CN112216354A - Intelligent dosing system and method based on CFD numerical simulation and machine learning - Google Patents
Intelligent dosing system and method based on CFD numerical simulation and machine learning Download PDFInfo
- Publication number
- CN112216354A CN112216354A CN202010980741.6A CN202010980741A CN112216354A CN 112216354 A CN112216354 A CN 112216354A CN 202010980741 A CN202010980741 A CN 202010980741A CN 112216354 A CN112216354 A CN 112216354A
- Authority
- CN
- China
- Prior art keywords
- machine learning
- data
- database
- model
- sedimentation tank
- 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.)
- Pending
Links
- 238000010801 machine learning Methods 0.000 title claims abstract description 89
- 238000004088 simulation Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000004062 sedimentation Methods 0.000 claims abstract description 78
- 238000006243 chemical reaction Methods 0.000 claims abstract description 75
- 239000003814 drug Substances 0.000 claims abstract description 38
- 230000004927 fusion Effects 0.000 claims abstract description 12
- 238000005457 optimization Methods 0.000 claims abstract description 11
- 238000004458 analytical method Methods 0.000 claims abstract description 10
- 238000012821 model calculation Methods 0.000 claims abstract description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 60
- 229910019142 PO4 Inorganic materials 0.000 claims description 20
- 239000010452 phosphate Substances 0.000 claims description 20
- 239000010865 sewage Substances 0.000 claims description 16
- 238000004519 manufacturing process Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000005189 flocculation Methods 0.000 claims description 8
- NBIIXXVUZAFLBC-UHFFFAOYSA-K phosphate Chemical compound [O-]P([O-])([O-])=O NBIIXXVUZAFLBC-UHFFFAOYSA-K 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 229940079593 drug Drugs 0.000 claims description 7
- 238000013499 data model Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000005054 agglomeration Methods 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 3
- 230000008878 coupling Effects 0.000 claims description 3
- 238000010168 coupling process Methods 0.000 claims description 3
- 238000005859 coupling reaction Methods 0.000 claims description 3
- 239000012530 fluid Substances 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims description 3
- 238000001556 precipitation Methods 0.000 claims description 3
- 238000003756 stirring Methods 0.000 claims description 3
- 239000007787 solid Substances 0.000 claims 1
- 230000000694 effects Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000016615 flocculation Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 3
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 description 2
- 239000003795 chemical substances by application Substances 0.000 description 2
- 238000005345 coagulation Methods 0.000 description 2
- 230000015271 coagulation Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 229910052698 phosphorus Inorganic materials 0.000 description 2
- 239000011574 phosphorus Substances 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000008394 flocculating agent Substances 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000010802 sludge Substances 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
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
- G16C20/70—Machine learning, data mining or chemometrics
-
- 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
- C02F1/008—Control or steering systems not provided for elsewhere in subclass C02F
-
- 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
-
- 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
- G16C20/10—Analysis or design of chemical reactions, syntheses or processes
Landscapes
- Chemical & Material Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Hydrology & Water Resources (AREA)
- Environmental & Geological Engineering (AREA)
- Organic Chemistry (AREA)
- Water Supply & Treatment (AREA)
- Theoretical Computer Science (AREA)
- Crystallography & Structural Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Analytical Chemistry (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Activated Sludge Processes (AREA)
Abstract
The invention discloses an intelligent dosing system and method based on CFD numerical simulation and machine learning, the system comprises an operational analysis module, an optimization control module and a reaction sedimentation tank, the operational analysis module comprises a historical operation database, a CFD simulation database, a model sample database and a machine learning model, the historical operation database comprises historical operation data of the reaction sedimentation tank under different parameters, the CFD simulation database comprises simulation data of the reaction sedimentation tank obtained by model calculation, the historical operation data and the CFD simulation data are subjected to data fusion in the model sample database to obtain model sample data, the machine learning model trains the model sample data to obtain dosing quantity parameters in the reaction sedimentation tank, and the optimization control module doses a medicament in the reaction sedimentation tank according to the dosing quantity parameters. According to the method, the historical data set is expanded in the CFD numerical simulation result, the dimensionality of machine learning input data is increased, a better prediction model is established by machine learning, and accurate addition of the medicament is realized.
Description
Technical Field
The invention belongs to an intelligent dosing system in the technical field of water treatment, and particularly relates to an intelligent dosing system and method based on CFD numerical simulation and machine learning.
