A method and system for detecting concentration of suspended solids in canal water samples

Abstract

The application relates to the technical field of water quality detection, and discloses a detection method and system for the concentration of suspended solids in a canal water sample, which comprises the following steps: acquiring the environmental parameters of the canal water body in real time, synchronously collecting the light scattering intensity signals and acoustic echo attenuation signals of the canal water body through a near-infrared optical sensor and an acoustic Doppler current profiler, calculating an initial suspended solid concentration value based on the light scattering intensity signals and the acoustic echo attenuation signals, correcting the initial suspended solid concentration value through a dynamic compensation algorithm according to the environmental parameters of the canal water body to generate pre-calibration concentration data, inputting the pre-calibration concentration data into a pre-trained machine learning model to output a final suspended solid concentration value, and wherein the pre-trained machine learning model comprises a random forest regression model and a gradient boosting decision tree model. The scheme can improve the detection accuracy under complex hydrological conditions and the real-time response capability of the model under a disturbance scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of water quality detection, and particularly relates to a detection method and system for suspended matter concentration in canal water samples. BACKGROUND

[0002] The detection of suspended matter concentration in canal water bodies is a key link in environmental monitoring and water ecological management. The traditional method is difficult to overcome the combined interference of optical and acoustic signals in dynamic water environment, especially in the scene of frequent navigation disturbance and variable hydrological conditions, the sensor measurement value is easily affected by factors such as temperature fluctuation, ion concentration mutation, water flow turbulence, etc., resulting in insufficient reliability of monitoring data in actual application, which restricts the accuracy of water quality evaluation and the scientificity of pollution prevention and control decision.

[0003] At present, the traditional scheme only compensates for the single influence of temperature or conductivity with a fixed coefficient, which cannot cope with the multi-parameter coupling interference of water temperature change, ion strength fluctuation, sudden water flow disturbance, etc., resulting in compensation failure under complex working conditions; the conventional algorithm lacks self-adaptive adjustment mechanism under disturbance working conditions such as rainstorm and ship navigation, and the model output lags behind the rapid change of hydrological conditions.

[0004] Therefore, it is urgent to develop a detection method and system for suspended matter concentration in canal water samples, which can improve the detection accuracy under complex hydrological conditions and the real-time response ability of the model under disturbance scene. SUMMARY

[0005] In order to solve the above technical problems, the application provides a detection method and system for suspended matter concentration in canal water samples, which can improve the detection accuracy under complex hydrological conditions and the real-time response ability of the model under disturbance scene.

[0006] The application provides a detection method for suspended matter concentration in canal water samples, which comprises the following steps:

[0007] S1, real-time acquisition of the environmental parameters of the canal water body, synchronous acquisition of the light scattering intensity signal and acoustic echo attenuation signal of the canal water body by a near-infrared optical sensor and an acoustic Doppler current profiler;

[0008] S2, calculation of the initial suspended matter concentration value based on the light scattering intensity signal and the acoustic echo attenuation signal;

[0009] S3, correction of the initial suspended matter concentration value by a dynamic compensation algorithm according to the environmental parameters of the canal water body, to generate pre-calibration concentration data;

[0010] S4, input of the pre-calibration concentration data into a pre-trained machine learning model to output the final suspended matter concentration value; wherein the pre-trained machine learning model includes a random forest regression model and a gradient boosting decision tree model.

[0011] Further, in S2, the calculation formula of the initial suspended substance concentration value is as follows:

[0012] ;

[0013] Wherein, C init represents the initial suspended substance concentration value, k1 represents the optical scattering coefficient, k2 represents the acoustic attenuation coefficient, I scatter represents the light scattering intensity signal, a measured represents the measured acoustic attenuation value, a water represents the water background attenuation value.

[0014] Further, the calculation formula of the water background attenuation value is as follows:

[0015] ;

[0016] Wherein, b represents the base constant, β represents the temperature sensitive coefficient of acoustic wave attenuation, T represents the water temperature of the canal water body, and f represents the acoustic wave frequency of the acoustic Doppler current profiler.

[0017] Further, in S1, the environmental parameters include the water temperature, conductivity and pH value of the canal water body.

