CN201449529U - Soda ash carbonization process control system based on neural network - Google Patents
Soda ash carbonization process control system based on neural network Download PDFInfo
- Publication number
- CN201449529U CN201449529U CN2009200566305U CN200920056630U CN201449529U CN 201449529 U CN201449529 U CN 201449529U CN 2009200566305 U CN2009200566305 U CN 2009200566305U CN 200920056630 U CN200920056630 U CN 200920056630U CN 201449529 U CN201449529 U CN 201449529U
- Authority
- CN
- China
- Prior art keywords
- neural network
- model
- network identifier
- input end
- output terminal
- 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.)
- Expired - Lifetime
Links
Landscapes
- Feedback Control In General (AREA)
Abstract
The utility model relates to a soda ash carbonization process control system based on a neural network, which comprises a neural network controller, a neural network identifier, a model base, a particle swarm optimizer, a data acquisition channel and a carbonization tower, wherein the data acquisition channel acquires process parameter values of the carbonization tower in real time for data preprocessing, and transmits processed data to the neural network identifier; the neural network identifier conducts modeling; models after modeling are emulated and revised, and then the neural network controller can read revised model parameters to generate control parameters; and simultaneously, the particle swarm optimizer conducts optimizing on control quantity to obtain a group of optimal control quantity, and then controls an actuator to move so as to complete the modeling and optimal control of the carbonization tower. The model base is used for storing the neural network identifier and establishing models of flow and temperature in the soda ash carbonization tower. The soda ash carbonization process control system can conduct modeling and optimal control on the carbonization tower, improve the automation level of units, increase the sodium conversion rate in the production process, and reduce the consumption.
Description
Technical field
The utility model relates to soda industry technological process and automation field, particularly a kind of soda carbonization technique control system based on neural network.
Background technology
Soda ash (Na2CO3) is a kind of important basic chemical raw materials, also be a kind of traditional chemical products of producing for many years and using history that have, it is widely used in all conglomeraties such as chemical industry, medicine, glass, metallurgy and papermaking, annual is in great demand, occupy important fundamental position in the development of the national economy, the Development of soda ash industry quality is directly connected to the raising of development and national economy and living standards of the people.
Soda Carbonization Process is the key link of alkali-making process, and mechanism is comparatively complicated.Carbonators is the core cell that whole soda ash is produced, various procedures such as existing chemical reaction, heat transfer, mass transfer, crystallization have the existence of gas, liquid, solid three-phase material again, and every tower is divided into cleaning and system alkali two states again, be existing continuous, complex process is intermittently arranged again.The quality that its operations index is finished is directly connected to product yield and quality, thereby influences the cost and the economic benefit of product.Therefore improve the controlling performance of each key process parameter in the tower, improve the quality of product and the important topic that output is Sodium Carbonate Plant.At present, though domestic each alkali factory carbonization process has adopted computer control system mostly, as DCS and PLC etc., but also rest on the hand/automatic control level of conventional instrument basically, still adopt single-circuit hand/control automatically for the reference mark, controlling performance depends on continued operation level and operator's artificial experience, control accuracy is comparatively coarse, make that the control accuracy of system is not high, and because the frequent artificial interference of operator makes that the fluctuation of system is bigger, thereby be difficult to guarantee the quality of full production and product, utilization ratio of raw materials, the utilization factor of sodium, the conversion ratios of carbon etc. are generally not high.
Therefore, to the modeling and optimization of the key process parameter of Soda Carbonization Process, be in order can be that the production of carbonators is more steady, fluctuation is few, reduce equipment loss, reduce working strength of workers, increase the output of soda ash, its final goal is the utilization factor for the conversion ratio that can improve carbon, sodium, improve the output and the quality of soda ash, improve the utilization factor of load, reduce cost, reduce energy consumption, thereby increase economic benefit.
The utility model content
The purpose of this utility model is to overcome the deficiencies in the prior art, design a kind of can be to the modeling and optimization of the key process parameter of Soda Carbonization Process, and reduce cost, reduce energy consumption, thereby increase the soda carbonization technique control system based on neural network of economic benefit.
