CN116721711B - Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving - Google Patents
Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving Download PDFInfo
- Publication number
- CN116721711B CN116721711B CN202310724171.8A CN202310724171A CN116721711B CN 116721711 B CN116721711 B CN 116721711B CN 202310724171 A CN202310724171 A CN 202310724171A CN 116721711 B CN116721711 B CN 116721711B
- Authority
- CN
- China
- Prior art keywords
- data
- electrostatic
- sensor
- solid
- concentration
- 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
- 239000007787 solid Substances 0.000 title claims abstract description 172
- 239000002245 particle Substances 0.000 title claims abstract description 129
- 238000000034 method Methods 0.000 title claims abstract description 67
- 239000000428 dust Substances 0.000 title claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 73
- 230000006698 induction Effects 0.000 claims abstract description 72
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 32
- 238000011176 pooling Methods 0.000 claims abstract description 22
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 230000004913 activation Effects 0.000 claims abstract description 16
- 230000008859 change Effects 0.000 claims abstract description 14
- 239000000463 material Substances 0.000 claims abstract description 14
- 230000008569 process Effects 0.000 claims abstract description 13
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 41
- 238000000691 measurement method Methods 0.000 claims description 10
- 238000011144 upstream manufacturing Methods 0.000 claims description 9
- 238000005259 measurement Methods 0.000 claims description 7
- 238000012544 monitoring process Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 230000003111 delayed effect Effects 0.000 claims description 3
- 238000009434 installation Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 2
- 239000012530 fluid Substances 0.000 claims description 2
- 238000013480 data collection Methods 0.000 claims 2
- 230000002265 prevention Effects 0.000 abstract description 8
- 238000004134 energy conservation Methods 0.000 abstract description 2
- 230000007613 environmental effect Effects 0.000 abstract description 2
- 230000003068 static effect Effects 0.000 description 28
- 238000004519 manufacturing process Methods 0.000 description 14
- 230000000694 effects Effects 0.000 description 7
- 239000003245 coal Substances 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000005457 optimization Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000002485 combustion reaction Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 239000002817 coal dust Substances 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000008014 freezing Effects 0.000 description 3
- 238000007710 freezing Methods 0.000 description 3
- 239000007789 gas Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000010291 electrical method Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 235000013312 flour Nutrition 0.000 description 1
- 239000010881 fly ash Substances 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 238000000227 grinding Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000009774 resonance method Methods 0.000 description 1
- 238000000790 scattering method Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 238000003325 tomography Methods 0.000 description 1
- 230000005514 two-phase flow Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Crystallography & Structural Chemistry (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Chemical & Material Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Measuring Volume Flow (AREA)
Abstract
本发明实施例公开的基于数据驱动的粉尘气固流固体颗粒浓度预测方法及系统，涉及节能环保技术领域。包括：获得粉尘运动静电感应及温度、湿度和内外气压差数据；通过互相关法计算固体颗粒的流速，根据输入物料质量和流速计算得到对照固体颗粒浓度数据；对获得的所有数据做归一化预处理，得到训练样本数据和测试样本数据；将所述训练样本数据输入卷积神经网络进行训练，得到浓度预测模型；以及，将所述测试样本数据输入所述浓度预测模型，输出得到固体颗粒浓度值和动态变化曲线图；所述卷积神经网络含有线性卷积模块、偏置常数向量、非线性激活函数、池化层和防过拟合正则化模块。便于快速准确地预测出粉尘等气固两相物中固体颗粒物浓度。
The data-driven dust gas solid flow solid particle concentration prediction method and system disclosed in the embodiments of the present invention relate to the technical field of energy conservation and environmental protection. Including: obtaining dust movement electrostatic induction and temperature, humidity and internal and external air pressure difference data; calculating the flow rate of solid particles through the cross-correlation method, and calculating the control solid particle concentration data based on the input material mass and flow rate; doing normalization pre-processing of all the data obtained Process to obtain training sample data and test sample data; input the training sample data to a convolutional neural network for training to obtain a concentration prediction model; and input the test sample data to the concentration prediction model and output to obtain the solid particle concentration value and dynamic change curve; the convolutional neural network contains a linear convolution module, a bias constant vector, a nonlinear activation function, a pooling layer and an overfitting prevention regularization module. It is convenient to quickly and accurately predict the concentration of solid particles in gas-solid two-phase materials such as dust.
Description
Technical Field
The invention belongs to the technical field of energy conservation and environmental protection, and particularly relates to a dust gas-solid flow solid particle concentration prediction method and system based on data driving.
Background
The concentration of solid particles is one of the important parameters describing the flow characteristics of the gas-solid two-phase flow. The method can accurately and reliably predict the concentration of the solid particles on line, greatly improve the production efficiency, reduce the energy consumption, help enterprises to optimize the industrial production process and realize cost reduction and synergy.
