CN114380379B - Chemical adding control method and system for slime water - Google Patents
Chemical adding control method and system for slime water Download PDFInfo
- Publication number
- CN114380379B CN114380379B CN202210162651.5A CN202210162651A CN114380379B CN 114380379 B CN114380379 B CN 114380379B CN 202210162651 A CN202210162651 A CN 202210162651A CN 114380379 B CN114380379 B CN 114380379B
- Authority
- CN
- China
- Prior art keywords
- dosing
- concentration
- data
- model
- feeding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- C—CHEMISTRY; METALLURGY
- C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F1/00—Treatment of water, waste water, or sewage
- C02F1/52—Treatment of water, waste water, or sewage by flocculation or precipitation of suspended impurities
- C02F1/5209—Regulation methods for flocculation or precipitation
-
- C—CHEMISTRY; METALLURGY
- C02—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F—TREATMENT OF WATER, WASTE WATER, SEWAGE, OR SLUDGE
- C02F1/00—Treatment of water, waste water, or sewage
- C02F1/52—Treatment of water, waste water, or sewage by flocculation or precipitation of suspended impurities
- C02F1/54—Treatment of water, waste water, or sewage by flocculation or precipitation of suspended impurities using organic material
- C02F1/56—Macromolecular compounds
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B11/00—Automatic controllers
- G05B11/01—Automatic controllers electric
- G05B11/36—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential
- G05B11/42—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential for obtaining a characteristic which is both proportional and time-dependent, e.g. P.I., P.I.D.
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to a chemical adding control method and a chemical adding control system for slime water, belonging to the technical field of automatic chemical adding control for slime water; the technical problem to be solved is as follows: the improvement of a chemical adding control method of slime water is provided; the technical scheme for solving the technical problems is as follows: the method comprises the following steps: s1: constructing an empirical model: summarizing long-term working experience of a preschool learner and field workers to obtain an empirical mathematical model serving as an empirical model for dosing control; s2: constructing a prediction model: learning historical data by using an LSTM network to obtain a prediction model, and performing model optimization according to real-time data; s3: and (3) dosing control: detecting and acquiring real-time data of a coal slime water feed through a sensor, transmitting the data to a controller, and dynamically adding medicine according to an empirical model; inputting prediction input data detected by a sensor into a prediction model to obtain a prediction overflow concentration, and realizing advanced adjustment of the dosage according to the difference between the prediction overflow concentration and the set overflow concentration; the invention is applied to adding chemicals into slime water.
Description
Technical Field
The invention discloses a chemical adding control method and system for slime water, and belongs to the technical field of automatic chemical adding control for slime water.
Background
The concentration of the coal slime water in a coal preparation plant is a key link for realizing solid-liquid separation, and is also a previous process of filter pressing, the concentration is extremely important in the whole closed circulation flow of washing water, and because the property of the coal slime water determines that the coal slime water is difficult to settle by gravity, a flocculating agent needs to be added into the coal slime water to form large flocs, and then the flocs are settled by gravity. At present, a coal preparation plant mainly adopts a mode of manually adding chemicals, a certain amount of flocculant is added into a quantitative clear water bucket by a worker and stirred to prepare a flocculant solution with a certain concentration, and then the solution is added into a feeding pipeline of a concentration tank. The flocculant is added completely depending on the experience of workers, the flocculant can not be dynamically added according to the feeding condition, and the adjustment of the medicament quantity in the coal preparation plant is mostly operated according to the experience after a post driver observes and observes the overflow water condition by naked eyes, the flocculation sedimentation process has the characteristics of large inertia and large lag, the operation workers are difficult to dynamically adjust the medicament quantity in real time according to the overflow condition, the error of manual operation is large, the medicament adjustment lag exists, the concentration effect of the coal slime water is influenced, the waste of the medicament is caused, the overflow water can not meet the use requirement of the circulating water, the direct discharge can pollute the environment and waste water resources.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to solve the technical problems that: provides an improvement of a chemical feeding control method of slime water.
In order to solve the technical problems, the invention adopts the technical scheme that: a chemical adding control method for slime water comprises the following steps:
s1: constructing an empirical model: summarizing long-term working experience of a preschool scholars and field workers to obtain an empirical mathematical model as an empirical model for dosing control;
s2: constructing a prediction model: learning historical data by using an LSTM network to obtain a prediction model, and performing model optimization according to real-time data;
s3: and (3) dosing control: detecting and acquiring real-time data of a coal slime water feeding material through a sensor, transmitting the data to a controller, and dynamically adding medicine according to an empirical model;
and inputting the prediction input data detected by the sensor into a prediction model to obtain the predicted overflow concentration, and realizing the advanced adjustment of the dosage according to the difference between the predicted overflow concentration and the set overflow concentration.
