CN-121974455-A - Sediment prediction and mud discharge control method for sedimentation tank driven by visual monitoring and model

CN121974455ACN 121974455 ACN121974455 ACN 121974455ACN-121974455-A

Abstract

Aiming at the problems of inaccurate sediment monitoring, rough sediment discharge control, low dosing precision and the like of the traditional sedimentation tank, the invention provides a sediment prediction and sediment discharge control method and system driven by a visual monitoring and model. The method comprises the steps of visual monitoring, collecting sediment pool images through an underwater camera device, identifying sediment interface characteristics, constructing a sediment thickness distribution thermodynamic diagram by combining an inverse distance weighted interpolation method, collecting parameters, acquiring running parameters such as inflow water flow, raw water turbidity and pH value in real time, carrying out model driving prediction, establishing a coupled multivariable BP neural network model, a sediment yield model and a multidimensional coupled sediment distribution prediction model, carrying out intelligent sediment discharge control, generating optimized instructions of sediment discharge frequency, quantity and area based on prediction results, and controlling sediment discharge operation. The invention realizes automatic sediment monitoring and sediment discharge control refinement, improves the precipitation efficiency and reduces the operation cost.

Inventors

ZENG WU
LIU HAIHUA
SHI ZHAN
CAI ZHENYIN
WANG XUAN
ZHAO YAN
HE JUNGUO

Assignees

江苏江南水务股份有限公司
广州大学

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (8)

1. A method for predicting sediment and controlling sediment discharge of a sedimentation tank driven by visual monitoring and models is characterized by comprising the following steps: the method comprises the steps of collecting visual images of sedimentation areas in a sedimentation tank in real time through an image collecting and analyzing module, identifying substrate sludge interface characteristics, obtaining real-time substrate sludge thicknesses of a plurality of measurement areas, and constructing a substrate sludge thickness distribution thermodynamic diagram; collecting the inflow water flow, the raw water turbidity, the pH value, the water temperature and the actual coagulant addition amount in real time through an operation parameter collecting module; through the data processing and modeling unit, based on the collected operation parameters and image data, the following models are built and operated: Coupling a multi-variable BP neural network model, and predicting the optimal coagulant addition amount; The mud production model dynamically estimates the total mud production amount based on the turbidity of raw water, the dosage and the inflow water flow; The multidimensional coupling sediment distribution prediction model outputs sediment thickness distribution thermodynamic diagrams at future time points; and generating an optimal sludge discharge control instruction according to the predicted sediment distribution thermodynamic diagram through the intelligent sludge discharge control module, and controlling sludge discharge operation.
2. The method of claim 1, wherein the image acquisition and analysis module comprises: At least one high resolution underwater camera device; a light source system; an image processing unit; A graduated scale or a preset area arranged in the sedimentation tank.
3. The method according to claim 1, wherein the sediment thickness distribution thermodynamic diagram is constructed by a spatial interpolation method, the spatial interpolation method is an inverse distance weighted interpolation method, and the calculation formula is: ; Wherein, the For the predicted bed mud thickness for the target grid points, The bottom mud thickness of the ith measurement point, For the distance between the grid point and the measurement point, The preset value is 2 for the distance weight power.
4. The method according to claim 1, wherein the input variables of the BP neural network model comprise real-time raw water turbidity, pH value, water temperature and inflow water flow, and the output is the optimal coagulant addition amount; The model is trained through historical data, input data is preprocessed through a maximum and minimum normalization method in the training process, a hidden layer uses a Sigmoid activation function, an output layer uses an identity function, a loss function is Mean Square Error (MSE), and an optimization algorithm is a gradient descent method.
5. The method of claim 1, wherein the mud production model is calculated by the formula: ; Wherein, the In order to produce the total amount of mud, In order to achieve the water inflow rate, For the turbidity of the raw water, In order to administer the amount of drug to be administered, 、 In order for the conversion coefficient to be a function of, For the density of the coagulant, For mud production coefficient, the mud production coefficient is calibrated regularly through actual measurement.
6. The method of claim 1, wherein the input variables of the multidimensional coupled sediment distribution prediction model comprise: accumulating the treated water quantity or the average inflow water quantity; Average raw water turbidity and other water quality parameters; Predicting or averaging the dosage; A floc particle size distribution spectrum; Information of a hydraulic flow velocity field in the sedimentation tank; time variable.
7. A system for implementing the method of any one of claims 1 to 6, comprising: the image acquisition and analysis module acquires visual images of sedimentation areas in the sedimentation tank in real time, identifies the interface characteristics of the sediment, acquires the real-time sediment thickness of a plurality of measurement areas, and constructs a sediment thickness distribution thermodynamic diagram; The operation parameter acquisition module is used for acquiring the water inlet flow, the raw water turbidity, the pH value, the water temperature and the actual coagulant addition amount in real time; The data processing and modeling unit is used for establishing and operating a BP neural network model, a mud yield model and a multidimensional coupling sediment distribution prediction model which are coupled with multiple variables based on the acquired operation parameters and image data; the intelligent sludge discharge control module generates an optimal sludge discharge control instruction according to the predicted sediment distribution thermodynamic diagram and controls sludge discharge operation; all modules are connected through communication, so that data real-time transmission and control instruction issuing are realized.
8. The system of claim 7, wherein the intelligent sludge discharge control module automatically generates optimized control instructions for sludge discharge frequency, sludge discharge amount and sludge discharge area according to the predicted sludge distribution thermodynamic diagram, and sends the optimized control instructions to the sludge discharge equipment for execution.

