CN-121978045-A - Intelligent dosing control method based on water quality on-line monitoring

CN121978045ACN 121978045 ACN121978045 ACN 121978045ACN-121978045-A

Abstract

The invention belongs to the field of wastewater treatment and industrial automatic control, and particularly relates to an intelligent dosing control method based on water quality on-line monitoring. The method comprises the steps of capturing a water body spectrum fingerprint spectrum by utilizing a full spectrum scanning module, obtaining a characteristic signal sequence, introducing a chemical informatics molecular algorithm to perform pretreatment and characteristic enhancement, identifying molecular fingerprint information, constructing a convolutional neural network model to decouple chemical oxygen demand, ammonia nitrogen, total phosphorus and water body toxicity indexes from spectrum signals, realizing multi-parameter parallel inversion, calculating dosing amount by an intelligent decision algorithm based on inversion indexes and spectrum dynamic trend, and driving accurate dosing of a medicament. By fusing full spectrum scanning and deep learning, the invention realizes high-precision anti-interference water quality perception, can identify potential risks and perform preventive intervention, and improves the intelligent level and economy of dosing control.

Inventors

BAI WENLONG
WANG JING
LIN HAO
Qiao le
SUN LIRU
GAO SHENGBIN

Assignees

内蒙古东源环保科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. The intelligent dosing control method based on water quality on-line monitoring is characterized by comprising the following steps of capturing a spectrum fingerprint of a water body to be treated by utilizing a full spectrum scanning module, and acquiring transmission spectrum data of the water body within a preset wave band range through high frequency to obtain an original spectrum characteristic signal sequence containing water body substance composition information; a molecular algorithm introduced into chemical informatics carries out pretreatment and feature enhancement on the original spectrum feature signal sequence, a correlation mapping between spectrum features and molecular structure features is established, and molecular fingerprint information in a water body is identified through a multidimensional feature extraction means; Simultaneously decoupling three potential water toxicity indexes of chemical oxygen demand, ammonia nitrogen, total phosphorus and water comprehensive biotoxicity, heavy metal toxicity and degradation-resistant toxic organic matter toxicity from complex spectrum signals through a constructed convolutional neural network model, identifying potential toxic substances by analyzing residual distribution of a real-time spectrum fingerprint spectrum and a preset standard spectrum library, and quantifying the three potential water toxicity indexes by adopting a relative toxicity equivalent method, so that parallel inversion and real-time quantification of multiple water quality parameters are realized; Based on the multi-parameter water quality index obtained by inversion and the dynamic change trend of the spectrum fingerprint, calculating the target dosing amount through an intelligent decision algorithm, and driving an executing mechanism to finish dosing of the medicament.
2. The intelligent dosing control method based on water quality on-line monitoring according to claim 1, wherein in the step of capturing the spectral fingerprint of the water body to be treated by using the full spectrum scanning module, before the original spectral characteristic signal sequence is collected, the operation of light path calibration and background subtraction is performed, wherein a preset standard reference liquid is filled into a measuring pool by controlling the switching of a multi-way valve, and the standard reference liquid is measured to obtain the standard spectral signal intensity; Collecting the spectrum signal intensity of an actual water sample flowing through a measuring pool in real time; and carrying out logarithmic conversion treatment on the actual water sample spectrum signal intensity and the reference spectrum signal intensity, namely calculating a common logarithmic value of quotient obtained by dividing the reference spectrum signal intensity by the actual water sample spectrum signal intensity to obtain a corresponding absorbance value, thereby generating an absorbance spectrogram covering an ultraviolet band and a visible light band.
3. The intelligent dosing control method based on water quality on-line monitoring according to claim 1, wherein the step of preprocessing the original spectrum characteristic signal sequence comprises the steps of firstly calculating average spectrums of all sample spectrums of a current batch; For each sample spectrum, establishing a linear regression model between the sample spectrum and the average spectrum, and calculating to obtain a linear regression coefficient containing a slope term and an intercept term; correcting the sample spectrum by using the linear regression coefficient, subtracting the intercept term from the sample spectrum and dividing the intercept term by the slope term, and eliminating baseline translation and offset caused by suspended particulate scattering in the water body.
4. The intelligent dosing control method based on-line water quality monitoring according to claim 1, wherein the step of preprocessing the original spectrum characteristic signal sequence further comprises smoothing denoising, namely setting a sliding window with a fixed length on a spectrum curve, performing low-order polynomial least squares fitting on spectrum data points in the sliding window, replacing an original value of a center point of the sliding window with a fitting value, and retaining morphological characteristics of spectrum characteristic peaks while suppressing high-frequency random noise.