The construction steps of the prediction model in the step S2 are as follows:
s2.1: obtaining historical data which comprises a feeding flow Q, a feeding density S, a dosing amount V, an overflow concentration L of the slime water in a concentration tank and an actual overflow concentration L' of the slime water at t + delta t at the moment t, wherein t is the current moment, and delta t is the difference between the measured actual overflow concentration and the dosing time;
s2.2: removing and interpolating abnormal values of historical data, and finally performing normalization processing on the abnormal values;
s2.3: and dividing historical data into prediction model input data and prediction model output data, inputting the historical data into a prediction model to train the model, and continuously optimizing the model.
The input data of the prediction model comprise a feeding flow Q, a feeding concentration S, a dosing amount V and a coal slime overflow concentration L at the time t after data processing;
the output data of the prediction model comprises the actual coal slime water overflow concentration L' at t + delta t after data processing;
the prediction model is obtained by utilizing LSTM to train according to input and output;
the model optimization is to incorporate the data obtained in real time into the historical data, and continuously update the historical data, so as to continuously improve the accuracy of the prediction model.
The LSTM neural network structure of the prediction model comprises an input layer, a first LSTM neural network layer, a first Dropout layer, a second LSTM neural network layer, a second Dropout layer, a full connection layer and an output layer which are sequentially connected, training is completed through an Adam optimization algorithm, and evaluation indexes of prediction effects adopt root mean square errors and average absolute percentage errors.
The LSTM neural network layer comprises a forgetting gate f t And input gate i t And an output gate o t The calculation formula is as follows:
f t ＝σ(W f ·[h t-1 ，x t ]+b f )；
i t ＝σ(W i ·[h t-1 ，x t ]+b i )；
o t ＝σ(W o [h t-1 ，x t 」+b o )；
in the above formula: h is a total of t-1 Is the output vector at the last moment, x t Is the input vector of the current moment, sigma is a Sigmoid activation function, W f And b f Network parameter vector, W, for forgetting gates f Is a first weight coefficient, b f Is a first bias coefficient; w is a group of i And b i As a network parameter vector of the input gate, W i Is a second weight coefficient, b i Is a second bias coefficient; w is a group of o And b o To output the network parameter vector, W, of the gate o Is a third weight coefficient, b o Is the third bias coefficient.
The step S3 of dynamically adding medicine according to the empirical model comprises the following steps:
inputting the dry coal slime as an empirical model to obtain a coarse chemical adding amount for dynamic chemical adding control, wherein the dry coal slime is obtained by the following steps:
obtaining real-time data of coal slime water feeding flow: flow detection is carried out by adopting a Doppler ultrasonic flowmeter, and real-time data is transmitted to a controller;
obtaining real-time data of coal slime water feeding concentration: firstly, detecting the feeding density by using a differential pressure type densimeter, and then converting the feeding density to obtain the feeding concentration, wherein the conversion formula is as follows:
in the formula, beta x The ore pulp density, beta, and S are the dry coal slime density and the feed concentration respectively;
and (3) calculating the dry coal slime amount, wherein the formula is as follows:
f(t)＝Q*S；
wherein f (t) is the amount of fed dry coal slime at the time t.
Obtaining the output medicine quantity according to the empirical model, wherein the calculation formula of the output medicine quantity is as follows:
in the formula, c 1 Is when 0 < f(t) is not more than Q1S 1, (f (t) -c'). Times.k is the dosage when Q1S 1 is not more than f (t) is not more than Q2S 2, c 2 And when f (t) > Q2S 2, H (t) is the dosage at the time of t, Q1S 1 is the amount of dry coal slime added in the first stage, the value is 70t/H, Q2S 2 is the amount of dry coal slime added in the second stage, the value is 110t/H, c' is a constant, the value is 5, k is a dosing coefficient, and the value range is 70-120 g/ton of dry coal slime.
The step S3 of adopting a prediction model to adjust the dosage in advance comprises the following steps:
inputting the real-time measured feeding flow Q, feeding concentration S, dosing quantity V and coal slime overflow concentration L which are subjected to data processing at the time t into a prediction model so as to obtain an overflow concentration predicted value L' at the time t + delta t, converting the deviation delta L of the overflow concentration predicted value and a set value into a dosing correction quantity e, inputting the dosing correction quantity e into a PID controller to obtain a corrected value delta u of the dosing quantity of the flocculating agent, summing the corrected value delta U with the coarse dosing quantity u output by an empirical model, correcting the coarse dosing quantity, and realizing advanced regulation dosing; selecting a language variable of the dosing correction quantity E as E;
and (3) obtaining the flocculant PAM regulating quantity under feedback control through PID control as follows:
D(t)＝Kc[E(t)-E(t-1)]+KiE(t)+Kd[E(t)-2E(t-1)+E(t-2)]；
in the formula, kc, ki and Kd are respectively a proportional coefficient, an integral coefficient and a differential coefficient, E (t) is the correction deviation of the dosage at the moment t, and the expression is as follows:
Kc(t)＝E(t)-E(t-1)；
Ki(t)＝E(t)；
Kd(t)＝E(t)-2E(t-1)+E(t-2)；
E(t)＝λ(L″-Ls)；
wherein Ls is an overflow concentration set value, and the lambda dosing correction coefficient.
