CN101187803A - Ammonia converter production optimization method based on data excavation technology - Google Patents
Ammonia converter production optimization method based on data excavation technology Download PDFInfo
- Publication number
- CN101187803A CN101187803A CNA2007101718090A CN200710171809A CN101187803A CN 101187803 A CN101187803 A CN 101187803A CN A2007101718090 A CNA2007101718090 A CN A2007101718090A CN 200710171809 A CN200710171809 A CN 200710171809A CN 101187803 A CN101187803 A CN 101187803A
- Authority
- CN
- China
- Prior art keywords
- data
- ammonia
- production
- data mining
- optimization
- 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.)
- Granted
Links
Images
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
A production optimization method of an ammonia synthesizer which is based on the data mining technology belongs to a research frontier of the cross discipline field such as chemical engineering, system engineering, data mining and the like. The invention combines production characteristics of ammonia synthesizers, in accordance with parameter data and analytical measurement data of history production technology of the ammonia synthesizers which are gathered by the distributed control system, namely DCS, production and optimization target of the ammonia synthesizers and mathematical models between the relative technological parameters are established by utilizing the data mining technology. And real time data which are produced by the ammonia synthesizers are input in the mathematical models which are established, production status of the ammonia synthesizers are assured by the models, and models establishing and maintenance, working order diagnosing, target predicting, and optimum operation guide of ammonia synthesizers are achieved. And the ammonia synthesizers are operated under optimum state, and production capability and control level of the device are improved. The method has the advantages of safety and reliability, small risk, low cost and easy pauperization.
Description
Technical field
The present invention relates to a kind of ammonia converter production optimization method based on data mining technology, the research forward position that belongs to interdisciplinary fields such as chemical engineering, systems engineering and data mining contains that production process optimization model and model maintenance, production status are monitored in real time, production status real-time diagnosis and processing parameter optimization.
Technical background
Oil and chemical enterprise are the foundation leg industries of China, occupy critical role in Chinese national economy.But compare with the world petrochemical industry production level, oil of China and chemical industry also have no small gap.For example, the required cash operation cost of one ton of product of every processing has reflected enterprise's production cost level, oil refining cash operation cost in 2003, world average level is 117 yuan/ton, advanced international standard is 86 yuan/ton, and China is about 130 yuan/ton, exceeds 11% and 51% respectively.The cash operation cost of ethylene production, China is 142 dollars/ton, exceeds 24% than advanced international standard, exceeds 5% than the Asian-Pacific area.Therefore, how to utilize industrial optimisation technique to raise labour productivity and resource utilization, promote the profitability and the competitiveness of China's oil and chemical industry comprehensively, crucial meaning is arranged for the sustainable development of China's oil and chemical industry.
Set about aspects such as the production level that promotes enterprise can the slave unit transformation, process modification, can obtain extraordinary effect though facts have proved these measures, and the cycle is long, investment is big.In contrast to this, utilize control technology and computer software technology that production operation is optimized, be easy to implement, instant effect, rate of return on investment height, just more and more obtain the attention of industry.In recent years, (Distributed Control System DCS) has been widely used in the big-and-middle-sized petrochemical equipment of China to Distributed Control System (DCS), for pilot and the domestic and international new technology of popularization lay the foundation.Existing in the world at present kind of petrochemical industry Optimization Software surplus company has released 30 more than 20, application spreads all over main petrochemical equipment, wherein advanced control (Advanced Process Control, APC) technology is implemented decompression as usual, catalytic cracking, catalytic reforming (CR), hydrocracking, polypropylene, tygon or the like at tens process units of China.According to the investigation result of Chemshare company, the gain on investments ratio of implementing advanced control on existing DCS system-based is 1: 4, and the gain on investments of controlling implement device real-time optimization on the basis the advanced person is 1: 4 than also.Therefore, advanced control and real-time optimization control enhancing efficiency by relying on tapping internal latent power effect are obvious.
In recent years, paid much attention to abroad based on the industrial optimisation technique of data mining, the case of application is increasing.It is a great challenge that data mining technology is introduced chemical production field.The production run of large and medium-sized enterprise such as modern chemical industry, oil refining comprises complicated physics, chemical change, and these variations must just can make the comprehensive benefit of production reach optimum by the monitoring to the kinds of processes index.Chemical process breaks down unavoidably sometimes, can in time find when breaking down, thereby correctly diagnose its reason to eliminate fault, also will be by summing up the rule that fault occurs, set up the mathematical model of discovery, tracing trouble, thereby smooth handling failure returns to normal production.These all need extracting useful information from large quantities of complex datas, set up the mathematical model of reflection objective law.Adopting data mining technology is one of necessary means of setting up mathematical model.Data mining technology be used for produce optimizing can control with the advanced person, real-time optimization control complements one another, bring out the best in each other.Chemical process relates to physics, the chemical change of many complexity, usually is difficult to set up model by mechanism, even if set up model, its precision is also very low, and model can only be used for showing the cardinal principle variation tendency of production, produces and can't be used for instructing.In addition, there is many variable factors and the interference (variation of feedstock property, equipment state, operation operating mode in the industrial processes, the interference of production environment and production system self), mathematical model is normally set up under a certain specified conditions, thereby only among a small circle, be suitable for, in the changeable production of actual complex, be difficult to use.Development along with computer science and process system engineering, the industrial processes automaticity is more and more higher, industrial production data is gathered and the more and more economic facility of storage, for a medium scale petrochemical iy produced device, the instrument item of its DCS system about 500 points of counting are if per minute is preserved a production data, so, 700,000 production datas are just arranged every day, can reach 2.5 hundred million data in 1 year.These data recording the feature of industrial processes, performance, variation etc., be the essence reflection of process units.Utilize data mining technology, can from industrial production data, seek rule and find knowledge, and instruct the production run of enterprise, thereby reach the optimization production process, make performance of enterprises maximization with these knowledge.
