points. ,include: A quantitative mixing network is used to mix water samples, standard solutions, and blank solutions or diluents in a controllable volume ratio to generate an unspecified reference point and at least two mixtures with different spiking ratios. These mixtures are then sequentially introduced into an in-situ flow-through detection chamber as the spiking increment concentration at the spiking gradient points. ; When using two-point spiking, obtain the corrected detection signal from the unspiked water sample. The detection signal after single quantitative spiking correction ,in, , The first The original output signals collected by each detection channel under conditions of unspecified water sample and spiked mixture; When using multi-point scaling, generate The spiking increment concentration at each spiking gradient point For each scalated gradient point, the first sample is collected. The raw output signal of each detection channel and based on the zero-point baseline The detection signal of the corresponding channel response after zero-point calibration The expression is: ； in, After zero-point calibration, the first The detection channel corresponds to the first spiking increment concentration The detection signal.

5. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, Step S4 includes: Based on detection signal Constructing regression data pairs The least squares or robust regression model is used to fit the linear model to obtain the channel sensitivity used for online updating of the calibration model parameters of the corresponding detection channel. And the unspecified response item of the current water sample used for converting the current water sample index values. The calculation formula is: ； ； in, For the first The detection channel corresponds to the first spiking increment concentration The mean of the calibrated detection signal; This represents the average concentration of the spiked increment; This is the zero-point correction response term for the current water sample under unspecified conditions; The first result is obtained by conversion based on the calibration model parameters. Channel estimates for each detection channel The expression is: ； The drift consistency penalty is characterized by a combination of the zero-point baseline change rate and the calibration parameter change rate. The expression is: ； in, , This is the drift penalty weighting coefficient, used to adjust the relative contributions of the zero-baseline drift term and the channel sensitivity drift term, and satisfies... , ; It is a time variable; The mutation penalty is obtained by using a sliding window first-order difference and threshold discrimination or robust residual discrimination based on Huber loss. .

6. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, Step S5 includes: Calculate the channel uncertainty for each detection channel. The calculation formula is: ； in, Zero-point noise is obtained from robust variance estimation of the blank / steady-state sequence; The regression residuals are from the most recent calibrated regression. and , which is the dimension conversion coefficient for the state term; The channel-level reliability of each channel is calculated based on the drift consistency penalty and the mutation penalty. The calculation formula is: ； in, This is the amount of the drift consistency penalty. This is the mutation penalty amount; Based on the channel uncertainty of each channel With channel-level credibility Determine the fusion weights The expression is: ； According to the fusion weight Channel estimates The fusion value is obtained by weighted fusion. The expression is: ； Introducing a quantity that includes neighbor station consistency penalty and the consistency penalty for physical and mechanistic constraints Consistency constraints are used to calculate the reliability of current data quality assessment and self-calibration trigger determination. The expression is: ； in, This is the amount of the drift consistency penalty. This is the mutation penalty amount obtained based on short-term mutation characteristics; This is the consistency penalty for physical and mechanistic constraints, obtained based on physical boundaries, mechanistic constraints, and cross-consistency. This refers to the neighbor station consistency penalty amount obtained based on the robust consistency of data from adjacent stations within the same water body. These are the weighting coefficients, and .

7. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 6, characterized in that, The neighbor station consistency penalty Robust consistency statistics on the same ecological indicator obtained from adjacent sites or multiple probes at the same cross section include: Construct a sample set of adjacent sites with the same ecological indicators And using its median Interquartile range function Construct an acceptable range and calculate the deviation, expressed as: ； Among them, fusion value The output of the indicators after consistency constraint fusion; It is a set of samples of the same ecological indicators from adjacent sites or multiple probes at the same cross section within a preset time window; The interquartile range of the sample set; To prevent small positive numbers with a denominator of zero; The neighbor station consistency penalty For credibility only The auxiliary items for calculating and generating status codes do not directly replace the fused value. When local pollution, stormwater inflow, spatial gradient of algal blooms, or spatial differences in hydrodynamics are identified, a spatial difference status code is generated or the consistency penalty for neighboring stations is reduced. The weight.

8. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 7, characterized in that, The consistency penalty for physical and mechanistic constraints Used to characterize fusion value The degree of deviation from physical boundary / mechanism constraints and the degree of cross-consistency between different detection mechanism channels of the same ecological indicator, including the range boundary constraint penalty term. Physical boundary penalty term corresponding to ecological indicators and / or inter-channel consistency penalty ; The consistency penalty for physical and mechanistic constraints for and / or inter-channel consistency penalty The operation or weighted combination, and trimmed to Intervals are used for confidence calculations; The range boundary constraint penalty item The calculation formula is: ； in, and These are the lower and upper limits of the reasonable range for ecological indicators; To prevent small positive numbers from being divided by zero; The physical boundary penalty item Including the introduction of temperature-based With air pressure saturated dissolved oxygen As a physical consistency constraint, the expression is: ； in, The dissolved oxygen concentration at the fusion output; This is the saturation tolerance coefficient; To prevent small positive numbers from being divided by zero; Inter-channel consistency penalty Channel estimates used to characterize the same ecological indicator under two or more different detection mechanism channels. The degree of deviation between them is expressed as: ； in, and These are channel estimates obtained for the same ecological indicator under two different detection mechanism channels; , These are respectively the channel estimates. , The corresponding channel uncertainty variance; To prevent small positive numbers with a denominator of zero.

9. The method for in-situ detection and self-calibration of water quality-ecological indicators according to claim 1, characterized in that, The conditions for triggering the self-calibration decision include: The reliability of the output is triggered according to a preset cycle and / or when any ecological indicator is detected or verified in its most recent cycle. Below the threshold The sensitivity change rate obtained by checking the calibration solution. Exceeding the threshold Or, inter-channel consistency penalty Exceeding the threshold Or the calibration validity period expires and triggers; under non-triggered conditions, maintain low-frequency zero-point calibration, blank flushing, or calibration solution verification; in, , These are the sensitivity check values ​​and their changes; Maintain low-frequency zero-point calibration or perform only blank flushing or calibration fluid checks under non-triggered conditions.

10. An apparatus for implementing the in-situ detection and self-calibration method for water quality-ecological indicators according to any one of claims 1-9, characterized in that, The device includes: In-situ flow-through detection chamber is used to introduce the water sample to be tested and provide a stable detection space; A multi-source sensor array is disposed within or connected to the in-situ flow-through detection cavity to obtain at least two types of raw output signals with different mechanisms. The microfluidic self-calibration module includes a blank liquid storage unit, a standard liquid storage unit, a quantitative transport unit, a microvalve assembly, and a quantitative mixing network. It is used to achieve switching between blank liquid, water sample, and standard liquid, zero-point calibration, and the addition of standard to the same water sample to form at least two spiking gradient points of incremental spiking concentration. The data processing module is electrically or communicatively connected to the multi-source sensor array and the microfluidic self-calibration module, and is used to execute the in-situ detection and self-calibration method of water quality-ecological indicators, and output the fusion value, confidence level and status code. The communication output module is used to send the fusion value, credibility, and status code to the host platform.

Images