A chemical feeding control system for coal slime water comprises a data acquisition module, a control module and a chemical feeding pump, wherein the data acquisition module comprises a flowmeter, a densimeter and a concentration meter in a concentration tank which are arranged on a coal slime water feeding pipeline and is used for acquiring flow, density and concentration data in the coal slime water concentration and chemical feeding control process in real time; the feeding pump is used for feeding a flocculating agent PAM into the feeding pipe; the control module is respectively connected with the data acquisition module and the dosing pump and comprises an industrial computer, an industrial switch, a controller, a touch screen display and a frequency converter; the touch screen display can display data collected in real time, and the dosing can be manually adjusted through the touch screen display; an empirical model program is stored in the controller, and the chemical dosing pump is controlled to perform coarse chemical dosing by regulating and controlling the output frequency of the frequency converter; the industrial computer is stored with an LSTM program, is connected with a dosing pump through an industrial exchanger, a controller and a frequency converter, and corrects the coarse dosing amount through PID control.
The flowmeter is a Doppler ultrasonic flowmeter, the densimeter is a differential pressure densimeter, the densimeter is a scattering concentration meter, and the dosing pump is a diaphragm dosing pump.
Compared with the prior art, the invention has the beneficial effects that: the system acquires real-time online data by collecting various data in the coal slime water treatment process, and the real-time online data is used as a prediction basis for the coal slime water overflow concentration prediction value, and an empirical model is adopted for feedforward control, so that the medicine can be dynamically added according to the feeding condition in real time, the structure is simple, and the labor intensity is reduced; the method adopts the LSTM neural network for prediction, can quickly and real-timely obtain the overflow concentration in advance, adjusts the adding amount of the flocculating agent through PID control, solves the lag problem of traditional manual adjustment, reduces the medicine consumption and simultaneously ensures that the concentration of the overflow water stably reaches the standard.
Drawings
The invention is further described below with reference to the accompanying drawings:
FIG. 1 is a schematic diagram of the system of the present invention;
FIG. 2 is a schematic diagram of a control method of the present invention;
FIG. 3 is a logical block diagram of the LSTM neural network of the present invention;
FIG. 4 is a schematic diagram of a prediction model according to the present invention;
FIG. 5 is a graph illustrating predicted results according to the present invention;
in the figure: 1 is an industrial computer, 2 is an industrial exchanger, 3 is a controller, 4 is a frequency converter, 5 is a touch screen display, 6 is a flowmeter, 7 is a densimeter, 8 is a buffer barrel, 9 is a dosing pump, 10 is a concentration meter, 11 is a flocculant solution storage barrel, and 12 is a concentration tank.
Detailed Description
As shown in figures 1 to 5, the chemical feeding control method and the chemical feeding control system for the coal slime water solve the problems that a flocculating agent cannot be dynamically fed and hysteresis exists in regulation.
The purpose of the invention can be realized by the following technical scheme:
a chemical feeding control method for slime water comprises the following steps:
firstly, establishing an empirical model: summarizing long-term working experience of a preschool learner and field workers to obtain an empirical mathematical model serving as an empirical model for dosing control;
secondly, constructing a prediction model: learning historical data by using an LSTM network to obtain a prediction model, and performing model optimization according to real-time data;
thirdly, controlling the dosing: real-time data of the coal slime water feeding material is obtained through sensor detection and transmitted to a PLC (programmable logic controller), and the medicine is dynamically added according to an empirical model; and inputting the prediction input data detected by the sensor into a prediction model to obtain the predicted overflow concentration, and realizing the advanced adjustment of the dosage according to the difference between the predicted overflow concentration and the set overflow concentration.
