CN118270897B - Treatment method for low-temperature Gao Zhuo heavy metal sewage leaked from tailing pond - Google Patents

Abstract

The invention provides a method for treating low-temperature Gao Zhuo heavy metal sewage leaked from a tailing pond, which comprises the following steps: magnetizing the tailing sewage by a magnetizing module according to the intensity of the magnetic field and the magnetizing time; conveying the magnetized tailing sewage to a mixing tank, and adding magnetic seeds, flocculant and coagulant into the mixing tank according to the addition amount of flocculant, the addition amount of coagulant and the addition amount of magnetic seeds; training a first neural network model according to actual values of sedimentation time, final turbidity of supernatant fluid and final concentration of heavy metal when sedimentation of the sedimentation tank is completed; and adjusting the treatment parameters of the next tailing sewage according to the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model. The low-temperature Gao Zhuo heavy metal sewage leaked from the tailing pond is effectively treated, the adding amount is saved, and the self-adaptive adjustment is performed according to the result predicted by the neural network.

Description

A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond

Technical Field

The invention relates to the technical field of water pollution control, and in particular to a method for treating low-temperature, high-turbidity, heavy metal sewage leaking from a tailings pond.

Background Art

Magnetic separation technology refers to a separation technology that uses the difference in magnetic potential of elements or components to treat substances with the help of an external magnetic field to achieve an enhanced separation process. According to the different external magnetic field sources, magnetic separation can be divided into permanent magnetic separation, electromagnetic separation, and superconducting magnetic separation; according to different application environments, it can be divided into wet magnetic separation and dry magnetic separation; according to different magnetic reactors, it can be divided into traditional magnetic separation, magnetic disk separation, high gradient magnetic separation, and open gradient magnetic separation. Due to its fast and efficient separation effect, especially with the development of superconducting magnet technology and high gradient magnetic technology, magnetic separation has been widely used in tailings sorting, slag recovery, kaolin decolorization and other fields. Unlike conventional water treatment technology, magnetic separation technology uses magnetic field force to directly act on pollutants or target impurities, thereby separating pollutants from the raw water system, without affecting the water body, and without chemical and biological reactions. It has significant advantages such as no secondary pollution, fast separation speed, and less land occupation. Magnetic separation has become a promising treatment technology in the field of water treatment and has broad application potential in water treatment.

High gradient magnetic separation technology refers to filling a certain amount of magnetically sensitive medium in a magnetic separator, causing the magnetic field around the magnetic medium to be alienated, generating a higher magnetic gradient, maximizing the magnetic field force, and thus improving the separation rate and efficiency. However, an important challenge in the application of magnetic separation technology in the field of water treatment is that most of the pollutants in wastewater are non-magnetic and cannot be directly removed by the magnetic field, so the magnetic carrier is a key factor affecting the promotion and application of magnetic separation technology in the field of water treatment. Therefore, according to the presence/absence of magnetic carriers and the types of magnetic carriers, the application of magnetic separation technology in water treatment can be divided into the following categories: direct magnetic separation, magnetic flocculation, magnetic adsorption and magnetic catalysis. At present, there are two main ways to treat sewage using magnetic separation technology: 1. Adding magnetic seeds to sewage, magnetic seeds are added to wastewater, and the addition of coagulants and coagulants allows pollutants to combine with magnetic media to form magnetic floccules. Under the action of the magnetic field, these magnetic floccules can be attracted and separated from the water. This method requires the addition of magnetic seeds, so the cost is high, and the treated magnetic medium needs to be recovered or processed, otherwise it will cause secondary pollution. 2. Magnetized sewage: The sewage is pre-magnetized, which changes the pollutants and water molecules in the water. After adding coagulants and coagulants, these magnetized pollutants aggregate to form magnetic floccules and can be separated from the water, but this method is less effective for non-magnetic or weakly magnetic pollutants.

Summary of the invention

The purpose of the present invention is to provide a method for treating low-temperature, high-turbidity heavy metal sewage leaked from a tailings pond, which can effectively treat low-temperature, high-turbidity heavy metal sewage leaked from a tailings pond through magnetization treatment and magnetic flocculation, saving the dosage of flocculants and coagulants. At the same time, the prediction results of the first neural network model can help the treatment personnel to make necessary adjustments according to the prediction results before the treatment begins, such as increasing or decreasing the amount of treatment agent added, changing the magnetic field strength or adjusting the magnetization time, to ensure that the treatment results achieve the expected effect, thereby ensuring that the actual sewage treatment process can meet environmental protection requirements and actual work needs.

A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond, comprising:

The tailings wastewater is magnetized according to the magnetic field strength and magnetization time;

The tailings sewage is transported to the mixing tank, and magnetic seeds, flocculants and coagulants are added to the mixing tank according to the flocculant dosage, coagulant dosage and magnetic seed dosage;

After the mixing reaction time, the magnetic flocs in the mixing pool are salvaged through the magnetic enrichment module, and the remaining tailings wastewater in the mixing pool is transferred to the sedimentation tank;

Dehydrating and separating the magnetic flocs through a magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, cyclically adding the magnetic seeds to a mixing tank, and transmitting the tailings wastewater to a sedimentation tank;

Detect the actual values of sedimentation time, final turbidity of supernatant and final concentration of heavy metals when sedimentation in sedimentation tank is completed;

The first neural network model is trained according to the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed;

The treatment parameters of the next tailings wastewater treatment are adjusted according to the sedimentation time, final turbidity of the supernatant and final concentration of heavy metals output by the first neural network model.