Taking a thermal power plant as an example, the concentration of coal dust in a pipeline has an important influence on the combustion condition of a boiler. The variation of the concentration of the pulverized coal directly affects the uniformity of the distribution of the pulverized coal in the pipeline and the stable combustion of the pulverized coal in the boiler. The pulverized coal with proper concentration is conveyed in the pipeline, so that the pulverized coal can be combusted efficiently, the energy utilization rate is improved, and the emission of fly ash and greenhouse gases is reduced. Therefore, the concentration of pulverized coal solid particles in a powder conveying pipe of a boiler combustion system is an important parameter in the pneumatic conveying process. The reasonable concentration can effectively improve the utilization efficiency of the coal dust and reduce the emission of pollutants.
Although the particulate matters conveyed by different pneumatic conveying systems are different, the technical principle of measuring the concentration of the particulate matters is similar. Many measurement methods such as attenuation method, scattering method, resonance method, electrical method, tomography method, electrostatic method, microwave method, etc. have been proposed. These methods have advantages and disadvantages and different applicable conditions. Existing sensors include acoustic sensors, optical sensors, digital image sensors, capacitive sensors, electrostatic sensors, microwave sensors, radiation sensors, and the like. However, due to the complex dynamic state of gas-solid flow, the concentration is difficult to accurately measure by the measurement principle of each sensor.
Many factors affect the validity of the measurement data. The velocity and concentration of solid particles in pneumatic conveying pipelines are affected by the gas velocity, fluid concentration, pipeline orientation and particle characteristics. And particle density, particle size and shape also affect particle flow characteristics. Thus, accurately measuring solid particle concentrations on-line is challenging and more tricky to predict. It is particularly important to provide a reliable method for online measurement and prediction of the concentration of solid particles such as dust.
Disclosure of Invention
Therefore, the embodiment of the invention aims to provide a method and a system for predicting the concentration of dust gas-solid flow solid particles based on data driving, which are convenient for rapidly and accurately predicting the concentration of solid particles in gas-solid two-phase matters such as dust and the like.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
the method for predicting the concentration of the dust gas-solid flow solid particles based on data driving is characterized by comprising the following steps of:
s10, acquiring dust movement static induction data by using a static induction measurement method, and correspondingly acquiring temperature data, humidity data and internal and external air pressure difference data in a pipeline through a temperature sensor, a humidity sensor and an air pressure sensor;
s20, calculating the flow velocity of the solid particles by a cross-correlation method, and calculating to obtain concentration data of the reference solid particles according to the mass of the input material and the flow velocity;
s30, carrying out normalization pretreatment on the temperature data, the humidity data and the internal and external air pressure difference data obtained in the step S10 and the control solid particle concentration data obtained in the step S20 to obtain training sample data and test sample data;
s40, inputting the training sample data into a convolutional neural network for training to obtain a concentration prediction model;
inputting the test sample data into the concentration prediction model, and outputting to obtain a solid particle concentration value and a dynamic change curve chart; the convolutional neural network comprises a linear convolutional module, a bias constant vector, a nonlinear activation function, a pooling layer and an overfitting prevention regularization module.
Optionally, in step S10, the obtaining dust movement electrostatic induction data by using an electrostatic induction measurement method includes: a first electrostatic sensor is arranged at the upstream in the pipeline, and a second electrostatic sensor is arranged at the downstream in the pipeline; the first electrostatic sensor and the second electrostatic sensor are identical in structure and are adjacent to each other in the position in the pipeline;
acquiring a first static induction signal output by a first static sensor and acquiring a second static induction signal output by a second static sensor;
in step S20, the calculating the solid particle transport speed by the cross-correlation method includes: determining the delay time of solid particles passing through between the first electrostatic sensor and the second electrostatic sensor according to the cross correlation between the first electrostatic induction signal and the second electrostatic induction signal;
and calculating and determining the flow rate of the solid particles according to the delay time.
Optionally, in step S20, the determining a delay time of the solid particles passing between the first electrostatic sensor and the second electrostatic sensor according to the cross correlation between the first electrostatic sensing signal and the second electrostatic sensing signal includes: according to the first electrostatic induction signal and the second electrostatic induction signal, calculating a correlation coefficient of the first electrostatic induction signal and the second electrostatic induction signal based on a first formula, wherein the correlation coefficient characterizes the cross correlation between the first electrostatic induction signal and the second electrostatic induction signal, and the first formula is as follows:the x (N) is a first static induction signal output by an upstream first static sensor, y (n+k) is a second static induction signal delayed by K time units by a downstream second static sensor, N is the number of the first static induction signals output by the upstream first static sensor, K is the number of preset delay data, K and N are variables and are positive integers, so that K is more than or equal to 0 and less than or equal to K-1, N is more than or equal to 0 and less than or equal to N-K, R xy (k) Representing the correlation coefficient;
according to a second formula, adjusting the value of k to change k from small to large to determine a correlation coefficient R xy (k) The corresponding k value at maximum, and is denoted as k 0 The method comprises the steps of carrying out a first treatment on the surface of the The second formula is
According to said k 0 And calculating the data acquisition interval time to obtain the delay time of the solid particles passing through the first electrostatic sensor and the second electrostatic sensor.
Optionally, in step S20, the determining the flow rate of the solid particles according to the delay time calculation includes: calculating the flow velocity of the solid particles according to a third formula based on the delay time and the distance between the first electrostatic sensor and the second electrostatic sensor; the third formula isL is the distance between the first electrostatic sensor and the second electrostatic sensor, k 0 T is the delay time of particles passing between adjacent sensors, and T is the data acquisition interval time.