The specific process is as follows:
empirical model construction
Converting the step change of the feeding flow Q and the concentration S into dry coal slime, wherein the amount of the dry coal slime is in direct proportion to the addition amount of the flocculating agent and the flocculating agent under the condition of not considering other factors, but the flocculating agent is added at the fastest speed when the concentration and the flow of a feeding pipe are large in the actual addition of the medicament; when the concentration and the flow of the feeding pipe are both small, the adding speed of the flocculating agent is slowest; when the concentration and the flow of the feeding pipe are in a medium degree, the adding speed of the flocculating agent is also in a medium speed, so that an empirical model is summarized through long-term working experience of forelearners and field workers, the dry coal slime is input as a prediction model, and the crude adding quantity is obtained to carry out dynamic adding control. Because the dry coal slime quantity has no direct detection instrument, the dry coal slime quantity is obtained by adopting a method of indirect measurement for calculating the concentration and the flow of the coal slime water, and the method comprises the following specific steps:
1. coal slime water feeding flow real-time data acquisition: flow detection is carried out by adopting a Doppler ultrasonic flowmeter, and real-time data is transmitted to a PLC (programmable logic controller);
2. obtaining real-time data of coal slime water feeding concentration: the method is characterized in that the method is limited by a detection sensor, a pressure difference type densimeter is adopted to detect the feeding density, and then the feeding concentration is obtained through conversion according to the feeding density, wherein the conversion formula is as follows:
in the formula, beta x The ore slurry density, beta the dry coal slime density and S the feed concentration.
3. And (3) calculating the dry coal slime quantity, wherein the formula is as follows:
f(t)＝Q*S；
wherein f (t) is the amount of fed dry coal slime at the time t.
Obtaining the output drug quantity according to an empirical model, wherein the output drug quantity is as follows:
in the formula, c 1 When f (t) is more than 0 and less than or equal to Q 1 *S 1 The dose of (f (t) -c'). Times.k is when Q 1 *S 1 ＜f(t)≤≤Q 2 *S 2 The amount of the drug added in time, c 2 When f (t) > Q 2 *S 2 The dosage of the medicine is H (t), and the dosage of the medicine at the time t is H (t). Q1 and S1 take the value of 70t/h, Q2 and S2 take the value of 110t/h, c' takes the value of 5, k is a dosing coefficient, and the value range is 70-120 g/ton of dry coal slime.
(II) construction of prediction model
The specific steps of the construction of the prediction model in the second step are as follows:
1) Obtaining historical data which comprises a feeding flow Q, a feeding density S, a dosing amount V, an overflow concentration L of the slime water in a concentration tank and an actual overflow concentration L' of the slime water at t + delta t at the moment t, wherein t is the current moment, and delta t is the difference between the measured actual overflow concentration and the dosing time;
2) Removing and interpolating abnormal values of the historical data, and finally performing normalization processing on the abnormal values, wherein the abnormal value removal is to prevent the abnormal values from influencing the prediction precision of the model; the normalization is to eliminate the large difference of the numerical dimensions of different variables, which may have adverse effect on the model effect when the network is trained and accelerate the network training speed;
3) Dividing historical data into prediction model input data and prediction model output data, and inputting the prediction model input data and the prediction model output data into a prediction model to train the model; the input data of the prediction model comprise a feeding flow Q, a feeding concentration S, a dosing amount V and a coal slime water overflow concentration L at the time t after data processing; the output of the prediction model comprises the actual coal slime overflow concentration L' at t + delta t after data processing; the prediction model is obtained by utilizing LSTM according to input and output training; the model optimization is to incorporate the data obtained in real time each time into the historical data and continuously update the historical data, thereby continuously improving the accuracy of the prediction model.
The method specifically comprises the following steps: 1. data processing
In the coal slime water treatment process, fine-grained coal is increased due to the complex and changeable geological structure of underground mining and the rapid improvement of the coal mining mechanization degree. And the content of quartz, calcite, highly dispersed kaolinite, montmorillonite, illite and other clay minerals with strong hydrophilicity is increased, so that a highly suspended and dispersed coal slime water system with complex components and stable properties is formed. In industrial production, a flocculating agent is added into coal slurry water to promote solid particles to settle. The type and the composition of the medicine and the concentration set value L of the overflow water s Under the relatively fixed condition, the factors that influence the actual detected value of the overflow concentration of the concentration tank include: the feed flow Q, the feed concentration S, the dosing quantity V, the overflow concentration L of the coal slurry in the concentration tank and the like. Due to the large difference of the numerical dimensions of different variables, the network may be trained while matching the modesThe effect of the model has adverse effect and the introduction of abnormal data can reduce the prediction accuracy of the model, so that data processing is required, and the specific steps are as follows:
1) Data cleaning: to eliminate noisy data, the outlier data values are typically cleaned. According to 3 σ In principle, assuming that sample data only contains random errors, calculating standard deviation according to the following formula, then obtaining a variable interval through probability, calling data exceeding the interval as an abnormal value, removing the abnormal value, and filling missing values through linear interpolation;
let n sample data x 1 ，x 2 ，x 3 ，...x n Average value of
The standard deviation formula is as follows:
e.g. deviation of sample data xi
Satisfy the requirements of
Then x is i Deleted as an abnormal value.