Ammonia synthesizer is the crucial workshop section of ammonia synthesizing industry, realize purpose energy-conservation, volume increase, and except improving technological process, device structure, the process control level and the operational administrative level of making great efforts the raising ammonia synthesizer are very important approach.Adopted Distributed Control System (DCS) DCS more at large on the present large and medium-sized ammonia synthesizer, and accumulated abundant production data, for optimal control provides possibility.But because digestion, absorption and independent development insufficient investment, DCS still is mainly used in data acquisition, shows and print and single loop control.In recent years, domestic and international many workers have carried out a large amount of research to the simulation of ammonia synthesizer with optimization, and have developed corresponding software, and these softwares mostly adopt simplified model or mechanism model to carry out the synthetic ammonia simulative optimization.Also has a certain distance apart from the production practical application.And the ammonia synthesizer production status is subjected to influence of various factors such as temperature, pressure, flow, gas componant in the actual production process, and processing parameter is rule of thumb determined basically at present, and it is very big to optimize potentiality.
Summary of the invention
The present invention has mainly solved the ammonia converter production optimization problem based on data mining technology.The present invention is in conjunction with the production characteristics of ammonia synthesizer, ammonia synthesizer production history data are placed in the higher dimensional space investigate, carry out a series of feature extraction by various data mining technologies, foundation is in order to describe the mathematical model of production law, and through forming a plurality of figures after the dimension-reduction treatment, show the rule that is hidden in the data mid-deep strata with brand-new visual angle, realize that finally ammonia synthesizer modeling and model maintenance, operating mode diagnosis and target prediction, Optimizing operation instruct.The process and the result of whole data mining have used visualization technique, and production engineer's data mining of participating in the overall process combines oneself knowhow and process knowledge and data mining technology.Based on the block diagram of the ammonia converter production optimization method of data mining technology referring to Fig. 1.Ammonia converter production optimization method based on data mining technology is made up of five aspects: (1) data acquisition: the production history technological parameter data of gathering ammonia synthesizer; (2) data pre-service: the production history data are removed noise processed, for data mining provides true and reliable data; (3) data mining: from the production history data, find knowledge and seek rule, from the data digging method storehouse, call relevant method according to the requirement of optimization problem and the situation of production data, set up the mathematical model between ammonia converter production optimization target and relevant processing parameter; (4) visual analyzing: the sample point in the higher dimensional space is mapped on the plane after by dimensionality reduction so as image, directly perceived, investigate and optimize the distribution of distinguishing from various visual angles, provide important references for the data mining expert seeks the optimization rule; (5) generate the optimization result: produce various forms of optimization conclusions.
The present invention is based on the ammonia converter production optimization method of data mining technology, carry out as follows.
1, ammonia synthesizer production history data acquisition
Ammonia converter production optimization method based on data mining technology, at first will determine production optimization aim T (synthetic ammonia output, ammonia net value or ton ammonia virgin gas consumption), T is subjected to M processing parameter (temperature, pressure, flow, flow velocity, gas componant, synthetic ammonia quality) X
1, X
2..., X
MInfluence, one group of homology discrete data set { P of Que Dinging therefrom
iBe called one of this optimization aim and describe sample set.P wherein
i={ X
I1, X
I2..., X
IM, T
i(i=1,2 ..., N) be called a sample.Gather the DCS system processing parameter historical data that is total to N sample in a period of time.For the data file that guarantees to gather comprises necessary quantity of information, sample number N is abundant, promptly will satisfy: N/M 〉=20.According to production optimization aim T, sample is divided into " 1 ", " 2 " two classes then, " 1 " class sample is called excellent class sample, and " 2 " class sample is called poor class sample.
2, ammonia synthesizer production history data pre-service
Ammonia converter production optimization method based on data mining technology, on the basis of gathering the production history data, utilize data mining technology, the production history data are carried out data file assessment (hyperpolyhedron model evaluation, arest neighbors leaving-one method criterion, the non-linear regression criterion), data structure analysis (differentiate by topological classification, the neighbour analyzes, near-linear is analyzed, time series analysis etc.), screening sample (cut apart by the subspace, the dead band is cut apart, outlier deletion etc.), correlation analysis (single-factor analysis, double factor is analyzed, multifactor optimization), independent variable screening (ballot method, the entropy method, the hyperpolyhedron method), finally obtain training sample set and test sample book collection.Partial analysis method and the method brief description used see Table 1.