Optionally, in the step S40, inputting the training sample data into a convolutional neural network for training includes: carrying out one-dimensional convolution on the training sample data by utilizing the linear convolution module, wherein the convolution kernel size is constant;
adding the bias constant vector before the output of the linear convolution module is activated;
applying an activation function to the output of the linear convolution module and outputting the activation function to a pooling layer;
carrying out pooling by adopting a maximum pooling function at a pooling layer, wherein the size of a pooling window is constant;
setting 20% weight freezing by using the overfitting prevention regularization module to prevent network overfitting;
and optimizing the network training parameters through loss function constraint to obtain the concentration prediction model.
Optionally, the optimizing the network training parameter through the constraint of the loss function includes: optimizing network training parameters by adopting a root mean square error loss function; wherein the root mean square error loss function is:
wherein n is the number of sample data of the first electrostatic sensor and the second electrostatic sensor, y i For the control solid particle concentration value, < >>Indicating a predicted value of the solid concentration.
Optionally, in step S30, normalizing the temperature data, the humidity data, and the inside-outside air pressure difference data obtained in step S10, and the control solid particle concentration data obtained in step S20 includes:
using the following maximum and minimum normalization functions
Eliminating the dimensional influence of the different parameters to obtain normalized data; wherein F (x) is a value normalized by x and is located in interval [0,1 ]]；x min Is the minimum value of the sequence; x is x max Is the maximum value of the sequence.
Optionally, in S20, the calculating the reference solid particle concentration data according to the input material mass and the flow rate includes:
multiplying the flow rate by the cross-sectional area of the pipeline to obtain the volume of solid particles passing through the pipeline in unit time;
calculating to obtain the concentration data of the control solid particles based on a fourth formula according to the mass of the material and the volume of the solid particles; wherein the fourth formula is:
P s to approximate the instantaneous solid particle concentration, m is the mass of solid particles passing through the cross section per unit time, A p V is the flow rate of the solid particles, which is the cross-sectional area of the conduit.
Optionally, the installation distance between the first electrostatic sensor and the second electrostatic sensor is 5-20cm, the first electrostatic sensor and the second electrostatic sensor adopt annular electrodes, the outer diameter of the annular electrodes is 10-30 mm, the inner diameter of the annular electrodes is 5-15 mm, and the surface area of the electrodes is larger than 50cm 2 The first electrostatic sensor and the second electrostatic sensor are connected with the signal processor through the signal acquisition card and are used for measuring and monitoring electrostatic induction signals in the gas-solid flow process in real time.
In a second aspect, the present invention also provides a dust gas-solid flow solid particle concentration prediction system based on data driving, which is characterized by comprising: the data acquisition unit is used for acquiring dust movement static induction data by utilizing a static induction measurement method and correspondingly acquiring temperature data, humidity data and internal and external air pressure difference data in the pipeline by a temperature sensor, a humidity sensor and an air pressure sensor;
the first calculation unit is used for calculating the flow velocity of the solid particles through a cross-correlation method and calculating to obtain reference solid particle concentration data according to the mass of the input material and the flow velocity;
the data processing unit is used for carrying out normalization pretreatment on the obtained temperature data, humidity data and internal and external air pressure difference data and the obtained control solid particle concentration data to obtain training sample data and test sample data;
the data training unit is used for inputting the training sample data into a convolutional neural network for training to obtain a concentration prediction model; the convolutional neural network comprises a linear convolutional module, a bias constant vector, a nonlinear activation function, a pooling layer and an overfitting prevention regularization module;
and the concentration prediction unit is used for inputting the test sample data into the concentration prediction model and outputting a solid particle concentration value and a dynamic change curve chart.
According to the data-driven dust gas-solid flow solid particle concentration prediction method and system, the concentration prediction model is built by training the convolutional neural network based on production environment data and deep learning, so that the method and system have strong adaptability and generalization capability, the concentration of solid particles in gas-solid two-phase matters such as dust can be conveniently and accurately predicted, and the online monitoring and change situation tracking of the concentration of the gas-solid flow solid particles such as dust are realized.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an embodiment of a method for predicting the concentration of solid particles in a dust gas-solid stream based on data driving according to the present invention.
FIG. 2 is a schematic diagram showing the layout structure of a dust gas-solid flow solid particle concentration prediction system based on data driving on pipelines in a production environment according to an embodiment of the invention.
FIG. 3 is a schematic diagram of an embodiment of a concentration-predicted convolutional neural network according to the present invention;
FIG. 3a is a schematic illustration of the fully connected layers of the convolutional neural network of FIG. 3;
FIG. 3b is a schematic diagram of an embodiment of a Dropout layer in the convolutional neural network of FIG. 3;
FIG. 4 is a graph of average absolute error of a convolutional neural network during training in accordance with one embodiment of the present invention;
FIG. 5 is a graph of a loss function of a convolutional neural network during training in accordance with one embodiment of the present invention;
FIG. 6 is a graph comparing the predicted effects of a predictive model trained in one embodiment of the invention;
fig. 7 is a schematic block diagram of an embodiment of a data-driven dust gas-solid flow solid particle concentration prediction system according to the present invention.
