System and method for predicting adding amount of potassium permanganate

Abstract

The invention discloses a system and a method for predicting the adding amount of potassium permanganate, and relates to a technology for predicting the content of divalent manganese ions in water and calculating the adding amount of potassium permanganate by using oxidation-reduction potential, which is particularly suitable for a manganese ion removal process in water treatment. According to the method, the back propagation neural network model is adopted, due to the superiority of the back propagation neural network model in nonlinear mapping and real-time prediction, the relationship between the oxidation-reduction potential and the concentration of the divalent manganese ions can be accurately established, the water treatment process can be optimized, the prediction precision can be improved, the waste of chemical reagents can be reduced, the water treatment cost can be reduced, and the water treatment effect can be improved.

Description

Technical Field

[0001] The invention relates to the technical field of wastewater treatment, and in particular to a system and method for predicting the dosage of potassium permanganate, which is particularly suitable for a manganese ion removal process in water treatment. Background Art

[0002] Manganese ions are one of the common pollutants in water treatment. Excessive divalent manganese ions can have adverse effects on water quality and even cause harm to human health. In traditional water treatment, potassium permanganate is usually added to oxidize and remove divalent manganese ions in water, but how to accurately calculate the required amount of potassium permanganate is still a problem that needs to be solved.

[0003] At present, the concentration of divalent manganese ions in water is usually measured by chemical analysis, spectroscopy, titration, etc. These methods are complicated, time-consuming, and not suitable for real-time monitoring in practical applications. In water treatment, redox potential, as an indicator reflecting the redox state of water, has been widely used in water quality monitoring. Studies have shown that there is a certain correlation between redox potential and the concentration of certain ions in water, but how to accurately convert the redox potential value into the concentration of divalent manganese ions and further calculate the required potassium permanganate dosage, the existing technology has not yet provided an effective solution.

[0004] Therefore, a system and method for predicting the concentration of divalent manganese ions in water by using redox potential and calculating the dosage of potassium permanganate is developed, which can improve the prediction accuracy, reduce the waste of chemical reagents, reduce water treatment costs, and improve water treatment effects. Summary of the invention

[0005] The present invention provides a system for predicting the dosage of potassium permanganate, comprising: Redox potential sensor, used to measure the redox potential of water samples; The data acquisition module uses mean filtering to denoise the redox potential data; A back propagation neural network processing unit, which establishes a back propagation neural network prediction model based on the denoised redox potential data to predict the concentration of divalent manganese ions; The dosage calculation module calculates the required potassium permanganate concentration based on the predicted divalent manganese ion concentration and the preset dosage ratio coefficient. The calculation formula is: ;in, C KMnO4 is the required potassium permanganate concentration. C Mn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, kis the proportionality factor, and the range of the proportionality factor is 2.0 to 2.5; Output module, showing real-time prediction of divalent manganese ion concentration and potassium permanganate dosage.

[0006] Furthermore, the output module includes a display screen for displaying the divalent manganese ion concentration and the amount of potassium permanganate added, and a control interface. The output module is connected to a water treatment control system via the control interface, and is connected to a dosing device via the water treatment control system.

[0007] Furthermore, the back propagation neural network processing unit includes a training module and a prediction module. The training module is used to train the back propagation neural network model using the denoised redox potential data, and the prediction module is used to make predictions based on the trained model.

[0008] Furthermore, the back-propagation neural network model is trained using the back-propagation algorithm, and the training process is optimized using the following error function: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples.

[0009] The present invention also provides a method for predicting the dosage of potassium permanganate, which uses a system for predicting the dosage of potassium permanganate, and the method comprises: Measure the redox potential of water samples; The redox potential data were denoised using mean filtering; The denoised redox potential data was used as input and the divalent manganese ion concentration in water was used as output to train and establish a back-propagation neural network prediction model. According to the established back propagation neural network prediction model, the concentration of divalent manganese ions in water is predicted; Calculate the amount of potassium permanganate required based on the predicted divalent manganese ion concentration.

[0010] Furthermore, the amount of potassium permanganate required to be added is calculated using the following formula: ;in, C KMnO4 is the required potassium permanganate concentration. C Mn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportional coefficient of dosage, and the range of the proportional coefficient is 2.0 to 2.5.

[0011] Furthermore, the structure of the back-propagation neural network includes: Input layer: The number of nodes is 1, and the input is the denoised redox potential data; Hidden layer: The number of nodes is 20, which is determined by mean square error optimization; Output layer: The number of nodes is 1, and the output is the predicted divalent manganese ion concentration.

[0012] Furthermore, the back-propagation neural network model is trained using the back-propagation algorithm, and the training process is optimized using the following error function: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples.