[0018] Further, in S3, according to the environmental parameters of the canal water body, the initial suspended substance concentration value is corrected by a compensation algorithm to generate the pre-calibration concentration data, including:

[0019] S31, according to the water temperature of the canal water body, the initial suspended substance concentration value is linearly compensated by the water temperature to obtain the compensated concentration data;

[0020] The calculation formula of the linear compensation of water temperature is as follows:

[0021] ;

[0022] Wherein, C temp represents the compensated concentration data, β t represents the water temperature compensation factor, T represents the water temperature of the canal water body, and T ref represents the reference water temperature.

[0023] S32, when the conductivity is greater than the preset conductivity and the pH value is less than the preset pH value, the compensated concentration data is corrected by ion interference to obtain the pre-calibration concentration data; otherwise, the compensated concentration data is directly output as the pre-calibration concentration data.

[0024] The calculation formula of the ion interference correction is as follows:

[0025] ;

[0026] ;

[0027] wherein, C cal represents the pre-calibration concentration data, K ion represents the ion interference correction factor, alpha represents the ion interference sensitivity coefficient, EC represents the conductivity of the canal water body, EC ref represents the reference conductivity.

[0028] Further, in S4, the pre-calibration concentration data is input into the pre-trained machine learning model, and the final suspended matter concentration value is output, including:

[0029] S41, acquiring turbidity time series data and three-dimensional flow velocity vector of the canal water body;

[0030] S42, determining the working condition mode based on the turbidity time series data and the three-dimensional flow velocity vector of the canal water body;

[0031] S43, according to the working condition mode, inputting the pre-calibration concentration data into the corresponding pre-trained machine learning model.

[0032] Further, in S42, the working condition mode is determined based on the turbidity time series data and the three-dimensional flow velocity vector of the canal water body, including:

[0033] determining the turbidity change slope according to the turbidity time series data, and determining the flow velocity fluctuation intensity according to the three-dimensional flow velocity vector;

[0034] if the turbidity change slope is greater than or equal to a first preset value, and / or the flow velocity fluctuation intensity is greater than or equal to a second preset value, then the disturbance working condition mode is determined;

[0035] otherwise, the normal working condition mode is determined.

[0036] Further, in S43, according to the working condition mode, the pre-calibration concentration data is input into the corresponding pre-trained machine learning model, including:

[0037] if it is the normal working condition mode, then the pre-calibration concentration data is input into the corresponding pre-trained random forest regression model;

[0038] if it is the disturbance working condition mode, then the pre-calibration concentration data is input into the corresponding pre-trained gradient boosting decision tree model.

[0039] The present application also provides a detection system for the concentration of suspended matter in canal water samples, which is used to execute the above-mentioned detection method for the concentration of suspended matter in canal water samples, and the system includes the following modules:

[0040] a data acquisition module for acquiring environmental parameters of the canal water body in real time, and synchronously collecting light scattering intensity signals and acoustic echo attenuation signals of the canal water body through a near-infrared optical sensor and an acoustic Doppler current profiler;

[0041] An initial value calculation module is configured to calculate an initial suspended matter concentration value based on the light scattering intensity signal and the acoustic echo attenuation signal;

[0042] A correction module is configured to correct the initial suspended matter concentration value by a dynamic compensation algorithm according to the environmental parameters of the canal water body to generate pre-calibration concentration data;

[0043] An output module is configured to input the pre-calibration concentration data into a pre-trained machine learning model to output a final suspended matter concentration value; wherein the pre-trained machine learning model includes a random forest regression model and a gradient boosting decision tree model.