In fig. 6, color lines are used to distinguish between the contrast effects, and black lines are not easily visualized.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
It should be understood that the described embodiments are merely some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
FIG. 1 is a flow chart of an embodiment of a method for predicting concentration of solid particles in a dust gas-solid stream based on data driving according to the present invention; FIG. 2 is a schematic diagram of an embodiment of the sensor layout of the present invention on a pipeline in a production environment; FIG. 3 is a schematic diagram of a concentration prediction convolutional neural network according to an embodiment of the present invention. Referring to fig. 1 to 3, a method S for predicting concentration of solid particles in a dust gas-solid stream based on data driving according to an embodiment of the present invention includes the steps of:
s10, acquiring dust movement static induction data by using a static induction measurement method, and correspondingly acquiring temperature data, humidity data and internal and external air pressure difference data in the pipeline by using a temperature sensor, a humidity sensor and an air pressure sensor.
Wherein the concentration of the gas-solid flow solid particles is also referred to herein as the concentration of the solid particles.
In some embodiments, as shown in fig. 2, in step S10, the obtaining dust movement electrostatic induction data using the electrostatic induction measurement method includes: a first electrostatic sensor 1 is arranged at the upstream in the pipeline, and a second electrostatic sensor 2 is arranged at the downstream in the pipeline; wherein the first electrostatic sensor 1 and the second electrostatic sensor 2 have the same structure and are adjacent to each other in the position in the pipeline; a first electrostatic induction signal output by the first electrostatic sensor 1 is acquired, and a second electrostatic induction signal output by the second electrostatic sensor 2 is acquired.
When the electrostatic sensor is installed, the distance between the two adjacent first electrostatic sensors and the second electrostatic sensor is L, the distance is not too large, and the arrangement of the distance is preferable to ensure that the gas-solid flow through the pipeline is similar.
Specifically, two capacitive plates can be selected as the electrostatic induction sensor, and an amplifying circuit is arranged at the output end of the capacitive plates and connected to the data acquisition card through the amplifying circuit for converting capacitance change into voltage signals.
The first electrostatic sensor 1 and the second electrostatic sensor 2 should be disposed on a gas-solid flow transmission channel, for example, on an inner wall of a pipe in fig. 2, so as to ensure that a gas-solid flow such as dust stably flows through the measurement area of the electrostatic sensor.
In some embodiments, the installation distance between the first electrostatic sensor 1 and the second electrostatic sensor 2 is 5-20cm, the first electrostatic sensor 1 and the second electrostatic sensor 2 adopt annular electrodes, the outer diameter of the annular electrodes is 10-30 mm, the inner diameter of the annular electrodes is 5-15 mm, and the surface area of the electrodes is larger than 50cm 2 The first electrostatic sensor 1 and the second electrostatic sensor 2 are connected with the signal processor through the signal acquisition card and are used for measuring and monitoring gas in real timeStatic induction signal in the solid flow process.
The electrostatic induction signal may include, for example: electrostatic induction characteristic parameters such as electrostatic charge amount, etc.
Meanwhile, a temperature sensor, a humidity sensor, an air pressure sensor and the like are arranged in the pipeline to obtain temperature, humidity and internal and external air pressure difference data of the pipeline. In some embodiments, the operating environment temperature range is-40-100deg.C, the relative humidity range is 0-95%, and the pipeline pressure range is 0.05-1.6MPa.
In this embodiment, the actual production environment data is collected and used for training the prediction model, and because the actual production environment data is collected in a real complex environment and contains the influence of various noise and interference factors, the model trained by using the data can better adapt to an actual scene, and the prediction result can be more accurate and reliable.
S20, calculating the flow velocity of the solid particles through a cross correlation method, and calculating according to the mass of the input material and the flow velocity to obtain concentration data of the reference solid particles.
In this embodiment, the key point of the cross correlation method is that the electrostatic induction values of the gas-solid flow passing through a short distance are very similar. Therefore, by calculating the short-distance similar static value acquisition interval time, the average speed of the gas-solid flow passing through the short distance L can be obtained.
Specifically, in step S20, the calculating the solid particle transport speed by the cross-correlation method includes: determining the delay time of solid particles passing through between the first electrostatic sensor and the second electrostatic sensor according to the cross correlation between the first electrostatic induction signal and the second electrostatic induction signal; and calculating and determining the flow rate of the solid particles according to the delay time.
In some embodiments, in S20, the calculating the reference solid particle concentration data from the input material mass and flow rate comprises: s21, multiplying the flow velocity by the cross-sectional area of the pipeline to obtain the volume of solid particles passing through the pipeline in unit time; s22, calculating to obtain the concentration data of the control solid particles based on a fourth formula according to the mass of the materials and the volume of the solid particles; wherein the fourth formula is:
P s to approximate the instantaneous solid particle concentration, m is the mass of solid particles passing through the cross section per unit time, A p V is the flow rate of the solid particles, which is the cross-sectional area of the conduit.
