Doc spectral monitoring method and apparatus

Abstract

The application relates to the technical field of water quality monitoring, and discloses a DOC spectrum monitoring method and device. The method comprises the following steps: pretreating a water sample through a filtering device to eliminate the interference of turbidity on spectrum detection; conveying the pretreated water sample to a detachable flow cell, and collecting the ultraviolet spectrum of the water sample through an independently arranged light source spectrometer; determining a target characteristic wavelength subset from the ultraviolet spectrum; inputting the target characteristic wavelength subset into a pre-trained integrated meta model to obtain a predicted value of the DOC concentration in the water sample; wherein the integrated meta model is constructed by adopting a stacking strategy, comprises multiple base models and a meta model, and the meta model is used for learning the prediction results of the base models and performing fusion output; and according to the state of a monitoring system or a preset maintenance period, a backwashing unit is triggered to perform a pre-cleaning operation before the flow cell is disassembled and maintained, so as to provide a technical scheme which takes into account the installation convenience, monitoring accuracy and operation stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water quality monitoring, and in particular to a DOC spectrum monitoring method and device. BACKGROUND

[0002] Under the background of accelerating the construction of smart water, as the core of the urban water supply system, the water quality monitoring of water purification plants, especially the monitoring of dissolved organic carbon (DOC), is particularly important. DOC not only relates to the safety of pipe network water quality (such as providing nutrition for bacteria reproduction to cause the growth of biofilm), but also directly affects the efficiency of water purification process (such as guiding the addition of coagulant to reduce cost). Due to the advantages of rapidity, non-destructiveness, low cost and on-line monitoring, the spectrum method has attracted widespread attention in DOC monitoring and become an important technical means for water quality monitoring in waterworks.

[0003] However, the inventors have found that the existing DOC detection technology at least has the following technical problems:

[0004] Firstly, only a few water purification plants have installed UV254 instruments capable of monitoring single wavelength to attempt to invert DOC through UV254, but this method has an essential defect: such instruments can only monitor the characteristic absorption at 254 nm wavelength, while due to the differences in molecular structure (such as macromolecular and small molecular forms) of organic matter in water, some organic matter has no significant absorption characteristics in the UV254 band, resulting in the same UV254 value but different laboratory detection results of DOC in actual measurement, which is difficult to accurately reflect the dynamic changes of water body DOC and cannot provide effective data support for the whole process regulation of smart waterworks;

[0005] Secondly, the scattering of light caused by turbidity, suspended particles and the like in the water sample at the raw water end of the waterworks will interfere with the ultraviolet-visible spectrum signal. The existing technology usually detects turbidity by measuring the visible spectrum and compensates and corrects the ultraviolet-visible spectrum signal by means of turbidity scattering; however, the inventors have found that the raw water used in waterworks is usually clean water with turbidity maintained at a low level of 3-5 NTU, and the commonly used turbidity scattering compensation method has limitations under this condition.

[0006] The above problems restrict the application effect of the spectrum method in the online monitoring of DOC in waterworks, and there is an urgent need for a technical solution that can balance monitoring accuracy and running stability. SUMMARY

[0007] An object of the present application is to provide a DOC spectrum monitoring method and device, aiming to provide a technical solution that balances installation convenience, monitoring accuracy and running stability.

[0008] To achieve the above object, some embodiments of the present application provide the following aspects:

[0009] In a first aspect, some embodiments of the present application provide a DOC spectral monitoring method, the method comprising: pretreating a water sample by a filtering device to eliminate the interference of turbidity on spectral detection; delivering the pretreated water sample to a detachable flow cell, and collecting an ultraviolet spectrum of the water sample by an independently arranged light source spectrometer; determining a target characteristic wavelength subset from the ultraviolet spectrum; inputting the target characteristic wavelength subset into a pre-trained integrated meta-model to obtain a predicted value of the DOC concentration in the water sample; wherein the integrated meta-model is constructed using a Stacking strategy, including multiple base models and a meta-model, the meta-model being used to learn the prediction results of the base models and output the results after fusion; and triggering a backwashing unit to perform a pre-cleaning operation before disassembling and maintaining the flow cell according to the monitoring system state or a preset maintenance period.

[0010] In a second aspect, some embodiments of the present application also provide a DOC spectral monitoring device, the device comprising: a pretreatment unit including a filtering device for pretreating a water sample to eliminate turbidity interference; a monitoring unit including a detachable flow cell and an independently arranged light source spectrometer, the light source spectrometer being used to collect an ultraviolet spectrum of the water sample in the flow cell; a backwashing unit for performing a pre-cleaning operation before disassembling and maintaining the flow cell; a processing unit for determining a target characteristic wavelength subset from the ultraviolet spectrum and inputting the subset into an integrated meta-model to obtain a predicted value of the DOC concentration; and a control system for triggering the backwashing unit according to the monitoring state or a preset maintenance period.

[0011] Compared with related technologies, in the scheme provided by the embodiments of the present application, the detachable flow cell is connected with the independent light source spectrometer through the quick-release optical interface, so that the device is more convenient to install, disassemble and maintain in the scene where the tap water plant needs to build a pipeline to introduce raw water, meeting the needs of the flow cell design for convenient installation and monitoring. Because the filtering device of the pretreatment unit can effectively eliminate the interference of turbidity and particulate matter on the DOC spectral monitoring, a purer water sample basis is provided for monitoring, and the monitoring accuracy is improved. At the same time, the ultrasonic cleaning transducer on the outer wall of the flow cell cooperates with the backwashing unit to clean the filtering device and the quartz light window directionally, and to remove pollutants in time, thus ensuring the long-term stable operation of the monitoring equipment, reducing the manual maintenance cost, and realizing the effective combination of convenient installation and accurate and stable monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0012] One or more embodiments are illustrated by way of example in the figures that form a part of this patent document, these example are not intended to limit the embodiments, elements having the same reference numbers in the figures indicate like elements unless otherwise expressly stated, the figures in the drawings do not constitute a proportional limitation.

[0013] Figure 1An exemplary flow chart of a DOC spectral monitoring method provided for some embodiments of the present application;

[0014] Figure 2 An exemplary diagram of effects based on prior art provided for some embodiments of the present application;

[0015] Figure 3 An exemplary diagram of effects based on the DOC spectral monitoring method and device provided for some embodiments of the present application;

[0016] Figure 4 An exemplary diagram of extracting key features using multi-dimensional strategy for spectral samples provided for some embodiments of the present application;

[0017] Figure 5 An exemplary diagram of constructing a meta-model using Stacking ensemble learning strategy based on key spectral features provided for some embodiments of the present application;

[0018] Figure 6 An exemplary diagram of effects of a single base model provided for some embodiments of the present application;

[0019] Figure 7 An exemplary diagram of effects of the meta-model provided for some embodiments of the present application;

[0020] Figure 8 An exemplary diagram of application of a DOC spectral monitoring device provided for some embodiments of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0022] First Embodiment

[0023] The first embodiment of the present application relates to a DOC spectral monitoring method. As shown in Figure 1 The method can include the following steps:

[0024] Step S101, pretreating the water sample through a filtering device to eliminate the interference of turbidity on spectral detection;

[0025] Step S102, delivering the pretreated water sample to a detachable flow cell, and collecting the ultraviolet spectrum of the water sample through an independently set light source spectrometer;

[0026] Step S103, determining a target characteristic wavelength subset from the ultraviolet spectrum;

[0027] Step S104, inputting the target characteristic wavelength subset into a pre-trained integrated meta-model to obtain a predicted value of the DOC concentration in the water sample; wherein the integrated meta-model is constructed using a Stacking strategy and includes multiple base models and a meta-model, and the meta-model is used to learn the prediction results of the base models and output the results after fusion;

[0028] Step S105, triggering the backwashing unit to perform a pre-cleaning operation before disassembling and maintaining the flow cell according to the monitoring system state or a preset maintenance period.

[0029] The above will be described in detail as follows.

[0030] For step S101, the filter device may be, but is not limited to, an MBR membrane, a stainless steel filter membrane, a ceramic filter membrane, and the like. Different filter membranes can be selected according to different use scenarios to remove interfering substances such as suspended particulate matter (e.g., silt particles), colloids, and microbial floccules in the water sample through the retention effect of the filter membrane, thereby eliminating the interference of turbidity on subsequent DOC spectrum monitoring from the source and providing a low-impurity and low-interference water sample basis for spectrum detection.

[0031] For step S102, the flow cell uses a stable structure design and optical optimization to provide a uniform light path environment for spectrum detection and avoid signal fluctuations caused by water flow disturbance or the pool body itself. The light source spectrometer is connected to the flow cell through a quick-release optical interface, which can not only ensure the convenience of equipment integration but also reduce interference between devices. For example, in actual application, the light source spectrometer can collect ultraviolet spectra in the range of 190-850 nm, which covers the characteristic absorption interval of DOC molecules and can effectively reflect the content information of DOC in the water sample.

[0032] For step S103, the target characteristic wavelength subset is determined according to the collected ultraviolet spectrum, which can achieve the purpose of eliminating redundant information. For example, considering that the 190-850 nm spectral range contains 2048 pixel points, which are prone to calculation redundancy due to excessive data volume, a PCA-GA hybrid intelligent algorithm can be used: first, compress the data dimension through PCA to reduce noise interference, and then optimize and select through a genetic algorithm to obtain the most significant wavelength points (e.g., 254 nm, 272 nm, etc.) that respond to DOC, which form the target characteristic wavelength subset. By using the PCA-GA hybrid intelligent algorithm, the defects of insufficient local search capability and unstable results of a single algorithm can be overcome, and it is ensured that the selected characteristic wavelengths can effectively reflect the change rule of DOC.

[0033] For step S104, the absorbance data of the target characteristic wavelengths 254 nm, 272 nm, etc. screened out can be input into the pre-trained integrated meta-model, and the DOC concentration in the water sample can be quickly output, realizing accurate quantitative conversion from the spectral signal to the concentration.

[0034] It can be understood that, due to the integrated meta-model provided in the embodiment, the prediction results of a plurality of base models are aggregated first, and then the optimal fusion strategy is learned through the meta-model, so that the complementarity of different models in feature extraction and pattern recognition can be fully utilized, thereby improving the accuracy of DOC concentration prediction. Especially suitable for water quality detection scenarios with complex data distribution, nonlinear correlation between features and concentration, etc. It can not only reduce the prediction fluctuation of a single model, but also compensate for the deviation of each base model through the meta-model.