Preferably, the actual values of the sedimentation time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed include:

The settling velocity, supernatant turbidity and heavy metal concentration of the sedimentation tank are input into the second neural network model;

The second neural network model outputs a second predicted value of the processing criterion parameter;

If the second predicted value of the processing standard parameter does not meet the preset standard, a warning signal is generated;

Additional dosing of flocculants and coagulants into the mixing tank and/or sedimentation tank based on early warning signals;

The tailings wastewater treatment measures and tailings wastewater treatment results adjusted according to the second predicted value are recorded and added to the training samples of the first neural network model.

Preferably, the training of the first neural network model according to the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed comprises:

Inputting the flocculant dosage, the coagulant dosage, the magnetic seed dosage, the slurry concentration, the magnetic field intensity and the magnetization time into the first neural network model;

The first neural network model outputs the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals;

The total loss function is calculated based on the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed;

Iterate the training until the total loss function meets the preset requirements.

Preferably, after detecting the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed, it also includes:

If the final turbidity and final heavy metal concentration of the supernatant meet the preset standards, the supernatant is discharged from the sedimentation tank; otherwise, the supernatant is returned to the magnetization module and mixed with the sewage for magnetization treatment.

Preferably, the total loss function is calculated based on the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling of the sedimentation tank is completed, including:

The total loss function includes mean square error and regularization loss;

The total loss function is expressed as:

The mean square error is expressed as:

Where n refers to the number of samples, assuming is the predicted value of the model, is the actual value;

The regularization loss is expressed as:

in represents the weight, is the regularization parameter.

Preferably, the iterative training stops when the total loss function meets the preset requirements, and includes:

Calculate the partial derivatives of the total loss function with respect to each weight and bias via backpropagation;

Set the total loss function to , the output of the jth neuron in the lth layer is a, then the partial derivative is expressed as ;

For each hidden layer, the chain rule is used to calculate the partial derivative of the loss function with respect to the input of each neuron in the current hidden layer:

in, represents the sum of all neurons in the l+1th layer, represents the input of the jth neuron in the lth layer, that is, the signal received by the neuron, which is specifically a linear combination of the output of the neurons in the previous layer and the corresponding weights plus the bias; represents the output of the jth neuron in the lth layer;

Get the partial derivatives of the loss function with respect to each weight and bias:

in, is the weight of the connection from the i-th neuron in the (l-1)th layer to the j-th neuron in the lth layer; is the bias of the jth neuron in the lth layer, that is, the threshold of the neuron;

Based on the calculated partial derivatives, each weight and bias parameter in the first neural network model is updated using the gradient descent method;

The training is repeated until the performance of the first neural network model reaches the preset requirement or the number of iterations reaches a threshold.

Preferably, setting the flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field strength and magnetization time includes:

Collect actual input parameter data during the historical treatment process, the input parameters include: flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field strength and magnetization time, and the corresponding treatment standard parameter results, the treatment standard parameter results include: sedimentation time, supernatant final turbidity and heavy metal final concentration;

Using the trained first neural network model to predict the input parameters, and comparing the predicted results with the actual processing standard parameter results to obtain the prediction error;

Calculate the correlation coefficient between each input parameter and the prediction error. The larger the absolute value of the correlation coefficient, the stronger the relationship between the input parameter and the prediction error, and the more significant the adjustment of the current parameter is needed.

Minimizing the prediction error is defined as the optimization goal, and the input parameters are optimized using a genetic algorithm;

During the optimization process, the adjustment range of each input parameter is proportional to the correlation coefficient of the prediction error. The larger the correlation coefficient, the larger the adjustment range of the parameter;

The optimized input parameter data is fed back to the processing process, and the iterative process is continued to achieve adaptive adjustment of parameters.

Preferably, the actual values of the sedimentation time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed include:

Install an ultrasonic sensor on the top of the sedimentation tank;

The transmitter and receiver of the ultrasonic sensor are directed toward the sediment surface, a number of ultrasonic sampling data are continuously acquired according to a preset sampling frequency, and the sedimentation velocity is calculated according to the ultrasonic sampling data;

Install a turbidity meter and heavy metal analyzer at the outlet of the supernatant in the sedimentation tank;

The turbidity meter and heavy metal analyzer monitor the turbidity and heavy metal concentration of the supernatant respectively;

The ultrasonic sampling data includes a number of distance data and sampling timestamps. The collection principle of the distance data is expressed as follows:

D = 0.5 * T * C

Where D represents the distance from the sensor to the solid-liquid separation interface, T represents the time difference from ultrasonic emission to reception, and C represents the propagation speed of ultrasonic waves in the supernatant;

The settling velocity is expressed as:

V _i = (D _i+1 -D _i )/t _i+1 -t _i ,

Wherein, _Vi represents the sedimentation velocity from time _ti to ti ₊₁ .