In some embodiments, in step S20, the determining a delay time for solid particles to pass between the first electrostatic sensor and the second electrostatic sensor according to the cross correlation between the first electrostatic sensing signal and the second electrostatic sensing signal includes: according to the first electrostatic induction signal and the second electrostatic induction signal, calculating a correlation coefficient of the first electrostatic induction signal and the second electrostatic induction signal based on a first formula, wherein the correlation coefficient characterizes the cross correlation between the first electrostatic induction signal and the second electrostatic induction signal, and the first formula is as follows:the x (N) is a first static induction signal output by an upstream first static sensor, y (n+k) is a second static induction signal delayed by K time units by a downstream second static sensor, N is the number of the first static induction signals output by the upstream first static sensor, K is the number of preset delay data, K and N are variables and are positive integers, so that K is more than or equal to 0 and less than or equal to K-1, N is more than or equal to 0 and less than or equal to N-K, R xy (k) Representing the correlation coefficient;
according to a second formula, adjusting the value of k to change k from small to large to determine a correlation coefficient R xy (k) The corresponding k value at maximum, and is denoted as k 0 The method comprises the steps of carrying out a first treatment on the surface of the The second formula is
According to said k 0 And calculating the data acquisition interval time to obtain the delay time of the solid particles passing through the first electrostatic sensor and the second electrostatic sensor.
In other embodiments, in step S20, the determining the flow rate of the solid particles according to the delay time calculation includes:
calculating the flow velocity of the solid particles according to a third formula based on the delay time and the distance between the first electrostatic sensor and the second electrostatic sensor; the third formula isL is the distance between the first electrostatic sensor and the second electrostatic sensor, k 0 T is the delay time of particles passing between adjacent sensors, and T is the data acquisition interval time.
In the embodiment, the flow velocity of the solid particles (which is exactly similar to the solid particles and is actually gas-solid flow) is determined through calculation based on the double electrostatic sensors and the time delay, so that the device has the advantages of simple structure, high precision, wide application and the like, and has wide application prospects in the field of dust flow velocity monitoring and control in industrial processes.
S30, carrying out normalization pretreatment on the temperature data, the humidity data and the internal and external air pressure difference data obtained in the step S10 and the control solid particle concentration data obtained in the step S20 to obtain training sample data and test sample data.
It can be understood that, because the dimension units of the original data collected by each sensor are inconsistent and do not have the condition of uniform dimension analysis, normalization pretreatment is needed to be carried out on the obtained data such as temperature, humidity, internal and external air pressure difference, static induction value, solid particle concentration and the like, so that further analysis can be facilitated.
In some embodiments, in step S30, normalizing the temperature data, the humidity data, and the inside-outside air pressure difference data obtained in step S10, and the control solid particle concentration data obtained in step S20 includes: using the following maximum and minimum normalization functions
Eliminating the dimensional influence of the different parameters to obtain normalized data; wherein F (x) is a value normalized by x and is located in interval [0,1 ]]；x min Is the minimum value of the sequence; x is x max Is the maximum value of the sequence.
After all the data obtained in the previous step are normalized, a part of the data is used as training sample data for training a prediction model; part of the data is used as verification sample data for model selection and super-parameter tuning; in addition, a part of the data is reserved as test sample data and used for testing and evaluating the prediction accuracy of the prediction model obtained through training.
S40, inputting the training sample data into a convolutional neural network for training to obtain a concentration prediction model;
inputting the test sample data into the concentration prediction model, and outputting to obtain a solid particle concentration value and a dynamic change curve chart; wherein the convolutional neural network comprises a linear convolutional module, a bias constant vector, a nonlinear activation function, a pooling layer and an overfitting prevention regularization module,
preferably, the linear convolution module comprises a plurality of layers.
It can be understood that the convolution operation only covers the local area of the input at a time, and the linear convolution module is arranged to be multi-layer, so that the receptive field of the whole network can be gradually expanded by stacking a plurality of convolution layers, and the method can receive more global information, is favorable for judging the whole input and improves the prediction accuracy of the concentration prediction model.
It can be understood that the convolutional neural network prediction is based on the comparison relationship between the solid particle concentration training value and the input data of the temperature, humidity, air pressure and static value, please refer to fig. 3, the training feature extraction parameters are continuously optimized and adjusted through the loss constraint function, so as to achieve the learning purpose, and after the parameters meeting the loss requirement are obtained, the network model obtained through training can automatically perform concentration prediction analysis on the gas-solid flow solid particle concentration according to the input data, so that the accurate prediction of the gas-solid flow solid particle concentration is realized.
Illustratively, as shown in FIG. 3a, x i To normalize input data, including temperature, humidity, air pressure difference, static induction signal (also called static induction value) A, static induction signal B, throughAnd finally obtaining output data y by multilayer convolution calculation, namely the gas-solid flow solid particle concentration value to be predicted. The specific convolution formula is:
y＝∑…∑σ(W i ×X+b i )
equations 1-6 are neural network convolution processes, x= (X1, X2, X3, X4, X5) is normalized input data vector, W i To be trained and learned for parameter matrix b i For the bias vector, sigma (·) is a nonlinear activation function, the invention adopts a ReLu activation function, which can improve the gradient vanishing problem, and the specific formula is:
σ(x)＝max(0,x)
overfitting often occurs during convolutional neural network learning due to excessive dependence on excessive parameters. The Dropout method is a common and effective regularization technology, as shown in fig. 3b, in the convolutional neural network in the embodiment of the invention, by adopting the Dropout overfitting prevention regularization module, some parameters are disabled (frozen) according to a certain probability and are temporarily discarded from the current neural network, so that certain local characteristic data are not relied on, and the model generalization capability is stronger.