[0035] For step S105, for the maintenance requirements of the flow cell, before disassembly and maintenance, the pre-cleaning operation of the backwashing unit can be triggered according to the monitoring system state feedback or the preset maintenance period. In this way, the pollutants (such as residual organic matter and small particles) on the inner wall of the flow cell and the surface of the filter device can be removed, so as to avoid the influence of residual pollutants on the performance of the equipment or the subsequent detection accuracy during the maintenance process. For example, the specific cleaning strategy can be: applying 0.1-0.3 MPa low pressure and 5-10 L / min large flow rate backwashing to the filter device to loosen the surface of the membrane; applying 0.3-0.5 MPa high pressure and 1-3 L / min small flow rate backwashing to the flow cell, and cooperating with an ultrasonic vibrator to enhance the cleaning effect. After cleaning, the effect is verified by collecting the blank spectrum of the flow cell. If the absorbance is lower than a threshold value (such as 0.02), it is determined that the cleaning is up to standard, otherwise, secondary backwashing is triggered to ensure that the equipment can maintain stable detection accuracy after maintenance.

[0036] It can be understood that, in the related art, existing water quality instruments are mostly fixed installation modes. In the scenario of introducing raw water from each node of a water plant to a monitoring point through complex pipelines, the installation process is tedious, and the pipeline needs to be interrupted for disassembly and maintenance in the later period, which seriously affects the monitoring efficiency and flexibility. The core components of the monitoring equipment are easily affected by the attachment of pollutants. The existing cleaning methods mostly rely on manual intervention, which not only has high maintenance cost, but also easily causes the equipment to run stably due to the delay in cleaning, thereby affecting the long-term monitoring accuracy.

[0037] Further, in the raw water detection scenario of a water plant, the existing technology usually adopts the method of visible segment spectrum turbidity measurement-turbidity scattering compensation to correct the ultraviolet-visible spectrum signal, so as to eliminate the light scattering interference caused by turbidity and suspended particles in the water sample. However, research shows that this method has limitations under low-concentration dissolved organic carbon (DOC) and low-turbidity water quality conditions. For example, Figure 2 As shown in Figure 2The x-axis represents wavelength (nm) and the y-axis represents absorbance (Abs). After performing two UV-Vis absorption spectral measurements on the same raw water sample, it was found that turbidity scattering only interfered with the visible spectrum. In the UV spectral region (especially the area marked by the red box in the figure), the curves of raw water spectrum one and raw water spectrum two almost overlapped, showing no significant difference. This indicates that in low-turbidity water, the effect of turbidity scattering on the UV spectrum is negligible. If the overall UV-Vis spectrum is still compensated based on changes in the visible spectrum, it will not only fail to improve detection accuracy but will also introduce additional interference, destroying the integrity of the original UV signal. Therefore, traditional turbidity scattering spectral compensation methods are ineffective under low-concentration DOC and low-turbidity water conditions, and a more suitable solution is urgently needed.

[0038] Therefore, to address the failure of traditional turbidity scattering compensation methods under low DOC concentration and low turbidity water conditions, it is necessary to develop a DOC spectral monitoring device integrating particulate matter filtration functionality. For example... Figure 3 As shown, Figure 3 The x-axis represents wavelength (nm) and the y-axis represents absorbance (Abs). The pretreatment unit in this scheme performs deep filtration on the water sample, effectively trapping suspended particulate matter and colloids. After eliminating light scattering interference, the spectral signal reaches a stable state (comparing the "spectrum before filtration" and the "spectrum after filtration," the signal stability across the entire wavelength range, especially in the ultraviolet region, is significantly improved after filtration). Based on this, by combining multi-method spectral feature extraction with a stacking strategy for multivariate model integration, not only is interference from visible spectrum correction in the ultraviolet region avoided, but accurate detection of DOC in water is also achieved, successfully overcoming the limitations of existing technologies in low signal-to-noise ratio spectral detection scenarios.

[0039] It is not difficult to see that the embodiments of this application take full-spectrum detection of DOC as the core technology. By covering the entire ultraviolet-visible spectrum, it comprehensively captures the absorption characteristics of organic matter with different structures, fundamentally breaking through the limitations of single-wavelength monitoring methods such as UV254 instruments, and becoming a key technology for smart water plants to realize real-time monitoring and precise control of DOC in various process sections.

[0040] Given the failure of traditional turbidity scattering compensation methods in low-concentration DOC and low-turbidity water, this embodiment provides a device with particulate matter filtration capabilities. This filtration device performs advanced treatment on the water sample through pretreatment unit components, efficiently trapping suspended particulate matter and colloids, eliminating light scattering interference, and significantly improving the signal-to-noise ratio of spectral detection. Figure 2 As shown, the spectral signal of the water sample processed by the device tends to be stable, successfully removing particulate matter interference and providing pure sample conditions for the full-spectrum method to accurately capture the characteristic spectrum of DOC.

[0041] In addition, the present scheme realizes synergistic effect through multi-dimensional technical innovation: the design of detachable flow cell and independent light source spectrometer not only guarantees the stability of the light path but also facilitates equipment maintenance, ensuring the reliability of full-spectrum acquisition; the strategy of screening characteristic wavelengths combined with Stacking integrated meta-model significantly improves the accuracy and robustness of low-concentration DOC detection; the automatic backwashing mechanism triggered by timing or state avoids secondary pollution and performance degradation by removing pollutants before maintenance. Finally, the deep integration of full-spectrum method and filtering device not only solves the problems of existing technologies but also promotes the DOC monitoring in intelligent water plants towards high precision, high stability, and low maintenance cost.

[0042] Second embodiment

[0043] The second embodiment of the present application relates to a DOC spectral monitoring method. The second embodiment is an improvement based on the first embodiment, and the specific improvement is that in the present embodiment, a specific implementation of determining a target characteristic wavelength subset from the ultraviolet spectrum is provided.

[0044] Specifically, the determination of the target characteristic wavelength subset from the ultraviolet spectrum, i.e., step S103, can include:

[0045] Step S1031, performing principal component analysis on the ultraviolet spectrum data to extract principal component factors with a cumulative contribution rate meeting a preset threshold;

[0046] Step S1032, dynamically adjusting the weight coefficients of the principal component factors according to the correlation between each principal component factor and the DOC concentration to form weighted principal component factors;

[0047] Step S1033, determining the target characteristic wavelength subset according to the weighted principal component factors and a target genetic algorithm.

[0048] For step S1031, for example, the covariance matrix of each wavelength point data can be calculated based on the collected original spectrum data (such as absorbance values of 2048 wavelength points in the range of 190-850 nm). The covariance matrix is used to quantify the correlation of fluctuations of different wavelength data and reflects the internal structure of the spectrum signal. For example, for 2048 wavelength points, a 2048x2048 covariance matrix can be generated, and the larger the element value in the matrix, the higher the correlation degree of the signals of the corresponding two wavelength points. Then, based on the covariance matrix, PCA operation is performed to select principal component factors with a cumulative contribution rate greater than a preset threshold (such as 85%). If the cumulative contribution rate of the first five principal component factors reaches 88%, these five factors are selected as the basis for subsequent analysis, realizing dimension reduction from 2048-dimensional original data to 5-dimensional principal components, while retaining core information and eliminating redundancy.

[0049] For step S1032, for example, the weight coefficients of the principal component factors can be dynamically adjusted according to the correlation between the principal component factors and the DOC concentration to form weighted principal component factors. The dynamic weighting mechanism provided in this step can make the subsequent feature screening more focused on the spectral region sensitive to the DOC. Taking the five principal components with a cumulative contribution rate of 88% as an example, after weighted calculation, principal component 1 obtains a higher weight due to high correlation, and the spectral features corresponding to principal component 1 will be given priority in subsequent screening, thereby improving the relevance of the feature wavelength subset to the DOC concentration.

[0050] For step S1033, for example, the weighted principal component factors can be taken as input, the population size is set to 50, the iteration number is set to 100 times, the crossover probability is set to 0.8, the mutation probability is set to 0.05, the target genetic algorithm (GA) is used to take the prediction accuracy of the feature wavelength subset to the DOC concentration as the fitness function, and the wavelength points with insignificant response to the DOC are gradually eliminated, and the sensitive wavelength points are retained. For example, after iterative optimization, the target genetic algorithm can screen out the wavelengths of 254 nm, 272 nm, 350 nm and the like that are most significant to the DOC absorption signal from 2048 wavelength points to form a target feature wavelength subset. In this step, through the combination strategy of PCA+GA, the search dimension of the target genetic algorithm can be reduced (from 2048 dimensions to 5 dimensions constrained by principal components), and the global search capability of the target genetic algorithm can be used to avoid the loss of local information caused by PCA dimension reduction, thereby effectively solving the problem that a single algorithm is prone to local optimization.

[0051] Optionally, in some embodiments, the weight coefficients of the principal component factors are dynamically adjusted according to the correlation between the principal component factors and the DOC concentration to form weighted principal component factors, i.e., step S1032 can include the following steps:

[0052] Step S10321, calculating the Pearson correlation coefficient between each principal component factor and the DOC concentration reference value to obtain a correlation quantization index;

[0053] Step S10322, dynamically generating an adaptive weight coefficient of each principal component factor through a preset nonlinear mapping function based on the correlation quantization index; wherein the principal component factor with higher correlation to the DOC concentration is assigned a larger weight;

[0054] Step S10323, weighting and fusing the original principal component factor matrix and the adaptive weight coefficient matrix to construct a weighted principal component factor.

[0055] For step S10321, for example, the linear correlation between the principal component factors and DOC concentration can be quantified using the Pearson correlation coefficient. Specifically, assuming five principal component factors (PC1-PC5) are extracted using PCA, each principal component factor is a linear combination of the original spectral data, reflecting a specific spectral characteristic pattern. By calculating the Pearson correlation coefficient between each principal component factor and the measured DOC concentration reference value, a quantitative index characterizing the correlation between the two is obtained. For example, the correlation coefficient between principal component PC1 and DOC concentration is calculated to be 0.92, indicating a strong positive correlation between PC1 and DOC concentration; the correlation coefficients for principal components PC2 are 0.85, PC3 is 0.71, PC4 is 0.53, and PC5 is 0.38; the closer the absolute value of the correlation coefficient is to 1, the stronger the characterization ability of the principal component factor for DOC concentration.