Table 1 data analysing method and brief description
| Analytical approach | Brief description |
| Data analysis | Data analysing method based on mathematical statistics |
| Descriptive statistic | The calculating and the four kinds of statistical graphs commonly used that contain the basic statistics amount. |
| Trend analysis | Can show the trend map of a plurality of variablees in groups, and have functions such as rearrangement. |
| Correlation analysis | Show in the matrix diagram mode and correlativity between the variable to show the distribution of all kinds of samples in the drawings simultaneously. |
| Level line is analyzed | It is the analysis tool that concerns between three variablees of research and the target.Can select wherein three from all variablees arbitrarily, figure will show related between three variablees and the target. |
3, the foundation of ammonia converter production optimization model and visual analyzing
Ammonia converter production optimization method based on data mining technology, carry out data mining according to the production history data of ammonia synthesizer DCS system acquisition being carried out the training sample set that pre-service obtains, set up the ammonia converter production optimization model, be used to instruct the production Optimizing operation.Core based on the ammonia converter production optimization method of data mining technology is a data digging method, this data digging method integrated use several different methods such as mathematical statistics, pattern-recognition, machine learning, and the method for our original creation, and designed a rational flow chart of data processing (based on the ammonia converter production optimization method flow chart of data processing of data mining technology referring to Fig. 2), make these methods form an organic whole, the mathematical model of ammonia synthesizer production is instructed in final foundation, thereby reaches the purpose of optimization production.In the actual production, have only good model just to have good production to optimize effect, when the equipment of process units or technological process take place than cataclysm, the deterioration in accuracy of Optimization Model, influence the accuracy rate of operating mode diagnosis and target prediction, thereby influence the effect that Optimizing operation instructs.At this moment, should consider modeling again.For this reason, can carry out the model maintenance operation based on the ammonia converter production optimization method of data mining technology, the user can rebulid new model as required.For image, the scheme that is optimized intuitively, adopted visualization technique based on the ammonia converter production optimization method of data mining technology.Partial data method for digging and visual analyzing technology and brief description see Table 2.
Table 2 data digging method and visual analyzing technology and brief description
| Data digging method and visual analyzing technology | Brief description |
| Pattern-recognition | This module is the data digging method based on pattern-recognition, is applicable to have strongly connected complicated applications occasion between multivariate, the variable. |
| Characteristic pattern | Multivariate space originally obtains a two-dimensional figure behind the mode identification technology dimensionality reduction, Here it is characteristic pattern.The regularity of distribution that can show the different sample points of two classes from the characteristic pattern is research and the important figure of finding to optimize rule. |
| The operating mode control chart | The one dimension figure that constitutes by two proper vectors respectively.Can monitor the operating mode of production run. |
| T square of figure | According to given quantity of information, constitute new statistic by front several features vector, be called the T square value.T square of figure is the one dimension figure, can be used to monitor production status. |
| SPE figure | The SPE value can be used to weigh model error.SPE figure can find the abnormal conditions in the production run. |
| The analog meter chart | Represent each variable with the mode of analogue instrument, link the variation of emulation production status, study the effect of each variable visual in imagely by " instrument " and characteristic pattern.This module is that this optimization system is peculiar. |
| Load diagram | Show the effect of each variable with graphics mode, Ben Tu and characteristic pattern contrast are used, in order to study each variable role. |
| The variable weight map | Show the size of each variable with histogrammic mode to the production effect. |
| Contribution plot 1 | Behind a given sample point, show contribution and the direction of each variable with graphics mode to sample point institute corresponding states, be the effective tool that research causes certain specific operation reason. |
| |
Behind given two sample points, be presented in the variation of two sample point institute corresponding statess with the mode of figure, the contribution of each variable and direction are the effective tools that research causes the working conditions change reason. |
4, production Optimization Model checking
Based on the ammonia converter production optimization method of data mining technology, after obtaining the production Optimization Model, the model that utilizes the test sample book set pair to be set up is verified, to investigate the reliability and stability of model.If model stability is reliable, promptly can be used for diagnosis of ammonia synthesizer production status and Operating Guideline.
5, online operating mode diagnosis and Operating Guideline
Ammonia converter production optimization method based on data mining technology, on the ammonia converter production optimization model based of having set up, gatherer DCS system real time data, can realize: diagnosis of (1) real-time working condition and target prediction: the real-time working condition diagnosis is inferred the operating mode of production according to the measured value of the real-time technological parameter of DCS system and the production Optimization Model of customization exactly, so that problem in time finding to produce, adjust manufacturing parameter, make operating mode keep good state.If manually import the measured value of some technological parameter, system will be according to the production Optimization Model automatic forecasting desired value of definite value.(2) Optimizing operation instructs: if current working is in the state of " poor ", how to regulate technological parameter, make operating mode forward the state of " good " to? the factor that influences optimization aim more for a long time, regulate which parameter? what are regulated? this is the difficult problem that the production operation personnel are faced, present most of enterprise only depends on workman's experience, goes back the effective aid decision making instrument of neither one.Ammonia converter production optimization method based on data mining technology will provide detailed Operating Guideline for operating personnel, by online version analog meter chart, tell you should adjust that Several Parameters, and wherein each parameter adjustment is to what.Ammonia converter production optimization method based on data mining technology also provides various figures, helps operating personnel to analyze operating mode, and the decision operational motion is observed the effect after operating.See Table 3 based on online operating mode diagnosis of the ammonia converter production optimization method of data mining technology and the description of Operating Guideline major function.