In some embodiments, the overfitting of model training may be reduced by controlling the model complexity of the fully connected layers primarily.
In some embodiments, in the S40, inputting the training sample data into a convolutional neural network for training includes: s41, carrying out one-dimensional convolution on the training sample data by using the linear convolution module, wherein the linear convolution module comprises a plurality of convolution layers, for example, 3-5 convolution layers, and the convolution kernel size is constant, preferably 3.
S42, adding the bias constant vector before the output of the linear convolution module is activated;
s43, applying an activation function to the output of the linear convolution module and outputting the activation function to a pooling layer;
s44, carrying out pooling by adopting a maximum pooling function at a pooling layer, wherein the size of a pooling window is constant;
s45, setting 20% -50% weight freezing by using the overfitting prevention regularization module, and preventing network overfitting;
s46, optimizing network training parameters through loss function constraint to obtain the concentration prediction model.
In this embodiment, by setting a plurality of convolution layers, features of input training sample data can be automatically learned, advanced feature expression is extracted, nonlinearity and robustness of a model are further increased through activation function and maximum pooling, and weight freezing is further used, so that the problem of network overfitting can be alleviated, generalization capability of a prediction model obtained through training is improved, and finally, optimal parameters of the model can be found through optimizing a loss function, which more accords with characteristics of training data than parameters initialized randomly, and is beneficial to improving prediction effects of the obtained concentration prediction model.
To optimize the network training parameters in some embodiments, in step S46, the constraint by the loss function includes: optimizing network training parameters by adopting a root mean square error loss function; wherein the root mean square error loss function is:
wherein n is the number of sample data of the first electrostatic sensor and the second electrostatic sensor, y i For the control solid particle concentration value, < >>Indicating a predicted value of the solid concentration.
In this embodiment, by optimizing the network parameters by using the root mean square error loss function, the model can be assisted to find the optimal parameters, and the prediction accuracy of the concentration prediction model obtained by training is improved.
In general, the effect and optimization objective of convolutional neural networks are achieved through a loss function, which can represent the size of the gap between the current predicted data and the actual data. The smaller the loss value, the better the neural network performance. In the convolution process, the back propagation mechanism is continuously utilized to continuously adjust the value of the neural network parameter, so that the loss value is reduced, and the purpose of training the learning parameter is achieved.
In this embodiment, the convolutional neural network is subjected to parameter optimization training by using the mean square error (Mean Square Error, MSE) as a loss function of the neural network model, so that a prediction model with higher stability and generalization capability can be obtained. Furthermore, MSE is able to evaluate the extent to which data changes, and if the MSE value is smaller, the accuracy of the description of the data by the constructed predictive neural network model is better.
In order to determine whether the prediction accuracy and performance of the model obtained by the optimization training meet the application requirements of the gas-solid flow particle concentration measurement engineering, in some embodiments, after the concentration prediction model is obtained, two evaluation indexes, namely an average absolute error (MeanAbsolute Error, MAE) and a root mean square error (Relative Mean Square Error, RMSE), are adopted to evaluate the model. Specifically, model evaluation was performed according to the following average absolute error and root mean square error functions:
and->
Wherein: n is the number of data; y is i Is the true value of the concentration;is the average of the true values; />Indicating a predicted concentration value.
In this embodiment, by adopting the above model evaluation scheme, it can be determined whether the model obtained by training meets the engineering use requirement, and whether further optimization and improvement are required, so as to ensure the prediction accuracy of the model in the subsequent engineering application.
For example, the above function gave a MAE of 0.0123 and REMS of 0.001. Wherein, MAE can measure the average error between model predictive value and true value, RMSE considers the standard deviation of the data set, can measure the relative error of the model; the predictive performance of the model can be comprehensively reflected by combining the two indexes. And determining whether it can be used in engineering applications to provide more accurate predictions of gas-solid flow solids concentration. In some embodiments, referring to FIG. 4, each point in the MAE curve represents the MAE value of the model after the corresponding training step is completed. The smaller the MAE value, the higher the prediction accuracy of the representation model, and the closer to the true value. In this embodiment, as can be seen from the graph, according to the training scheme of the concentration prediction model provided by the embodiment of the invention, the MAE curve is wholly reduced and finally converged to a relatively stable smaller value, which indicates that the performance of the model is continuously improved and finally reaches a relatively optimized state.
In the model training process, MAE curves are generally used together with training Loss curves, and model training effects are evaluated from two aspects. Referring to FIG. 5, the Loss value decreases, indicating that the model parameters are approaching the optimum.