2) Normalization treatment: as each parameter of the slime water treatment process data has different dimensions and dimension units, different parameters can be influenced mutually. In order to eliminate the influence of dimension between parameters, the sewage data needs to be normalized to ensure that each parameter is in the same order, which is beneficial to the dimensionless weighting of the data and the rapid convergence of the loss function of the following algorithm, and the data normalization formula is as follows:
wherein x * Denotes the data after normalization, x i Representing data before normalization, x max Is the maximum value of the sample data, x min Is the minimum value of the sample data.
2. LSTM network construction
The invention adopts matlab platform to build LSTM neural network for predicting overflow concentration, the LSTM neural network prediction model comprises: the training of the input layer, the first LSTM neural network layer, the first Dropout layer, the second LSTM neural network layer, the second Dropout layer, the full-connection layer and the output layer which are sequentially connected is completed through an Adam optimization algorithm, and the evaluation index of the prediction effect adopts a root mean square error and an average absolute percentage error. Wherein, as shown in fig. 3, the LSTM cell unit is internally provided with a plurality of thresholds including a forgetting gate f t And an input gate i t And an output gate o t The mathematical expression is as follows:
f t ＝σ(W f .[h t-1 ，x t ]+b f )
i t ＝σ(W i ·[h t-1 ，x t ]+b i )
o t ＝σ(W o [h t-1 ，x t 」+b o )；
wherein h is t-1 Is the output vector at the previous moment, x t Is the input vector of the current moment, sigma is Sigmoid activation function, W f And b f Network parameter vector, W, for forgetting gate f Is a first weight coefficient, b f Is a first bias coefficient; w i And b i As a network parameter vector of the input gate, W i Is a second weight coefficient, b i Is a second bias coefficient; w o And b o As a vector of network parameters of the output gate, W o Is a third weight coefficient, b o Is the third bias coefficient.
As shown in fig. 4, the neural network structure of the prediction model in the present invention includes a first LSTM neural network, a first Dropout layer, a second LSTM neural network, and a second Dropout layer, which are connected in sequence. In the embodiment, the matlab is used as a modeling environment to model and learn and train the LSTM neural network; the evaluation index of the prediction effect adopts a root mean square error and an average absolute percentage error, and the training of the LSTM neural network is completed through an adam optimization algorithm, so that the Dropout layer can prevent overfitting, and the generalization capability of the model is improved.
(III) control of dosing
1. Coarse dosing control
And in the third step, dynamic dosing is realized by the empirical model, wherein a coarse dosing amount u is obtained by converting the real-time feeding flow Q and the feeding concentration S measured by the sensors into the dry coal slime amount, the coarse dosing amount u is converted into an output frequency which is used as the input of a frequency converter, and the frequency of a dosing pump is controlled by the frequency converter.
The data input by the empirical model comprises a feeding flow Q and a feeding concentration S; as shown in fig. 1, the concentration meter and the flow meter collect data, transmit the data to the PLC in which an empirical model program is pre-stored, obtain the coarse dosage through the empirical model, and control the output power of the frequency converter through the PLC, thereby achieving the real-time adjustment of the dosage.
2. Predictive feedback regulation control
The third step of adjusting the dosage in advance is to input the feed flow Q, the feed concentration S, the dosage V and the coal slime overflow concentration L which are measured in real time at the time t and are subjected to data processing into a prediction model so as to obtain an overflow concentration predicted value L' at the time t + delta t, change the deviation delta L of the overflow concentration predicted value and a set value into a dosage correction quantity e, input the dosage correction quantity e into a PID controller so as to obtain a corrected value delta u of the dosage of the flocculating agent, sum the corrected value delta u with the coarse dosage u output by an empirical model, and correct the coarse dosage so as to realize the adjustment of dosage in advance; selecting a language variable of the dosing correction quantity E as E;
the PID comprises a proportional feedback characteristic Kc, an integral feedback characteristic Ki and a differential feedback characteristic Kd, and the expression of the PID is as follows:
Kc(t)＝E(t)-E(t-1)
Ki(t)＝E(t)
Kd(t)＝E(t)-2E(t-1)+E(t-2)。
the data input by the prediction model comprises a feeding flow Q, a feeding concentration S, a feeding amount V and an overflow concentration L of the coal slurry in the concentration tank at the time t, the data needs to be processed firstly, and then the data is input into the prediction model, and the prediction result is shown in fig. 5.
The control process is as shown in fig. 2, and according to an overflow concentration predicted value L ″ output by the prediction model and a set value Ls, a dosage correction deviation E (t) can be calculated, and the specific formula is as follows:
E(t)＝λ(L″-Ls)；
wherein E (t) is a dosing amount correction quantity converted by the difference value of the actual overflow concentration predicted value L' and the overflow water concentration set value Ls at the time t.