Description

Sediment prediction and mud discharge control method for sedimentation tank driven by visual monitoring and model Technical Field The invention belongs to the field of water treatment, and relates to a method for periodically monitoring and predicting trend of sediment distribution prediction of a advection sedimentation tank by utilizing a visual monitoring and data model and a mud discharge control method. Background The problems of inaccuracy and non-comprehensiveness of sediment distribution monitoring exist in the existing advection sedimentation tank management and sediment discharge control technology. (1) The traditional method relies on manual periodic measurement or simple experience estimation, is time-consuming and labor-consuming and low in precision, and cannot acquire the flocculating constituent settlement characteristics and the sediment thickness/distribution condition in the sedimentation tank in real time, so that the dynamic change of the sediment is reflected. (2) The mud discharging operation of the existing water plant often depends on empirical judgment, timing mud discharging or simple feedback of single parameters (such as out-water turbidity exceeding standard). The control mode lacks accurate prediction of sediment distribution areas and sediment amount, causes lag and rough sediment discharge operation, is easy to cause incomplete sediment discharge, causes excessive sediment accumulation in local areas, influences sediment efficiency and effluent quality, or excessively discharges the sediment, wastes a large amount of clean water resources and increases the running cost of water plants. (3) The traditional dosing control is mostly based on a simple turbidity-dosage relationship or delayed effluent quality feedback, so that the accurate dosing under multivariable coupling cannot be realized, and the flocculation effect is affected. The traditional dosing control can not realize accurate dosing under multivariable coupling, and the dosing amount, the mud amount and the hydraulic sedimentation process can not form an optimized closed loop. In conclusion, the invention aims to solve the key problems of difficult acquisition of sediment distribution information of the sedimentation tank, lack of scientific basis for mud discharge control decision and rough process management in the prior art, so as to realize fine and intelligent management of the advection sedimentation tank. Disclosure of Invention In order to solve the problems in the background technology, the invention provides a method for predicting sediment and controlling the sediment discharge of a sedimentation tank driven by visual monitoring and a model. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a sediment prediction and sediment discharge control method for a sedimentation tank driven by visual monitoring and models comprises the following steps: the method comprises the steps of collecting visual images of sedimentation areas in a sedimentation tank in real time through an image collecting and analyzing module, identifying substrate sludge interface characteristics, obtaining real-time substrate sludge thicknesses of a plurality of measurement areas, and constructing a substrate sludge thickness distribution thermodynamic diagram; collecting the inflow water flow, the raw water turbidity, the pH value, the water temperature and the actual coagulant addition amount in real time through an operation parameter collecting module; through the data processing and modeling unit, based on the collected operation parameters and image data, the following models are built and operated: Coupling a multi-variable BP neural network model, and predicting the optimal coagulant addition amount; The mud production model dynamically estimates the total mud production amount based on the turbidity of raw water, the dosage and the inflow water flow; The multidimensional coupling sediment distribution prediction model outputs sediment thickness distribution thermodynamic diagrams at future time points; and generating an optimal sludge discharge control instruction according to the predicted sediment distribution thermodynamic diagram through the intelligent sludge discharge control module, and controlling sludge discharge operation. Further, the image acquisition and analysis module includes: At least one high resolution underwater camera device; a light source system; an image processing unit; A graduated scale or a preset area arranged in the sedimentation tank. Further, the sediment thickness distribution thermodynamic diagram is constructed by a spatial interpolation method, wherein the spatial interpolation method is an inverse distance weighted interpolation method, and the calculation formula is as follows: ; Wherein, the For the predicted bed mud thickness for the target grid points,The bottom mud thickness of the ith measurement point,For the distance between the grid po