In combination with fig. 4 and fig. 5, in the convolutional neural network training process provided by the embodiment of the invention, the Loss curve and the MAE curve synchronously drop, which indicates that optimization of model parameters directly drives improvement of prediction accuracy, and the model training is ideal.
As shown in fig. 6, in order to further test the prediction performance of the model, test sample data is input into the above-mentioned trained obtained concentration prediction model, a solid particle concentration value and a dynamic change curve chart are output, and by comparing with a control solid particle concentration value (true value), as can be seen from the figure, there is little difference, and it is proved that the prediction accuracy of the concentration prediction model obtained based on the actual production environment data training in the embodiment of the present invention is higher.
Therefore, the data-driven dust gas-solid flow solid particle concentration prediction method provided by the embodiment of the invention is used for constructing a concentration prediction model through training based on the production environment data and the deep learning convolutional neural network, has stronger adaptability and generalization capability, can be well applied to dust and other gas-solid flow solid particle concentration measurement engineering through evaluation and verification, is convenient for rapidly and accurately predicting the concentration of solid particles in dust and other gas-solid two-phase matters, and realizes on-line monitoring and change situation tracking of dust and other gas-solid flow solid particle concentration.
Referring to fig. 7, in some embodiments, based on the foregoing technical concept of the method for predicting the concentration of the solid particles in the gas-solid stream, the present invention further provides a system 200 for predicting the concentration of the solid particles in the gas-solid stream of dust based on data driving, which includes: a data acquisition unit 210, configured to obtain electrostatic induction data of dust movement by using an electrostatic induction measurement method, and obtain temperature data, humidity data, and internal and external air pressure difference data in the pipeline by using a temperature sensor, a humidity sensor, and an air pressure sensor;
a first calculating unit 220, configured to calculate a flow rate of the solid particles by a cross-correlation method, and calculate a reference solid particle concentration data according to the input material mass and the flow rate;
the data processing unit 230 is configured to perform normalization preprocessing on the obtained temperature data, humidity data, and internal and external air pressure difference data, and the obtained reference solid particle concentration data, to obtain training sample data and test sample data;
the data training unit 240 is configured to input the training sample data into a convolutional neural network for training, so as to obtain a concentration prediction model; the convolutional neural network comprises a linear convolutional module, a bias constant vector, a nonlinear activation function, a pooling layer and an overfitting prevention regularization module;
and a concentration prediction unit 260, configured to input the test sample data into the concentration prediction model, and output a solid particle concentration value and a dynamic variation graph.
The system of this embodiment may be used to implement the technical solution of the method embodiment shown in fig. 1, and its implementation principle and technical effects are similar to those of the foregoing method embodiment, and will not be repeated here, but reference may be made to each other.
In addition, it is to be understood that the system shown in fig. 7 may also be used to perform other embodiments of the foregoing method embodiments, and on the premise of clarity and brevity, the remaining embodiments will not be described in detail, and may be referred to each other.
According to the dust gas-solid flow solid particle concentration prediction system based on data driving, through the mutual cooperation and the synergistic effect of the program function modules, the concentration prediction model is built based on the training of the production environment data and the deep learning convolutional neural network, so that the concentration prediction model has strong adaptability and generalization capability, the concentration of solid particles in gas-solid two-phase matters such as dust can be predicted conveniently and rapidly, and the online monitoring and change situation tracking of the concentration of the gas-solid flow solid particles such as dust are realized.
In summary, it can be seen that the method and the system for predicting the concentration of the solid particles in the dust gas-solid flow provided by the embodiment of the invention adopt actual production environment data to drive learning training to obtain a concentration prediction model, avoid the problem that a physical prediction model cannot adapt to the unstable prediction error caused by the dynamic change of the gas-solid flow along with the environment and equipment, and improve the accuracy of a prediction result.
Furthermore, in the process of realizing the aim of the invention, the problem of the deficiency of training control data is solved by adopting a cross-correlation method, and the effect of real-time quick dynamic prediction can be improved.