Background
A chemical coagulation method by adding a phosphorus removal agent is one of effective methods for removing micro suspended matters and colloidal impurities in sewage. As the quality and the quantity of the inlet water fluctuate greatly, the discharge of some sewage treatment plants operated by experience is difficult to reach the standard stably. Under the condition, a plurality of sewage treatment plants mostly adopt a method of overdosing medicaments to achieve the purpose of reaching the discharge standard. However, the excessive addition of the medicament not only increases the medicament cost, but also causes the problem that the sludge yield is increased and the corresponding treatment cost is increased. In more serious cases, the excessive addition of the medicament can cause secondary pollution to water quality. Therefore, the accurate addition of the phosphorus removal agent is very important for the standard discharge of a sewage plant and the reduction of the operation cost.
Currently, intelligent dosing systems for water treatment are classified into two types: the first type is realized by an automatic control mode, for example, a PLC controller in a patent CN209526334U adds medicine according to the temperature, pressure and liquid level of a medicine adding box, and a controller in a patent CN109592760A controls the medicine adding according to the flow rate and turbidity of inlet and outlet water, however, the automatic control method is limited by the problems of nonlinearity, uncertainty, time lag, more variables and the like existing in the sewage treatment process, and the efficient and stable control of the automatic control method is difficult to realize; the second type is realized by means of intelligent algorithms, such as neural network model algorithm of patent CN110981021A, dosing amount algorithm developed by patents CN106227251A and CN106348408A, and the influence of more variables on the dosing amount of the medicament can be considered, so that the conversion from automation to intelligence is realized. However, the above intelligent dosing systems and methods do not consider the influence of the flow field characteristics in the reaction settling tank on the dosing and flocculation effects.
The current sewage treatment plants are basically designed and operated according to empirical formulas and parameters, and the flow and mixing conditions in the reaction settling tank are not considered. However, the flow in the reaction sedimentation tank is not uniform, such as vortex flow, flow dead zone, short-circuit flow and the like, which seriously restricts the dosing effect of coagulation and flocculation, so that the dosing amount is increased or the effluent quality is difficult to reach the standard. Meanwhile, for the condition of severe fluctuation of water quality and water quantity, the flow field change in the reaction sedimentation tank is large, and the adding amount of the medicament is lack of regulation and control basis at the moment.
Therefore, in combination with the above-mentioned technical problems, a new technical solution is needed.
Disclosure of Invention
The invention aims to provide an intelligent dosing system and method based on CFD numerical simulation and machine learning, wherein the CFD numerical simulation method considers the influence of flow characteristics on dosing effect, supplements the integrity of a model sample database, and improves the accuracy and reliability of machine learning; meanwhile, data fusion is carried out on the CFD simulation data and the operation data, the CFD simulation data and the operation data are provided for machine learning to carry out training and deep learning, and finally optimal parameters are provided for an intelligent dosing system in the water treatment process.
In order to achieve the object, according to one aspect of the present invention, the present invention provides an intelligent dosing system based on CFD numerical simulation and machine learning, which includes an operational analysis module, an optimization control module, and a reaction sedimentation tank, wherein the operational analysis module includes a historical operation database, a CFD simulation database, a model sample database, and a machine learning model, the historical operation database includes historical operation data of the reaction sedimentation tank accumulated by sewage production operation under different parameters, the CFD simulation database includes simulation data of the reaction sedimentation tank obtained by model calculation, the historical operation database and the CFD simulation database are respectively configured to be capable of sending the historical operation data and the simulation data to the model sample database, and the model sample database is configured to perform data fusion on data information respectively sent by the historical operation database and the CFD simulation database And obtaining model sample data, wherein the machine learning model is configured to train the model sample data, the machine learning model is further configured to calculate and obtain dosing quantity parameters in the reaction sedimentation tank, and the optimization control module is configured to dose a medicament in the reaction sedimentation tank according to the dosing quantity parameters output by the machine learning.
In a further embodiment, the intelligent dosing operation analysis module further comprises a real-time operation database, the real-time operation database comprises real-time operation data of the reaction sedimentation tank, and the machine learning model is configured to match the real-time operation data with model sample data in a similarity manner, so as to continuously modify the machine learning model on line.
In a further embodiment, the system further comprises a medicine storage device, the optimization control module comprises a controller and a metering pump, the metering pump can be configured to feed the medicine in the medicine storage device into the reaction sedimentation tank, the controller can be configured to call the medicine feeding amount parameter output by the machine learning module, and the quantity of the medicine fed into the reaction sedimentation tank by the metering pump is controlled according to the medicine feeding amount parameter.
In a further embodiment, the historical operating data includes dosage, inflow water flow, inflow turbidity, inflow phosphate concentration, outflow water flow, outflow turbidity, and outflow phosphate concentration in the reaction sedimentation tank during the sewage production operation.