[0013] Furthermore, the redox potential measurement range is 880 mV to 991 mV, corresponding to a divalent manganese ion concentration in water of 0.2 mg / L to 1.5 mg / L.

[0014] Furthermore, mean filtering uses a sliding average method with a window size of 3 to remove noise from the redox potential data.

[0015] The beneficial effects of the present invention are: The present invention can accurately predict the concentration of divalent manganese ions in water by utilizing a back propagation neural network model, thereby avoiding errors in traditional methods; based on real-time monitoring data of redox potential, the required dosage of potassium permanganate can be calculated in real time, the water treatment process is optimized, and excessive dosage is reduced; by accurately predicting the concentration of divalent manganese ions, the dosage of potassium permanganate can be accurately controlled, thereby avoiding waste of chemical reagents and reducing water treatment costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a line graph of predicted concentration and true value; Figure 2 A flow chart of a system for predicting potassium permanganate dosage; Figure 3 A schematic diagram of the back-propagation neural network structure; Figure 4 This is a comparison chart between the test set prediction results and the standard curve; Figure 5 This is a comparison table of 10 sets of predicted values ​​and true values; Figure 6 It is a schematic diagram of the process steps and corresponding equipment; Figure 7 Schematic diagram of back-propagation neural network parameters. DETAILED DESCRIPTION

[0017] The technical solution of the present invention will be described below in conjunction with the accompanying drawings of the present invention, but the described embodiments are only part of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0018] like Figure 1 Shown are the experimental data of the present invention.

[0019] The experimental design is as follows: Under ideal conditions, 27 sets of water sample data were collected: the oxidation-reduction potential (ORP) was 880-991mV, the divalent manganese ion Mn 2+ The concentration is 0.2-1.5 mg / L. After the redox potential is denoised by mean filtering, it is input into the back propagation neural network for training.

[0020] Prediction and results: The trained neural network model is used to predict the redox potential data in the actual water sample and accurately predict the divalent manganese ion Mn in the water. 2+ According to the prediction results, the dosage of potassium permanganate was obtained through the calculation formula.

[0021] Effect evaluation: By comparing with the traditional chemical analysis method, the error of the back propagation (BP) neural network prediction model is within 5%, with high prediction accuracy. It can realize real-time monitoring and dosage control, significantly improve the efficiency of water treatment, and reduce the waste of chemical reagents.

[0022] like Figures 1 to 7 As shown, the present invention provides a system for predicting the dosage of potassium permanganate, comprising: Oxidation reduction potential sensor (ORP sensor), used to measure the oxidation reduction potential of water samples; The data acquisition module uses mean filtering to denoise the redox potential data; A back propagation neural network processing unit, which establishes a back propagation neural network prediction model based on the denoised redox potential data to predict the concentration of divalent manganese ions; The dosage calculation module calculates the required potassium permanganate concentration based on the predicted divalent manganese ion concentration and the preset dosage ratio coefficient. The calculation formula is: ;in, C KMnO4 is the required potassium permanganate concentration. C Mn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportionality factor, and the range of the proportionality factor is 2.0 to 2.5; The output module displays the real-time predicted divalent manganese ion concentration and potassium permanganate dosage. The output module includes a display screen for displaying the divalent manganese ion concentration and the potassium permanganate dosage, and a control interface. The output module is connected to a water treatment control system through the control interface, and is connected to a dosing device through the water treatment control system.

[0023] The back propagation neural network processing unit includes a training module and a prediction module. The training module is used to train the back propagation neural network model using the denoised redox potential data, and the prediction module is used to make predictions based on the trained model. Specifically, the back-propagation neural network model is trained by the back-propagation algorithm, and the training process is optimized by the following error function: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples.

[0024] As a preferred embodiment, an oxidation-reduction potential sensor (ie, an ORP sensor) measures the oxidation-reduction potential value in a water sample in real time as a divalent manganese ion Mn 2+ The redox potential sensor has a measurement range of 880-2000 mV and an accuracy of ±1 mV, and is suitable for divalent manganese ions Mn in water. 2+ Ideal water quality conditions with a concentration of 0.2-1.5 mg / L. The redox potential sensor is installed in the water treatment pipeline to ensure that representative water samples can be collected.

[0025] As a preferred embodiment, the data acquisition module converts the analog signal of the redox potential sensor into a digital signal and performs denoising. The data acquisition module collects redox potential data through an analog-to-digital converter (ADC) and uses a mean filter to remove noise. The mean filter uses a window size of 3. The data acquisition module hardware uses an acquisition card with a sampling frequency of 1 Hz to ensure real-time performance.