Furthermore, by adopting the scheme provided by the embodiment of the invention, higher prediction precision can be obtained, so that data support is provided for the aspects of safe production, efficiency improvement, dust emission reduction and the like of dust processing enterprises such as coal dust combustion, stone grinding dust, flour processing dust control and the like, a digital production foundation for the enterprises is laid, and the improvement of production quality and efficiency is facilitated.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), or the like.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the scope of the present invention should be included in the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310724171.8A CN116721711B (en) | 2023-06-19 | 2023-06-19 | Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310724171.8A CN116721711B (en) | 2023-06-19 | 2023-06-19 | Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116721711A CN116721711A (en) | 2023-09-08 |
| CN116721711B true CN116721711B (en) | 2023-12-01 |
Family
ID=87865813
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310724171.8A Active CN116721711B (en) | 2023-06-19 | 2023-06-19 | Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116721711B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119007418B (en) * | 2024-10-24 | 2025-03-04 | 杭州蜂鸟智能机器人有限公司 | Dust concentration early warning method, computer storage medium and terminal equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102338732A (en) * | 2011-07-06 | 2012-02-01 | 浙江大学 | Method for measuring fluidization parameter of gas-solid fluidized bed |
| CN104655395A (en) * | 2015-01-16 | 2015-05-27 | 华北电力大学 | Device and method for detecting flow parameters of powder fluid in rectangular tube based on electrostatic induction |
| CN109523013A (en) * | 2018-10-15 | 2019-03-26 | 西北大学 | A kind of air particle pollution level estimation method based on shallow-layer convolutional neural networks |
| CN111351540A (en) * | 2020-03-23 | 2020-06-30 | 浙江大学 | Method and system for detecting particle mass flow rate during pneumatic conveying |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019209723A1 (en) * | 2018-04-26 | 2019-10-31 | Becton, Dickinson And Company | Characterization and sorting for particle analyzers |
| CN112766549B (en) * | 2021-01-07 | 2024-12-24 | 清华大学 | Air pollutant concentration forecasting method, device and storage medium |
-
2023
- 2023-06-19 CN CN202310724171.8A patent/CN116721711B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102338732A (en) * | 2011-07-06 | 2012-02-01 | 浙江大学 | Method for measuring fluidization parameter of gas-solid fluidized bed |
| CN104655395A (en) * | 2015-01-16 | 2015-05-27 | 华北电力大学 | Device and method for detecting flow parameters of powder fluid in rectangular tube based on electrostatic induction |
| CN109523013A (en) * | 2018-10-15 | 2019-03-26 | 西北大学 | A kind of air particle pollution level estimation method based on shallow-layer convolutional neural networks |
| CN111351540A (en) * | 2020-03-23 | 2020-06-30 | 浙江大学 | Method and system for detecting particle mass flow rate during pneumatic conveying |
Non-Patent Citations (2)
| Title |
|---|
| 《Signal decoupling and analysis from inner flush-mounted electrostatic sensor for detecting pneumatic conveying particles》;Chao Wang等;《Powder Technology》;全文 * |
| 《气固循环流化床颗粒浓度波动信号的预测》;李晓祥 等;《过程工程学报》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116721711A (en) | 2023-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Prediction model of natural gas pipeline crack evolution based on optimized DCNN-LSTM | |
| CN102072922B (en) | Particle swarm optimization neural network model-based method for detecting moisture content of wood | |
| Zhang et al. | Early fault detection method of rolling bearing based on MCNN and GRU network with an attention mechanism | |
| CN109492708B (en) | A missing data interpolation method for pipeline magnetic flux leakage detection based on LS-KNN | |
| CN110782067B (en) | A Prediction Method of Sintering End Point Fluctuation Range Based on Fuzzy Information Granulation | |
| CN116721711B (en) | Method and system for predicting concentration of dust gas-solid flow solid particles based on data driving | |
| Jiang et al. | A multisensor cycle-supervised convolutional neural network for anomaly detection on magnetic flux leakage signals | |
| CN112465239B (en) | Desulfurization system operation optimization method based on improved PSO-FCM algorithm | |
| Shi et al. | Prediction method of ball valve internal leakage rate based on acoustic emission technology | |
| CN114417688A (en) | Valve internal leakage rate detection method based on acoustic emission | |
| CN118656656B (en) | A method and system for accurately quantifying sewage outlet flow based on a neural network model | |
| CN116305985A (en) | A local intelligent ventilation method based on multi-sensor data fusion | |
| Cui et al. | Application of rotating machinery fault diagnosis based on deep learning | |
| Peng et al. | Remaining useful life prognosis of turbofan engines based on deep feature extraction and fusion | |
| Hu et al. | Status detection from spatial-temporal data in pipeline network using data transformation convolutional neural network | |
| Wan et al. | Prediction method of large-diameter ball valve internal leakage rate based on CNN-GA-DBN | |
| Asilian et al. | Application of Levenberg-Marquardt Backpropagation Algorithm in Artificial Neural Network for Self-Calibration of Deflection Type Wheatstone Bridge Circuit in CO Electrochemical Gas Sensor | |
| CN119844716A (en) | Water supply and drainage pipeline leakage position positioning method and system | |
| Abbas et al. | Mass flow measurement of pneumatically conveyed solids through multi-modal sensing and machine learning | |
| Liang et al. | A Multi-Sensor Detection Method Based on WGAN-GP and Attention-Bi-GRU for Well Control Pipeline Defects | |
| Miao et al. | Research on abnormal data detection of gas boiler supply based on deep learning network | |
| Zeng et al. | Mass flow rate measurement of pneumatically conveyed solids in a square-shaped pipe through multisensor fusion and data-driven modeling | |
| Lyu et al. | Interval prediction of remaining useful life based on convolutional auto-encode and lower upper bound estimation | |
| CN117541125A (en) | Intelligent food quality detection method and device | |
| CN113899809B (en) | In-pipeline detector positioning method based on CNN classification and RNN prediction |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20230908 Assignee: Shaanxi Jinyu Testing Technology Co.,Ltd. Assignor: TAIYUAN University OF SCIENCE AND TECHNOLOGY Contract record no.: X2024980018527 Denomination of invention: Data driven prediction method and system for solid particle concentration in dust gas-solid flow Granted publication date: 20231201 License type: Exclusive License Record date: 20241014 |