A system for treating low-temperature, high-turbidity, heavy metal wastewater leaked from a tailings pond, comprising:

A magnetization module is used to magnetize the tailings wastewater according to the magnetic field strength and magnetization time;

The sedimentation module is used to transport the tailings sewage to the mixing tank, and add magnetic seeds, flocculants and coagulants into the mixing tank according to the flocculant dosage, coagulant dosage and magnetic seed dosage;

The first separation module is used to salvage the magnetic flocs in the mixing pool through the magnetic enrichment module after the mixing reaction time, and transfer the remaining tailings wastewater in the mixing pool to the sedimentation tank;

The second separation module is used to dehydrate and separate the magnetic flocs through the magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, and the magnetic seeds are circulated to the mixing tank, and the tailings wastewater is transmitted to the sedimentation tank;

A detection module is used to detect the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed;

A prediction module is used to train a first neural network model based on actual values of settling time, final turbidity of supernatant and final concentration of heavy metals when settling in the sedimentation tank is completed;

The regulating module is used to adjust the treatment parameters of the next tailings wastewater according to the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model.

An electronic device comprises: a processor and a memory, wherein the memory is used to store computer program codes, and the computer program codes comprise computer instructions. When the processor executes the computer instructions, the electronic device executes a method for treating low-temperature, high-turbidity, heavy metal wastewater leaked from a tailings pond.

This application can effectively treat low-temperature, high-turbidity, heavy metal sewage leaking from tailings ponds through magnetization treatment and magnetic flocculation, saving the dosage of flocculants and coagulants. At the same time, the prediction results of the first neural network model can help treatment personnel make necessary adjustments based on the prediction results before the treatment begins, such as increasing or decreasing the amount of treatment agent added, changing the magnetic field strength or adjusting the magnetization time, to ensure that the treatment results achieve the expected effect, thereby ensuring that the actual sewage treatment process can meet environmental protection requirements and actual work needs.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

Fig. 1 is a schematic diagram of the overall process of the present invention;

FIG2 is a schematic diagram of a sedimentation tank sewage parameter detection process of the present invention;

FIG3 is a schematic diagram of a first neural network training process of the present invention;

FIG. 4 is a schematic diagram of the hardware structure of an electronic device of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

At present, there are two main ways to treat sewage using magnetic separation technology: 1. Adding magnetic seeds to sewage. Magnetic seeds are added to wastewater. The addition of coagulants and coagulant aids allows pollutants to combine with magnetic media to form magnetic floccules. Under the action of the magnetic field, these magnetic floccules can be attracted and separated from the water. This method requires the addition of magnetic seeds, so the cost is relatively high, and the treated magnetic media needs to be recovered or processed, otherwise it will cause secondary pollution. 2. Magnetized sewage. The sewage is pre-magnetized, so that both the pollutants and water molecules in the water change. After adding coagulants and coagulants, these magnetized pollutants aggregate to form magnetic floccules and can be separated from the water, but this method is slightly less effective for non-magnetic or weakly magnetic pollutants.

This application can effectively treat low-temperature, high-turbidity, heavy metal sewage leaking from tailings ponds through magnetization treatment and magnetic flocculation, saving the dosage of flocculants and coagulants. At the same time, the prediction results of the first neural network model can help treatment personnel make necessary adjustments based on the prediction results before the treatment begins, such as increasing or decreasing the amount of treatment agent added, changing the magnetic field strength or adjusting the magnetization time, to ensure that the treatment results achieve the expected effect, thereby ensuring that the actual sewage treatment process can meet environmental protection requirements and actual work needs.

Example 1

A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond, comprising:

S100, magnetizing the tailings wastewater according to the magnetic field intensity and magnetization time;

The specific steps of magnetization include: pre-treating the sewage to remove most of the suspended matter and solid particles in the tailings sewage; selecting the magnetization device, designing and selecting a suitable magnetization device according to the actual treatment requirements and the requirements of the magnetic field, including the magnet, the shape of the magnetic field, the strength of the magnetic field, etc. Magnetization treatment, passing the pre-treated sewage through the magnetization device, making it flow in a strong magnetic field, and being subjected to the effect of the magnetic field to achieve the decomposition and precipitation of pollutants. Terminal treatment: After the sewage has been magnetized, it undergoes subsequent terminal treatments such as reaction, precipitation, filtration, etc. according to different treatment requirements.

S200, transporting the tailings wastewater to a mixing tank, and adding magnetic seeds, flocculants and coagulants into the mixing tank according to the flocculant dosage, coagulant dosage and magnetic seed dosage;

Magnetic flocculation includes inorganic flocculants and organic flocculants. Among them, inorganic flocculants include inorganic flocculants and inorganic polymer flocculants, and organic flocculants include synthetic organic polymer flocculants, natural organic polymer flocculants and microbial flocculants. The process principle of magnetic flocculation technology is to add magnetic seeds to the traditional flocculation mixing sedimentation process to enhance the flocculation effect, form high-density flocs and increase the specific gravity of the flocs, so as to achieve the purpose of efficient pollution removal and rapid sedimentation. According to the ionic polarity and metallic properties of the magnetic seeds, as the nucleus of the flocculants, the flocculation and binding ability of suspended pollutants in the water is greatly enhanced, and the amount of flocculants used is reduced. The specific gravity of the flocs mixed with magnetic seeds increases, which can make the flocs settle quickly.

S300, after the mixing reaction time, the magnetic flocs in the mixing pool are salvaged through the magnetic enrichment module, and the remaining tailings wastewater in the mixing pool is transferred to the sedimentation tank;

It takes a certain mixing reaction time to completely separate the pollutants in the tailings wastewater. Once the pollutants in the tailings wastewater are separated from the water and precipitated, the magnetic flocs can be salvaged from the tailings wastewater through the magnetic enrichment module, and the remaining tailings wastewater that has not been treated will enter the next treatment process.