[0026] As a preferred embodiment, the back propagation neural network processing unit (ie, the BP neural network processing unit) predicts the denoised redox potential data and calculates the divalent manganese ion concentration CMn in the water. 2+ The present invention uses the back propagation neural network model (BP neural network model) because it has strong nonlinear fitting ability and can accurately capture the redox potential 880-991 mV and divalent manganese ion Mn 2+ The complex relationship between concentrations of 0.2-1.5 mg / L. Through training, a stable mapping model is established.

[0027] Implementation: Use Python to build a back propagation neural network model (BP neural network model) based on the TensorFlow framework. The data input is the redox potential value, and the output is the predicted divalent manganese ion concentration CMn 2+ .

[0028] like Figure 3 and Figure 7 As shown, the structure of the back propagation neural network includes: Input layer: The number of nodes is 1, and the input is the denoised redox potential data; Hidden layer: The number of nodes is 20, which is determined by mean square error optimization, and the Sigmoid activation function is used to enhance the nonlinear fitting ability; Output layer: The number of nodes is 1, and the output is the predicted divalent manganese ion Mn 2+ concentration.

[0029] As a preferred embodiment, the dosage calculation module calculates the amount of potassium permanganate to be added according to the predicted divalent manganese ion concentration. The required potassium permanganate concentration is calculated by a preset dosage ratio coefficient. C KMnO4.

[0030] formula: ; in, C KMnO4 is the required potassium permanganate concentration, CMn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportionality coefficient, and the range of the proportionality coefficient is 2.0 to 2.5. After the dosage is calculated, the result is transmitted to the dosing device through the control system to adjust the amount of potassium permanganate added.

[0031] As a preferred embodiment, the output module displays the real-time prediction results and calculation results, and feeds the data back to the water treatment control system. The output device includes a display screen and a control interface. The display screen is used for operators to view the predicted concentration and dosage. The control interface is used to connect to the automatic control dosing system, and the dosing device is adjusted by the automatic control system to accurately control the dosage of potassium permanganate. The control system automatically adjusts the supply of potassium permanganate according to the set threshold value to avoid excessive addition.

[0032] like Figures 1 to 7 The present invention also provides a method for predicting the dosage of potassium permanganate, using the system for predicting the dosage of potassium permanganate, the method comprises: Measure the redox potential of water samples; The redox potential data were denoised using mean filtering; The denoised redox potential data was used as input and the divalent manganese ion concentration in water was used as output to train and establish a back-propagation neural network prediction model. According to the established back propagation neural network prediction model, the concentration of divalent manganese ions in water is predicted; Calculate the amount of potassium permanganate required based on the predicted divalent manganese ion concentration.

[0033] The amount of potassium permanganate required to be added is calculated using the following formula: ;in, C KMnO4 is the required potassium permanganate concentration, CMn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportional coefficient of dosage, and the range of the proportional coefficient is 2.0 to 2.5.

[0034] Before measuring water samples, water samples need to be collected first: under ideal conditions (without considering interference from turbidity, temperature, pH, etc.), an oxidation-reduction potential sensor is installed at the inlet of the water treatment pipeline to collect oxidation-reduction potential data of 27 groups of water samples. The data is transmitted in real time through the data acquisition module.

[0035] When the data acquisition module performs denoising, the collected redox potential data is denoised using mean filtering, with a window size of 3, that is, taking the average of three consecutive data points. For example, for the redox potential sequence [947,948, 946], the denoised value is 947 mV. This method effectively smooths the noise and ensures the stability of the input data.

[0036] Training and optimization of back-propagation neural network models: Training data, use the known redox potential (ORP) and the corresponding divalent manganese ion concentration dataset for training. The dataset should cover the concentration range of divalent manganese ions in water (0.2-1.5 mg / L) and the corresponding redox potential values.

[0037] The network structure adopts a back-propagation neural network structure with one hidden layer: the number of input layer nodes is 1 (ORP value); the number of output layer nodes is 1 (predicted divalent manganese ion concentration); the hidden layer has 20 nodes, and the Sigmoid activation function is used to enhance the nonlinear fitting ability.

[0038] Training process: The back propagation algorithm and gradient descent method are used to optimize the weights. The learning rate is set to 0.01 to ensure convergence stability. The error function formula is: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples, 27.

[0039] Stop condition: error E<0.001 or 500 iterations.

[0040] The results show that after training, the average prediction error of the back-propagation neural network model for 27 sets of data is 2.8%.

[0041] Furthermore, to prevent overfitting, the early stopping method is used, that is, the training is stopped when the error of the validation set no longer decreases. Five validation sets are randomly selected from the 27 sets of data.

[0042] Furthermore, in order to improve the effect of back propagation neural network training, the input data can be normalized. The normalization method used in this experiment is Min-Max normalization: scaling the data to the [0,1] interval.