[0056] For step S10322, for example, the correlation index can be converted into adaptive weight coefficients through a preset nonlinear mapping function, wherein the higher the correlation, the greater the weight. The nonlinear mapping function can be an exponential function or a sigmoid function, and this embodiment does not specifically limit it.

[0057] In some embodiments, the following formula may be used:

[0058]

[0059] in, For the first The correlation coefficients of the principal components To adjust parameters (such as) ), The corresponding weight has a value range of (0,1). It is the natural constant (also known as the Euler number).

[0060] For example, regarding the correlation coefficient Substituting into the formula, we get ; correspond , correspond The final weight vector is [0.99, 0.98, 0.93, 0.75, 0.54], which shows that the principal components with high correlation are given greater weight.

[0061] For step S10323, for example, the original principal component factor matrix and the adaptive weight coefficients can be multiplied by matrix multiplication to achieve weighted fusion. Assume the original principal component factor matrix is... ,in, For the sample size, Principal component number, such as The weight coefficient matrix is The weighted component factor matrix WPC is calculated as follows: WPC = PC × W.

[0062] For example, in the original principal component matrix, the principal component values ​​of a certain sample are... The weight vector is: W=[0.99,0.98,0.93,0.75,0.54] T Where the superscript T is the transpose symbol, indicating that the row vector is converted into a column vector; the weighted calculation is as follows: =0.8×0.99+0.6×0.98+( )×0.93+0.2×0.75+0.1×0.54=0.792+0.588-0.279+0.15+0.054=1.305; In other words, the obtained weighted principal component factor strengthens the contribution of highly correlated principal components (such as PC1 and PC2) to the results, weakens the influence of low-correlation principal components (such as PC5), and makes the spectral features more focused on the signal strongly correlated with DOC concentration.

[0063] The target genetic algorithm can be a genetic algorithm already existing in related technologies. Optionally, in some other embodiments, determining the target feature wavelength subset based on the weighted principal component factors and the target genetic algorithm, i.e., step S1033, may include the following steps:

[0064] Step S10331: Determine the wavelength selection population according to the binary encoding rules;

[0065] Step S10332: Calculate the prediction error of each wavelength subset in the population based on the partial least squares regression method, and use the reciprocal of the prediction error as the individual fitness value;

[0066] Step S10333: Iterative evolution begins from the initial population. In each iteration, selection, crossover, and mutation operations are performed, and after selection, inferior solutions with lower fitness are accepted with a preset probability. In the iterative evolution process, a multi-objective optimization method is adopted to simultaneously pursue higher prediction accuracy and fewer features.

[0067] Step S10334: When the number of iterations reaches the preset maximum value or the optimal solution no longer improves after several consecutive generations, stop the iteration and select the target feature wavelength subset with the highest prediction accuracy and the fewest feature count from the Pareto front corresponding to the final generation population.

[0068] For step S10331, an exemplary embodiment can follow the coding rule of "one bit for one wavelength": each wavelength in the range of 190-850 nm (a total of 2048 wavelengths) corresponds to one binary bit, "1" representing the selection of the wavelength, and "0" representing the non-selection; a complete binary string (such as "100...110") represents a specific wavelength selection combination, referred to as an "individual" in the population. For example, if 190 nm corresponds to the first bit of the binary string and 200 nm corresponds to the 11th bit, then the first bit "1" in the binary string indicates the selection of 190 nm, and the 11th bit "0" indicates the non-selection of 200 nm. By randomly generating 50 such 2048-bit binary strings, an initial wavelength selection population can be formed, covering 50 different wavelength combination schemes.

[0069] For step S10332, an exemplary embodiment can parse the corresponding wavelength subset for each "individual" (binary string) in the population, train a PLSR model using the spectral data of the wavelength subset and the measured DOC concentration, and calculate the prediction error (such as the root mean square error RMSE) of the PLSR model; then take the inverse of the error as the individual fitness value - the smaller the error, the higher the fitness value, representing the better prediction performance of the wavelength subset.

[0070] For example, the wavelength subset corresponding to a certain individual is 254 nm, 278 nm, and 350 nm. After training the PLSR model using the absorbance data of these wavelengths, the prediction error RMSE is calculated to be 0.08 mg / L, and the fitness value of the individual is 1 / 0.08 = 12.5. If another individual has an RMSE of 0.1 mg / L, its fitness value is 10. It can be seen that the wavelength subset of the former is better.

[0071] For step S10333, an exemplary embodiment can achieve population iterative optimization through selection, crossover, and mutation operations, while introducing a "accepting inferior solution" mechanism and a multi-objective optimization strategy to balance search efficiency and global optimality. The specific process can include:

[0072] Selection operation: select individuals according to the fitness value from high to low (such as retaining the top 30%), as parents to participate in reproduction;

[0073] Crossover operation: randomly pair the parents with a preset crossover probability (such as 0.8), exchange part of the binary bits (such as exchanging bits 500-1000), and generate offspring individuals;

[0074] Mutation operation: randomly flip part of the binary bits of the offspring (such as "1" to "0") with a preset mutation probability (such as 0.01) to avoid the population falling into local optimum;

[0075] Accepting inferior solution: select individuals with lower fitness values with a small probability (such as 5%) after selection to increase population diversity;

[0076] Multi-objective optimization: simultaneously pursue "higher prediction accuracy" (larger fitness value) and "fewer features" (lower proportion of "1"s in binary string) in each iteration to avoid excessive reliance on redundant wavelengths.

[0077] For example, after the initial population is selected, 15 high-fitness individuals are retained, randomly paired, and then crossed at a probability of 0.8 to generate 50 offspring by flipping 1 bit at a probability of 0.01. At the same time, 2 individuals with lower fitness are retained, and the next generation population is finally formed. Through 30 iterations, the population gradually evolves towards "high accuracy + few features".

[0078] For step S10334, for example, after the termination condition is met, the optimal feature wavelength subset is selected from the final population. The termination condition can be that the number of iterations reaches a preset maximum value (e.g., 50 generations), or the fitness value of the optimal solution does not significantly improve (change < 0.1) for 10 consecutive generations. At this time, the target genetic algorithm can select the individual with the highest prediction accuracy and the fewest features from the "Pareto frontier" corresponding to the final population (i.e., the optimal solution set that cannot optimize both objectives simultaneously).

[0079] For example, after 50 iterations, the Pareto frontier contains 3 candidate solutions: solution A (accuracy 98%, feature number 8), solution B (accuracy 99%, feature number 5), and solution C (accuracy 97%, feature number 3). Among them, solution B simultaneously satisfies "highest accuracy" and "fewer features", and the binary string corresponding to solution B is analyzed. The wavelengths corresponding to "1" are 254 nm, 272 nm, 310 nm, 350 nm, and 410 nm, which form the target feature wavelength subset.

[0080] In this embodiment, by introducing the poor solution acceptance strategy of the simulated annealing mechanism into the traditional genetic algorithm iteration, the target genetic algorithm can jump out of the local optimum with a certain probability, effectively overcoming the premature convergence defect of the traditional genetic algorithm. At the same time, the multi-objective optimization framework is used to simultaneously optimize the prediction accuracy and the number of features, abandoning the limitations of the traditional single-objective genetic algorithm that requires manual setting of weights or priorities. The Pareto solution set that best balances prediction ability and model complexity is automatically found, and then the optimal feature subset is automatically output based on the iteration stopping condition, realizing the selection of a smaller number of DOC sensitive wavelength combinations with more stable prediction performance and easier actual deployment from high-dimensional spectral features.

[0081] It can be found that, in the embodiment of the application, the core factor with the strongest explanation of spectral variation is extracted through principal component analysis, and the weight of each principal component is dynamically adjusted according to the actual correlation with the DOC concentration, thereby breaking through the limitation of traditional PCA that only takes the variance contribution rate as a fixed weight; the weighting method strengthens the influence of the principal components closely related to the DOC concentration and weakens irrelevant noise interference, thereby providing a more accurate and more discriminative feature space for the target genetic algorithm search; on this basis, the target genetic algorithm can more efficiently screen out a feature wavelength subset with fewer quantities and better prediction performance, and can improve the pertinence of feature selection and the accuracy and robustness of the DOC concentration prediction model.

[0082] Third embodiment

[0083] The third embodiment of the application relates to a DOC spectrum monitoring method. The seventh embodiment is an improvement based on the first embodiment, and the specific improvement lies in that in the embodiment, a training method of the integrated meta-model is provided.

[0084] Specifically, the training method of the integrated meta-model can include:

[0085] Step S201, dividing a training data set into a training set and a validation set;

[0086] Step S202, training a plurality of different base models using the training set;

[0087] Step S203, predicting the validation set using each base model, and taking the prediction result as a new feature;

[0088] Step S204, training a meta-model by taking the new feature as input, and learning the optimal fusion strategy of the prediction result of each base model by the meta-model.

[0089] First, key features can be extracted using a multi-dimensional strategy for spectral samples. For example, as shown in FIG. 1, N spectral features significantly associated with DOC can be selected by SelectKBest combined with F test; M spectral features critical to DOC response can be selected by partial least squares (PLS) weight coefficient test; and P spectral features critical to DOC prediction can be selected by Random Forest feature importance test; then K most representative important spectral features can be selected by integrating the results of the three methods through a spectral feature fusion link to provide input for subsequent models, so that the limitations of a single method can be overcome. Figure 4 Further, as shown in FIG. 2, the feature selection process can be further improved by combining the above-mentioned multi-dimensional strategy with a genetic algorithm.

[0090] Figure 5 ​The Stacking ensemble learning process shown can be based on the screened key spectral features to build meta-models to exert the synergistic advantages of multiple models, and the specific steps are as follows:

[0091] For step S201, the original training data set can be divided into a training set (the "training sample" set containing the "training 1", "training 2", etc. identifiers in the figure is the input data of the base model training stage) and a validation set; there is also an independent test set (the "test sample" in the figure, used to verify the performance of the integrated model). When dividing, random sampling or stratified sampling according to the DOC concentration interval can be used, for example, a stratified sampling ratio of 7:3 or 8:2 is selected.

[0092] Suppose the original training data set contains 1000 samples (each sample contains the absorbance value of the key spectral features and the corresponding DOC concentration true value, i.e. the result corresponding to the "label" in the figure), and 8:2 stratified sampling is used, the samples are distributed according to the DOC concentration interval (0-1mg / L, 1-3mg / L, 3-5mg / L), 800 samples can be obtained to form the training set (composed of "training 1", "training 2", etc.), and 200 samples of the validation set (used to generate new features), to ensure that the training set and the validation set are consistent in concentration distribution and feature fluctuation trend.