S400, the magnetic flocs are dehydrated and separated through a magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, the magnetic seeds are circulated and added to the mixing tank, and the tailings wastewater is transmitted to the sedimentation tank;

The magnetic seeds in the magnetic flocs are separated and recycled by using the rare earth permanent magnetic drums according to the characteristics of the magnetic seeds themselves and then recycled in the tailings sewage treatment system, so as to achieve the purpose of highly purifying the effluent and reducing the sewage treatment costs.

S500, detecting the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed;

The sedimentation time of the sedimentation tank is calculated from the time when the tailings wastewater is completely transferred to the sedimentation tank, that is, when all the tailings wastewater is transferred to the sedimentation tank, the sedimentation time of the sedimentation tank is calculated. Because in the process of tailings wastewater transfer, the movement of the upper tailings wastewater will affect the sedimentation effect of the lower tailings wastewater, so when the tailings wastewater is still, that is, when the tailings wastewater is completely transferred to the sedimentation tank, the sedimentation time is calculated. The final turbidity of the supernatant is obtained by sampling the supernatant in multiple areas of the sedimentation tank, and then calculating the average turbidity of the supernatant in each area as the final turbidity of the supernatant. The calculation of the final concentration of heavy metals requires sampling the tailings wastewater after precipitation, and then filtering and diluting it, and using atomic absorption spectrometry and inductively coupled plasma emission spectrometry to determine the final concentration of heavy metals in the tailings wastewater.

S600, training a first neural network model according to actual values of the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals when the settling in the sedimentation tank is completed;

The output of the first neural network model is the actual values of the sedimentation time when the sedimentation in the sedimentation tank is completed, the final turbidity of the supernatant and the final concentration of heavy metals. The input is the magnetic field strength and magnetization time during the magnetization process, the flocculant dosage in the mixing tank, the coagulant dosage, the magnetic seed dosage and the mixing reaction time. The first neural network is used to predict the control parameters in the entire tailings wastewater treatment process.

S700, adjusting the treatment parameters of the next tailings wastewater according to the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model.

This application can effectively treat low-temperature, high-turbidity, heavy metal sewage leaking from tailings ponds through magnetization treatment and magnetic flocculation, saving the dosage of flocculants and coagulants. At the same time, the prediction results of the first neural network model can help treatment personnel make necessary adjustments based on the prediction results before the treatment begins, such as increasing or decreasing the amount of treatment agent added, changing the magnetic field strength or adjusting the magnetization time, to ensure that the treatment results achieve the expected effect, thereby ensuring that the actual sewage treatment process can meet environmental protection requirements and actual work needs.

Preferably, S500, detecting the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed includes:

S510, inputting the settling velocity, supernatant turbidity and heavy metal concentration of the sedimentation tank into the second neural network model;

The sedimentation rate of the sedimentation tank is determined by the radius and depth of the sedimentation tank, and different types of tailings wastewater have different sedimentation rates. The supernatant turbidity indicates the turbidity of the supernatant liquid of the tailings wastewater, and the heavy metal concentration indicates the content of heavy metals in the tailings wastewater after the tailings wastewater is treated.

S520, the second neural network model outputs a second predicted value of the processing standard parameter;

As a preferred embodiment, the present invention adopts an RNN model as the architecture of the second neural network. The RNN model is a deep learning model suitable for processing data with sequence or order dependencies. The RNN model retains a hidden state when processing each input, and the hidden state will be passed to the next time step so that the RNN model can remember previous information.

The RNN model includes an input layer, an output layer and a hidden layer. In this application, the output layer uses a softmax function to normalize the data, and tanh is used as the activation function in the hidden layer.

S530, if the second predicted value of the processing standard parameter does not meet the preset standard, generating a warning signal;

The preset standard is the type and weight of flocculants and coagulants expected to be put into the mixing tank and/or sedimentation tank. The preset standard is set according to the treatment standards of the general industry. The second predicted value of the treatment standard parameter is the type and weight of flocculants and coagulants to be put into the mixing tank and/or sedimentation tank predicted by the RNN model. If the second predicted value of the treatment standard parameter output by the RNN model does not meet the preset standard, it means that the type and weight of flocculants and coagulants required to be put into the mixing tank and/or sedimentation tank do not reach the amount required to treat tailings wastewater, and the dosage of flocculants and coagulants needs to be increased.

S540, additionally adding flocculants and coagulants to the mixing tank and/or the sedimentation tank according to the early warning signal;

The information in the early warning signal includes the amount and type of additional flocculants and coagulants that need to be added to the mixing tank and/or sedimentation tank. The tailings wastewater treatment staff then adjusts the amount of flocculants and coagulants added to the mixing tank and/or sedimentation tank in real time based on the information in the early warning signal.

S550, recording the tailings wastewater treatment measures and tailings wastewater treatment results adjusted according to the second prediction value, and adding them to the training samples of the first neural network model.

Although more training samples will lead to longer training time, for some complex problems, such as in this application, the first neural network needs to predict the sedimentation time when the sedimentation of the sedimentation tank is completed, the final turbidity of the supernatant and the final concentration of heavy metals based on the input parameters into the treatment pool. If there are not enough training samples, it will easily lead to defects in the prediction of the first neural network. At this time, more training samples can make the training of the first neural network more perfect, make up for the defects and deficiencies, and make the prediction results of the first neural network more accurate.