[0044] The embodiments of the present application have the following technical effects:

[0045] The present application improves the signal analysis accuracy through the joint calculation of optical scattering and acoustic attenuation, and breaks through the limitation of single compensation by constructing a dynamic compensation chain for water temperature linear compensation and ion interference correction, and by layering correction for temperature, conductivity, pH and other multi-parameter coupling effects. Through double signal cooperation and environmental parameter hierarchical compensation, the data reliability under complex hydrological conditions is effectively improved. Based on the working condition mode recognition of turbidity time sequence and flow velocity vector, the adaptive switching of random forest regression and gradient boosting decision tree is driven to realize dynamic optimization of the model. Based on the intelligent working condition recognition of flow velocity fluctuation intensity and turbidity slope, the real-time response capability of the model in the disturbance scene such as dense navigation period and rainstorm impact is ensured. The adaptive switching mechanism of the machine learning model can significantly reduce the misjudgment risk in extreme working conditions, and provides stable technical support for long-term monitoring of canal water quality. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0047] Figure 1 is a flow chart of a method for detecting suspended matter concentration in a canal water sample provided by an embodiment of the present application;

[0048] Figure 2 is a logic diagram of a method for correcting an initial suspended matter concentration value by a dynamic compensation algorithm provided by an embodiment of the present application;

[0049] Figure 3 is a structural diagram of a detection system for suspended matter concentration in a canal water sample provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application.

[0051] The embodiment of the present application provides a detection method for the concentration of suspended solids in a canal water sample, Figure 1 is a flow chart of the detection method for the concentration of suspended solids in a canal water sample provided by the embodiment of the present application, and the method comprises the following steps: Figure 1

[0052] S1, real-time acquisition of environmental parameters of the canal water body, synchronous acquisition of light scattering intensity signals and acoustic echo attenuation signals of the canal water body by a near-infrared optical sensor and an acoustic Doppler current profiler.

[0053] In some embodiments, the environmental parameters can include water temperature, conductivity and pH value of the canal water body, etc., wherein the water temperature parameter can be measured in real time by directly contacting the water body with an immersion type temperature sensor, and the change of the water temperature will simultaneously affect the Brownian motion intensity of the suspended particles and the acoustic wave propagation attenuation characteristics; the conductivity parameter can be measured in situ by using an electrode type sensor, and the solubility ion concentration is indirectly reflected by analyzing the conductivity of the water body, and the parameter is coupled with the surface charge distribution of the suspended particles; the pH value parameter can be obtained by using a glass electrode sensor, and the hydrogen ion activity level of the water body is reflected, and the stability and aggregation form of the suspended solids colloid will be changed in the acidic or alkaline environment.

[0054] The near-infrared optical sensor is used for emitting near-infrared light and receiving scattered light generated by the suspended particles in the water body, and the light scattering intensity signal is generated. The acoustic Doppler current profiler synchronously emits ultrasonic waves into the water body, and captures the acoustic echo attenuation signal caused by the particles, so as to ensure that the two signals are accurately matched in time to avoid sampling delay problems. The data integrity is enhanced by complementing the optical and acoustic signals, thereby providing a basis for subsequent calculation.

[0055] S2, based on the light scattering intensity signal and the acoustic echo attenuation signal, the initial suspended solids concentration value is calculated.

[0056] In some embodiments, the calculation formula of the initial suspended solids concentration value is as follows:

[0057] ;

[0058] Wherein, C init represents the initial suspended solids concentration value, k1 represents the optical scattering coefficient, k2 represents the acoustic attenuation coefficient, I​scatter represents the light scattering intensity signal, a measured represents the measured acoustic attenuation value, a water represents the water background attenuation value;

[0059] The calculation formula of the water background attenuation value is as follows:

[0060] ;

[0061] wherein b represents a base constant, the base constant represents a basic sound energy attenuation level of pure water under standard temperature conditions at a specific frequency, can be determined by laboratory, β represents a temperature sensitivity coefficient of sound wave attenuation, is used to describe the blocking effect of the increase of water molecule thermal motion caused by temperature change on the energy transmission of sound wave, and the physical meaning is that the increase of molecular collision frequency caused by temperature rise increases the sound energy loss rate, T represents the water temperature of the canal water body, and f represents the frequency of the acoustic Doppler current profiler. High-frequency sound waves are easily affected by the relaxation effect of water molecules, and low-frequency sound waves are more easily scattered by bubbles. The frequency parameter realizes the targeted correction of the background attenuation value. Through dynamic modeling of the inherent acoustic characteristics of the water body, the additional attenuation caused by suspended solids and the water background absorption are accurately distinguished, the systematic interference of temperature and sound frequency on the acoustic signal is eliminated, the coupling effect of environmental variables on the acoustic signal is weakened, and the decoupling accuracy of the acoustic signal is improved.