[0093] For step S202, the "training sample" (samples with "training 1", "training 2", etc. identifiers) of the training set can be used to train three types of base models (such as PLS, RF, XGBoost) with diversity and complementarity, such as Figure 5 The "training sample" modules corresponding to the PLS, RF, and XGBoost on the left side are shown), which can capture different dimensional feature relationships in the data using the characteristics of each algorithm. Among them, for the PLS model, the absorbance of the key spectral features in the "training sample" can be used as the input, and the DOC concentration ("label") can be used as the output to build a linear mapping relationship; for the RF model, multiple decision trees can be integrated to learn the non-linear relationship between the spectral features and the DOC concentration; for the XGBoost model, the decision tree can be optimized based on the gradient boosting strategy, which is sensitive to local feature differences, to capture subtle non-linear correlations. Iterative optimization of the three models on the "training sample", such as adjusting the principal component number of PLS, adjusting the tree depth of RF, and adjusting the learning rate of XGBoost, can obtain three trained base models (corresponding to Figure 5 The "prediction" module results generated by the PLS, RF, and XGBoost models in the middle "training sample").

[0094] For step S203, exemplarily, the trained base models can be used to make predictions on a validation subset (a subset divided from the training set “training samples”) in the training set: the predicted values of each sample in the validation subset by the PLS model form a column vector A1, the predicted values by the RF model form a column vector A2, and the predicted values by the XGBoost model form a column vector A3; then, the three column vectors A1, A2, and A3 are transversely spliced to form a new feature matrix of “sample number x 3”, at this time, each row of the new feature matrix corresponds to a sample in the validation subset, and is composed of the corresponding numerical values of the sample in A1, A2, and A3, that is, the new feature vector of the sample, which is uniformly denoted as [A1, A2, A3].

[0095] For example, it is assumed that the validation subset includes 3 samples, namely sample 1, sample 2, and sample 3: the predicted values of the 3 samples by the PLS model are as follows: sample 1 is predicted to be A1=1.1 mg / L, sample 2 is predicted to be A1=1.3 mg / L, and sample 3 is predicted to be A1=1.2 mg / L, which form a column A1:

[0096] A1= ;

[0097] Similarly: A2= (predicted values of the 3 samples by the RF model, column vector);

[0098] A3= (predicted values of the 3 samples by the XGBoost model, column vector);

[0099] The new feature matrix obtained after splicing is as follows:

[0100] At this time, the first row of the matrix is the new feature vector [A1, A2, A3] of sample 1, which is [1.1, 1.0, 1.2], and the same applies to the other samples. That is, the column is “the prediction of all samples by a single base model”, and the row after splicing is “the prediction of a single sample by all base models”, so each row of the new feature matrix is naturally the new feature vector [A1, A2, A3] of the corresponding sample. For example, if a sample in the validation subset corresponds to 1.3 mg / L in A1, 1.4 mg / L in A2, and 1.3 mg / L in A3, then the corresponding row of the sample in the new feature matrix is its new feature vector [A1, A2, A3] = [1.3, 1.4, 1.3].

[0101] For step S204, exemplary, the new feature matrix can be input, the validation subset DOC concentration true value ("label") is output, the meta model is trained, the meta model learns how to optimally fuse each sample [A1, A2, A3] (i.e. PLS, RF, XGBoost prediction results of the sample), get fusion strategy (such as weight distribution, error correction, etc.).

[0102] As shown in Figure 5 When predicting the test set samples, first get: PLS prediction value of each sample of the test set constitutes column B1, RF prediction value constitutes column B2, and XGBoost prediction value constitutes column B3; The combination of B1, B2 and B3 values of each test sample forms a vector [B1, B2, B3], which is input into the meta model. The meta model applies the trained fusion strategy (such as the learned weighting rule based on [A1, A2, A3]) to [B1, B2, B3] to output the final ensemble prediction result. Assuming that linear regression is selected as the meta model, the new feature matrix is input, and the validation subset measured DOC concentration is output, the fusion formula is trained:

[0103] Final prediction value = 0.3xA1 (PLS prediction value in [A1, A2, A3]) + 0.4xA2 (RF prediction value in [A1, A2, A3]) + 0.3xA3 (XGBoost prediction value in [A1, A2, A3]);

[0104] That is, for test samples, the meta model will substitute [B1, B2, B3] into this logic, (such as [B1, B2, B3] = [1.5, 1.3, 1.6] for a certain test sample, the final prediction value = 0.3x1.5 + 0.4x1.3 + 0.3x1.6), to get the ensemble prediction result.

[0105] In summary, it can be determined that A1, A2, A3 are the prediction result columns of the base model on the training set verification subset (each column corresponds to the prediction value of all verification samples of the base model, which is a column vector); the A1, A2, A3 three column vectors are transversely spliced to form a new feature matrix of “sample number x 3”; each row of the new feature matrix corresponds to the numerical combination of A1, A2, A3 of a sample in the verification subset, that is, the new feature vector [A1, A2, A3] of the sample. Correspondingly, the logic of B1, B2, B3 is consistent with that of A: first as the prediction result column of the base model on the test set sample (column vector), and then the B1, B2, B3 numerical combination of each test sample forms a vector [B1, B2, B3], which is the input of the meta-model applying the fusion strategy. The meta-model first takes the new feature matrix (containing [A1, A2, A3] of all verification samples) as the input and the DOC true value of the verification subset as the output to learn the fusion rule; then the fusion rule is applied to [B1, B2, B3] of the test set, and finally the integrated prediction result is output.

[0106] Compared with a single model, in the present application, multiple base models (PLS, RF, XGBoost) are cooperated, and trained by using training samples (including “training 1”, “training 2” and the like), which can reduce the risk of overfitting of a single model to training data and improve the generalization ability; the fitting advantages of different base models to data are different (for example, PLS is good at linear correlation, and RF and XGBoost are good at nonlinear correlation), and the complementarity of A1, A2, A3 can compensate for local prediction errors after integration; the learning ability of the integrated multiple models can more fully capture the complex relationship between spectral features and DOC concentration; the fusion strategy of the meta-model (such as weighting or modification of A1, A2, A3 results) can weaken the influence of abnormal prediction of individual base model on the final result (B1, B2, B3 and the like), so that the detection result is more stable.

[0107] Referring to FIG. 1, Figure 6 The single base model effect is shown, and it can be seen that the training / test results of PLS (test R 2 = 0.9030, training R 2 = 0.9963), RF (test R 2 = 0.8537, training R 2 = 0.9802), and XGBoost (test R 2 = 0.8977, training R 2 = 0.9852) are relatively dispersed, and the fitting degree is limited; referring to FIG. 2, Figure 7 It can be seen that in the effect of the integrated model, the training R 2 = 0.9822, test R 2=0.9709, the predicted values of the training set and the test set are closer to the actual values, the fitting degree is significantly improved, and the superiority of the integration strategy is verified.

[0108] It should be noted that the present embodiment can also be improved on the basis of the second embodiment.

[0109] It can be found that, in the embodiment of the application, through the cooperation of multi-method feature extraction and the Stacking integration strategy, on the one hand, a plurality of heterogeneous base models can be used to learn diversified feature representations from training data, effectively capturing the internal laws of data at different levels; on the other hand, the prediction results of each base model on the validation set are used as meta-features to construct a more hierarchical input space to fuse the prediction advantages of different models; and then through the meta-model learning the optimal fusion strategy of these prediction results, the deviation or overfitting problem of a single model can be overcome, and the generalization ability, robustness and overall prediction accuracy of the DOC concentration prediction model can be improved. This method is not only conducive to retaining the unique advantages of each base model (such as PLS being good at linear fitting and RF being good at capturing nonlinear relationships), but also eliminates the limitations of a single model through the meta-model, achieving the purpose of high-precision prediction of the DOC concentration.

[0110] Fourth embodiment

[0111] The fourth embodiment of the application relates to a DOC spectrum monitoring method. The fourth embodiment is an improvement on the basis of the first embodiment, and the specific improvement lies in that in the present embodiment, a dynamic calibration and flow cell maintenance calibration strategy for the integrated meta-model is provided to ensure that the model still maintains high-precision prediction ability after hardware state fluctuation or maintenance.

[0112] Specifically, in some embodiments, the method can further include:

[0113] Step S301, according to the hardware state parameters monitored by the control system, and in combination with the measured values of the standard water samples, at least one of the following calibration operations is performed: according to the measured values of a single set of standard water samples, correcting the output deviation of the integrated meta-model; according to the gradient concentration standard sample set, retraining the correction curve of the integrated meta-model; when the prediction error continuously exceeds the preset threshold, triggering the reconstruction of the integrated meta-model;

[0114] Step S302, for the re-calibration after the flow cell maintenance, the standard water sample is pre-filtered by the backflushing unit, the interference factors are removed by the pretreatment unit, and the pure standard sample spectrum data are obtained for the calibration of the integrated meta-model.

[0115] For step S301, for example, the integrated meta-model can be adjusted according to the hardware state parameters monitored by the control system (such as flow cell tightness, optical interface alignment accuracy, light source stability, etc.) and the measured values of the standard water sample, to solve the model deviation problem caused by changes in water sample properties or fluctuations in hardware state. Specifically, it can include the following three ways:

[0116] When the hardware state is stable (such as no obvious mechanical looseness, optical path deviation), but the measured value of a single set of standard water sample has a fixed deviation from the model predicted value, the deviation can be eliminated by correcting the output intercept or coefficient of the integrated meta-model. For example, if the standard water sample with a concentration of 1.0 mg / L has a model predicted value of 1.2 mg / L (fixed deviation +0.2 mg / L), the modified predicted value can be made consistent with the measured value by adjusting the fusion formula of the meta-model (such as changing the original formula intercept term from 0.3 to 0.1).

[0117] When the hardware state fluctuates slightly or the water sample matrix changes gradually (such as seasonal changes causing changes in the composition of background organic matter), and the prediction error of the gradient concentration standard sample set (such as 0.5 mg / L, 1.0 mg / L, 1.5 mg / L, 2.0 mg / L) shows systematic deviation, the integrated meta-model correction curve can be retrained using the measured values and spectral data of the four gradient standard samples. For example, if the original model predicts a value of 0.65 mg / L for a 0.5 mg / L standard sample and a value of 1.8 mg / L for a 2.0 mg / L standard sample (overall high), the model parameters of the base model and the meta-model can be optimized using the measured values and spectral data of the four gradient standard samples to control the prediction error of the new correction curve to within 5%.