In this embodiment, the second neural network model is used to predict the treatment standard parameters in real time based on the sedimentation velocity, supernatant turbidity and heavy metal concentration, so as to realize early warning analysis and real-time adjustment of the reaction process and ensure that the subsequent sewage treatment plan can be adjusted in time. At the same time, all adjustment measures and results will be recorded and fed back to the data center to timely eliminate mechanical failures or optimize the first neural network model by increasing the training samples of the first neural network model.

Preferably, referring to FIG3 , S600 , training the first neural network model according to the actual values of the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals when the settling in the sedimentation tank is completed includes:

S610, inputting the flocculant dosage, the coagulant dosage, the magnetic seed dosage, the slurry concentration, the magnetic field intensity and the magnetization time into the first neural network model;

The flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field intensity and magnetization time are all inputs of the first neural network.

S620, the first neural network model outputs the sedimentation time, the final turbidity of the supernatant, and the final concentration of heavy metals;

The sedimentation time, final turbidity of the supernatant and final concentration of heavy metals were the outputs of the first neural network.

S630, calculating a total loss function according to the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals when the settling of the sedimentation tank is completed;

S640, iterative training stops when the total loss function meets the preset requirements.

The loss function is used to measure the difference or error between the prediction result of the first neural network model and the actual result. The loss function is a non-negative real-valued function, usually expressed as L(Y, f(x)), where Y is the actual value and f(x) is the predicted value of the model. The smaller the loss function, the better the robustness of the model. In machine learning, the loss function is the core part of the empirical risk function and an important part of the structural risk function. The structural risk function includes the empirical risk term and the regularization term. Its purpose is to control the complexity of the model through the regularization term while reducing the empirical risk to prevent overfitting. The loss function selects different loss functions according to different learning tasks (such as classification and regression). For example, mean square error (MSE) is a commonly used regression loss function; in classification problems, cross entropy loss or hinge loss may be used. The loss function is not only the optimization target in the model training process, but also an important indicator for evaluating model performance.

Preferably, an MLP model is used as the architecture of the first neural network model;

The input layer nodes of the MLP model are set to 6, and the output layer nodes are set to 3.

The input layer is responsible for receiving external input information and converting it into a format that the neural network can process. The input layer can receive different types of raw data or pre-processed data. For example, in the field of image recognition, the input layer can receive raw three-dimensional colorful images; in the field of audio recognition, it can receive two-dimensional waveform data after Fourier transformation; in natural language processing, the input layer can receive sentence vectors represented in one dimension. The role of the input layer is to convert external input data into a form that the model can process. For example, when processing images, the input layer will receive the pixel values of the image, and when processing text, it will receive the characters or word vectors of the text. These raw data are converted and processed by the neural network to finally get useful output. The input data must be numerical, and non-numerical content needs to be converted into numerical values. The process of processing data before inputting it into the neural network is called data processing. For example, image data is usually converted into matrix pixel data for neural network processing. The input layer may also perform some pre-processing operations, such as normalization and de-meaning, to improve the performance of the model.

The output layer is the end point of the neural network model. It is responsible for converting the processing results of the neural network into a format that can be understood by humans or other organisms and outputting them. The setting of the output layer depends on the type of task faced by the neural network. For example, in image recognition tasks, the output layer may output objects or scenes that can be recognized by humans; in language processing tasks, the output layer may output sentences or paragraphs. The output layer is composed of several neurons, each of which represents an object, and the additional value it outputs represents the probability that the object belongs to a specific category.

In this application, the input layer nodes are set to 6, which means that the input parameters provide comprehensive information about the treatment process. The MLP model performs nonlinear transformations through hidden layers to learn the high-level correlations between these input parameters and treatment effects (such as settling time, final turbidity of the supernatant, and final concentration of heavy metals). The output layer nodes are set to 3, corresponding to the three key indicators of treatment effect: settling time, final turbidity of the supernatant, and final concentration of heavy metals. The goal of MLP training is to predict these three outputs as accurately as possible. After the model is trained, the input parameters can be input into the model in order to minimize the settling time, final turbidity of the supernatant, and final concentration of heavy metals.

Preferably, S630, calculating the total loss function according to the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals when the settling of the sedimentation tank is completed includes:

The total loss function includes mean square error and regularization loss;

The total loss function is expressed as:

The mean square error is expressed as:

Where n refers to the number of samples, assuming is the predicted value of the model, is the actual value;

The regularization loss is expressed as:

in represents the weight, is the regularization parameter.

The mean square error is a commonly used indicator in prediction evaluation. It is used to evaluate the fit of the first neural network model on the given data. It is calculated as the average of the squares of the differences between the model predictions and the actual observations. The smaller the value of the mean square error, the more accurate the prediction of the first neural network model. It should be noted that the mean square error cannot be equal to 0, because this will cause the model to overfit the data and may ignore some important factors, resulting in large fluctuations in practical applications.