[0062] The optical scattering coefficient k1 is calibrated in the laboratory, which represents the light scattering response intensity caused by unit concentration of suspended solids under specific water quality conditions; the acoustic attenuation coefficient k2 is also calibrated in the laboratory, which describes the absorption characteristics of suspended solids to sound energy. The water background attenuation value represents the inherent absorption of pure water to sound wave, which needs to be deducted from the measured total attenuation to avoid systematic error. This separation process can clearly distinguish the contribution of suspended solids and the acoustic interference of water molecules themselves.

[0063] The light scattering intensity signal collected in real time is multiplied by the optical scattering coefficient to obtain the light signal contribution value; at the same time, the net attenuation value of the acoustic signal is calculated, that is, the water background interference is eliminated by subtracting the water background attenuation value from the measured acoustic attenuation value, and then multiplied by the acoustic attenuation coefficient to obtain the acoustic signal contribution value; finally, the two types of signal contribution values are weighted and fused to generate the initial suspended solids concentration value. Through this dual-mode signal cooperative mechanism, when one type of signal is distorted due to environmental factors (such as optical signal saturation caused by high turbidity), the other signal can provide compensatory data support, thereby significantly improving the anti-interference ability of the initial calculation result. The optical scattering intensity signal reflects the size distribution of particulate matter, and the acoustic echo attenuation signal indicates the density characteristics of suspended solids. By weighting and fusing the two signals to output the initial concentration value, the deviation caused by relying on a single signal is avoided, so as to improve the robustness of the initial result.

[0064] S3, correcting the initial suspended matter concentration value by a dynamic compensation algorithm according to the environmental parameters of the canal water body, to generate pre-calibration concentration data.

[0065] In some embodiments, Figure 2 is a logic diagram of a method for correcting an initial suspended matter concentration value by a dynamic compensation algorithm, as shown in Figure 2 S3 includes the following sub-steps:

[0066] S31, performing water temperature linear compensation on the initial suspended matter concentration value according to the water temperature of the canal water body, to obtain compensated concentration data;

[0067] The calculation formula of the water temperature linear compensation is as follows:

[0068] ;

[0069] wherein C temp represents the compensated concentration data, β t represents a water temperature compensation factor, i.e., the response sensitivity of the suspended matter concentration to temperature change, T represents the water temperature of the canal water body, and T ref represents a reference water temperature.

[0070] Exemplarily, in a low-temperature environment, the particle agglomeration effect is intensified, the optical scattering signal is enhanced, and the acoustic attenuation is weakened. At this time, the compensation algorithm automatically adjusts the output weight according to the positive and negative values of the temperature difference, so that the compensated concentration data is closer to the real distribution state of the particulate matter.

[0071] S32, when the electrical conductivity is greater than a preset electrical conductivity and the pH value is less than a preset pH value, performing ion interference correction on the compensated concentration data to obtain pre-calibration concentration data; otherwise, directly outputting the compensated concentration data as the pre-calibration concentration data;

[0072] The calculation formula of the ion interference correction is as follows:

[0073] ;

[0074] ;

[0075] wherein C cal represents the pre-calibration concentration data, K ion represents an ion interference correction factor, and a represents an ion interference sensitive coefficient, EC represents the electrical conductivity of the canal water body, and EC ref represents a reference electrical conductivity.

[0076] Specifically, if the conductivity detection value continues to exceed the threshold value and the pH value is lower than the lower limit value, it indicates that the water body has a complex scenario of superimposed high ionic strength and acidic conditions. At this time, the ion interference sensitive coefficient is applied, which is based on the logarithmic response relationship between the ion interference sensitive coefficient to make the correction factor nonlinearly increase with the conductivity. In the calculation formula of ion interference correction, the reference conductivity represents the non-interference reference state, and the actual conductivity deviates from the reference by a certain value, which is amplified by the sensitive coefficient to generate a correction strength parameter. Finally, the temperature compensation result is multiplied by the correction factor to output the pre-calibration concentration data; if there is no complex interference condition, the compensated concentration data is directly transmitted.