[0118] When the prediction error continuously exceeds the preset threshold (such as 5 consecutive errors > 20%, or a single error > 30%), and the hardware state has no significant abnormalities, it is determined that the original model cannot adapt to the current data distribution, triggering a comprehensive reconstruction of the integrated meta-model. For example, if the prediction error for a 1.0 mg / L standard sample for three consecutive days is 22%, 25%, and 28% (all exceeding the 20% threshold), the training set and validation set can be re-divided, the feature wavelength subset can be reused, and all base models and meta-models can be retrained to adapt to significant changes in water sample properties.

[0119] For step S302, for example, to address possible contamination (such as residual contaminants) or optical path deviation (such as optical alignment error after reinstallation) introduced by the detachable maintenance of the flow cell, the purity of the standard water sample spectral data can be ensured through preprocessing, providing a reliable benchmark for the calibration of the integrated meta-model.

[0120] The specific process can include: calling the backwashing unit to perform pre-filtering on the standard water sample (such as trapping small particles and colloids by the filtering device, and removing air bubbles by the defoaming device), and then further eliminating turbidity, color and other interference by the pretreatment unit; then collecting the spectrum data (such as the absorbance of characteristic wavelengths such as 254 nm and 272 nm) of the pure standard sample after processing, and using the data to calibrate the integrated element model to correct the prediction deviation caused by the hardware state after maintenance. For example, after the flow cell is reinstalled after maintenance, the baseline of the spectrum may be shifted due to the residual trace amount of organic matter in the inner wall. At this time, the standard water sample with a concentration of 1.0 mg / L is pre-filtered (the MBR membrane traps impurities with a particle size greater than 0.1 μm, and the defoaming device eliminates air bubbles in the light path), and the absorbance at 254 nm is 0.6 (0.68 without filtering, containing interference); using the pure spectrum data to calibrate the model, the original prediction value of 0.9 mg / L is corrected to 1.0 mg / L, ensuring that the model still maintains high accuracy after maintenance.

[0121] It should be noted that the present embodiment can also be improved on the basis of the second embodiment and / or the third embodiment.

[0122] It can be found that, in the embodiments of the present application, by monitoring the hardware state in real time and combining the measured value of the standard water sample, different calibration strategies such as single-point correction, curve re-fitting or model reconstruction can be adaptively selected according to different situations such as model output deviation, correction curve deviation or serious performance degradation, effectively solving the prediction misalignment problem caused by hardware drift or water quality change; further, in the calibration after maintenance, the standard water sample can be pre-filtered by the backwashing unit to eliminate the interference such as particles or air bubbles that it may contain itself, ensuring that the spectrum data used for calibration is pure and reliable, thereby significantly improving the accuracy of the calibration operation, the stability of the long-term operation of the system and the prediction reliability under different working conditions.

[0123] The step division of the above methods is only for the purpose of clear description, and when implemented, one step can be combined or some steps can be split and decomposed into multiple steps, as long as the same logical relationship is included, and all are within the protection scope of the present application; adding irrelevant modifications or introducing irrelevant designs in the algorithm or flow, but not changing the core design of the algorithm and flow, are within the protection scope of the present application.

[0124] Fifth Embodiment

[0125] The fifth embodiment of the present application relates to a DOC spectrum monitoring device. The device is used to implement the method described in any one or more of the first to fourth embodiments, and can include a pretreatment unit, a monitoring unit, a backwashing unit, a processing unit and a control system.

[0126] The pre-processing unit comprises a filtering device for pre-processing the water sample to eliminate turbidity interference.

[0127] The monitoring unit comprises a detachable flow cell and an independently arranged light source spectrometer, and the light source spectrometer is used to collect the ultraviolet spectrum of the water sample in the flow cell.

[0128] The backwashing unit is used to perform pre-cleaning operation before the flow cell is detached for maintenance.

[0129] The processing unit is used to determine a target characteristic wavelength subset from the ultraviolet spectrum and input the subset into the integrated meta-model to obtain a DOC concentration prediction value.

[0130] The control system is used to trigger the backwashing unit according to the monitoring state or a preset maintenance period.

[0131] The above units are described in detail as follows.

[0132] Specifically, the pre-processing unit is built-in with a filtering device, which is specially designed to solve the problem of turbidity and particulate matter light scattering noise covering the DOC spectrum signal in low-concentration DOC monitoring. It can be understood that in the low-DOC concentration scenario, the light scattering noise caused by suspended particles and colloids in water will far exceed the DOC spectrum signal itself, resulting in the failure of conventional turbidity compensation algorithm. In this embodiment, the filtering device can perform deep filtration on the inlet water through its precise membrane filtration function, directly intercepting particulate matter and colloids, removing turbidity interference from the source, avoiding light scattering noise from drowning the DOC spectrum signal, and being conducive to improving the accuracy of subsequent DOC spectrum detection.

[0133] For example, the membrane pore size of the filtering device can be 0.1-0.2 μm.

[0134] Specifically, the monitoring unit is composed of a detachable flow cell and an independent light source spectrometer, and realizes the dual functions through the optimization of optical design and mechanical structure.

[0135] Exemplarily, the main body of the detachable flow cell is made of stainless steel, and both ends are sealed with high-transmittance quartz glass. The transmittance of the quartz glass is greater than or equal to 90% in the wavelength range of 190-850 nm. The light path coaxiality deviation between the light source, the flow cell and the monitor is less than or equal to 0.1 mm through precise mechanical structure control, so as to reduce the influence of optical deviation on monitoring. The outer wall of the flow cell is integrated with an ultrasonic cleaning vibrator for removing contaminants on the surface of the quartz optical window. In some examples, the frequency of the vibrator is 40 kHz, the power is between 50-100 W, and the distance between the vibrator and the quartz optical window on the inner wall of the flow cell is 1-2 mm. The system can start ultrasonic cleaning at a preset period. Through high-frequency vibration, the biological membrane and bacteria are stripped to prevent them from adhering and affecting the light transmittance. The detachable structure facilitates the staff to carry out deep cleaning or maintenance, which is conducive to ensuring the long-term monitoring stability.

[0136] The light source spectrometer is detachably connected with the flow cell through a quick-release optical interface. This design is mainly aimed at the biological growth problem caused by the residual DOC in continuous monitoring. It can be understood that, since the DOC provides nutrition for bacteria, biological growth is easy to occur in the monitoring cavity during period monitoring, and the biological growth adheres to the monitoring light window, causing the ultraviolet spectrum absorption to be enhanced, which interferes with the accuracy of DOC monitoring. In the embodiment, the detachable design of the light source spectrometer and the flow cell can facilitate the staff to detach the two, so as to thoroughly maintain and clean the flow cell, which can fundamentally avoid the influence of biological adhesion on monitoring and ensure the accuracy of the results.

[0137] Exemplarily, the quick-release optical interface between the light source spectrometer and the flow cell has a sealing and waterproof function. While ensuring the stable alignment of the light path during monitoring and the non-leakage of the water path, the quick-release optical interface can realize the quick disassembly and assembly of the two, which can meet the sealing and optical stability requirements during monitoring and provide convenience for daily maintenance.

[0138] Specifically, the backwashing unit is connected with the pretreatment unit and the monitoring unit through pipelines and cooperates with the ultrasonic cleaning vibrator of the monitoring unit to realize directional cleaning of the filtering device and the quartz optical window.

[0139] Exemplarily, when the backwashing unit performs pulse flushing on the filtering device, the backwashing unit can provide a pulse pressure of 0.1-0.3 MPa through a booster pump, each flushing lasts for 10-30 seconds, and the interval is 1-2 hours for repetition, so as to efficiently remove the particulate matter and colloid intercepted on the membrane surface and maintain the filtering precision. When the ultrasonic cleaning vibrator cleans the quartz window, the backwashing unit can inject a small amount of clean water into the flow cell, which cooperates with the ultrasonic vibration to timely remove the stripped contaminants, thereby improving the cleaning effect. It can be seen that through the linkage cleaning mechanism, the backwashing unit can provide a stable operating environment for the pretreatment and monitoring links.

[0140] Specifically, the processing unit is responsible for screening a target feature wavelength subset from the ultraviolet spectrum and inputting the subset into a pre-trained ensemble meta-model to calculate and output a predicted value of the DOC concentration in the water sample through the model.

[0141] Specifically, the control system, as the intelligent core of the device, can have multiple control functions. For example, it can automatically trigger the backwashing unit to perform pre-cleaning operation according to the hardware state parameters (such as the pollution degree of the flow cell, the filtration pressure of the membrane assembly) collected by the monitoring system or the preset maintenance period, to ensure the stable operation state of the equipment; it can perform dynamic calibration operation in combination with the measured value of the standard water sample and the hardware state, including correcting the output deviation of the ensemble meta-model according to a single set of standard water samples, retraining the correction curve using a set of gradient concentration standard samples, and triggering model reconstruction when the prediction error is continuously out of limit, to ensure the monitoring accuracy; and it can also coordinate the linkage operation of the pretreatment unit, the monitoring unit, the backwashing unit and the processing unit, for example, after the flow cell is maintained, the control system controls the backwashing unit to perform pre-filtration treatment on the standard water sample to ensure the purity of the calibration data, and simultaneously schedules the processing unit to complete the model calibration process.

[0142] It can be found that, compared with the related art, in view of the actual situation that the water quality instrument of a waterworks is usually fixedly installed in a water room, and the plant needs to build a pipeline to introduce raw water of each node into each detection instrument for monitoring, the applicant creatively designs a flow cell type full-spectrum water quality monitoring instrument to facilitate installation and monitoring. Specifically, in the scheme provided in the embodiments of the present application, a detachable flow cell is used and connected with an independent light source spectrometer through a quick-release optical interface, so that the device is more convenient to install, detach and maintain in the scene where the waterworks needs to build a pipeline to introduce raw water, meeting the needs of the flow cell type design for convenient installation and monitoring; because the filtering device of the pretreatment unit can effectively eliminate the interference of turbidity and particulate matter on DOC spectrum monitoring, a purer water sample basis is provided for monitoring, and the monitoring accuracy is improved; at the same time, because the ultrasonic cleaning vibrator on the outer side wall of the flow cell cooperates with the backwashing unit, the filtering device and the quartz light window can be cleaned directionally, and pollutants can be removed in time, so that the long-term stable operation of the monitoring equipment is ensured, the artificial maintenance cost is reduced, and the effective combination of convenient installation and accurate and stable monitoring is achieved.