Regularization is a technique used to improve machine learning models. Its main purpose is to reduce overfitting and improve the generalization ability of the model. In machine learning, overfitting refers to the phenomenon that the model performs well on the training data but performs poorly on new data (test data). Regularization is achieved by adding a regularization term to the loss function of the model. This regularization term penalizes the size or number of model parameters, thereby encouraging the model to choose simpler and more generalizable hypotheses. Common regularization methods include L1 regularization and L2 regularization, which are implemented by adding the sum of the absolute values or the sum of squares of the model parameters to the loss function, respectively. L1 regularization tends to make the model parameters sparse, that is, many parameters become zero, which helps feature selection and reduces the complexity of the model. L2 regularization makes the parameters smaller, but not completely zero, which helps reduce the instability and risk of overfitting of the model. Regularization loss is to regularize the loss function and add an additional term after the loss function, that is, L1 regularization or L2 regularization.

Preferably, S640, iterative training stops when the total loss function meets a preset requirement, including:

S641, calculate the partial derivative of the total loss function with respect to each weight and bias by backpropagation;

Set the total loss function to , the output of the jth neuron in the lth layer is a, then the partial derivative is expressed as ;

For each hidden layer, the chain rule is used to calculate the partial derivative of the loss function with respect to the input of each neuron in the current hidden layer:

in, represents the sum of all neurons in the l+1th layer, represents the input of the jth neuron in the lth layer, that is, the signal received by the neuron, which is specifically a linear combination of the output of the neurons in the previous layer and the corresponding weights plus the bias; represents the output of the jth neuron in the lth layer;

Get the partial derivatives of the loss function with respect to each weight and bias:

in, is the weight of the connection from the i-th neuron in the (l-1)th layer to the j-th neuron in the lth layer; is the bias of the jth neuron in the lth layer, that is, the threshold of the neuron;

S642, updating each weight and bias parameter in the first neural network model using a gradient descent method according to the calculated partial derivatives;

Specifically, for each weight and bias parameter, its updated value is equal to the original value minus the learning rate multiplied by the partial derivative of the parameter. The goal of this step is to make the value of the next loss function smaller by changing the value of each weight and bias parameter. Gradient descent is an optimization algorithm used to find the local minimum of a function. It is used to find model parameters that minimize the objective function in machine learning and artificial intelligence. The basic idea of gradient descent is to gradually adjust the model parameters (or independent variables) through continuous iterations so that the value of the objective function (usually the error function or loss function) decreases. In each iteration, the algorithm calculates the gradient of the objective function at the current point (that is, the direction in which the function value changes fastest), and then moves a small step in the negative gradient direction, that is, approaches the local minimum of the objective function. The key to this method is to use the gradient information of the objective function to guide the next moving direction and step size. There are several variants of gradient descent, such as stochastic gradient descent and adaptive gradient descent, which are used to process large data sets and sparse data, respectively. Gradient descent can also be contrasted with gradient ascent, which moves in the positive direction of the gradient to find the local maximum of the function.

S643, repeat the training until the performance of the first neural network model reaches a preset requirement or the number of iterations reaches a threshold.

Preferably, setting the flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field strength and magnetization time includes:

Collect the actual input parameter data during the historical treatment process, including: flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field strength and magnetization time, and the corresponding treatment standard parameter results, including: sedimentation time, supernatant final turbidity and heavy metal final concentration;

Using the trained first neural network model to predict the input parameters, and comparing the predicted results with the actual processing standard parameter results to obtain the prediction error;

Calculate the correlation coefficient between each input parameter and the prediction error. The larger the absolute value of the correlation coefficient, the stronger the relationship between the input parameter and the prediction error, and the more significant the adjustment of the current parameter is needed.

Minimizing the prediction error is defined as the optimization goal, and the input parameters are optimized using a genetic algorithm;

During the optimization process, the adjustment range of each input parameter is proportional to the correlation coefficient of the prediction error. The larger the correlation coefficient, the larger the adjustment range of the parameter;

The optimized input parameter data is fed back to the processing process, and the iterative process is continued to achieve adaptive adjustment of parameters.

This embodiment collects the actual input parameter data and the corresponding processing parameter results in the historical processing process, calculates the correlation coefficient between each input parameter and the prediction error based on the prediction error, and uses the genetic algorithm to achieve global optimization and avoid falling into the local optimal solution, thereby adaptively adjusting the input parameters and improving the stability and accuracy of the processing effect.

Preferably, referring to FIG. 2 , S500 , detecting the actual values of the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals when the settling in the sedimentation tank is completed includes:

S510, install an ultrasonic sensor on the top of the sedimentation tank;

S520, the transmitter and the receiver of the ultrasonic sensor are directed toward the sediment surface, a plurality of ultrasonic sampling data are continuously acquired according to a preset sampling frequency, and a sedimentation velocity is calculated according to the ultrasonic sampling data;

The main components of ultrasonic sensors include a piezoelectric crystal that can convert electrical energy into ultrasonic waves (mechanical waves) and can convert received sound waves back into electrical energy again. These sensors are usually used to detect physical quantities such as the position, size, speed or liquid level of an object. Based on the propagation characteristics of sound waves in air, liquid or solid, the distance between the sensor and the target object can be calculated by measuring the time difference between the emission and reception of ultrasonic waves.

S530, installing a turbidity meter and a heavy metal analyzer at the outlet of the supernatant of the sedimentation tank;

S540, turbidity meter and heavy metal analyzer monitor supernatant turbidity and heavy metal concentration, respectively;

A certain volume of liquid is used in a sedimentation tank and then subjected to a digestion treatment, wherein all the substances of the heavy metal elements to be measured are oxidized into ionic states, and the heavy metal ions to be measured are complexed with the color developer to form a complex of a specific color. At a certain wavelength, the complex has a maximum absorption, and the absorbance is linearly correlated with the concentration of the object to be measured. The concentration of the heavy metal to be measured can be calculated from the absorbance.