In a further embodiment, the simulation data includes flow field information, floc settling efficiency, and effluent suspension in the reactive sedimentation tank calculated from the actual geometry of the reactive sedimentation tank, inlet conditions of the influent water, and water quality.
In a further embodiment, the real-time operation data includes the dosing amount, the water inlet flow rate, the water inlet turbidity, the water inlet phosphate concentration, the water outlet flow rate, the water outlet turbidity and the water outlet phosphate concentration of the reaction sedimentation tank during real-time operation.
According to another aspect of the present invention, the present invention provides an intelligent dosing method based on CFD numerical simulation and machine learning, comprising the following steps: s1: recording the dosage, inflow flow, inflow turbidity, inflow phosphate radical concentration, outflow flow, outflow turbidity and outflow phosphate radical concentration of a reaction sedimentation tank under different parameters in the sewage production operation, and constructing a historical operation database; s2: establishing a physical model of the reaction sedimentation tank according to the actual geometric structure of the reaction sedimentation tank, the inlet condition of inlet water and the water quality condition, then respectively simulating according to different flow rates, stirring speeds and suspended matter concentrations, calculating to obtain flow field information, floc sedimentation efficiency and effluent suspended matters in the reaction sedimentation tank, and establishing a CFD simulation database; s3: performing data fusion on the historical operation database and the CFD simulation database to construct a big data model sample database; s4: training model sample data in the model sample database through a machine learning model, and calculating to obtain a dosing amount parameter in a reaction sedimentation tank; s5: according to the real-time operation data in the reaction sedimentation tank obtained by a real-time operation database, the machine learning model is configured to carry out similarity matching on the real-time operation data and model sample data, and the machine learning model is continuously corrected on line to obtain an optimal dosing quantity parameter; s6: and calling the optimal drug adding quantity parameter obtained by the machine learning model by the controller, and controlling the quantity of the drug added into the reaction sedimentation tank by the metering pump according to the drug adding quantity parameter.
In a further embodiment, in step S2, the physical model is constructed as follows: and (3) performing numerical simulation on the flocculation process in the reaction sedimentation tank by adopting a floc agglomeration and crushing submodel of a double-fluid model coupling particle group balance method, revealing the direct relation between the flow condition and the sedimentation efficiency, and obtaining sample data beyond historical data.
In a further embodiment, in step S2, the flow field information includes whether a vortex, a dead flow zone, or a short-circuit flow condition exists in the reaction sedimentation tank.
In a further embodiment, in step S4, the machine learning model trains model sample data from a model sample database using an artificial neural network algorithm; in step S5, when similarity matching is performed on the real-time operation data and model sample data, if the similarity is high, the machine learning model directly outputs a dosing amount parameter; and if the matching degree is low, calculating corresponding dosing quantity parameters under the same working condition by CFD, transmitting the dosing quantity parameters to the machine learning model for output, then transmitting the real-time running data with the low matching degree to the model sample database for data fusion to obtain new model sample data, and then training the new model sample data by the machine learning model to update the machine learning model.
Compared with the prior art, the intelligent control system combining CFD numerical simulation and machine learning is adopted, the relation between the flow field information and the flocculation effect in the reaction sedimentation tank is given in the mechanism through CFD numerical simulation, the original historical operation data set is expanded through CFD numerical simulation, the dimensionality of machine learning model input data is increased, the machine learning is facilitated to establish a better prediction model, and accurate adding of the medicament is achieved.
Drawings
FIG. 1 is a schematic diagram of the system of the intelligent dosing system based on CFD numerical simulation and machine learning;
FIG. 2 is a flow diagram of an offline learning process of the machine learning model of the present application;
FIG. 3 is a flow chart of online prediction and model update of a machine learning model in the present application.
Detailed Description
To further clarify the technical measures and effects of the present invention adopted to achieve the intended purpose, the following detailed description is given of specific embodiments, structures, characteristics and effects according to the present invention with reference to the accompanying drawings and preferred embodiments.
Referring to fig. 1 to 3, fig. 1 is a schematic system diagram of an intelligent dosing system based on CFD numerical simulation and machine learning according to the present application; FIG. 2 is a flow diagram of an offline learning process of the machine learning model of the present application; FIG. 3 is a flow chart of online prediction and model update of a machine learning model in the present application.