Through PID control, the adjustment quantity of the flocculating agent PAM under feedback control can be obtained as follows:
D(t)＝Kc[E(t)-E(t-1)]+KiE(t)+Kd[E(t)-2E(t-1)+E(t-2)]；
in the formula, kc, ki and Kd are respectively a proportional coefficient, an integral coefficient and a differential coefficient.
And summing the correction quantity and the coarse dosing quantity of the flocculating agent, converting and inputting the sum into a frequency converter, and realizing the advanced adjustment of the flocculating agent PAM.
As shown in figure 1, the invention also provides a coal slime water dosing system which comprises an acquisition module, a control module and a dosing pump, wherein the acquisition module comprises a flowmeter 6, a densimeter 7 and a concentration meter 10, the flowmeter 6 and the densimeter 7 are arranged in the feeding pipe, and the concentration meter 10 is arranged in the concentration tank 12 and connected with the controller 3 and used for acquiring relevant data of a coal slime water treatment process and used as a basis for dynamically dosing and adjusting the dosing amount in advance. The dosing pump 9 is connected with the control module of the frequency converter 4 and is used for dosing. The control module comprises an industrial computer 1, an industrial switch 2, a PLC controller 3 and a frequency converter 4, wherein an empirical model program is stored in the PLC, the frequency output by the frequency converter 4 is regulated to control a dosing pump 9 to perform coarse dosing, an LSTM program is stored in the industrial computer 1, the dynamic dosing and advanced adjustment of the dosing quantity provided by the invention are realized by acquiring acquired real-time data as input data and executing the program, and finally the frequency converter 4 is connected with the dosing pump 9 to control the dynamic dosing and adjustment of a flocculant PAM in a slime water concentration tank through the frequency control dosing pump. Wherein flowmeter 6 is doppler ultrasonic flowmeter, densimeter 7 is the differential densimeter, concentration meter 10 is astigmatic formula concentration, dosing pump 9 is the diaphragm dosing pump.
It should be noted that, regarding the specific structure of the present invention, the connection relationship between the modules adopted in the present invention is determined and can be realized, except for the specific description in the embodiment, the specific connection relationship can bring the corresponding technical effect, and the technical problem proposed by the present invention is solved on the premise of not depending on the execution of the corresponding software program.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. A chemical adding control method for slime water is characterized by comprising the following steps: the method comprises the following steps:
s1: constructing an empirical model: summarizing long-term working experience of a preschool learner and field workers to obtain an empirical mathematical model serving as an empirical model for dosing control;
s2: constructing a prediction model: learning historical data by using an LSTM network to obtain a prediction model, and performing model optimization according to real-time data;
s3: controlling the dosing: detecting and acquiring real-time data of a coal slime water feed through a sensor, transmitting the data to a controller, and dynamically adding medicine according to an empirical model;
inputting prediction input data detected by a sensor into a prediction model to obtain a prediction overflow concentration, and realizing advanced adjustment of the dosage according to the difference between the prediction overflow concentration and the set overflow concentration;
the construction steps of the prediction model in the step S2 are as follows:
s2.1: obtaining historical data including a feeding flow Q, a feeding density S, a dosing amount V, an overflow concentration L of the slime water in the concentration tank at time t, and an actual overflow concentration L' of the slime water at time t +/t, wherein t is the current time, and t is the difference between the measured actual overflow concentration and the dosing time;
s2.2: abnormal value elimination and interpolation are carried out on historical data, and finally normalization processing is carried out on the abnormal value elimination and interpolation;
s2.3: and dividing historical data into prediction model input data and prediction model output data, inputting the historical data into a prediction model to train the model, and continuously performing model optimization.
2. The chemical feeding control method for slime water as claimed in claim 1, wherein the chemical feeding control method comprises the following steps: the input data of the prediction model comprise a feeding flow Q, a feeding concentration S, a dosing amount V and a coal slime overflow concentration L at the time t after data processing;
the output data of the prediction model comprises the actual coal slime overflow concentration L' at t +/t after data processing;
the prediction model is obtained by utilizing LSTM according to input and output training;
the model optimization is to incorporate the data obtained in real time each time into the historical data and continuously update the historical data, thereby continuously improving the accuracy of the prediction model.
3. The chemical feeding control method for slime water as claimed in claim 2, wherein: the LSTM neural network structure of the prediction model comprises an input layer, a first LSTM neural network layer, a first Dropout layer, a second LSTM neural network layer, a second Dropout layer, a full connection layer and an output layer which are sequentially connected, training is completed through an Adam optimization algorithm, and evaluation indexes of prediction effects adopt root mean square errors and average absolute percentage errors.