[0077] The initial concentration value is corrected based on the environmental parameters, and the dynamic compensation algorithm is executed. The influence of water temperature on suspended matter concentration is reflected in that the change of water temperature may cause particles to aggregate or disperse. The compensation algorithm uses a linear model to adjust the initial value according to the measured water temperature, and the reference water temperature is set as a fixed reference point to ensure that the temperature drift is smoothly inhibited. Through the cooperative analysis of conductivity and pH parameters, it is judged whether to execute ion interference correction. When the conductivity is detected to be abnormally high or the pH is low, the algorithm automatically applies the correction factor to offset the masking effect of ion strength mutation on the signal, and generates pre-calibration concentration data. Through the hierarchical compensation mechanism, the data accuracy is improved, the environmental interference coupling problem is solved, and reliable intermediate data can be generated under normal and extreme conditions.

[0078] Specifically, water temperature fluctuations directly affect the kinetic properties of particles, and the combination of high conductivity and low pH changes the surface potential of suspended matter, causing optical scattering signal to be high and acoustic signal phase to shift. The dynamic compensation algorithm separates the independent action paths of temperature and ion interference, avoiding both insufficient single parameter compensation and overcompensation. This embodiment can significantly reduce the risk of measurement distortion under seasonal temperature fluctuations or sudden industrial pollution scenarios, such as the spring snowmelt period, when low temperature and snowmelt salt input occur simultaneously. The two-stage compensation can simultaneously suppress temperature drift and chloride ion interference, ensuring data output stability and providing high-quality input for machine learning models.

[0079] S4, input the pre-calibration concentration data into the pre-trained machine learning model to output the final suspended matter concentration value.

[0080] The pre-trained machine learning model includes a random forest regression model and a gradient boosting decision tree model. Model training is based on historical water sample data sets covering different hydrological conditions to ensure generalization ability. The historical water sample data set used for pre-training needs to cover at least the following multi-dimensional features: hydrological condition range: water temperature gradient 5-35℃ (step 5℃), conductivity range 100-1500μS / cm (including salinity mutation scenario), pH fluctuation range 6.0-8.5, turbidity dynamic range 10-500NTU, disturbance event sample: ship navigation disturbance (flow velocity fluctuation intensity ≥0.3m / s), and the like. 2 / s 2 ), storm runoff impact (turbidity change slope ≥ 15 NTU / min), simultaneously acquiring optical scattering intensity, acoustic attenuation value, three-dimensional flow velocity vector and laboratory standard weight method concentration value, not less than 50 groups of effective samples for each type of hydrological condition; the random forest regression model is good at processing stable state, and the gradient boosting decision tree model is suitable for high variability scenarios, according to different working conditions, the corresponding machine learning model is selected, the adaptability to complex water environment is enhanced, the influence of environmental factors interference is reduced, and the stability and repeatability of the canal water quality monitoring are improved.

[0081] In some embodiments, S4 comprises the following sub-steps:

[0082] S41, acquiring turbidity time series data and three-dimensional flow velocity vector of the canal water body.

[0083] Specifically, flow velocity mutation will stir up the bottom mud to increase turbidity, and abnormal change of turbidity is often accompanied by change of flow velocity mode in a specific direction. Turbidity time series data is acquired by continuous sampling of optical turbidity sensor at a fixed period, and the data sequence reflects the dynamic evolution process of water turbidity, and the change trend implies event information such as flow shock or external pollution input. Three-dimensional flow velocity vector is measured by acoustic Doppler current profiler in vertical profile layering, and the horizontal, vertical and vertical flow velocity components are obtained by beam array solution to form a spatial flow velocity distribution model.

[0084] S42, determining the working condition mode based on the turbidity time series data and the three-dimensional flow velocity vector of the canal water body.

[0085] Specifically, it comprises:

[0086] According to the turbidity time series data, the turbidity change slope is determined, and according to the three-dimensional flow velocity vector, the flow velocity fluctuation intensity is determined.