Examples
As shown in fig. 1, the present invention provides an intelligent dosing system based on CFD numerical simulation and machine learning, which includes an operation analysis module, an optimization control module, and a reaction sedimentation tank. The operation analysis module comprises a historical operation database, a CFD simulation database, a model sample database, a machine learning model and a time operation database. The historical operation database comprises historical operation data of the reaction sedimentation tank accumulated in the sewage production operation under different parameters, and the historical operation data comprises the dosage, the water inlet flow, the water inlet turbidity, the water inlet phosphate radical concentration, the water outlet flow, the water outlet turbidity and the water outlet phosphate radical concentration in the reaction sedimentation tank in the sewage production operation. The CFD simulation database comprises simulation data of the reaction sedimentation tank, wherein the simulation data are obtained through model calculation, and the simulation data comprise flow field information, floc sedimentation efficiency and water outlet suspended matters in the reaction sedimentation tank, which are obtained through calculation according to the actual geometric structure of the reaction sedimentation tank, the inlet condition of water inlet and the water quality condition. The historical operation database and the CFD simulation database are respectively configured and can send historical operation data and simulation data to the model sample database, the model sample database is configured and carries out data fusion on data information respectively sent by the historical operation database and the CFD simulation database to obtain model sample data, the machine learning model is configured and can train the model sample data, the machine learning model is further configured and calculates to obtain dosing quantity parameters in the reaction sedimentation tank, and the optimization control module is configured and can dose a medicament in the reaction sedimentation tank according to the dosing quantity parameters output by the machine learning. The real-time operation database comprises real-time operation data of the reaction sedimentation tank, and the real-time operation data comprises the dosage, the water inlet flow, the water inlet turbidity, the water inlet phosphate radical concentration, the water outlet flow, the water outlet turbidity and the water outlet phosphate radical concentration of the reaction sedimentation tank during real-time operation. The machine learning model is configured to match the real-time running data with model sample data in similarity, and continuously corrects the machine learning model on line.
In a further embodiment, the system further comprises a medicine storage device, the optimization control module comprises a controller and a metering pump, the metering pump can be configured to feed the medicine in the medicine storage device into the reaction sedimentation tank, the controller can be configured to call the medicine feeding amount parameter output by the machine learning module, and the quantity of the medicine fed into the reaction sedimentation tank by the metering pump is controlled according to the medicine feeding amount parameter.
According to another aspect of the present invention, the present invention provides an intelligent dosing method based on CFD numerical simulation and machine learning, comprising the following steps:
s1: recording the dosage, water inlet flow, water inlet turbidity, water inlet phosphate radical concentration, water outlet flow, water outlet turbidity and water outlet phosphate radical concentration of a reaction sedimentation tank under different parameters in the sewage production operation, and constructing a history operation database;
s2: establishing a physical model of the reaction sedimentation tank according to the actual geometric structure of the reaction sedimentation tank, the inlet condition of inlet water and the water quality condition, wherein the physical model is established as follows: and (3) performing numerical simulation on the flocculation process in the reaction sedimentation tank by adopting a flocculation agglomeration and fragmentation submodel of a double-fluid model coupling particle group balance method, revealing the direct relation between the flow condition and the sedimentation efficiency, and obtaining sample data beyond historical data. After the physical model is established, simulation is respectively carried out according to different flow rates, stirring speeds and suspended matter concentrations, flow field information, floc precipitation efficiency and effluent suspended matters in the reaction sedimentation tank are obtained through calculation, and a CFD simulation database is established, wherein the flow field information comprises whether eddy current, flow dead zone or short-circuit flow conditions exist in the reaction sedimentation tank.
Wherein, the CFD numerical simulation solves the flow and mass transfer process through the following equation:
conservation of mass equation:
conservation of momentum equation:
solving the growth and crushing process of the suspended substances in the sewage after adding the flocculating agent through a group equilibrium equation:
s3: performing data fusion on the historical operation database and the CFD simulation database to construct a big data model sample database, wherein the data of the data model sample database mainly comprises information such as dosage, inflow water flow, inflow water turbidity, inflow water phosphate radical concentration, outflow water flow, outflow water turbidity, outflow water phosphate radical concentration, flow field information, precipitation efficiency, outflow water suspended matters and the like;
s4: training the model sample data in the model sample database through a machine learning model, and calculating to obtain the dosing quantity parameter in the reaction sedimentation tank. As shown in fig. 2, which is an offline learning portion of the machine learning model. The method specifically comprises the following steps:
firstly, initializing a weight value and a threshold value of a model, providing a learning mode parameter, namely a learning rate, to a current network, calculating input and output values of each unit in a middle layer according to input data, and finally calculating the input and output values of each unit in an output layer until one forward calculation is completed. And secondly, calculating the final output value and the real value to obtain the correction error of each unit of the output layer, and then calculating the correction error of each unit of the middle layer in a back propagation manner. And comparing the error with an error threshold, if the error is larger than the threshold, reversely adjusting the connection weight from the middle layer to the output layer and the output threshold of each unit of the output layer, then continuously adjusting the connection weight from the input layer to the middle layer and the output threshold of each unit of the middle layer, namely updating the parameters, and finally repeating the first step. In order to improve the generalization ability of the model, the learning rate is adjusted until the optimal network structure is reached, and the cycle learning is repeated until the convergence requirement (the learning times or the precision is reached) is reached.