Ultrasonic sampling data includes several distance data and sampling timestamps. The collection principle of distance data is expressed as follows:

D = 0.5 * T * C

Where D represents the distance from the sensor to the solid-liquid separation interface, T represents the time difference from ultrasonic emission to reception, and C represents the propagation speed of ultrasonic waves in the supernatant;

The sedimentation velocity is expressed as:

V _i = (D _i+1 -D _i )/t _i+1 -t _i ,

Wherein, _Vi represents the sedimentation velocity from time _ti to ti ₊₁ .

If the final turbidity and final heavy metal concentration of the supernatant meet the preset standards, the supernatant is discharged from the sedimentation tank; otherwise, the supernatant is returned to the magnetization module to be mixed with the sewage for magnetization treatment. This application ensures that the final discharge substances meet environmental protection standards through real-time monitoring of the discharge quality of the supernatant and reflux treatment.

Example 2

A system for treating low-temperature, high-turbidity, heavy metal wastewater leaked from a tailings pond, comprising:

A magnetization module is used to magnetize the tailings wastewater according to the magnetic field strength and magnetization time;

The sedimentation module is used to transport the tailings sewage to the mixing tank, and add magnetic seeds, flocculants and coagulants into the mixing tank according to the flocculant dosage, coagulant dosage and magnetic seed dosage;

The first separation module is used to salvage the magnetic flocs in the mixing pool through the magnetic enrichment module after the mixing reaction time, and transfer the remaining tailings wastewater in the mixing pool to the sedimentation tank;

The second separation module is used to dehydrate and separate the magnetic flocs through the magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, and the magnetic seeds are circulated to the mixing tank, and the tailings wastewater is transmitted to the sedimentation tank;

A detection module is used to detect the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed;

A prediction module is used to train a first neural network model based on actual values of settling time, final turbidity of supernatant and final concentration of heavy metals when settling in the sedimentation tank is completed;

The regulating module is used to adjust the treatment parameters of the next tailings wastewater according to the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model.

Example 3

An electronic device includes: a processor and a memory, the memory is used to store computer program codes, the computer program codes include computer instructions, when the processor executes the computer instructions, the electronic device executes a method for treating low-temperature and high-turbidity heavy metal wastewater leaked from a tailings pond.

Referring to FIG4 , the electronic device 2 includes a processor 21, a memory 22, an input device 24, and an output device 23. The processor 21, the memory 22, the input device 24, and the output device 23 are coupled via a connector, and the connector includes various interfaces, transmission lines, or buses, etc., which are not limited in the embodiments of the present invention. It should be understood that in various embodiments of the present invention, coupling refers to mutual connection in a specific manner, including direct connection or indirect connection through other devices, for example, through various interfaces, transmission lines, buses, etc.

The processor 21 may be one or more graphics processing units (GPUs). When the processor 21 is a GPU, the GPU may be a single-core GPU or a multi-core GPU. Optionally, the processor 21 may be a processor group consisting of multiple GPUs, and the multiple processors are coupled to each other via one or more buses. Optionally, the processor may also be other types of processors, etc., which are not limited in the embodiments of the present invention.

The memory 22 can be used to store computer program instructions and various computer program codes including program codes for executing the scheme of the present invention. Optionally, the memory includes but is not limited to random access memory (RAM), read-only memory (ROM), erasable programmable read only memory (EPROM), or portable read only memory (CD-ROM), which is used for related instructions and data.

The input device 24 is used to input data and/or signals, and the output device 23 is used to output data and/or signals. The output device 23 and the input device 24 can be independent devices or an integrated device.

This application can effectively treat low-temperature, high-turbidity, heavy metal sewage leaking from tailings ponds through magnetization treatment and magnetic flocculation, saving the dosage of flocculants and coagulants. At the same time, the prediction results of the first neural network model can help the treatment personnel to make necessary adjustments according to the predicted results before the treatment begins, such as increasing or decreasing the amount of treatment agent added, changing the magnetic field strength or adjusting the magnetization time, to ensure that the treatment results achieve the expected effect, thereby ensuring that the actual sewage treatment process can meet environmental protection requirements and actual work needs.

The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims (6)