[0087] Specifically, the turbidity change slope is obtained by differential calculation of time series data, indicating the turbidity change rate per unit time, and high slope value indicates sudden disturbance events such as rainstorm scouring or ship stirring. The flow velocity fluctuation intensity is calculated by three-dimensional flow velocity vector variance, which reflects the intensity of water flow turbulence.

[0088] If the turbidity change slope is greater than or equal to the first preset value, and / or the flow velocity fluctuation intensity is greater than or equal to the second preset value, the disturbance working condition mode is determined; otherwise, the normal working condition mode is determined.

[0089] Wherein, the first preset value is a turbidity change slope threshold value, used to determine the critical value of the water body sudden disturbance event, and the second preset value is a flow velocity fluctuation intensity threshold value, used to quantify the critical value of the water flow turbulence intensity; Exemplarily, according to the turbidity time series data of the historical disturbance events (such as ship navigation, rainstorm runoff) of the canal, the distribution range of the turbidity change slope when the event occurs is calculated, and the turbidity change slope corresponding to the minimum detectable event intensity of the sudden change of the suspended matter concentration in the canal water body is selected as the first preset value; The three-dimensional flow velocity vector data of the ship navigation and rainstorm runoff is obtained by the acoustic Doppler current profiler, and the flow velocity variance is calculated, and the critical energy density corresponding to the sudden change of the water body kinetic energy can be determined by the turbulence energy spectrum analysis. The fluctuation intensity is taken as the second preset value.

[0090] That is, when the turbidity change slope is continuously in a high state, or the flow velocity fluctuation intensity shows a severe fluctuation feature, the disturbance working condition determination flag is triggered; otherwise, the normal working condition mode is maintained. The traditional method only relies on the absolute value of turbidity, which is easy to ignore the slow-changing pollution event, and the isolated flow velocity monitoring cannot distinguish natural turbulence from artificial disturbance. Exemplarily, the embodiment can trigger accurate determination by combining turbidity micro-change (such as propeller stirring bottom mud) and vertical flow velocity mutation in the ship low-speed navigation scene through the double-channel cooperative verification mechanism; When the canal water body is in the stable hydrological state of dry season, the turbidity curve is gently superimposed with low turbulence intensity, and the normal working condition flag is maintained.

[0091] S43, input the concentration data before calibration into the corresponding pre-trained machine learning model according to the working condition mode.

[0092] Specifically includes:

[0093] If it is a normal working condition mode, the concentration data before calibration is input into the corresponding pre-trained random forest regression model;

[0094] If it is a disturbance working condition mode, the concentration data before calibration is input into the corresponding pre-trained gradient boosting decision tree model.

[0095] In some embodiments, the random forest regression model is triggered in the normal working condition mode, which is generated by training historical normal hydrological data set, and its internal structure contains a parallel operation mechanism of multiple decision trees; each decision tree constructs branch rules based on optical acoustic signal characteristics, such as light scattering intensity signal interval division, acoustic attenuation value threshold judgment, etc. The model reduces the sensitivity of a single tree to noise data by integrating learning strategies to average the output results of multiple trees. When the pre-calibration concentration data is input, the model first analyzes the optical signal feature branch to determine the interval range of the light scattering intensity signal; then it enters the acoustic attenuation value branch, and combines the temperature-compensated acoustic attenuation to make a second classification; finally, the multi-path decision result is gathered to the regression output layer to generate the concentration prediction value. The model has obvious advantages in stable hydrological conditions, such as when the turbidity fluctuates gently in the canal dry season, the model can accurately capture the linear correlation characteristics of optical acoustic signals.