In order to realize above-mentioned technical purpose, the utility model comprises following technical characterictic: a kind of soda carbonization technique control system based on neural network comprises nerve network controller, neural network identifier, model bank, particle swarm optimization device, data acquisition channel and carbonators;
The output terminal of described carbonators is connected with the neural network identifier input end by data acquisition channel; The neural network identifier output terminal is connected with the input end of nerve network controller, and the neural network identifier output terminal is connected with the input end of nerve network controller by model bank; The nerve network controller output terminal is connected with particle swarm optimization device input end, and the nerve network controller output terminal is connected with particle swarm optimization device input end by data acquisition channel; The output terminal of particle swarm optimization device is connected with the input end of carbonators.
Further, in order to realize the emulation correction to modeling process, described neural network identifier comprises neural net model establishing module, model emulation module, model editing module; The output terminal of data acquisition channel is connected with the input end of neural net model establishing module; Data acquisition channel is connected with the model emulation module; Model emulation module output terminal is connected with the neural net model establishing module; Neural net model establishing module output terminal is connected with the input end of model editing module and the input end of model bank.
Further, described data acquisition channel comprises acquisition module, the data preprocessing module that connects successively.
The utility model compared with prior art, have following beneficial effect: the utility model control system comprises neural network identifier, nerve network controller and particle swarm optimization device, control theory, Optimum Theory have been made full use of, neural network, System Discrimination, particle swarm optimization, intelligent search algorithm etc., the carbonization technique process is realized detection, control, modeling, optimization, scheduling, management and decision-making, reach the integrated technology that increases output, improves the quality, reduces purposes such as consumption.
Description of drawings
Accompanying drawing 1 is the theory diagram of this modeling and optimization control system;
Accompanying drawing 2 is the neural network identifier structural drawing in the native system.
Embodiment
Principle of work of the present utility model is by the analysis to the carbonators reaction mechanism, with the black-box modeling principle, uses the neural net model establishing algorithm, according to the historical data of carbonators operation, sets up the nonlinear model of carbonators feature (temperature and flow).According to the network model of setting up, use the particle swarm optimization algorithm, to the output temperature optimizing, obtain the optimal value of one group of flow, and with the setting value of this optimal value as controller, the control executing mechanism action realizes the optimal control to carbonization process.
Accompanying drawing 1 is the theory diagram of native system, comprises nerve network controller 1, neural network identifier 2, model bank 3, particle swarm optimization device 4, data acquisition channel 5 and carbonators 6; The output terminal of described carbonators 6 is connected with neural network identifier 2 input ends by data acquisition channel 5; Neural network identifier 2 output terminals are connected with the input end of nerve network controller 1, and neural network identifier 2 output terminals are connected with the input end of nerve network controller 1 by model bank 3; Nerve network controller 1 output terminal is connected with particle swarm optimization device 4 input ends, and nerve network controller 1 output terminal is connected with particle swarm optimization device 4 input ends by data acquisition channel 5; The output terminal of particle swarm optimization device 4 is connected with the input end of carbonators 6.
The course of work of system is: data acquisition channel 5 is gathered the process parameter value of carbonators 6 in real time, carry out the data pre-service, comprise and promptly carry out filtering and normalized, give neural network identifier 2 with the data transfer after handling, neural network identifier 2 carries out modeling, after the model process emulation correction after the modeling, nerve network controller 1 promptly can read revised model parameter, generate controlled variable, 4 pairs of controlled quentity controlled variables of particle swarm optimization device are carried out optimizing simultaneously, after obtaining one group of optimum control amount, the modeling and optimization control to carbonators is finished in the control executing mechanism action.Model bank 3 storage inside be flow and temperature model in the soda ash carbonators of setting up with neural network identifier 2.
Accompanying drawing 2 is the neural network identifier structural drawing in the native system, comprises neural net model establishing module 22, model emulation module 23, model editing module 24; The output terminal of data acquisition channel 5 is connected with the input end of neural net model establishing module 22; Data acquisition channel 5 is connected with model emulation module 23; Model emulation module 23 output terminals are connected with neural net model establishing module 22; Neural net model establishing module 22 output terminals are connected with the input end of model editing module 24 and the input end of model bank 3.