The online operating mode diagnosis of table 3 and Operating Guideline major function and brief description
| The function title | Brief description |
| The operating mode diagnosis | Production data is directly imported Optimization Model, device is carried out the diagnosis of online in real time operating mode. |
| Online trend map | Each point of online trend map has been represented production status at that time, and the broken line of being made up of some points has been represented the process that production status changes in certain period.As time goes on, new production status point constantly refreshes broken line, will occur a constantly broken line of wriggling on characteristic plane.Online trend map has shown the track that production run changes, rectangular box among the figure is expressed as optimizes the zone, or the production control district, observe production status point and whether be in the boxed area, just can infer whether production is in the desired control state.In case run out of excellent district, illustrate to produce and left the optimization operating mode, need adjust the production operation parameter. |
| On-line Control figure | On-line Control figure develops from characteristic trend figure, be similar to the control chart (being called " stopping Hart figure " again) of quality management, form by two width of cloth figures, the longitudinal axis is respectively two proper vectors of characteristic pattern, transverse axis is the time, and two horizontal lines up and down in the figure are respectively the up-and-down boundary in excellent district.Because this optimization system is got two proper vectors (first feature and second proper vector), therefore must observe corresponding therewith two On-line Control figure simultaneously. |
| The analog meter chart | The analog meter chart is by the computer simulation based on Optimization Model, investigates complicated relation between each variable and the optimization aim intuitively, visually.The rectangular area on the right is that the left side is the bar shaped button of correlated variables, changes the size of each variable in the optimization district that characteristic pattern obtains in the analog meter chart, and the bead of analog meter chart will move inside and outside rectangle region. |
| Optimize prompting | Optimize prompting and can tell how the production technology personnel regulate current variable, make the concrete operations scheme of optimizing the district of adjusting to of producing. |
| Operating Guideline | Optimization system instructs for the user provides production operation, tell the production operation personnel: how is present production status? if do not produce and optimizing the district, then further remind operating personnel should regulate which parameter, transfer, what are regulated toward what direction. |
| Generate form | Real time data derive make the production technology personnel can the derived data storehouse in the production data of a certain period. |
The ammonia converter production optimization method that the present invention is based on data mining technology is used for the production optimization of synthetic ammonia, can not change (or very much not changing) equipment not doing (or doing less) experiment, does not disturb optimization production under the prerequisite of production.Said method is after putting into operation on the device, and further steadily and optimized the production operation of synthetic ammonia, the up-to-standard rate of ammonia reaches 100%, and hydrazine yield has increased by 6%.This method has satisfied process engineering technology and operating personnel's actual needs, and optimization method is reliable, stable after tested.
Successful Application based on the ammonia converter production optimization method of data mining technology, DCS system data resource is fully used, the data mining technology successful Application is produced reality to ammonia synthesizing industry, and achievement of the present invention can be generalized on the ammonia synthesizer of the same type.
Description of drawings
Fig. 1: based on the ammonia converter production optimization method block diagram of data mining technology
Fig. 2: based on the ammonia converter production optimization method flow chart of data processing of data mining technology
Fig. 3: based on the ammonia converter production optimization method through engineering approaches embodiment of data mining technology
Fig. 4: No. 1 synthetic tower production Optimization Model pattern-recognition perspective view
Fig. 5: No. 2 synthetic tower production Optimization Model pattern-recognition perspective views
Fig. 6: No. 3 synthetic tower production Optimization Model pattern-recognition perspective views
Embodiment
The present invention is successful Application on Zhanhua branch office of Yunnan Yun Wei Group Co.,Ltd ammonia synthesizer, well solved engineering construction problem (the through engineering approaches embodiment is referring to Fig. 3), realized that ammonia synthesis process units production Optimization Model is set up, real-time working condition is diagnosed, Optimizing operation instructs.After the ammonia synthesizer optimization operation, not only stablized the quality index of synthetic ammonia preferably, and synthetic ammonia output has improved 6%.Now be illustrated with following indefiniteness embodiment.
One: No. 1 synthesizer production of embodiment Optimization Model
1, optimization aim variable and optimization independent variable: to go into the fresh tolerance of tower (FI0811a, Nm
3/ h) be target variable, and related technical parameters such as virgin gas hydrogen richness (AI0811c, %), cold shock three flows (FI0805a, Nm
3/ h), one section temperature in (TI0801_2a of synthetic tower, ℃), synthetic tower second stage exit temperature (TI0801_5a, ℃), three sections temperature (TIC08004a_PV of synthetic tower, ℃), three sections temperature (TI0801_12a of synthetic tower, ℃), four sections temperature (TI0801_16a of synthetic tower, ℃), 42 technological parameters of useless pot inlet air temperature degree (TI0803a) and water cooler outlet temperature (TI0807a, ℃) or the like are that independent variable is analyzed.Fresh tolerance is greater than 26500Nm
3The sample of/h is excellent class sample (being defined as 1 class), smaller or equal to 26500Nm
3/ h sample is difference class sample (being defined as 2 classes).
2, data set: gather No. 1 tower of ammonia synthesizer DCS system production data of 10 on July 6,12 o'clock to 2006 on the 23rd April in 2006, per 30 minutes samples, training set effective sample number is 3548.