4. The chemical feeding control method for slime water as claimed in claim 3, wherein: the LSTM neural network layer comprises a forgetting gate f t And input gate i t And an output gate o t The calculation formula is as follows:
in the above formula: h is t-1 Is the output vector at the previous moment, x t Is the input vector of the current moment, sigma is Sigmoid activation function, W f And b f Network parameter vector, W, for forgetting gate f Is a first weight coefficient, b f Is a first bias coefficient; w is a group of i And b i As a network parameter vector of the input gate, W i Is a second weight coefficient, b i Is a second bias coefficient; w o And b o As a vector of network parameters of the output gate, W o Is a third weight coefficient, b o Is the third bias coefficient.
5. The chemical feeding control method for slime water as claimed in claim 1, wherein the chemical feeding control method comprises the following steps: the step S3 of dynamically adding medicine according to the empirical model comprises the following steps:
inputting the dry coal slime as an empirical model to obtain a coarse chemical adding amount for dynamic chemical adding control, wherein the dry coal slime is obtained by the following steps:
coal slime water feeding flow real-time data acquisition: flow detection is carried out by adopting a Doppler ultrasonic flowmeter, and real-time data are transmitted to a controller;
obtaining real-time data of coal slime water feeding concentration: firstly, detecting the feeding density by using a differential pressure type densimeter, and then converting the feeding density to obtain the feeding concentration, wherein the conversion formula is as follows:
in the formula (I), the compound is shown in the specification,β x the density of the ore pulp is shown as,βdry coal slurry density is obtained, and S is feed concentration;
and (3) calculating the dry coal slime amount, wherein the formula is as follows:
f(t)=Q*S；
wherein f (t) is the fed dry coal slime amount at the time t;
obtaining the output medicine quantity according to the empirical model, wherein the calculation formula of the output medicine quantity is as follows:
in the formula, c 1 Is when 0<The dosage of f (t) is less than or equal to Q1S 1, (f (t) -c'). Times.k is when Q1S 1<f (t) is less than or equal to Q2S 2, c 2 When f (t)>The dosage of Q2S 2, H (t) is the dosage of t, Q1S 1 is the amount of dry coal slime added in the first stage, Q2S 2 is the amount of dry coal slime added in the second stage, c' is a constant, k is a dosing coefficient, and the value range is 70-120 g/ton of dry coal slime.
6. The coal slime water dosing control method according to claim 1, wherein the method comprises the following steps: the step S3 of adopting a prediction model to adjust the dosage in advance comprises the following steps:
inputting the real-time measured feeding flow Q, feeding concentration S, dosing quantity V and coal slime overflow concentration L which are subjected to data processing at time t into a prediction model to obtain an overflow concentration prediction value L '' at time t +/t, changing the deviation L of the overflow concentration prediction value and the set value into a dosing correction quantity e, inputting the dosing correction quantity e into a PID controller to obtain a corrected value u of the dosing quantity of the flocculant, summing the corrected value u with the coarse dosing quantity u output by the empirical model, and correcting the coarse dosing quantity to realize advanced regulation and dosing; selecting a language variable of the dosing correction quantity E as E;
and obtaining the adjustment quantity of the flocculating agent PAM under feedback control through PID control as follows:
D(t)＝Kc [E(t)-E(t-1)]+KiE(t)+Kd[E(t)-2E(t-1)+E(t-2)]；
in the formula, kc, ki and Kd are respectively a proportional coefficient, an integral coefficient and a differential coefficient, E (t) is the dosage correction deviation at the moment t, and the expression is as follows:
Kc(t)＝E(t)-E(t-1) ；
Ki(t)＝E(t) ；
Kd(t)＝E(t)-2E(t-1)+E(t-2) ；
E(t)＝λ（L''-Ls）；
wherein Ls is an overflow concentration set value, and the lambda dosing correction coefficient.
7. The utility model provides a coal slime water adds medicine control system which characterized in that: the system comprises a data acquisition module, a control module and a dosing pump, wherein the data acquisition module comprises a flowmeter, a densimeter and a concentration meter in a concentration tank which are arranged on a coal slime water feeding pipeline and is used for acquiring flow, density and concentration data in the coal slime water concentration dosing control process in real time; the dosing pump is used for dosing a flocculating agent PAM into the feeding pipe; the control module is respectively connected with the data acquisition module and the dosing pump and comprises an industrial computer, an industrial exchanger, a controller, a touch screen display and a frequency converter; the touch screen display can display data collected in real time, and the dosing can be manually adjusted through the touch screen display; an empirical model program is stored in the controller, and the medicine feeding pump is controlled to carry out coarse medicine feeding by regulating and controlling the output frequency of the frequency converter; the industrial computer stores an LSTM program, is connected with a dosing pump through an industrial switch, a controller and a frequency converter, and corrects the coarse dosing amount through PID control.