[0096] In some embodiments, the gradient boosting decision tree model is switched to the running state in the disturbance working condition mode. The model adopts a serialized decision tree construction method, and the post-tree continuously corrects the prediction residual of the pre-tree to form an iterative optimization chain. Disturbance scene data such as ship navigation stirring and storm runoff impact can be introduced in the model training stage, so that the decision tree learns the signal distortion compensation rules under high turbulence conditions. In addition to receiving pre-calibration concentration data, the model input also synchronously accesses real-time turbidity change slope and three-dimensional flow velocity vector direction characteristic parameters. The first decision tree generates a basic prediction value based on the turbidity change slope; the second tree corrects the optical signal distortion component for the flow velocity fluctuation intensity feature; and the subsequent trees optimize the acoustic signal phase shift error layer by layer. For example, when the vertical flow velocity component suddenly changes and causes bubble interference, the model automatically reduces the weight coefficient of the acoustic attenuation value through the bubble shielding effect rules learned from history. This chain correction mechanism is particularly critical in flood transit scenarios, and can effectively resist the blocking interference of the near-infrared light path by the wrapped debris.

[0097] The machine learning cluster in this embodiment keeps the random forest model running at low power consumption and high precision in the stable water flow condition in the dry season, and the gradient boosting decision tree model takes over the processing task in time when the disturbance occurs in the flood season. The double-model collaborative mechanism can provide adaptive technical support for the whole-cycle monitoring of the canal hydrology.

[0098] The application improves signal analysis accuracy through combined calculation of optical scattering and acoustic attenuation, performs layered correction on temperature, conductivity, pH and other multi-parameter coupling effects through the construction of a dynamic compensation chain for water temperature linear compensation and ion interference correction, breaks through the limitation of single compensation, effectively resists composite interference such as temperature change, high ion strength, acidic water body, and improves data reliability under complex hydrological conditions through double signal cooperation and environmental parameter hierarchical compensation; Based on the working condition mode recognition of turbidity time sequence and flow velocity vector, the adaptive switching of random forest regression and gradient boosting decision tree is driven to realize dynamic optimization of the model, and the intelligent working condition recognition based on flow velocity fluctuation intensity and turbidity slope ensures the real-time response ability of the model in disturbance scenes such as dense navigation period and rainstorm impact, and the adaptive switching mechanism of the machine learning model can significantly reduce the misjudgment risk of extreme working conditions, and provides stable technical support for long-term monitoring of canal water quality.

[0099] The embodiment of the application also provides a detection system for suspended matter concentration in a canal water sample, which is used for executing the detection method for suspended matter concentration in a canal water sample, Figure 3 is a structural diagram of a detection system for suspended matter concentration in a canal water sample provided by the embodiment of the application, referring to Figure 3 The system comprises the following modules:

[0100] The data acquisition module is used for acquiring the environmental parameters of the canal water body in real time, and synchronously collecting the light scattering intensity signals and acoustic echo attenuation signals of the canal water body through the near-infrared optical sensor and the acoustic Doppler current profiler.

[0101] The initial value calculation module is used for calculating the initial suspended matter concentration value based on the light scattering intensity signals and the acoustic echo attenuation signals.

[0102] The correction module is used for correcting the initial suspended matter concentration value through a dynamic compensation algorithm according to the environmental parameters of the canal water body, and generating pre-calibration concentration data.

[0103] The output module is used for inputting the pre-calibration concentration data into the pre-trained machine learning model, and outputting the final suspended matter concentration value; wherein the pre-trained machine learning model comprises a random forest regression model and a gradient boosting decision tree model.

[0104] The system embodiment corresponds to the above-mentioned method embodiment one by one, and will not be repeated here.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting the concentration of suspended solids in canal water samples, characterized in that, The method includes the following steps: S1. Real-time acquisition of environmental parameters of the canal water body: The light scattering intensity signal and acoustic echo attenuation signal of the canal water body are collected synchronously by a near-infrared optical sensor and an acoustic Doppler current profiler; Among them, the environmental parameters include the water temperature, conductivity and pH value of the canal water body. S2. Calculate the initial suspended matter concentration value based on the light scattering intensity signal and the acoustic echo attenuation signal; S3. Based on the environmental parameters of the canal water, the initial suspended solids concentration value is corrected using a dynamic compensation algorithm to generate pre-calibration concentration data; specifically including: S31. Perform linear water temperature compensation on the initial suspended solids concentration value based on the water temperature of the canal to obtain the compensated concentration data. The formula for calculating water temperature linear compensation is as follows: ； Among them, C init C represents the initial suspended solids concentration. temp This represents the concentration data after compensation, β t This represents the water temperature compensation factor, where T represents the water temperature of the canal. ref Indicates reference water temperature; S32. When the conductivity is greater than the preset conductivity and the pH value is less than the preset pH value, perform ion interference correction on the compensated concentration data to obtain the pre-calibration concentration data; otherwise, directly output the compensated concentration data as the pre-calibration concentration data. The calculation formula for ion interference correction is as follows: ； ； Among them, C cal K represents the concentration data before calibration. ion α represents the ion interference correction factor, α represents the ion interference sensitivity coefficient, and EC represents the conductivity of the canal water. ref Indicates reference conductivity; S4. Input the pre-calibration concentration data into a pre-trained machine learning model and output the final suspended matter concentration value; wherein, the pre-trained machine learning model includes a random forest regression model and a gradient boosting decision tree model.