3, data pre-service: the laggard line data file assessment of the sample of removal device start-stop car and load fluctuation (hyperpolyhedron model evaluation in above-mentioned training set, arest neighbors leaving-one method criterion, the non-linear regression criterion), data structure analysis (differentiate by topological classification, the neighbour analyzes, near-linear is analyzed, time series analysis etc.), screening sample (cut apart by the subspace, the dead band is cut apart, outlier deletion etc.), correlation analysis (single-factor analysis, double factor is analyzed, multifactor optimization), independent variable screening (ballot method, the entropy method, the hyperpolyhedron method) step, finally forming sample number is 1646, variable number is 9 a modeling sample collection.
4, set up the production Optimization Model: data mining results shows, influencing the main technologic parameters that No. 1 synthetic tower goes into the fresh tolerance of tower has: and the virgin gas hydrogen richness (AI0811c, %), cold shock three flows (FI0805a, Nm
3/ h), one section temperature in (TI0801_2a of synthetic tower, ℃), synthetic tower second stage exit temperature (TI0801_5a, ℃), three sections temperature (TIC08004a_PV of synthetic tower, ℃), three sections temperature (TI0801_12a of synthetic tower, ℃), four sections temperature (TI0801_16a of synthetic tower, ℃), useless pot inlet air temperature degree (TI0803a) and water cooler outlet temperature (TI0807a, ℃).Operation mode identification projection finds to have clear regularity (Fig. 4) in the hyperspace of being opened by above-mentioned 9 variablees.The horizontal stroke of standardized data among the figure (F (1)), ordinate (F (2)) equation are:
F(1)＝-0.05×(AI0811c+0.57×(FI0805a)+0.45×(TI0801_2a)-0.14(TI0801_5a)+0.15×(TIC0804a_PV)-0.48×(TI0801_12a)+0.34×(TI0801_16a)+027×(TI0803a)-0.08×(TI0807a)
F(2)＝-0.06×(AI0811c)-0.03×(FI0805a)+0.32×(TI0801_2a)+0.02×(TI0801_5a)+0.11×(TIC0804a_PV)+0.44×(TI0801_12a)-0.35×(TI0801_16a)+0.73×(TI0803a)+0.18×(TI0807a)
Total excellent class sample of sample area accounts for 49.45% among Fig. 4, and excellent class sample accounts for 95.38% in the excellent class sample area.Optimize in the district if make to produce to maintain, excellent class sample ratio can be improved 45.93%.
5, modelling verification: for the reliability of verification model, the production in we set up with the production data on July 6,10 o'clock 12 o'clock to 2006 on the 23rd April in 2006 model prediction on July 8,11: 30 0 o'clock to 2006 on the 7th July in 2006, totally 95 in sample utilizes the rate of accuracy reached 84.2% of the mathematical model prediction sample set sample class that data digging method obtains.Therefore, institute's established model can satisfy the production needs substantially.
6, online operating mode diagnosis and Operating Guideline: the real-time production data of ammonia synthesizer DCS system is imported in the model of having set up, realized online production status diagnosis, target prediction and production operation guidance.
Two: No. 2 ammonia converter production optimization models of embodiment
1, optimization aim variable and optimization independent variable: to go into the fresh tolerance of tower (FI0811b, Nm
3/ h) be the optimization aim variable, related technical parameters such as virgin gas hydrogen richness (AI0811c, %), one section temperature in (TI0801_7b of synthetic tower, ℃), synthetic tower one section outlet temperature (TIC0801_2b_PV, ℃), two sections temperature (TI0801_3b of synthetic tower, ℃), two sections temperature (TI0801_10b of synthetic tower, ℃), three sections temperature (TIC0801_12b_PV of synthetic tower, ℃), three sections temperature (TI0801_6b of synthetic tower, ℃), useless pot inlet air temperature degree (TI0804b) and water cooler outlet temperature (TI0807b, ℃) or the like totally 42 technological parameters are that independent variable is analyzed.Fresh tolerance is greater than 34000Nm
3The sample of/h is excellent class sample (being defined as 1 class), smaller or equal to 34000Nm
3/ h sample is difference class sample (being defined as 2 classes).
2, data set: gather No. 2 towers of ammonia synthesizer DCS system production data of 10 on July 6,12 o'clock to 2006 on the 23rd April in 2006, per 30 minutes samples, training set effective sample number is 3548.
3, data pre-service: in above-mentioned training set, carry out data file assessment (hyperpolyhedron model evaluation behind the sample of removal device start-stop car and load fluctuation successively, arest neighbors leaving-one method criterion, the non-linear regression criterion), data structure analysis (differentiate by topological classification, the neighbour analyzes, near-linear is analyzed, time series analysis), screening sample (cut apart by the subspace, the dead band is cut apart, the outlier deletion), correlation analysis (single-factor analysis, double factor is analyzed, multifactor optimization), independent variable screening (ballot method, the entropy method, the hyperpolyhedron method) step, finally forming sample number is 2088, variable number is 9 a modeling sample collection.