[0043] Integration and debugging of the device: The redox potential sensor, data acquisition module, back propagation neural network processing unit and dosage calculation module are interconnected through the Modbus protocol. The data flow is: Sensor outputs redox potential → acquisition module denoising → neural network prediction → dosage calculation → output control.

[0044] like Figure 5 As shown, in a laboratory environment, 27 sets of known Mn 2+ The system was verified by water samples of different concentrations. The results showed: ORP=880 mV, predicted Mn 2+ =0.21 mg / L (actual 0.20 mg / L), error 5%; ORP=947 mV, predicted Mn 2+ =0.50 mg / L (actual 0.50 mg / L), error 0%; ORP=991 mV, predicted Mn 2+ =1.48 mg / L (actual 1.50 mg / L), error 1.33% (not shown in the figure).

[0045] The average error is 2.8%, and the dosage control error is less than 5%, which meets the water treatment needs under ideal conditions.

[0046] The method provided by the present invention is not only applicable to the treatment of drinking water, but can also be widely used in the fields of industrial wastewater treatment, environmental water treatment, etc. By real-time prediction of the concentration of divalent manganese ions in water and accurate calculation of the dosage of potassium permanganate, the efficiency of the water treatment process can be greatly improved, resource utilization can be optimized, and environmental pollution can be reduced.

Claims

1. A system for predicting potassium permanganate dosage, characterized in that: include: Redox potential sensor, used to measure the redox potential of water samples; The data acquisition module uses mean filtering to denoise the redox potential data; A back propagation neural network processing unit, which establishes a back propagation neural network prediction model based on the denoised redox potential data to predict the concentration of divalent manganese ions; The dosage calculation module calculates the required potassium permanganate concentration based on the predicted divalent manganese ion concentration and the preset dosage ratio coefficient. The calculation formula is: ;in, C KMnO4 is the required potassium permanganate concentration. C Mn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportionality coefficient, and the range of the proportionality coefficient is 2.0 to 2.5; the output module displays the real-time predicted divalent manganese ion concentration and potassium permanganate dosage.

2. The system for predicting potassium permanganate dosage according to claim 1, characterized in that: The output module includes a display screen for displaying the concentration of divalent manganese ions and the dosage of potassium permanganate, and a control interface. The output module is connected to a water treatment control system via the control interface, and is connected to a dosing device via the water treatment control system.

3. The system for predicting potassium permanganate dosage according to claim 2, characterized in that: The back propagation neural network processing unit includes a training module and a prediction module. The training module is used to train the back propagation neural network model using the denoised redox potential data, and the prediction module is used to make predictions based on the trained model.

4. The system for predicting potassium permanganate dosage according to claim 3, characterized in that: The back propagation neural network model is trained using the back propagation algorithm, and the training process is optimized using the following error function: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples.

5. A method for predicting the dosage of potassium permanganate, characterized in that: Using the system for predicting potassium permanganate dosage according to claim 1, the method comprises: Measure the redox potential of water samples; The redox potential data were denoised using mean filtering; The denoised redox potential data was used as input and the divalent manganese ion concentration in water was used as output to train and establish a back-propagation neural network prediction model. According to the established back propagation neural network prediction model, the concentration of divalent manganese ions in water is predicted; Calculate the amount of potassium permanganate required based on the predicted divalent manganese ion concentration.

6. The method for predicting potassium permanganate dosage according to claim 5, wherein: The required amount of potassium permanganate is calculated using the following formula: ;in, C KMnO4 is the required potassium permanganate concentration. C Mn 2+ is the predicted divalent manganese ion concentration, V water is the volume of the water body, k is the proportional coefficient of dosage, and the range of the proportional coefficient is 2.0 to 2.

5.

7. The method for predicting potassium permanganate dosage according to claim 6, wherein: The structure of the back-propagation neural network includes: Input layer: The number of nodes is 1, and the input is the denoised redox potential data; Hidden layer: The number of nodes is 20, which is determined by mean square error optimization; Output layer: The number of nodes is 1, and the output is the predicted divalent manganese ion concentration.

8. The method for predicting potassium permanganate dosage according to claim 5, wherein: The back propagation neural network model is trained using the back propagation algorithm, and the training process is optimized using the following error function: ;in, is the real Mn 2+ concentration, is the predicted value, and N is the number of data samples.

9. The method for predicting potassium permanganate dosage according to claim 5, wherein: The redox potential measurement range is 880 mV to 991 mV, corresponding to the divalent manganese ion concentration in water of 0.2 mg / L to 1.5 mg / L.

10. The method for predicting potassium permanganate dosage according to claim 5, characterized in that: Mean filtering uses a sliding average method with a window size of 3 to remove noise from the redox potential data.

Images