The machine learning model uses an artificial neural network algorithm to train model sample data from a model sample database. In order to obtain an accurate model, an artificial neural network is subjected to algorithm improvement, the content of the artificial neural network comprises the steps of adopting a loss function of cross entropy and carrying out regularization treatment, considering a gradient descent method of momentum, and a hyperbolic tangent function is adopted as an activation function by a hidden layer and an output layer. Output net of ith node of hidden layer of network after initializing weights of each layeriComprises the following steps:
output O of i-th node of hidden layeriComprises the following steps:
wherein, tanh is an excitation function of hyperbolic tangent; m is the number of neurons of the input layer, namely the data of the big data model sample after data fusion; w is aijIs input intoThe weight value from the jth node of the layer to the ith node of the hidden layer; bjIndicating the threshold of the hidden layer.
Outbound net for k-th node of outbound layerkComprises the following steps:
output O of kth node of output layerkComprises the following steps:
where q is the predicted output layer dimension loss function:
wherein, y(j)k is the jth sample, and the actual value of the kth neuron of the output layer is output; λ is a regularization parameter, which acts to prevent overfitting; l is the number of neural network layers, slFor a certain layer of neuron number, the calculation is started from the input layer, i is 1. The goal of neural network training is to minimize the difference between the network output and the actual y value, and loss (w), and solve by using a gradient descent method considering momentum.
The hidden layer error is:
δ(2)=(w(2))Tδ(3).*g'(o(2));
the gradient of the change of the weight value of each layer is as follows:
Δ(1)=Δ(1)+δ(l+1)(net(1))T;
namely, it is
The calculation takes into account the gradient change of momentum:
where v is the change in gradient at the last iteration.
Updating the weight of the network:
where α is the learning rate.
The connection weights from the input layer to the hidden layer and from the hidden layer to the output layer are continuously updated through the formula, so that the algorithm error is converged to a designated value epsilon. At this time, the predicted dosing amount parameter in the reaction sedimentation tank can be obtained.
S5: and according to the real-time operation data in the reaction sedimentation tank obtained by the real-time operation database, the machine learning model is configured to carry out similarity matching on the real-time operation data and model sample data, and the machine learning model is continuously corrected on line to obtain the optimal dosing quantity parameter. As shown in fig. 3, the online prediction and model update portion of the machine learning model. When the real-time operation data on site is transmitted to the trained machine learning model, similarity matching is carried out on the real-time operation data and a big data model sample database, and if the similarity is high, such as 90% -97% or more, the machine learning model directly outputs a dosing quantity parameter; if the matching degree is low, such as 0% -90%, corresponding dosing quantity parameters under the same working condition are calculated through CFD simulation and are transmitted to the machine learning model for output, then real-time operation data when the matching degree is low are transmitted to the model sample database for data fusion to obtain new model sample data, then the machine learning model performs offline learning training on the new model sample data, and updates the machine learning model to provide a more accurate model for follow-up.
S6: and calling the optimal dosing quantity parameter obtained by the machine learning model by the controller, and controlling the quantity of the medicament added into the reaction sedimentation tank by the metering pump according to the dosing quantity parameter to realize the accurate addition of the medicament into the reaction sedimentation tank. And meanwhile, result data obtained under the optimal parameter condition is fed back to the real-time operation database to supplement the richness of the data, so that the training of machine learning is more accurate.
Compared with the prior art, the intelligent control system combining CFD numerical simulation and machine learning is adopted, the relation between the flow field information and the flocculation effect in the reaction sedimentation tank is given in the mechanism through CFD numerical simulation, the original historical operation data set is expanded through CFD numerical simulation, the dimensionality of machine learning model input data is increased, the machine learning is facilitated to establish a better prediction model, and accurate adding of the medicament is achieved.
As used herein, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, including not only those elements listed, but also other elements not expressly listed.
In this document, the terms front, back, upper and lower are used to define the components in the drawings and the positions of the components relative to each other, and are used for clarity and convenience of the technical solution. It is to be understood that the use of the directional terms should not be taken to limit the scope of the claims.