4, set up the production Optimization Model: data mining results shows, influencing the main technologic parameters that No. 2 synthetic towers go into the fresh tolerance of tower has: virgin gas hydrogen richness (AI0811c, %), one section temperature in (TI0801_7b of synthetic tower, ℃), synthetic tower one section outlet temperature (TIC0801_2b_PV, ℃), two sections temperature (TI0801_3b of synthetic tower, ℃), two sections temperature (TI0801_10b of synthetic tower, ℃), three sections temperature (TIC0801_12b_PV of synthetic tower, ℃), three sections temperature (TI0801_6b of synthetic tower, ℃), useless pot inlet air temperature degree (TI0804b) and water cooler outlet temperature (TI0807b, ℃) etc.Operation mode identification projection finds to have clear regularity (Fig. 5) in the hyperspace of being opened by above-mentioned 9 variablees.The horizontal stroke of standardized data among Fig. 8 (F (1)), ordinate (F (2)) equation are:
F(1)＝-0.03×(AI0811c)+0.18×(TIC0801_2b_PV)+0.23×(TI0801_3b)-0.14×(TI0801_6b)-0.13×(TI0801_7b)-0.47×(TI0801_10b)+0.27×(TIC0801_12b_PV)+0.76×(TI0804b)-0.01×(TI0807b)
F(2)＝0.13×(AI0811c)+0.08×(TIC0801_2b_PV)+0.22×(TI0801_3b)+0.50×(TI0801_6b)-0.29×(TI0801_7b)+0.27×(TI0801_10b)-0.62×(TIC0801_12b_PV)+0.36×(TI0804b)-0.06×(TI0807b)＜2.49
Total excellent class sample of sample area accounts for 48.28% among Fig. 5, and excellent class sample accounts for 95.80% in the excellent class sample area.Optimize in the district if make to produce to maintain, excellent class sample ratio can be improved 47.52%.
5, modelling verification: for the reliability of verification model, the production in we set up with the production data on July 6,10 o'clock 12 o'clock to 2006 on the 23rd April in 2006 model prediction on July 8,11: 30 0 o'clock to 2006 on the 7th July in 2006, totally 95 in sample utilizes the rate of accuracy reached 91.6% of the mathematical model prediction sample set sample class that data digging method obtains.Therefore, institute's established model has good robustness.
6, online operating mode diagnosis and Operating Guideline: the real-time production data of ammonia synthesizer DCS system is imported in the model of having set up, realized online production status diagnosis, target prediction and production operation guidance.
Three: No. 3 ammonia converter production optimization models of embodiment
1, optimization aim variable and optimization independent variable: to go into the fresh tolerance of tower (FI0811c, Nm
3/ h) be the optimization aim variable, related technical parameters such as virgin gas hydrogen richness (AI0811c, %), circulating air hydrogen richness (AI0802c, %), one section temperature in (TI0801_2c of synthetic tower, ℃), synthetic tower one section outlet temperature (TIC0801_4c_PV, ℃), two sections temperature (TIC0801_6c_PV of synthetic tower, ℃), three sections temperature (TIC0801_9c_PV of synthetic tower, ℃), three sections temperature (TI0801_12c of synthetic tower, ℃), secondary air outlet temperature (TI0816c) and water cooler outlet temperature (TI0820c, ℃) or the like 52 technological parameters are that independent variable is analyzed.Fresh tolerance is greater than 27500Nm
3The sample of/h is excellent class sample (being defined as 1 class), smaller or equal to 27500Nm
3/ h sample is difference class sample (being defined as 2 classes).
2, data set: gather No. 3 towers of ammonia synthesizer DCS system production data of 10 on July 6,12 o'clock to 2006 on the 23rd April in 2006, per 30 minutes samples, training set effective sample number is 3548.
3, data pre-service: in above-mentioned training set, carry out data file assessment (hyperpolyhedron model evaluation behind the sample of removal device start-stop car and load fluctuation successively, arest neighbors leaving-one method criterion, the non-linear regression criterion), data structure analysis (differentiate by topological classification, the neighbour analyzes, near-linear is analyzed, time series analysis), screening sample (cut apart by the subspace, the dead band is cut apart, the outlier deletion), correlation analysis (single-factor analysis, double factor is analyzed, multifactor optimization), independent variable screening (ballot method, the entropy method, the hyperpolyhedron method) step, finally forming sample number is 1866, variable number is 9 a modeling sample collection.
4, set up the production Optimization Model: data mining results shows, influencing the main technologic parameters that No. 3 synthetic towers go into the fresh tolerance of tower has: virgin gas hydrogen richness (AI0811c, %), circulating air hydrogen richness (AI0802c, %), one section temperature in (TI0801_2c of synthetic tower, ℃), synthetic tower one section outlet temperature (TIC0801_4c_PV, ℃), two sections temperature (TIC0801_6c_PV of synthetic tower, ℃), three sections temperature (TIC0801_9c_PV of synthetic tower, ℃), three sections temperature (TI0801_12c of synthetic tower, ℃), secondary air outlet temperature (TI0816c) and water cooler outlet temperature (TI0820c, ℃).Operation mode identification projection finds to have clear regularity (Fig. 9) in the hyperspace of being opened by above-mentioned 9 variablees.The horizontal stroke of standardized data among Fig. 9 (F (1)), ordinate (F (2)) equation are:
F(1)＝-0.15×(AI0811c)+0.13×(AI0802c)-0.42×(TI0801_2c)+0.03×(TIC0801_4c_PV)+0.28×(TIC0801_6c_PV)+0.63×(TIC0801_9c_PV)-0.42×(TI0801_12c)+0.30×(TI0816c)-0.21×(TI0820c)
F(2)＝-0.23×(AI0811c)+0.30×(AI0802c)-0.32×(TI0801_2c)+0.02×(TIC0801_4c_PV)+0.0003×(TIC0801_6c_PV)-0.27×(TIC0801_9c_PV)+0.59×(TI0801_12c)+0.39×(TI0816c)-0.42×(TI0820c)
Total excellent class sample of sample area accounts for 47.64% among Fig. 6, and excellent class sample accounts for 95.87% in the excellent class sample area.Optimize in the district if make to produce to maintain, excellent class sample ratio can be improved 48.23%.