[0143] It can be found that the present embodiment is a device embodiment corresponding to the first embodiment, and the present embodiment can be implemented in cooperation with the first embodiment. The related technical details mentioned in the first embodiment are still valid in the present embodiment. In order to reduce repetition, they will not be repeated here. Correspondingly, the related technical details mentioned in the present embodiment can also be applied in the first embodiment.

[0144] Sixth embodiment

[0145] The sixth embodiment of the present application relates to a DOC spectrum monitoring device. The sixth embodiment is an improvement based on the fifth embodiment, and the specific improvement is that in the present embodiment, the structure of the device is further refined, as shown in Figure 8 .

[0146] Specifically, the device can include:

[0147] An inlet pipeline is provided with an inlet ball valve 101 and an inlet electromagnetic valve 103 for controlling the input of raw water, which enters the pretreatment unit after passing through the inlet ball valve 101 and the inlet electromagnetic valve 103;

[0148] A pretreatment unit includes a water storage tank 201, a filter device 205, and a defoaming device 202 connected in sequence, the inlet of the water storage tank 201 is connected to the outlet of the inlet electromagnetic valve 103, the filter device 205 is used to remove suspended solids and colloids in the water sample, and the defoaming device 202 is used to eliminate air bubbles in the water sample to avoid light scattering interference on spectrum detection;

[0149] The monitoring unit is connected to the outlet of the defoaming device 202 through the inlet of the flow cell 203, and is connected to the outlet ball valve 102, and the detected water sample can be discharged through the outlet ball valve 102; the light source spectrometer is used to collect the ultraviolet-visible absorption spectrum of the water sample in the flow cell 203;

[0150] A purified water collection tank 204 is connected to the outlet of the flow cell 203 for storing the detected purified water to provide a clean water source for backwashing;

[0151] A backwashing unit includes a water pump 303, a booster pump 304, an air pump 305, a first backwashing electromagnetic valve 105, and a second backwashing electromagnetic valve 106; the inlet of the water pump 303 is connected to the water storage tank 201, and the outlet is connected to the defoaming device 202 through a pipeline for providing power for the water sample from the water storage tank 201 to the defoaming device 202 during regular detection; the inlet of the booster pump 304 is connected to the purified water collection tank 204, and the outlet is connected to the backwashing electromagnetic valve through a backwashing pipeline, which is connected to the filter device and the flow cell respectively for performing reverse flushing; the air pump 305 is connected to the backwashing pipeline for providing air-assisted backwashing;

[0152] Specifically, during regular detection, the water pump 303 is started to transport the water pretreated by the filter device 205 in the water storage tank 201 to the flow cell 203 through the defoaming device 202 in sequence to provide power for sampling and detection; during backwashing, the booster pump 304 is started to extract purified water from the purified water collection tank 204 to provide high-pressure power for backwashing; the air pump 305 is started to inject air into the backwashing pipeline to enhance the flushing impact force through "air-water mixing".

[0153] Exemplarily, the backwashing electromagnetic valve can include a first backwashing electromagnetic valve 105 and a second backwashing electromagnetic valve 106; the first backwashing electromagnetic valve 105 is opened during backwashing to deliver the high-pressure gas-water mixed fluid to the filter device 205, so as to realize reverse washing of the filter material; the second backwashing electromagnetic valve 106 is opened during normal detection to provide a path for the water pump 303 to deliver the water sample to the defoaming device 202, and is closed during backwashing to switch the pipeline to a "backwashing dedicated state";

[0154] An ultrasonic oscillator 302 is arranged at the filter device 205, and is used for ultrasonic cleaning of the surface of the filter device 205 during cleaning;

[0155] A liquid level sensor 301 is arranged in the water storage tank 201, and is used for detecting the liquid level of the water storage tank 201, and triggering a detection process to be started when the liquid level reaches a preset threshold;

[0156] A control system is electrically connected with each electromagnetic valve, water pump, sensor and ultrasonic oscillator. Specifically, the control system can be electrically connected with the water inlet ball valve 101, the feed electromagnetic valve 103, the water outlet ball valve 102, the first backwashing electromagnetic valve 105, the second backwashing electromagnetic valve 106, the water pump 303, the booster pump 304, the air pump 305, the ultrasonic oscillator 302, the liquid level sensor 301 and the light source spectrometer, and is used for realizing the method process control as described in any one or more of the first embodiment to the fourth embodiment.

[0157] Referring to Figure 8 The device is applied to the DOC spectrum monitoring device. Specifically, the device cooperates with the water inlet pipeline, the pretreatment unit, the monitoring unit and the backwashing unit to realize the whole process cooperation of water quality pretreatment-DOC spectrum monitoring-component self-cleaning maintenance.

[0158] Exemplarily, the water inlet pipeline realizes manual and automatic double control of raw water input through the water inlet ball valve 101 and the feed electromagnetic valve 103. The water inlet ball valve 101 is used as a manual main switch of the raw water access device, manually controls the opening / closing of the water inlet path, and meets the water path management requirements of equipment maintenance, emergency water cutoff and the like. The feed electromagnetic valve 103 is used as an automatic control valve, is kept open during normal detection, so that the raw water flows into the water storage tank 201 stably; and is automatically closed during the cleaning process, so as to cut off the connection between the raw water and the pretreatment unit, and avoid interference of the raw water on the cleaning process.

[0159] Exemplarily, the pretreatment unit can include a water storage tank 201, a filtering device 205, and a liquid level sensor 301. The water storage tank 201 temporarily stores raw water delivered through a water inlet pipeline, provides a stable water source for subsequent detection processes, and ensures detection continuity. The filtering device 205 filters the raw water in the water storage tank 201 to remove impurities such as suspended solids and colloids in the water sample, thereby eliminating turbidity interference on spectral detection from the source and providing a low-impurity water sample basis for the monitoring unit. The liquid level sensor 301 is installed in the water storage tank 201 and detects the liquid level in real time. When the liquid level meets the standard, it triggers the operation of subsequent components such as the water pump 303, which is a key signal source for starting the regular detection process.

[0160] Exemplarily, the device can further include a defoaming and spectral pretreatment unit. The defoaming and spectral pretreatment unit can be a defoaming device 202 that connects the pretreatment unit and the monitoring unit to ensure the stability of the optical environment for spectral detection. The defoaming device 202 receives the water sample processed by the filtering device 205, eliminates bubbles in the water sample (bubbles can cause light scattering and seriously interfere with spectral detection accuracy), and provides a bubble-free and optically uniform water sample for the monitoring unit.

[0161] Exemplarily, the monitoring unit includes a flow cell 203, a light source spectrometer, and a flow cell cooperation, which is not separately labeled in the figure and an outlet ball valve 102. The flow cell 203: the inlet is connected to the outlet of the defoaming device 202, and is the core place for real-time spectral detection such as DOC concentration detection, providing a stable space for the light source spectrometer to collect spectra. Light source spectrometer: collects the ultraviolet-visible absorption spectrum of the water sample in the flow cell 203, providing raw data support for subsequent characteristic wavelength screening + integrated element model prediction of DOC concentration. The outlet ball valve 102 controls the passage of the water after detection, discharges the water after detection, or guides it to the clean water collection tank for standby.

[0162] Exemplarily, the device can further include a clean water recovery unit, which can be a clean water collection tank 204. The inlet of the clean water collection tank 204 is connected to the outlet of the flow cell 203, and stores the detected clean water. This achieves water resource recycling and provides a clean water source for the backwashing unit to avoid introducing new pollution during backwashing.

[0163] Exemplary, the backwashing unit can include a water pump 303, a booster pump 304, an air pump 305, a first backwashing solenoid valve 105, and a second backwashing solenoid valve 106. The water pump 303 is activated during regular detection, and the water in the water storage tank 201 that has been pretreated by the filter device 205 is sequentially delivered to the flow tank 203 through the defoaming device 202 to provide power for the monitoring process. The booster pump 304 is activated during the cleaning process, and clean water in the clean water collection tank 204 is extracted to provide a high-pressure power source for backwashing. The air pump 305 is activated during the cleaning process, and air is injected into the backwashing pipeline to enhance the cleaning effect of backwashing through the impact of "air-water mixing", and to more thoroughly clean the pipeline and the filter device 205. The first backwashing solenoid valve 105 is opened during cleaning, and cooperates with the booster pump 304 and the air pump 305 to deliver backwashing water / air to the filter device 205 and other components to perform "reverse flushing". The second backwashing solenoid valve 106 is opened during regular detection to provide a path for the water pump 303 to deliver water samples to the defoaming device 202, and is closed during cleaning to switch the pipeline state to adapt to the backwashing process.

[0164] The device can also include an auxiliary cleaning and emptying unit, which includes an ultrasonic oscillator 302 and an emptying solenoid valve 104. The ultrasonic oscillator 302 is installed at the filter device 205 and is activated during cleaning to clean the surface of the filter device 205 through high-frequency ultrasonic vibration to shake off the fine contaminants adhered to the surface, thereby ensuring the long-term filtering efficiency of the filter device 205. The emptying solenoid valve 104 is opened at the later stage of the cleaning process to empty the water stored in the water storage tank 201 and the sludge generated during cleaning, and to discharge the residual materials after the device is cleaned, thereby preparing for the next round of detection / cleaning.

[0165] The working process of the device as shown in Figure 7 may include:

[0166] 1. Regular detection process: the water inlet ball valve 101 (manually opened and kept), the feeding solenoid valve 103, and the second backwashing solenoid valve 106 are opened; the emptying solenoid valve 104 and the first backwashing solenoid valve 105 are closed. After the liquid level sensor 301 detects that the liquid level of the water storage tank 201 meets the standard, the water pump 303 is activated, the raw water is filtered by the filter device 205 and defoamed by the defoaming device 202, and then enters the flow tank 203; after the light source spectrometer collects the spectrum, the water flows into the clean water collection tank 204, and finally is discharged or stored through the water outlet ball valve 102.