The features of the embodiments and embodiments described herein above may be combined with each other without conflict.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. An intelligent dosing system based on CFD numerical simulation and machine learning is characterized by comprising an operation analysis module, an optimization control module and a reaction sedimentation tank,
the operation analysis module comprises a historical operation database, a CFD simulation database, a model sample database and a machine learning model, the historical operation database comprises historical operation data of the reaction sedimentation tank accumulated through sewage production operation under different parameters, the CFD simulation database comprises simulation data of the reaction sedimentation tank obtained through model calculation, the historical operation database and the CFD simulation database are respectively configured and can send the historical operation data and the simulation data to the model sample database, the model sample database is configured and can carry out data fusion on data information respectively sent by the historical operation database and the CFD simulation database to obtain model sample data, the machine learning model is configured and can train the model sample data, and the machine learning model is also configured and can calculate to obtain quantity parameters in the reaction sedimentation tank,
and the optimization control module is configured to feed a medicament into the reaction sedimentation tank according to the dosing quantity parameter output by the machine learning.
2. The intelligent dosing system based on CFD numerical simulation and machine learning of claim 1 wherein the intelligent dosing operational analysis module further comprises a real-time operational database, the real-time operational database comprising real-time operational data of the reaction sedimentation tank, the machine learning model configured to enable similarity matching of the real-time operational data with model sample data, and to continuously modify the machine learning model on-line.
3. The intelligent dosing system based on CFD numerical simulation and machine learning of claim 1, further comprising a drug storage device, wherein the optimization control module comprises a controller and a metering pump, the metering pump can be configured to dose the drug in the drug storage device into the reaction sedimentation tank, the controller can be configured to call the dosing amount parameter output by the machine learning module, and control the amount of the drug dosed by the metering pump into the reaction sedimentation tank according to the dosing amount parameter.
4. The intelligent dosing system based on CFD numerical simulation and machine learning of claim 1, wherein the historical operating data includes dosing amount, inlet water flow, inlet water turbidity, inlet water phosphate concentration, outlet water flow, outlet water turbidity and outlet water phosphate concentration in the reaction sedimentation tank during sewage production operation.
5. The intelligent dosing system based on CFD numerical simulation and machine learning of claim 1, wherein the simulation data includes flow field information, floc precipitation efficiency and effluent suspended solids in the reaction settling tank calculated according to the actual geometry of the reaction settling tank, inlet conditions of inlet water and water quality.
6. The intelligent dosing system based on CFD numerical simulation and machine learning of claim 2, wherein the real-time operation data comprises dosing amount, inflow flow, inflow turbidity, inflow phosphate concentration, outflow flow, outflow turbidity and outflow phosphate concentration of the reaction sedimentation tank during real-time operation.
7. An intelligent dosing method based on CFD numerical simulation and machine learning is characterized by comprising the following steps:
s1: recording the dosage, inflow flow, inflow turbidity, inflow phosphate radical concentration, outflow flow, outflow turbidity and outflow phosphate radical concentration of a reaction sedimentation tank under different parameters in the sewage production operation, and constructing a historical operation database;
s2: establishing a physical model of the reaction sedimentation tank according to the actual geometric structure of the reaction sedimentation tank, the inlet condition of inlet water and the water quality condition, then respectively simulating according to different flow rates, stirring speeds and suspended matter concentrations, calculating to obtain flow field information, floc sedimentation efficiency and effluent suspended matters in the reaction sedimentation tank, and establishing a CFD simulation database;
s3: performing data fusion on the historical operation database and the CFD simulation database to construct a big data model sample database;
s4: training model sample data in the model sample database through a machine learning model, and calculating to obtain a dosing amount parameter in a reaction sedimentation tank;
s5: according to the real-time operation data in the reaction sedimentation tank obtained by a real-time operation database, the machine learning model is configured to carry out similarity matching on the real-time operation data and model sample data, and the machine learning model is continuously corrected on line to obtain an optimal dosing quantity parameter;
s6: and calling the optimal dosing quantity parameter obtained by the machine learning model by the controller, and controlling the quantity of the medicament added into the reaction sedimentation tank by the metering pump according to the dosing quantity parameter.
8. The intelligent dosing method based on CFD numerical simulation and machine learning of claim 7, wherein in step S2, the physical model is constructed as follows:
and (3) performing numerical simulation on the flocculation process in the reaction sedimentation tank by adopting a floc agglomeration and crushing submodel of a double-fluid model coupling particle group balance method, revealing the direct relation between the flow condition and the sedimentation efficiency, and obtaining sample data beyond historical data.