1. A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond, comprising: The magnetization module is used to magnetize the tailings wastewater according to the magnetic field strength and magnetization time; The tailings wastewater after magnetization treatment is transported to a mixing pool, and magnetic seeds, flocculants and coagulants are added to the mixing pool according to the flocculant dosage, coagulant dosage and magnetic seed dosage; After the mixing reaction time, the magnetic flocs in the mixing pool are salvaged through the magnetic enrichment module, and the remaining tailings wastewater in the mixing pool is transferred to the sedimentation tank; Dehydrating and separating the magnetic flocs through a magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, cyclically adding the separated magnetic seeds to a mixing tank, and transmitting the separated tailings wastewater to a sedimentation tank; Detect the actual values of sedimentation time, final turbidity of supernatant and final concentration of heavy metals when sedimentation in sedimentation tank is completed; The first neural network model is trained according to the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed; Adjust the treatment parameters of the next tailings wastewater according to the settling time, final turbidity of the supernatant and final concentration of heavy metals output by the first neural network model; The training of the first neural network model according to the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling of the sedimentation tank is completed comprises: Inputting the flocculant dosage, the coagulant dosage, the magnetic seed dosage, the slurry concentration, the magnetic field intensity and the magnetization time into the first neural network model; The first neural network model outputs the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals; The total loss function is calculated based on the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed; Iterate the training until the total loss function meets the preset requirements. 2. A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond according to claim 1, characterized in that after detecting the actual values of the sedimentation time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed, it also includes: If the final turbidity and final heavy metal concentration of the supernatant meet the preset standards, the supernatant is discharged from the sedimentation tank; otherwise, the supernatant is returned to the magnetization module and mixed with the sewage for magnetization treatment. 3. A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond according to claim 2, characterized in that the total loss function is calculated based on the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model and the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed, including: The total loss function includes mean square error and regularization loss; The total loss function is expressed as: The mean square error is expressed as: Where n refers to the number of samples, assuming is the predicted value of the model, is the actual value; The regularization loss is expressed as: in represents the weight, is the regularization parameter. 4. A method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond according to claim 3, characterized in that the iterative training stops when the total loss function meets the preset requirements and comprises: Calculate the partial derivatives of the total loss function with respect to each weight and bias via backpropagation; Set the total loss function to , the output of the jth neuron in the lth layer is a, then the partial derivative is expressed as ; For each hidden layer, the chain rule is used to calculate the partial derivative of the loss function with respect to the input of each neuron in the current hidden layer: in, represents the sum of all neurons in the l+1th layer, represents the input of the jth neuron in the lth layer, that is, the signal received by the neuron, which is specifically a linear combination of the output of the neurons in the previous layer and the corresponding weights plus the bias; represents the output of the jth neuron in the lth layer; Get the partial derivatives of the total loss function with respect to each weight and bias: ; in, is the weight of the connection from the i-th neuron in the (l-1)th layer to the j-th neuron in the lth layer; is the bias of the jth neuron in the lth layer, that is, the threshold of the neuron; Based on the calculated partial derivatives, each weight and bias parameter in the first neural network model is updated using the gradient descent method; The training is repeated until the performance of the first neural network model reaches the preset requirement or the number of iterations reaches a threshold. 5. The method for treating low-temperature, high-turbidity heavy metal wastewater leaked from a tailings pond according to claim 1, characterized in that the flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field intensity and magnetization time are set to include: Collect actual input parameter data during the historical treatment process, the input parameters include: flocculant dosage, coagulant dosage, magnetic seed dosage, slurry concentration, magnetic field strength and magnetization time, and the corresponding treatment standard parameter results, the treatment standard parameter results include: sedimentation time, supernatant final turbidity and heavy metal final concentration; Using the trained first neural network model to predict the input parameters, and comparing the predicted results with the actual processing standard parameter results to obtain the prediction error; Calculate the correlation coefficient between each input parameter and the prediction error. The larger the absolute value of the correlation coefficient, the stronger the relationship between the input parameter and the prediction error, and the more significant the adjustment of the current parameter is needed. Minimizing the prediction error is defined as the optimization goal, and the input parameters are optimized using a genetic algorithm; During the optimization process, the adjustment range of each input parameter is proportional to the correlation coefficient of the prediction error. The larger the correlation coefficient, the larger the adjustment range of the parameter; The optimized input parameter data is fed back to the processing process, and the iterative process is continued to achieve adaptive adjustment of parameters. 6. A system for treating low-temperature, high-turbidity, heavy metal wastewater leaked from a tailings pond, comprising: A magnetization module is used to magnetize the tailings wastewater according to the magnetic field strength and magnetization time; The sedimentation module is used to transport the tailings wastewater after magnetization treatment to the mixing tank, and add magnetic seeds, flocculants and coagulants into the mixing tank according to the flocculant dosage, coagulant dosage and magnetic seed dosage; The first separation module is used to salvage the magnetic flocs in the mixing pool through the magnetic enrichment module after the mixing reaction time, and transfer the remaining tailings wastewater in the mixing pool to the sedimentation tank; The second separation module is used to dehydrate and separate the magnetic flocs through the magnetic recovery module to obtain tailings wastewater, magnetic seeds and sludge, and the separated magnetic seeds are circulated and added to the mixing tank, and the separated tailings wastewater is transmitted to the sedimentation tank; A detection module is used to detect the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the sedimentation in the sedimentation tank is completed; A prediction module is used to train a first neural network model based on actual values of settling time, final turbidity of supernatant and final concentration of heavy metals when settling in the sedimentation tank is completed; The training of the first neural network model according to the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling of the sedimentation tank is completed comprises: Inputting the flocculant dosage, the coagulant dosage, the magnetic seed dosage, the slurry concentration, the magnetic field intensity and the magnetization time into the first neural network model; The first neural network model outputs the settling time, the final turbidity of the supernatant, and the final concentration of heavy metals; The total loss function is calculated based on the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model and the actual values of the settling time, the final turbidity of the supernatant and the final concentration of heavy metals when the settling in the sedimentation tank is completed; Iterate the training until the total loss function meets the preset requirements; The regulating module is used to adjust the treatment parameters of the next tailings wastewater according to the settling time, the final turbidity of the supernatant and the final concentration of heavy metals output by the first neural network model.