[0167] 2. Cleaning process: feed electromagnetic valve 103, second backwash electromagnetic valve 106, water pump 303 are closed; emptying electromagnetic valve 104, first backwash electromagnetic valve 105 are opened. The booster pump 304 is started to extract the water in the clean water collection tank 204, and the air pump 305 is started to inject air, and the reverse flushing of the filter device 205 and other components is realized; at the same time, the ultrasonic oscillator 302 starts to clean the surface of the filter device 205; after cleaning, the emptying electromagnetic valve 104 is opened to empty the residual in the water storage tank 201.

[0168] 3. Abnormal and maintenance process: if the liquid level sensor 301 detects that the liquid level of the water storage tank 201 is abnormal, such as too high / low, the protection mechanism of the control system can be triggered; the electromagnetic valve, water pump and other components can be realized by manual operation such as closing the water inlet ball valve 101 or control system instruction, realizing the maintenance passage control of the device.

[0169] It is not difficult to find that in the embodiment of the present application, the present application precisely controls the raw water input through the water inlet ball valve 101 and the feed electromagnetic valve 103, and stores the water in the water storage tank 201 to ensure the stability of the water source; the filter device 205 in the pretreatment unit effectively removes suspended solids and colloids, and the defoaming device 202 eliminates bubble interference, thereby ensuring the accuracy of spectral detection from the source; the monitoring unit uses an independent light source spectrometer to collect the ultraviolet spectrum of the pretreated water sample, thereby ensuring data reliability; the clean water source is stored in the clean water collection tank 204, and the backwash unit realizes air-water mixed backwash through the booster pump 304 and the air pump 305, and combines the ultrasonic oscillator 302 to deeply clean the surface of the filter device 205, thereby effectively solving the problems of membrane pollution and light window attachment; the control system cooperates with the liquid level sensor 301 and other components to intelligently trigger the detection and cleaning process, thereby realizing high precision, high stability and low maintenance cost of DOC monitoring.

[0170] Seventh embodiment

[0171] The seventh embodiment of the present application relates to a DOC spectral monitoring device. The seventh embodiment is an improvement based on the fifth embodiment, and the specific improvement is that in the present embodiment, the optical path length of the flow cell is 30-50 mm.

[0172] Specifically, in the present embodiment, the optical path length of the flow cell is set to 30-50 mm, which is a key parameter precisely designed for low-concentration DOC monitoring requirements combined with the Lambert-Beer law.

[0173] It can be understood that in the scene of a water treatment plant, the DOC concentration of treated raw water is usually less than 2 mg / L, which belongs to the low-level monitoring category. According to the Lambert-Beer law (A=abc), the absorbance (A) is related to the solution concentration (c), the optical path (b) and the molar absorption coefficient (a, unit: L / mol·cm). In the present embodiment, the optical path length of the flow cell is set to 30-50 mm, which is a key parameter precisely designed for low-concentration DOC monitoring requirements combined with the Lambert-Beer law. , the absorption ability of a substance to a specific wavelength of light) is proportional to the concentration of the substance. When the concentration of DOC is extremely low, its absorption signal to light is weak, and if the optical path is too short, it will result in too small absorbance value, which is easy to be overwhelmed by instrument noise and difficult to accurately capture the spectral characteristics of DOC; while the optical path is too long, although it can enhance the absorption signal, it may introduce additional interference due to the cumulative absorption of trace impurities in water, which affects the monitoring accuracy.

[0174] Based on this, the applicant selects an optical path of 30-50 mm, which can amplify the absorbance signal of low concentration DOC by appropriately increasing the thickness of the light transmission liquid layer to meet the monitoring sensitivity requirements, and can avoid the problem of interference amplification caused by excessively long optical path, so as to balance the signal strength and monitoring stability in low-level DOC monitoring.

[0175] It should be noted that the present embodiment can also be an improvement based on the sixth embodiment.

[0176] It can be found that, in the present embodiment, the optical path length of the flow cell is set to 30-50 mm, according to the Lambert-Beer law, which can form an appropriate absorbance signal with low concentration DOC (usually less than 2 mg / L): it can avoid the problem of weak signal and easy to be overwhelmed by noise caused by too short optical path, and can prevent the interference of impurity cumulative absorption caused by too long optical path, so as to enhance the recognition degree of low concentration DOC spectral signal, provide more reliable basic data for subsequent detection, and thus improve the accuracy and stability of detection.

[0177] Eighth Embodiment

[0178] The eighth embodiment of the present application relates to a DOC spectral monitoring device. The eighth embodiment is an improvement based on the fifth embodiment, and the specific improvement is that in the present embodiment, a specific implementation mode of the control system is provided.

[0179] Specifically, the control system can include:

[0180] The state monitoring module is used for real-time monitoring of the connection state of the flow cell and the sealing property of the optical interface.

[0181] The maintenance control module is used for automatically starting and stopping the monitoring process according to the monitoring state, and triggering the calibration program after maintenance.

[0182] The pollution identification module is used for identifying the pollution degree of the MBR membrane and the optical window.

[0183] The cleaning control module is used for generating cleaning instructions according to the pollution degree, and linkage control of the backwashing unit and the ultrasonic vibrator to perform cleaning operation.

[0184] Specifically, the state monitoring module is configured to monitor the connection stability (e.g., whether the mechanical connection is loose or misaligned) of the detachable flow cell and the quick-release optical interface and the sealing performance (e.g., whether there is water leakage or air leakage that may interfere with the optical path) of the optical interface in real time.

[0185] Specifically, the maintenance control module performs specific control actions according to the monitoring results of the state monitoring module. When the state monitoring module detects that the flow cell or the optical interface is disconnected, the maintenance control module can immediately suspend the monitoring process and activate the maintenance mode to avoid the device running under abnormal connection conditions and reduce the risk of failure. When the state monitoring module identifies that the flow cell is reinstalled in place and the optical interface connection is restored, the maintenance control module will automatically restart the monitoring process and trigger a calibration program (e.g., verify and correct the integrated element model with a standard water sample) to ensure that the monitoring accuracy after maintenance is not affected. The state monitoring module and the maintenance control module cooperate with each other to achieve full-process monitoring and intelligent response of the connection state of key components, providing reliable protection for the stable operation and detection accuracy of the device.

[0186] Specifically, the pollution identification module is configured to identify the pollution degree of the filter device and / or the quartz optical window of the flow cell according to a preset period, sensor feedback, or manual instructions. For example, the pollution identification module can obtain pollution information in three ways: (1) periodically evaluate the pollution state according to a preset period (e.g., based on the basic period of starting the ultrasonic cleaning vibrator every 2 hours); (2) capture pollution changes in real time by receiving sensor feedback data (e.g., MBR membrane pollution data monitored by the membrane pollution degree sensor); and (3) respond to manual instructions to meet the cleaning needs of manual intervention. By comprehensively analyzing these information, the pollution identification module can clearly distinguish whether the target cleaning object is the filter device, the flow cell, or both.

[0187] Specifically, the cleaning control module can include a cleaning instruction generation module and a cleaning execution control module. The cleaning instruction generation module is configured to generate a cleaning instruction according to the pollution degree. The cleaning instruction can include the target cleaning object and the corresponding pressure parameter and flow parameter. The cleaning execution control module is configured to control the backwashing unit and the ultrasonic cleaning vibrator to perform directional cleaning according to the target cleaning object and the corresponding pressure parameter and flow parameter in the cleaning instruction. When the target cleaning object is the filter device, the backwashing unit is applied to the filter device with the corresponding pressure parameter and flow parameter. When the target is the flow cell, the ultrasonic cleaning vibrator is activated, and the auxiliary cleaning water flow with the corresponding pressure parameter and flow parameter is applied to the flow cell.

[0188] Specifically, the cleaning instruction generation module is configured to generate specific executable cleaning instructions based on the pollution degree determined by the pollution identification module and the target cleaning object. The cleaning instructions include two major elements: one is the specific target cleaning object (filtration device or flow cell); the other is the pressure parameter and flow parameter matched with the target cleaning object. For example, according to the preset program, the filtration device cleaning corresponds to “low pressure and large flow” (such as 0.1-0.3 MPa pressure and 5-10 L / min flow), and the flow cell cleaning corresponds to “high pressure and small flow” (such as 0.3-0.5 MPa pressure and 1-3 L / min flow). By determining these parameters, the cleaning effect and equipment protection can be considered, and at the same time, the cleaning intensity suitable for different components can be ensured.

[0189] Specifically, the cleaning execution control module is configured to convert the cleaning instructions into specific hardware operations, and link the backwashing unit and the ultrasonic cleaning vibrator to realize directional cleaning.

[0190] When the target cleaning object is a filtration device, the cleaning execution control module can start the booster pump and the air compressor, and apply backwashing water flow and auxiliary gas to the filtration device according to the pressure parameter and flow parameter in the cleaning instructions, so as to efficiently remove the pollutants trapped on the membrane surface.

[0191] When the target cleaning object is a flow cell, the cleaning execution control module can first activate the ultrasonic cleaning vibrator of the outer wall of the flow cell (according to the preset period, such as 30 seconds each time), and control the booster pump to apply auxiliary cleaning water flow with corresponding pressure parameter and flow parameter to the flow cell, so as to cooperate with the ultrasonic vibration to completely strip and remove the pollutants on the surface of the quartz optical window.

[0192] Optionally, in some embodiments, the control system can further include:

[0193] The model management module is configured to monitor the prediction performance of the integrated meta-model, and trigger model updating when the performance decreases;

[0194] The adaptive calibration module is configured to automatically adjust the parameters of the integrated meta-model according to the changes of hardware state and water quality characteristics.

[0195] Specifically, the model management module can manage the full life cycle of the integrated meta-model and continuously monitor its predictive performance in actual applications. For example, the model prediction value can be compared with the actual detection data in real time to calculate indicators such as prediction error and coefficient of determination (R2) to determine whether the model performance has decreased. Once it is found that the prediction error continuously exceeds the preset threshold or the accuracy of the model on the test set continuously decreases, the module will immediately trigger the model updating mechanism. For example, a new training data set is automatically called to retrain the integrated meta-model, or a transfer learning strategy is used to optimize the model parameters to ensure that the model always maintains high accuracy and stability and adapts to the changing monitoring environment.