Its principle of work is: neural network identifier 2 is mainly used in foundation, emulation and the correction of nonlinear model. data acquisition channel 5 obtains stage casing airshed, hypomere airshed in the carbonization technique in real time, goes out the alkali flow, in and the measured data of discharge, tower middle part temperature (12 circle temperature, 17 circle temperature, 23 circle temperature) etc., pick out abnormal data, carry out pre-service again, sample data and test data are divided into groups according to 1: 1 ratio; Based on sample data, the neural net model establishing module 22 utilization neural net model establishing algorithms in the neural network identifier 2 are set up nonlinear mathematical model; Based on test data, the model that utilizes 23 pairs of neural networks of model emulation module to set up carries out emulation; Model editing module 24 utilizes the result of emulation that model is revised.
Control method of the present utility model comprises the steps:
The process parameter value that a, data acquisition channel 5 are gathered carbonators 6 is in real time carried out the data pre-service;
Data transfer after b, the processing is given neural network identifier 2, carries out modeling by neural network identifier 2, and the model after the modeling is through the emulation correction;
C, nerve network controller 1 read model parameter, generate controlled variable, and 4 pairs of controlled quentity controlled variables of particle swarm optimization device are carried out optimizing, obtain one group of optimum control amount after, the control executing mechanism action.
Further, described step b comprises the steps:
The neural net model establishing 22 of b1, neural network identifier 2 is set up mathematical model according to the data that obtain; The model that model emulation device 23 utilizes the data of acquisition that neural net model establishing 22 is set up carries out validation verification;
The nonlinear model that b2, model editing module 24 are set up neural net model establishing module 22 according to simulation result is revised;
B3, neural net model establishing module 22 are gone into model data store in the model bank 3.
Further, native system has two kinds of working methods, and promptly step c comprises the mode of working online and the mode that works offline.
When working online mode, the model parameters that nerve network controller 1 is set up in real time according to neural network identifier 2 generate controlled quentity controlled variables, and controlled quentity controlled variable is carried out optimizing at particle swarm optimization device 4, topworks's action of control tower during the fructufy of optimization.The model that should set up in real time is for determining the parameter of controller, simultaneously, obtain actual operation parameters in the production run by collection, utilize the flow process parameter value of particle swarm optimization algorithm optimizing optimum, controlled quentity controlled variable setting value as controller, the control executing mechanism action, the flow technological parameter of dynamic adjustments carbonization process.After each the calculating, only export current controlled quentity controlled variable and impose on real process.To constantly next, recomputate controlled quentity controlled variable according to new measurement data.
When working offline mode, neural network identifier 2 is not worked, nerve network controller 1 generates controlled quentity controlled variable according to the model parameter in the model bank 3, particle swarm optimization device 4 carries out optimizing according to the model of model bank 3 storages to controlled quentity controlled variable, the optimum control amount that optimizing is obtained is as the setting value of ANN (Artificial Neural Network) Control, topworks's action of control carbonators, thus realize offline optimization control.
The modeling and optimization control system of the soda carbonization technique process that native system provides can be carried out modeling and optimization control to the core cell-carbonators of soda ash production, improve the automatization level of unit, the steady production operating mode increases production run sodium conversion ratio, reduces consuming.
Claims (3)
1. the soda carbonization technique control system based on neural network is characterized in that: comprise nerve network controller (1), neural network identifier (2), model bank (3), particle swarm optimization device (4), data acquisition channel (5) and carbonators (6);
The output terminal of described carbonators (6) is connected with neural network identifier (2) input end by data acquisition channel (5); Neural network identifier (2) output terminal is connected with the input end of nerve network controller (1), and neural network identifier (2) output terminal is connected with the input end of nerve network controller (1) by model bank (3); Nerve network controller (1) output terminal is connected with particle swarm optimization device (4) input end, and nerve network controller (1) output terminal is connected with particle swarm optimization device (4) input end by data acquisition channel (5); The output terminal of particle swarm optimization device (4) is connected with the input end of carbonators (6).