9. The intelligent dosing method based on CFD numerical simulation and machine learning of claim 8, wherein in step S2, the flow field information includes whether there is a vortex, a dead flow zone or a short-circuit flow condition in the reaction sedimentation tank.
10. The intelligent dosing method based on CFD numerical simulation and machine learning of claim 8, wherein in step S4, the machine learning model uses an artificial neural network algorithm to train model sample data from a model sample database;
in step S5, when similarity matching is performed on the real-time operation data and model sample data, if the similarity is high, the machine learning model directly outputs a dosing amount parameter; and if the matching degree is low, calculating corresponding dosing quantity parameters under the same working condition by CFD, transmitting the dosing quantity parameters to the machine learning model for output, then transmitting the real-time operation data with the low matching degree to the model sample database for data fusion to obtain new model sample data, and then training the new model sample data by the machine learning model to update the machine learning model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010980741.6A CN112216354A (en) | 2020-09-17 | 2020-09-17 | Intelligent dosing system and method based on CFD numerical simulation and machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010980741.6A CN112216354A (en) | 2020-09-17 | 2020-09-17 | Intelligent dosing system and method based on CFD numerical simulation and machine learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112216354A true CN112216354A (en) | 2021-01-12 |
Family
ID=74050110
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010980741.6A Pending CN112216354A (en) | 2020-09-17 | 2020-09-17 | Intelligent dosing system and method based on CFD numerical simulation and machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112216354A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112733473A (en) * | 2021-01-14 | 2021-04-30 | 南京欧帕提亚信息科技有限公司 | Automatic and intelligent ship free rolling attenuation numerical simulation method based on CFD |
CN112947617A (en) * | 2021-02-01 | 2021-06-11 | 陕西中良智能科技有限公司 | Machine learning device, pH value intelligent fine-tuning control system and control method thereof |
CN113536698A (en) * | 2021-09-16 | 2021-10-22 | 大唐环境产业集团股份有限公司 | Method and device for establishing circulating water dosing model of thermal power plant |
CN113687040A (en) * | 2021-07-28 | 2021-11-23 | 广州杰赛科技股份有限公司 | Decision tree algorithm-based water plant real-time dosing quantity prediction method, device and medium |
CN113868967A (en) * | 2021-12-03 | 2021-12-31 | 中铁第一勘察设计院集团有限公司 | Structure monitoring data prediction method, device and storage medium |
CN114065949A (en) * | 2022-01-17 | 2022-02-18 | 心鉴智控(深圳)科技有限公司 | Intelligent water quality prediction dosing system based on historical data |
CN114117954A (en) * | 2021-11-08 | 2022-03-01 | 北京工业大学 | Dynamic real-time visualization method for three-dimensional reaction field in reactor |
CN114913927A (en) * | 2021-02-08 | 2022-08-16 | 欧阳德方 | Method for developing pharmaceutical composition and predicting and evaluating pharmaceutical composition |
CN115329661A (en) * | 2022-07-22 | 2022-11-11 | 上海环保(集团)有限公司 | Intelligent dosing model modeling, intelligent dosing system creating and dosing method |
CN117170334A (en) * | 2023-11-02 | 2023-12-05 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103011356A (en) * | 2012-08-15 | 2013-04-03 | 重庆水务集团股份有限公司 | Method for controlling automatic chemical dosing of high-turbidity water system |
CN107726358A (en) * | 2017-10-12 | 2018-02-23 | 东南大学 | Boiler Combustion Optimization System and method based on CFD numerical simulations and intelligent modeling |
-
2020
- 2020-09-17 CN CN202010980741.6A patent/CN112216354A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103011356A (en) * | 2012-08-15 | 2013-04-03 | 重庆水务集团股份有限公司 | Method for controlling automatic chemical dosing of high-turbidity water system |
CN107726358A (en) * | 2017-10-12 | 2018-02-23 | 东南大学 | Boiler Combustion Optimization System and method based on CFD numerical simulations and intelligent modeling |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112733473A (en) * | 2021-01-14 | 2021-04-30 | 南京欧帕提亚信息科技有限公司 | Automatic and intelligent ship free rolling attenuation numerical simulation method based on CFD |
CN112947617A (en) * | 2021-02-01 | 2021-06-11 | 陕西中良智能科技有限公司 | Machine learning device, pH value intelligent fine-tuning control system and control method thereof |
CN114913927A (en) * | 2021-02-08 | 2022-08-16 | 欧阳德方 | Method for developing pharmaceutical composition and predicting and evaluating pharmaceutical composition |