[0196] Specifically, when the hardware state changes (such as light path deviation caused by aging of the flow cell or performance degradation of the light source spectrometer) or the water quality characteristics fluctuate (such as sudden increase in raw water turbidity or change in organic matter composition), the adaptive calibration module can quickly capture the change information and optimize the parameters of the integrated meta-model based on a preset algorithm. For example, for the problem of hardware light path deviation, the weight coefficient related to the spectral feature wavelength in the model is adjusted; for the change in water turbidity, the compensation parameter of the model for background noise is optimized. Through this adaptive calibration mechanism, the integrated meta-model can quickly adapt to external environmental changes and continuously maintain high-precision DOC concentration prediction capability.

[0197] It should be noted that the embodiment can also be improved on the basis of the sixth and / or seventh embodiments.

[0198] It can be found that, in the embodiment, the control system is provided with the state monitoring module and the maintenance control module. The state monitoring module can monitor the connection state of the detachable flow cell and the quick-release optical interface in real time, and the maintenance control module can pause the monitoring process and activate the maintenance mode when the connection is disconnected according to the monitoring result, and automatically restart the monitoring process and trigger the calibration check after the connection is restored. Therefore, detection errors or faults caused by running of the equipment in an abnormal connection state can be avoided, the detection accuracy after maintenance is ensured, and the stability of the device operation and the reliability of the detection result are ensured.

[0199] It is worth mentioning that each module involved in each of the above embodiments is a logical module. In actual application, one logical unit can be one physical unit, or a part of one physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of the present application, units not closely related to solving the technical problems proposed in the present application are not introduced in the embodiment, but this does not mean that there are no other units in the embodiment.

[0200] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented by the apparatus and methods according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0201] The scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to encompass all variations falling within the meaning and scope of equivalents of the claims. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device in software or hardware. Terms such as "first," "second," etc., are used only for distinguishing descriptions and do not indicate any particular order, nor should they be construed as indicating or implying relative importance.

[0202] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily made by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims, and the above embodiments should be regarded as exemplary and non-limiting.

Claims

1. A method of DOC spectral monitoring, characterized in that, The method comprises: The water sample is pretreated by a filtering device to eliminate the interference of turbidity on spectral detection; The pretreated water sample is delivered to a detachable flow cell, and the ultraviolet spectrum of the water sample is collected by an independently arranged light source spectrometer; A target characteristic wavelength subset is determined from the ultraviolet spectrum; The target characteristic wavelength subset is input into a pre-trained integrated meta-model to obtain a predicted value of the DOC concentration in the water sample; wherein the integrated meta-model is constructed using a Stacking strategy and comprises multiple base models and a meta-model, and the meta-model is used to learn the prediction results of the base models and output the results after fusion; According to the state of the monitoring system or a preset maintenance period, a backwashing unit is triggered to perform a pre-cleaning operation before the flow cell is detached and maintained; The target characteristic wavelength subset is determined from the ultraviolet spectrum, which comprises: performing principal component analysis on the ultraviolet spectrum data to extract principal component factors with a cumulative contribution rate meeting a preset threshold; dynamically adjusting the weight coefficients of the principal component factors according to the correlation between each principal component factor and the DOC concentration to form weighted principal component factors; the weighted principal component factors are used to enhance the contribution degree of the components in the spectral characteristics that are closely related to the DOC concentration; and the target characteristic wavelength subset is determined according to the weighted principal component factors and a target genetic algorithm; The target characteristic wavelength subset is determined according to the weighted principal component factors and the target genetic algorithm, which comprises: determining a wavelength selection population according to a binary coding rule; calculating the prediction error of each wavelength subset in the population based on a partial least squares regression method, and taking the inverse of the prediction error as an individual fitness value; starting from an initial population to perform iterative evolution, performing selection, crossover and mutation operations in each iteration process, and accepting a poor solution with a worse fitness after selection with a preset probability; wherein, in the iterative evolution process, a multi-objective optimization method is used to simultaneously pursue higher prediction accuracy and fewer feature quantities; when the number of iterations reaches a preset maximum value or the optimal solution does not improve continuously for multiple generations, the iteration is stopped, and the target characteristic wavelength subset with the highest prediction accuracy and the fewest feature quantities is selected from the final generation population corresponding to the Pareto frontier; The target characteristic wavelength subset is determined according to the weighted principal component factors and the target genetic algorithm, and the weighted principal component factors are used as inputs.

2. The method of claim 1, wherein, The weight coefficients of the principal component factors are dynamically adjusted according to the correlation between each principal component factor and the DOC concentration to form weighted principal component factors, which comprises: The Pearson correlation coefficients between each principal component factor and the DOC concentration reference value are calculated to obtain a correlation quantization index; Based on the correlation quantization index, adaptive weight coefficients of each principal component factor are dynamically generated by a preset nonlinear mapping function; wherein the principal component factor with a higher correlation with the DOC concentration is assigned a larger weight; The original principal component factor matrix is weighted and fused with the adaptive weight coefficient matrix to construct a weighted principal component factor.

3. The method of claim 1, wherein, The training method of the integrated meta-model comprises: The training data set is divided into a training set and a validation set; The training set is used to train multiple different base models; The base models are used to predict the validation set, and the prediction results are used as new features; The meta-model is trained by taking the new features as input, and the optimal fusion strategy of the prediction results of the base models is learned through the meta-model.

4. The method according to any one of claims 1 to 3, characterized in that, The method further comprises: According to the hardware state parameters monitored by the control system, at least one of the following calibration operations is performed in combination with the measured values of the standard water samples: according to the measured values of a single set of standard water samples, the output deviation of the integrated meta-model is corrected; according to the gradient concentration standard sample set, the correction curve of the integrated meta-model is retrained; when the prediction error continuously exceeds the preset threshold, the reconstruction of the integrated meta-model is triggered; For the recalibration after the circulation cell is maintained, the standard water samples are pre-filtered by the backwashing unit, the interference factors are removed by the pretreatment unit, and the pure standard sample spectrum data are obtained for the calibration of the integrated meta-model.

5. A DOC spectral monitoring device, characterized in that The device is used to implement the method of any one of claims 1-4, and the device comprises: The pretreatment unit comprises a filtering device for pretreating the water samples to eliminate turbidity interference; The monitoring unit comprises a detachable circulation cell and an independently arranged light source spectrometer, and the light source spectrometer is used to collect the ultraviolet spectrum of the water samples in the circulation cell; The backwashing unit is used to perform pre-cleaning operation before the circulation cell is detached and maintained; The processing unit is used to determine a target feature wavelength subset from the ultraviolet spectrum and input the subset into the integrated meta-model to obtain a DOC concentration prediction value; The control system is used to trigger the backwashing unit according to the monitoring state or the preset maintenance period; The determination of the target feature wavelength subset from the ultraviolet spectrum comprises: performing principal component analysis on the ultraviolet spectrum data to extract principal component factors with a cumulative contribution rate meeting a preset threshold; according to the correlation between each principal component factor and the DOC concentration, the weight coefficient of the principal component factor is dynamically adjusted to form a weighted principal component factor; the weighted principal component factor is used to enhance the contribution degree of the components closely related to the DOC concentration in the spectral features; and the target feature wavelength subset is determined according to the weighted principal component factor and a target genetic algorithm; The determination of the target feature wavelength subset according to the weighted principal component factor and the target genetic algorithm comprises: determining a wavelength selection population according to a binary coding rule; calculating the prediction error of each wavelength subset in the population based on a partial least squares regression method, and taking the inverse of the prediction error as an individual fitness value; starting from an initial population, iterative evolution is performed, selection, crossover and mutation operations are performed in each iteration process, and a poor solution with worse fitness is accepted after selection with a preset probability; wherein, in the iterative evolution process, a multi-objective optimization method is used to simultaneously pursue higher prediction accuracy and fewer feature quantities; when the iteration number reaches a preset maximum value or the optimal solution does not improve continuously for multiple generations, the iteration is stopped and the target feature wavelength subset with the highest prediction accuracy and the fewest feature quantities is selected from the Pareto front corresponding to the final generation population; The device comprises:

6. The apparatus of claim 5, wherein, ​ Water inlet pipeline, provided with a water inlet ball valve and a feed electromagnetic valve, for controlling raw water input; The pretreatment unit comprises a water storage pool, a filter device and a defoaming device connected in sequence, the inlet of the water storage pool is connected with the outlet of the feed electromagnetic valve, the filter device is used for removing suspended solids and colloids in the water sample, and the defoaming device is used for eliminating air bubbles in the water sample; The monitoring unit is connected with the outlet of the defoaming device through the inlet of the flow cell, and is connected with a water outlet ball valve through the outlet, and the light source spectrometer is used for collecting the ultraviolet spectrum of the water sample in the flow cell; The purified water collecting pool is connected with the outlet of the flow cell through the inlet, and is used for storing the purified water after detection; The backwashing unit comprises a water pump, a booster pump and an air pump, the inlet of the water pump is connected with the water storage pool, the outlet is connected with the defoaming device through a pipeline, and the water pump is used for providing power for the water sample from the water storage pool to the defoaming device during normal detection; the inlet of the booster pump is connected with the purified water collecting pool, the outlet is connected with a backwashing electromagnetic valve through a backwashing pipeline, and the backwashing electromagnetic valve is connected with the filter device and the flow cell respectively, and is used for performing reverse flushing; the air pump is connected with the backwashing pipeline, and is used for providing air-assisted backwashing; The ultrasonic oscillator is arranged at the filter device, and is used for ultrasonic cleaning of the surface of the filter device during cleaning; The liquid level sensor is arranged in the water storage pool, and is used for detecting the liquid level and triggering the detection process; The control system is electrically connected with the electromagnetic valves, the water pump, the sensor and the ultrasonic oscillator.

7. The apparatus of claim 6, wherein, The control system comprises: The state monitoring module is used for monitoring the connection state of the flow cell and the sealing property of the optical interface in real time; The maintenance control module is used for automatically starting and stopping the monitoring process according to the monitoring state, and triggering the calibration program after maintenance; The pollution identification module is used for identifying the pollution degree of the MBR membrane and the optical window; The cleaning control module is used for generating a cleaning instruction according to the pollution degree, and is used for linkage control of the backwashing unit and the ultrasonic oscillator to perform cleaning operation.

8. The apparatus of claim 7, wherein, The control system further comprises: The model management module is used for monitoring the prediction performance of the integrated meta-model, and triggering model updating when the performance decreases; The adaptive calibration module is used for automatically adjusting the parameters of the integrated meta-model according to the change of hardware state and the change of water quality characteristics.

Images