5. The intelligent dosing control method based on-line water quality monitoring according to claim 1, wherein the characteristic enhancement process comprises derivative transformation and frequency domain decomposition of the spectrum signal, wherein the characteristic absorption peaks overlapped in the spectrum curve are amplified by calculating the first derivative and the second derivative of the absorbance spectrum; identifying a zero crossing point and a local extreme point in the second-order differential spectrum, and locking the central wavelength position of the molecular absorption peak; Decomposing a one-dimensional spectrum signal into a plurality of frequency scale subspace components comprising approximate components and detail components by using a discrete wavelet transformation technology, extracting characteristic energy values of frequency bands, and forming a molecular fingerprint vector; and when the cosine similarity of the included angle exceeds a preset threshold, identifying the pollution source category of the current water sample and calling a corresponding characteristic weight set.
6. The intelligent dosing control method based on water quality online monitoring according to claim 1, wherein the convolutional neural network model comprises an input layer, a plurality of alternately arranged convolutional layers and pooling layers, a shared full-connection layer and a plurality of independent output branches; The input layer receives the molecular fingerprint information; the convolution layer performs local feature extraction on input vectors through a preset number of convolution cores, and captures the peak position, peak width and peak intensity features in a spectrum curve; the pooling layer reduces the feature dimension by performing a maximum downsampling operation; After the full-connection layer is shared, independent regression branches are set for different water quality indexes, wherein the inversion branch for chemical oxygen demand is configured to assign a first weight to ultraviolet band characteristics, and the inversion branch for ammonia nitrogen is configured to assign a second weight to molecular vibration frequency multiplication characteristics in the near infrared region.
7. The intelligent dosing control method based on-line water quality monitoring according to claim 1, wherein the training and verification process of the convolutional neural network model comprises training with a large-scale standard water sample data set of known concentration, defining a total loss function as a weighted sum of chemical oxygen demand prediction error, ammonia nitrogen prediction error, total phosphorus prediction error and unknown toxicity prediction error, and adjusting a network weight by minimizing the total loss function and utilizing a back propagation algorithm; connecting data verification logic behind the output layer, establishing a logic constraint relation among water quality parameters, marking the current data as invalid when the multi-parameter water quality index obtained by inversion violates the logic constraint relation, and performing deduction and filling by utilizing the valid characteristic data of the last sampling period.
8. The intelligent dosing control method based on water quality on-line monitoring according to claim 1, wherein the intelligent decision algorithm combines feedback control and feedforward compensation control, wherein the feedforward compensation control calculates basic dosing quantity according to the chemical oxygen demand water inlet load, the ammonia nitrogen water inlet load and the total phosphorus water inlet load obtained by inversion and a preset stoichiometric ratio coefficient of chemical reaction; And the feedback control acquires water quality monitoring data at the outlet of the reaction tank in real time, calculates the deviation between the measured value of the outlet water quality and the set target value, and corrects the basic dosage by utilizing a proportional-integral-differential regulation algorithm, wherein the correction is equal to the sum of the product of a proportional gain and the deviation, the product of an integral gain and the integral of the accumulation time of the deviation, and the product of a differential gain and the change rate of the deviation.
9. The intelligent dosing control method based on-line monitoring of water quality according to claim 8, wherein the process of calculating the target dosing amount further comprises real-time monitoring of a change rate of a spectral fingerprint and load prediction, wherein a change slope of a spectral absorbance integral value is continuously calculated in a preset time window, when an absolute value of the change slope exceeds a preset sudden pollution early warning threshold value and the duration exceeds a preset duration, the system is switched from a normal mode to an emergency dosing mode, and the dosing amount is increased according to a preset enhancement proportion on the basis of the basic dosing amount; and (3) performing sequence modeling on the historical flow and concentration data by using a long-short-term memory network, predicting a pollutant load peak value in a preset time period in the future, and adjusting the output frequency of the executing mechanism in advance according to the pollutant load peak value.
10. The intelligent dosing control method based on-line water quality monitoring according to claim 1, further comprising the steps of performing an automatic cleaning cycle after a preset operation period is reached, stopping sampling and closing a water inlet valve, injecting a cleaning agent with a preset concentration into a measuring tank, starting an ultrasonic generator to generate cavitation effect, driving a mechanical scraping device to perform reciprocating scraping on a measuring window, removing cleaning waste liquid, flushing for a plurality of times by using ultrapure water, and re-performing reference spectrum calibration; acquiring a rotating speed signal and a pressure sensor signal of the actuating mechanism in real time, and automatically switching to a standby pump set when detecting that the pipeline pressure is abnormal or the rotating speed deviation exceeds a preset range; And transmitting the acquired spectrum data and inversion results to the cloud end through an Internet of things gateway, periodically optimizing parameters of the convolutional neural network model by using an incremental learning mechanism, and transmitting updated model weights back to the scene.