2. The method for detecting suspended solids concentration in canal water samples according to claim 1, characterized in that, In step S2, the initial suspended solids concentration is calculated using the following formula: ； Among them, C init Indicates the initial suspended matter concentration, k1 represents the optical scattering coefficient, k2 represents the acoustic attenuation coefficient, and I... scatter α represents the light scattering intensity signal. measured α represents the measured acoustic attenuation value. water This represents the background decay value of the water body.

3. The method for detecting suspended solids concentration in canal water samples according to claim 2, characterized in that, The formula for calculating the background attenuation value of the water body is as follows: ； Where b represents the ground state constant, β represents the temperature sensitivity coefficient of sound wave attenuation, T represents the water temperature of the canal, and f represents the sound wave frequency of the acoustic Doppler current profiler.

4. The method for detecting suspended solids concentration in canal water samples according to claim 1, characterized in that, In step S4, the pre-calibration concentration data is input into a pre-trained machine learning model, and the final suspended solids concentration value is output, including: S41. Obtain time-series data of turbidity and three-dimensional velocity vector of the canal water; S42. Determine the operating mode based on the turbidity time series data of the canal water and the three-dimensional flow velocity vector; S43. Based on the operating mode, input the pre-calibration concentration data into the corresponding pre-trained machine learning model.

5. The method for detecting suspended solids concentration in canal water samples according to claim 4, characterized in that, In step S42, the operating mode is determined based on the turbidity time-series data of the canal water and the three-dimensional flow velocity vector, including: The turbidity change slope is determined based on the turbidity time series data, and the flow velocity pulsation intensity is determined based on the three-dimensional flow velocity vector. If the turbidity change slope is greater than or equal to the first preset value, and / or the flow velocity pulsation intensity is greater than or equal to the second preset value, then it is determined to be a disturbance operating condition mode. Otherwise, it will be set to the normal operating mode.

6. The method for detecting suspended solids concentration in canal water samples according to claim 5, characterized in that, In step S43, according to the operating mode, the pre-calibration concentration data is input into the corresponding pre-trained machine learning model, including: If it is a normal operating mode, the concentration data before calibration is input into the corresponding pre-trained random forest regression model; If it is a perturbation mode, the pre-calibration concentration data is input into the corresponding pre-trained gradient boosting decision tree model.

7. A system for detecting the concentration of suspended solids in canal water samples, used to perform the method for detecting the concentration of suspended solids in canal water samples as described in any one of claims 1-6, characterized in that, The system includes the following modules: The data acquisition module is used to acquire environmental parameters of the canal water in real time. It synchronously collects light scattering intensity signals and acoustic echo attenuation signals of the canal water through a near-infrared optical sensor and an acoustic Doppler current profiler. The initial value calculation module is used to calculate the initial suspended matter concentration value based on the light scattering intensity signal and the acoustic echo attenuation signal; The correction module is used to correct the initial suspended solids concentration value based on the environmental parameters of the canal water body through a dynamic compensation algorithm, and generate the concentration data before calibration. The output module is used to input the pre-calibration concentration data into a pre-trained machine learning model and output the final suspended matter concentration value; wherein, the pre-trained machine learning model includes a random forest regression model and a gradient boosting decision tree model.

Images