5, online operating mode diagnosis and Operating Guideline: the real-time production data of ammonia synthesizer DCS system is imported in the model of having set up, realized online production status diagnosis, target prediction and production operation guidance.
Claims (7)
1. ammonia converter production optimization method based on data mining technology is characterized in that the concrete steps of this method are:
A. gather historical processing parameter data of ammonia synthesizer and analyzing test data;
The historical data of b. utilize gathering, the maintenance data digging technology is set up the mathematical model between ammonia converter production optimization target and relevant technological parameter;
C. will install the production real time data is input in the mathematical model of having set up;
D. determine the ammonia synthesizer production status according to this model, realize ammonia converter production optimization.
2. the ammonia converter production optimization method based on data mining technology as claimed in claim 1, it is characterized in that the historical processing parameter data described in the step a are device Distributed Control System (DCS) (Distributed ControlSystem, DCS) Ji Lu temperature, pressure, flow, flow velocity, gas componant data; The historical analysis test data is synthetic ammonia output, synthetic ammonia qualitative data.
3. the ammonia converter production optimization method based on data mining technology as claimed in claim 1, it is characterized in that the data mining technology described in the step b, comprise selection and sampling, data scrubbing and pre-service, data conversion and yojan, data digging method, model evaluation and explanation, result's report and use.
4. the ammonia converter production optimization method based on data mining technology as claimed in claim 1 is characterized in that the production optimization aim described in the step b is synthetic ammonia output, ammonia net value or ton ammonia virgin gas consumption.
5. the ammonia converter production optimization method based on data mining technology as claimed in claim 1 is characterized in that the production real time data described in the step c is the processing parameter data that synthetic ammonia installation DCS system gathers in real time.
6. the ammonia converter production optimization method based on data mining technology as claimed in claim 1 is characterized in that the ammonia converter production optimization described in the steps d, comprises that model maintenance, operating mode diagnosis, target prediction and Optimizing operation instruct.
7. the ammonia converter production optimization method based on data mining technology as claimed in claim 3 is characterized in that described data digging method is the hyperpolyhedron method, pattern-recognition is projecting method step by step, non-linear best projection regression modeling method, the variable screening technique, principal component analytical method, multiple differentiation vector method, albefaction linear mapping method, the offset minimum binary method, dress box method, the arest neighbors method, Artificial Neural Network, genetic algorithm, rough set method, traditional decision-tree, one or more method combinations in support vector machine method and the method for visualizing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101718090A CN101187803B (en) | 2007-12-06 | 2007-12-06 | Ammonia converter production optimization method based on data excavation technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101718090A CN101187803B (en) | 2007-12-06 | 2007-12-06 | Ammonia converter production optimization method based on data excavation technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101187803A true CN101187803A (en) | 2008-05-28 |
| CN101187803B CN101187803B (en) | 2011-07-20 |
Family
ID=39480236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2007101718090A Expired - Fee Related CN101187803B (en) | 2007-12-06 | 2007-12-06 | Ammonia converter production optimization method based on data excavation technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101187803B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542335A (en) * | 2011-06-16 | 2012-07-04 | 广州市龙泰信息技术有限公司 | Mixed data mining method |
| CN106527143A (en) * | 2016-12-07 | 2017-03-22 | 吉林师范大学 | SCR system urea injection control method based on data drive prediction control |
| CN106597853A (en) * | 2016-12-28 | 2017-04-26 | 中南大学 | Active dynamic regulation method in hydrocracking process |
| CN109376778A (en) * | 2018-10-09 | 2019-02-22 | 宁波大学 | A kind of failure modes diagnostic method based on characteristic variable weighting |
| CN109643085A (en) * | 2016-08-23 | 2019-04-16 | 埃森哲环球解决方案有限公司 | Real-time industrial equipment production forecast and operation optimization |
| CN112099435A (en) * | 2019-06-18 | 2020-12-18 | 发那科株式会社 | Diagnostic device and diagnostic method |
| CN112130453A (en) * | 2020-07-30 | 2020-12-25 | 浙江中控技术股份有限公司 | Control method and system for improving MCS production stability based on machine learning |
| WO2021007845A1 (en) * | 2019-07-16 | 2021-01-21 | 东北大学 | Cloud-edge collaborative forecasting system and method for aluminum oxide production indexes |
| CN113341888A (en) * | 2021-04-06 | 2021-09-03 | 普赛微科技(杭州)有限公司 | Multivariable process control method |
| CN117348503A (en) * | 2023-12-06 | 2024-01-05 | 山东三岳化工有限公司 | Propylene oxide production data monitoring system and method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4495960B2 (en) * | 2003-12-26 | 2010-07-07 | キヤノンＩｔソリューションズ株式会社 | Model creation device for the relationship between process and quality |