Description

Intelligent dosing control method based on water quality on-line monitoring Technical Field The invention belongs to the field of wastewater treatment and industrial automatic control, and particularly relates to an intelligent dosing control method based on water quality on-line monitoring. Background Along with the continuous progress of water treatment technology, an intelligent dosing control system for industrial wastewater and town sewage treatment has become a key for ensuring the effluent quality to reach the standard and reducing the running cost. Traditional water quality monitoring systems are often used as a core component of a water treatment process, and the dosing frequency of dosing equipment is guided through the real-time capture of key water quality parameters. Especially in the complex and changeable water quality environment, the real-time and accurate water quality data feedback directly influences the stability of the biochemical treatment unit and the economical efficiency of medicament addition, and high requirements are provided for the monitoring precision, the anti-interference capability and the response speed of the system. The automatic dosing technology based on the on-line sensor realizes the dynamic tracking and feedback control of specific water quality indexes through integrating an electrochemical or conventional optical sampling module. The technology aims to automatically adjust the dosage of the medicament according to fluctuation of water inflow load so as to replace the traditional manual experience value dosage mode. By establishing a mapping model between water quality parameters and dosing amount, the system attempts to minimize medicament waste and reduce negative effects on subsequent processes while ensuring treatment efficiency. The traditional sensor is single in monitoring index and is extremely easy to cross interference of physical characteristics such as turbidity, chromaticity and the like of the water body, so that the accuracy of sensing data is greatly reduced. Meanwhile, the existing dosing control logic mostly belongs to post-compensation dosing, lacks predictability of potential pollution risks, and is difficult to realize preventive dosing by capturing spectral characteristic changes before water quality indexes reach a concentration threshold. In addition, the conventional analysis method is difficult to efficiently decouple a plurality of chemical indexes and potential toxic substances from complex water body signals, so that the dosing decision lacks global property, and is difficult to cope with a nonlinear and sudden water quality degradation scene. Disclosure of Invention The invention aims to provide an intelligent dosing control method based on water quality on-line monitoring, which can effectively solve the problems in the background technology. Under a complex water treatment environment, the traditional water quality monitoring sensor is extremely easy to cross interference of physical characteristics such as turbidity, chromaticity and the like of a water body, so that accuracy of perceived data is greatly reduced, monitoring indexes are often limited to single parameters, and full coverage of complex pollution components cannot be realized. Meanwhile, the existing dosing control logic mostly belongs to post-compensation dosing, lacks predictability of potential pollution risks, and is difficult to realize preventive dosing by capturing spectral characteristic changes before water quality indexes reach a concentration threshold. In addition, conventional analysis methods are difficult to efficiently decouple multiple chemical indexes and potentially toxic substances from complex water signals, resulting in a lack of global dosing decisions. The invention aims to realize accurate perception and intelligent dosing control of water quality by multi-parameter inversion of fusion spectral fingerprint spectrum and chemical informatics. In order to achieve the purpose, the intelligent dosing control method based on water quality on-line monitoring comprises the following specific steps: Capturing a spectral fingerprint of a water body to be processed by utilizing a full spectrum scanning module, and acquiring transmission spectrum data of the water body within a preset wave band range through high frequency to acquire an original spectrum characteristic signal sequence containing water body substance composition information; step 2, introducing a molecular algorithm in chemical informatics to perform pretreatment and feature enhancement on the original spectrum feature signal sequence, establishing association mapping between spectrum features and molecular structure features, and identifying molecular fingerprint information in a water body through a multidimensional feature extraction means; Step 3, simultaneously decoupling three potential water toxicity indexes of chemical oxygen demand, ammonia nitrogen, total phosphorus, comprehensive