CN113687040A (en) * | 2021-07-28 | 2021-11-23 | 广州杰赛科技股份有限公司 | Decision tree algorithm-based water plant real-time dosing quantity prediction method, device and medium |
CN113536698B (en) * | 2021-09-16 | 2022-01-25 | 大唐环境产业集团股份有限公司 | Method and device for establishing circulating water dosing model of thermal power plant |
CN113536698A (en) * | 2021-09-16 | 2021-10-22 | 大唐环境产业集团股份有限公司 | Method and device for establishing circulating water dosing model of thermal power plant |
CN114117954A (en) * | 2021-11-08 | 2022-03-01 | 北京工业大学 | Dynamic real-time visualization method for three-dimensional reaction field in reactor |
CN114117954B (en) * | 2021-11-08 | 2023-04-18 | 北京工业大学 | Dynamic real-time visualization method for three-dimensional reaction field in reactor |
CN113868967A (en) * | 2021-12-03 | 2021-12-31 | 中铁第一勘察设计院集团有限公司 | Structure monitoring data prediction method, device and storage medium |
CN113868967B (en) * | 2021-12-03 | 2022-04-12 | 中铁第一勘察设计院集团有限公司 | Structure monitoring data prediction method, device and storage medium |
CN114065949A (en) * | 2022-01-17 | 2022-02-18 | 心鉴智控(深圳)科技有限公司 | Intelligent water quality prediction dosing system based on historical data |
CN115329661A (en) * | 2022-07-22 | 2022-11-11 | 上海环保(集团)有限公司 | Intelligent dosing model modeling, intelligent dosing system creating and dosing method |
CN115329661B (en) * | 2022-07-22 | 2023-06-23 | 上海环保(集团)有限公司 | Intelligent dosing model modeling, intelligent dosing system creation and dosing method |
CN117170334A (en) * | 2023-11-02 | 2023-12-05 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
CN117170334B (en) * | 2023-11-02 | 2024-03-08 | 钥准医药科技(启东)有限公司 | Intelligent control method and system for rapid drug fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112216354A (en) | Intelligent dosing system and method based on CFD numerical simulation and machine learning | |
CN103011356B (en) | Method for controlling automatic chemical dosing of high-turbidity water system | |
CN106873379B (en) | Sewage treatment optimal control method based on iterative ADP algorithm | |
CN102385315B (en) | Intelligent coagulation chemical dosing control system for water plant and control method thereof | |
CN103809557B (en) | A kind of sewage disposal process optimal control method based on neutral net | |
CN106802563B (en) | A kind of sewage procedure optimization control method based on drosophila optimization and LSSVM | |
CN107601632A (en) | A kind of coagulation Automatic Dosing control method and system | |
CN102616927B (en) | Adjusting method of technological parameters of sewage treatment and device | |
CN101825870B (en) | Method and system for controlling supply quantity of water-treatment flocculating agent | |
CN103570190B (en) | Method for controlling feeding amount of chemical phosphorus removal reagents of water reclamation plants based on fuzzy control | |
CN109052633B (en) | Intelligent sewage treatment stepless regulation and control system | |
CN114230110B (en) | Short-range intelligent phosphorus removal and medicine addition control method, equipment and system for sewage treatment | |
CN113705898A (en) | Method and system for predicting dosage of tap water coagulant | |
CN103744293A (en) | Waste water treatment monitoring method and system based on fuzzy neural network | |
CN102464426A (en) | Intelligent adding control method of chemical phosphorus removal agent for municipal wastewater plant and device thereof | |
CN112782975A (en) | Sewage treatment aeration intelligent control method and system based on deep learning | |
CN103810309A (en) | Soft measurement modeling method of A2O municipal sewage treatment process based on constraint theory | |
CN110357236B (en) | Sewage plant intelligent control method based on mutation inversion effluent prediction model | |
CN114573120B (en) | Multivariable control method, device and system for coal chemical industry grey water granulation and hardness removal | |
CN114861535A (en) | Water treatment process machine learning modeling method using CFD as data preprocessing | |
CN115759399A (en) | Dosing control method, device, equipment and storage medium | |
Wang et al. | Model predictive control for chlorine dosing of drinking water treatment based on support vector machine model | |
CN205856256U (en) | The intelligence control system of sewage treatment plant's key process parameter | |
CN1464346A (en) | Self-adaptive control administration apparatus and process thereof | |
KR20040010500A (en) | Method and device for pH control of water supply facilities using tuning method of 2-DOF PID controller by neural network |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210112 |
|
RJ01 | Rejection of invention patent application after publication |