| CN101075124A (en) * | 2007-06-15 | 2007-11-21 | 武汉钢铁(集团)公司 | Stanermo wind-cooling linear fuzzy controlling method and system |
-
2007
- 2007-12-06 CN CN2007101718090A patent/CN101187803B/en not_active Expired - Fee Related
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542335A (en) * | 2011-06-16 | 2012-07-04 | 广州市龙泰信息技术有限公司 | Mixed data mining method |
| US11264121B2 (en) | 2016-08-23 | 2022-03-01 | Accenture Global Solutions Limited | Real-time industrial plant production prediction and operation optimization |
| CN109643085A (en) * | 2016-08-23 | 2019-04-16 | 埃森哲环球解决方案有限公司 | Real-time industrial equipment production forecast and operation optimization |
| CN109643085B (en) * | 2016-08-23 | 2022-05-10 | 埃森哲环球解决方案有限公司 | Real-time industrial plant production prediction and operational optimization |
| CN106527143A (en) * | 2016-12-07 | 2017-03-22 | 吉林师范大学 | SCR system urea injection control method based on data drive prediction control |
| CN106527143B (en) * | 2016-12-07 | 2019-02-15 | 吉林师范大学 | SCR system method for urea injection control based on data-driven PREDICTIVE CONTROL |
| CN106597853A (en) * | 2016-12-28 | 2017-04-26 | 中南大学 | Active dynamic regulation method in hydrocracking process |
| CN106597853B (en) * | 2016-12-28 | 2019-05-31 | 中南大学 | One kind being hydrocracked active dynamic regulation method in process |
| CN109376778A (en) * | 2018-10-09 | 2019-02-22 | 宁波大学 | A kind of failure modes diagnostic method based on characteristic variable weighting |
| CN109376778B (en) * | 2018-10-09 | 2021-06-15 | 宁波大学 | Fault classification diagnosis method based on characteristic variable weighting |
| CN112099435A (en) * | 2019-06-18 | 2020-12-18 | 发那科株式会社 | Diagnostic device and diagnostic method |
| CN112099435B (en) * | 2019-06-18 | 2024-02-06 | 发那科株式会社 | Diagnostic device and diagnostic method |
| WO2021007845A1 (en) * | 2019-07-16 | 2021-01-21 | 东北大学 | Cloud-edge collaborative forecasting system and method for aluminum oxide production indexes |
| CN112130453A (en) * | 2020-07-30 | 2020-12-25 | 浙江中控技术股份有限公司 | Control method and system for improving MCS production stability based on machine learning |
| CN113341888A (en) * | 2021-04-06 | 2021-09-03 | 普赛微科技(杭州)有限公司 | Multivariable process control method |
| CN117348503A (en) * | 2023-12-06 | 2024-01-05 | 山东三岳化工有限公司 | Propylene oxide production data monitoring system and method |
| CN117348503B (en) * | 2023-12-06 | 2024-02-20 | 山东三岳化工有限公司 | Propylene oxide production data monitoring system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101187803B (en) | 2011-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101187803B (en) | Ammonia converter production optimization method based on data excavation technology | |
| US8670874B2 (en) | Method and apparatus for energy and emission reduction | |
| CN101619850B (en) | Dispatching method and dispatching system based on load online forecasting of thermoelectric power system | |
| CN101446827B (en) | Process fault analysis device of process industry system and method therefor | |
| CN109886430A (en) | A kind of equipment health state evaluation and prediction technique based on industrial big data | |
| JP6298214B2 (en) | System and method for maximizing the expected utility of signal injection test patterns within a utility grid | |
| US20230229124A1 (en) | Operation control system and a control method for a gas-steam combined cycle generator unit | |
| CN110161999A (en) | Coking intelligent manufacturing system based on big data | |
| CN107291830A (en) | A kind of creation method of equipment knowledge base | |
| Gong et al. | Energy efficiency evaluation in ethylene production process with respect to operation classification | |
| CN117434912B (en) | Method and system for monitoring production environment of non-woven fabric product | |
| CN117670378B (en) | Food safety monitoring method and system based on big data | |
| CN116184948A (en) | Intelligent monitoring disc for water plant and application system and method of early warning diagnosis technology | |
| CN114677025A (en) | Intelligent management system and management method for catalyst operation | |
| CN105900030A (en) | Method, system, and computer program product for analyzing production and/or process-engineering processes and/or process steps in a plant | |
| CN113177362A (en) | Furnace temperature prediction method and device based on furnace temperature soft measurement model | |
| EP3906447B1 (en) | Method, system and computer program product for evaluation of energy consumption in industrial environments | |
| CN117217553A (en) | Low-carbon operation method of cogeneration system | |
| CN111367255A (en) | Performance evaluation test system and method for multi-variable control system | |
| Appel et al. | Comprehensive Analysis Competence and Innovative Approaches for Sustainable Chemical Production | |
| Arakelyan et al. | Analysis of the DCS historical data for estimation of input signal significance | |
| Bagchi et al. | Data analytics and stochastic modeling in a semiconductor fab | |
| Mendia et al. | An Intelligent Procedure for the Methodology of Energy Consumption in Industrial Environments | |
| CN118089287B (en) | Water chiller energy efficiency optimizing system based on intelligent algorithm | |
| Saraswathi et al. | Comparative analysis of k-means and self organizing map clustering on boiler process data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110720 Termination date: 20131206 |