2. the soda carbonization technique control system based on neural network according to claim 1 is characterized in that: described neural network identifier (2) comprises neural net model establishing module (22), model emulation module (23), model editing module (24); The output terminal of data acquisition channel (5) is connected with the input end of neural net model establishing module (22); Data acquisition channel (5) is connected with model emulation module (23); Model emulation module (23) output terminal is connected with neural net model establishing module (22); Neural net model establishing module (22) output terminal is connected with the input end of model editing module (24) and the input end of model bank (3).
3. the soda carbonization technique control system based on neural network according to claim 2 is characterized in that: described data acquisition channel (5) comprises acquisition module, the data preprocessing module that connects successively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009200566305U CN201449529U (en) | 2009-05-15 | 2009-05-15 | Soda ash carbonization process control system based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009200566305U CN201449529U (en) | 2009-05-15 | 2009-05-15 | Soda ash carbonization process control system based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN201449529U true CN201449529U (en) | 2010-05-05 |
Family
ID=42554251
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009200566305U Expired - Lifetime CN201449529U (en) | 2009-05-15 | 2009-05-15 | Soda ash carbonization process control system based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN201449529U (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101598927B (en) * | 2009-05-15 | 2011-03-30 | 广东工业大学 | Control system of soda carbonization technique based on neural network and control method thereof |
-
2009
- 2009-05-15 CN CN2009200566305U patent/CN201449529U/en not_active Expired - Lifetime
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101598927B (en) * | 2009-05-15 | 2011-03-30 | 广东工业大学 | Control system of soda carbonization technique based on neural network and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101598927B (en) | Control system of soda carbonization technique based on neural network and control method thereof | |
| CN102360181B (en) | Low-temperature heat real-time optimization system based on general algorithm sequential quadratic programming (GA-SQP) mixed optimization strategy | |
| CN103742901B (en) | A kind of defining method of Properties of CFB power consumption analysis optimization target values system | |
| CN102156496B (en) | Blending control method for temperature of reactive kettle | |
| CN111077869A (en) | Optimization control method and system for big data intelligent control bag-type dust collector | |
| CN109185917B (en) | Boiler combustion state online diagnosis method and system based on flame intensity signal | |
| CN102809928A (en) | Control optimizing method for energy consumption of thermal equipment of industrial enterprise | |
| CN113837464A (en) | Load prediction method of cogeneration boiler based on CNN-LSTM-Attention | |
| CN102637020B (en) | A kind of steel rolling heating furnace integrated control system | |
| CN103088448B (en) | A kind of carbon fibre precursor jet stretch technique controlled based on data-driven cooperative intelligent | |
| CN104597755A (en) | Hydrometallurgical gold cyanide leaching optimization method | |
| CN106950824A (en) | Stalk fermentation alcohol fuel process feeding prediction control system and method based on fuzzy neural network | |
| CN109858125B (en) | Thermal power unit power supply coal consumption calculation method based on radial basis function neural network | |
| CN107942873A (en) | A kind of intelligent accounting of the operation cost of Furniture manufacture production line and monitoring method | |
| CN201449529U (en) | Soda ash carbonization process control system based on neural network | |
| CN206696629U (en) | A kind of pressurized water reactor core output control device | |
| CN112094050A (en) | Optical fiber preform rod tail wire drawing control system and method | |
| CN112023650A (en) | Full-working-condition self-adaptive pH adjusting system | |
| CN202600429U (en) | Automatic control system for extraction of traditional Chinese medicine | |
| CN114881207A (en) | Steam pressure prediction method and device based on LSTM deep circulation neural network | |
| CN115826648A (en) | Heating furnace trusteeship temperature control method and system based on cloud edge cooperation | |
| CN101604420A (en) | The method for building up of gas equilibrium model | |
| CN103439964A (en) | On-line updating system and method for OCSVM monitoring model | |
| CN111639742A (en) | System and method for diagnosing state fault of desulfurization and denitrification circulating pump | |
| CN105404146A (en) | Diagnosis method and system for furnace condition of calcium carbide furnace |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| AV01 | Patent right actively abandoned |
Granted publication date: 20100505 Effective date of abandoning: 20090515 |