8. The chemical feeding control system for slime water as claimed in claim 7, wherein: the flowmeter is a Doppler ultrasonic flowmeter, the densimeter is a differential densimeter, the densimeter is a scattering concentration meter, and the dosing pump is a diaphragm dosing pump.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210162651.5A CN114380379B (en) | 2022-02-22 | 2022-02-22 | Chemical adding control method and system for slime water |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210162651.5A CN114380379B (en) | 2022-02-22 | 2022-02-22 | Chemical adding control method and system for slime water |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114380379A CN114380379A (en) | 2022-04-22 |
| CN114380379B true CN114380379B (en) | 2023-01-20 |
Family
ID=81206189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210162651.5A Active CN114380379B (en) | 2022-02-22 | 2022-02-22 | Chemical adding control method and system for slime water |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114380379B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116434857B (en) * | 2023-03-24 | 2024-03-15 | 青岛思普润水处理股份有限公司 | Intelligent dosing setting method of sewage treatment system |
| CN117192973A (en) * | 2023-11-08 | 2023-12-08 | 山西智河水利科技有限公司 | Automatic yellow river mud water recovery system of control |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105676814A (en) * | 2016-01-11 | 2016-06-15 | 大唐环境产业集团股份有限公司 | SFLA-SVM-based digital water island online agent adding control method |
| CN111533290A (en) * | 2020-04-30 | 2020-08-14 | 重庆市水务资产经营有限公司 | Method for generating optimal operation plan library of sewage treatment process and applying complex scene |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2021149472A (en) * | 2020-03-18 | 2021-09-27 | 日新電機株式会社 | Device and method for predicting quality of discharged water |
| CN111994970B (en) * | 2020-07-31 | 2022-06-21 | 上海上实龙创智能科技股份有限公司 | LSTM-based dosing prediction method and dosing system for efficient sewage sedimentation tank |
| CN113433910A (en) * | 2021-06-30 | 2021-09-24 | 武汉理工大学 | Water purification plant intelligent dosing control system and method based on digital twin |
-
2022
- 2022-02-22 CN CN202210162651.5A patent/CN114380379B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105676814A (en) * | 2016-01-11 | 2016-06-15 | 大唐环境产业集团股份有限公司 | SFLA-SVM-based digital water island online agent adding control method |
| CN111533290A (en) * | 2020-04-30 | 2020-08-14 | 重庆市水务资产经营有限公司 | Method for generating optimal operation plan library of sewage treatment process and applying complex scene |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114380379A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114380379B (en) | Chemical adding control method and system for slime water | |
| CN106933099B (en) | Coal preparation plant thickener and filter press medicament adding cooperative control system | |
| CN103011356B (en) | Method for controlling automatic chemical dosing of high-turbidity water system | |
| WO2021057349A1 (en) | Intelligent control system and method for heavy medium separation process | |
| CN102385315B (en) | Intelligent coagulation chemical dosing control system for water plant and control method thereof | |
| CN110054274B (en) | Water purification flocculation precipitation dosing control method | |
| CN103344530B (en) | Soft measurement method for granularity of cement raw material grinded by ball mill | |
| CN114230110B (en) | Short-range intelligent phosphorus removal and medicine addition control method, equipment and system for sewage treatment | |
| CN101825870B (en) | Method and system for controlling supply quantity of water-treatment flocculating agent | |
| CN103744293A (en) | Waste water treatment monitoring method and system based on fuzzy neural network | |
| CN102122134A (en) | Method and system for wastewater treatment of dissolved oxygen control based on fuzzy neural network | |
| CN102169077B (en) | Hybrid intelligent soft measuring method of overflow granularity index in wet grinding process | |
| CN101607151B (en) | Embedded intelligent dosing controller and control method thereof | |
| CN103173584A (en) | Blast furnace burden-distribution control system with self-learning control function | |
| CN106745707A (en) | A kind of control system and control method of sewage disposal sludge material balance | |
| CN110090478B (en) | Intelligent control method for deep cone thickener in filling scene | |
| CN102008998A (en) | Automatic control device and automatic control method of hydrocyclone | |
| CN115108617B (en) | Method and system for adding medicine by coagulation | |
| CN113705898A (en) | Method and system for predicting dosage of tap water coagulant | |
| CN105301960A (en) | Method for controlling input amount of tap water flocculating agent | |
| CN103500280A (en) | Cell concentration online soft measurement method in glutamic acid fermentation process | |
| CN115318431A (en) | Control method and device for heavy-medium shallow slot system and processor | |
| CN114781249A (en) | High-density clarification tank dosage prediction and control method based on multidimensional scoring model | |
| KR20060092660A (en) | Automatic control device and method for wastewater treatment using fuzzy control | |
| CN113104945A (en) | Intelligent coal slime water dosing method and system based on floc characteristic feedback |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |