CN104517162A - On-line hardness forecasting method of continuous annealing product by means of integrated learning - Google Patents
On-line hardness forecasting method of continuous annealing product by means of integrated learning Download PDFInfo
- Publication number
- CN104517162A CN104517162A CN201410843307.8A CN201410843307A CN104517162A CN 104517162 A CN104517162 A CN 104517162A CN 201410843307 A CN201410843307 A CN 201410843307A CN 104517162 A CN104517162 A CN 104517162A
- Authority
- CN
- China
- Prior art keywords
- sample
- continuous annealing
- product hardness
- temper
- hardness
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Control Of Heat Treatment Processes (AREA)
- Manufacturing & Machinery (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Factory Administration (AREA)
Abstract
The invention discloses an on-line hardness forecasting method of a continuous annealing product by means of integrated learning, and belongs to the technical field of automatic control of continuous annealing production process in iron and steel enterprises. An integrated learning modeling method using LSSVM as a sub-learning machine is used by utilizing historical continuous annealing production data samples of enterprises, and hardness forecasting models of off-line products are respectively established for strip steel with different tempers; in practical production, the continuous annealing production process data is read in real time, and the hardness of the current strip steel product is forecasted in real time through the hardness forecasting models of off-line products established through the integrated learning; a test on the practical production data proves that the method can obviously improve the accuracy and the robustness of hardness forecasting results of the continuous annealing product, so that site operation personnel can master the quality of the current strip steel product in real time and can adjust timely according to situations; therefore, the deficiency of off-line detection large lag is made up, and the product quality, the production operation level and the economic benefit of the continuous annealing production line are improved.
Description
Technical field
The invention belongs to the automatic control technology field of iron and steel enterprise's continuous annealing production run, particularly a kind of continuous annealing product hardness Online integration study forecasting procedure.
Background technology
In the actual production process of iron and steel enterprise's cold rolling mill continuous annealing unit, the hardness of band steel weighs the core index of product quality and Instructing manufacture.In actual production process, the hardness of band steel also cannot realize on-line checkingi, and scene is head, portion by intercepting annealing rear band steel, then carries out test experiment analysis to measure the hardness being with steel, thus judges product quality situation.But, the quality information being obtained product by test experiment analysis generally all has regular hour hysteresis quality, that is only have after a period of time produced by band steel, the quality information that it is concrete could be obtained, and speed of production with steel in actual production quickly, in minutes just can complete annealing in process, thus the very large situation of band steel hardness fluctuations often occurs, cause producing the quality problems such as hardness even waste product not up to standard, have a strong impact on the economic benefit of cold rolling mill.
Paper " forecast of continuous annealing unit strip quality and process monitoring system design and implimentation [D] based on PLS " (Wang Yuan, Northeastern University, 2009) although propose a kind of based on offset minimum binary (Partial Least Squares for belt steel product hardness, PLS) data driven type modeling method, but the method proposed in this section of paper can not meet the needs of actual production process, main cause has: the information relevant to band steel hardness that (1) document is considered is less, only there are about 20, procedural information relevant to band steel hardness in actual production process then will reach 51, (2) the PLS method that proposes of the document is main or for the monitoring and fault diagnosis of continuous annealing production run, and PLS method belongs to a kind of linear regression method, the production run of reality is then nonlinear, thus causes the precision of PLS method not high, (3) to there is input item numerous and intercouple and the narrower problem of Output rusults scope for sample data, cause existing between sample that input item difference is comparatively large and Output rusults is identical or close, make traditional Data Modeling Method be difficult to obtain higher precision of prediction and robustness.
Summary of the invention
For the deficiency that prior art exists, the invention provides a kind of continuous annealing product hardness Online integration study forecasting procedure.
Technical scheme of the present invention is:
A kind of continuous annealing product hardness Online integration study forecasting procedure, comprises the steps:
Step 1: for each temper band steel, sample is produced in the continuous annealing of reading from database in the nearest m time period, and sample set is produced in the continuous annealing obtained containing B sample;
Step 2: sample set is produced in the continuous annealing for each temper band steel, carries out data processing;
Step 2.1 is normalized each data item in sample set respectively, to eliminate the impact that between different pieces of information item, dimension is different;
Step 2.2: adopt the gross error detection method based on cluster for the sample set after normalized, in Rejection of samples set, A comprises the sample of human error;
Step 2.2.1: calculate the minimum Eustachian distance d between each sample and other B-1 sample respectively
i(i=1,2 ..., B);
Step 2.2.2: the mean value d calculating the minimum Eustachian distance of all samples
avg, by d
avgq doubly as the first threshold;
Step 2.2.3: by the minimum Eustachian distance d in sample set and between other B-1 sample
ithe sample being greater than the first threshold value is rejected, and remaining sample is carried out cluster simultaneously, obtains P cluster;
Step 2.2.4: calculate the minimum Eustachian distance D between each cluster and other P-1 cluster respectively
u(u=1,2 ..., P);
Step 2.2.5: the mean value D calculating the minimum Eustachian distance of all clusters
avg, and by D
avgr doubly as the second threshold;
Step 2.2.6: the minimum Eustachian distance between other P-1 cluster is greater than the second threshold value and the cluster that in cluster, sample size is less than t and the sample that comprises thereof are rejected;
Step 3: according to the sample set after step 2 data processing, asks for pivot transition matrix M and adopts for often kind of temper band steel and set up its product hardness forecasting model based on the integrated study modeling method of LSSVM and form off-line continuously annealing steel strip product hardness forecasting model storehouse;
Step 3.1: utilize PCA method, carry out dimensionality reduction to the matrix that production process data all in the sample set after step 2 data processing are formed, keeping characteristics root is more than or equal to the pivot of 1, and the pivot of reservation is stored in pivot transition matrix M;
Step 3.2: according to the sample set after step 2 data processing, adopts the integrated study modeling method based on LSSVM, sets up its product hardness forecasting model for often kind of temper band steel, forms off-line continuously annealing steel strip product hardness forecasting model storehouse; The described method setting up its product hardness forecasting model for often kind of temper band steel, comprises the steps:
Step 3.2.1: for sample standard deviations all in the sample set after step 2 data processing give identical weighted value, i.e. the weight w of each sample
i=1/ (B-A);
Step 3.2.2: based on current sample weights, uses the sub-learning machine model of a LSSVM method establishment belt steel product hardness forecast, calculates the root-mean-square error of this model and the training error of each sample;
Step 3.2.3: according to the training error adjustment sample weights of each sample, method is: the weight of the sample that weight increases and training error is less of the sample first making training error larger reduces, obtain new sample weights, then new sample weights is normalized;
Step 3.2.4: repeated execution of steps 3.2.2 is to step 3.2.3, set up the sub-learning machine model of N number of continuously annealing steel strip product hardness forecast, and calculate the training error of the root-mean-square error of each sub-learning machine model and each sample corresponding to each sub-learning machine model;
Step 3.2.5: according to the root-mean-square error of each sub-learning machine model that step 3.2.4 calculates, calculate the weight of each sub-learning machine model;
Step 3.2.6: according to N number of sub-learning machine model and respective weight thereof, be an integrated study machine model by N number of sub-learning machine model integrated, this integrated study machine model is this kind of temper belt steel product hardness forecasting model, and it exports the weighting composite value of the belt steel product hardness predicted value for each sub-learning machine model output;
Step 3.2.7: by this kind of temper belt steel product hardness forecasting model stored in off-line continuously annealing steel strip product hardness forecasting model storehouse;
Step 4: according to the band steel temper treating that continuous annealing is produced, selects corresponding temper belt steel product hardness forecasting model from continuously annealing steel strip product hardness forecasting model storehouse, starts continuous annealing and produces;
Step 5: with the maximum sampling period T of continuous annealing production line up-sampling point
maxz doubly as the sampling period of belt steel product hardness forecasting model, Real-time Obtaining continuous annealing production process data, as the process input vector of this kind of temper belt steel product hardness forecasting model, record is carried out to the associated production status data that sampled point each on continuous annealing production line produces when every coiled strip steel product head passes simultaneously; Described production process data is each sampled point the associated production status data of generation and synthesis of corresponding band steel information when every coiled strip steel product head passes on continuous annealing production line; Described band steel information comprises himself specification information and hot rolling information (band steel is in the production information of hot-rolled process);
Step 6: adopt clustering method to carry out gross error detection to process input vector, if this process input vector comprises human error, then delete this process input vector, abandon the belt steel product hardness forecast in this cycle, and wait for the process input vector in next sampling period; Otherwise, then step 7 is entered into;
Step 7: utilize the pivot transition matrix M obtained in step 3, dimensionality reduction is carried out to current process input vector, obtain the process input vector represented with pivot;
Step 8: based on the process input vector after dimensionality reduction, uses this kind of temper belt steel product hardness forecasting model to forecast belt steel product hardness;
Step 9: after the continuous annealing of this coiled strip steel has been produced, intercepts band steel taping head part and carries out offline inspection, obtain this coiled strip steel product hardness actual value, line item of going forward side by side;
Step 10: product hardness actual value corresponding with it for the continuous annealing production process data of recorded every coiled strip steel carried out mating and synthesize, forms a new continuous annealing production sample; Each continuous annealing is produced sample and is comprised H data item; A described H data item comprises several relevant continuous annealing production status data item, several band steel information data item and belt steel product hardness actual values;
Step 11: upgrade continuous annealing and produce sample, method is: for a certain temper band steel, after the continuous annealing production sample collection in its nearest O time period completes, then sample being produced in the continuous annealing in this nearest O time period joins in enterprise database, sample is produced in the continuous annealing in the O time period the earliest in this database simultaneously and deletes;
Step 12: according to step 1 to the method for step 3, produce sample according to the continuous annealing after upgrading, respectively re-training and renewal are carried out to each temper belt steel product hardness forecasting model;
Step 13: when beginning is produced in new band steel continuous annealing, repeated execution of steps 4 to step 13, realizes the online forecasting of known temper belt steel product hardness.
Beneficial effect of the present invention: continuous annealing product hardness Online integration study forecasting procedure of the present invention uses LSSVM as sub-learning machine maker, and use integrated study framework by the result integration of each sub-learning machine, as final belt steel product hardness predicted value.Through the inspection of actual production data, method of the present invention can significantly improve precision and the robustness of continuous annealing product hardness forecast result, make site operation personnel can grasp the quality of current belt steel product in real time, and according to circumstances adjust in good time, compensate for the deficiency of offline inspection large time delay, thus help continuous annealing production line improves the quality of products, improves production operation level, increases economic benefit.
Accompanying drawing explanation
Fig. 1 is the continuous annealing product hardness Online integration study forecasting procedure principle schematic of one embodiment of the present invention;
Fig. 2 is the continuous annealing product hardness Online integration study forecasting procedure process flow diagram of one embodiment of the present invention;
Fig. 3 is the forecast surface chart of the continuous annealing product hardness Online integration study forecasting procedure of one embodiment of the present invention;
Fig. 4 is the continuous annealing product hardness integrated study predicted value of one embodiment of the present invention and the comparative result figure of corresponding test sample book;
Fig. 5 is the continuous annealing product hardness on-line monitoring surface chart of one embodiment of the present invention;
Embodiment
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
Continuous annealing production is one important procedure of iron and steel enterprise's cold rolling mill, as shown in Figure 1, continuous annealing production line can be divided into following 9 stages according to function: heating furnace (HF), soaking pit (SF), leer (SCF), 1# black furnace (1C), 1# overaging stove (1OA), 2# overaging stove (2OA), 2# cool furnace (2C), water quenching oven (WQ), planisher.In process of production, cold-strip steel is walked in each stove of production line with certain speed, make it complete the heat treatment process such as heating, cooling according to the annealing process route of setting, thus cancellation band steel is because of cold rolling caused internal stress, again after smooth, obtain high-quality band steel.
For in actual production process, the hardness of band steel also cannot realize the present situation of on-line checkingi, the invention provides a kind of continuous annealing product hardness Online integration study forecasting procedure.The principle of work of the method is as shown in Figure 1: first, utilize the history continuous annealing production data sample of enterprise, use with LSSVM (Least Square Support Vector Machine, least square method supporting vector machine) as the integrated study modeling method of sub-learning machine, set up its off-line product hardness forecasting model respectively for different temper band steel, and form off-line continuously annealing steel strip product hardness forecasting model storehouse; Secondly, in actual production process, according to treating that the band steel temper that continuous annealing is produced selects corresponding product hardness forecasting model from the continuously annealing steel strip product hardness forecasting model storehouse that off-line is set up; Then, obtain from the sampling of continuous annealing production run and control system each sampled point on continuous annealing production line every coiled strip steel product head through time the associated production status data that produces go forward side by side line number Data preprocess and be with steel information to carry out synthesis to obtain continuous annealing Real-time Production Process data; Then, clustering method is used to carry out gross error detection to continuous annealing Real-time Production Process data; For the continuous annealing Real-time Production Process data not comprising human error, through PCA (Principle Component Analysis, pivot analysis) after method dimension-reduction treatment, pass to selected product hardness forecasting model, thus the belt steel product hardness predicted value under exporting current working to site operation personnel, and then be convenient to site operation personnel and grasp product quality in time and carry out corresponding operating conditions adjustment according to current belt steel product hardness predicted value.In addition, the production process data at every coiled strip steel place of taking the lead can store, when belt steel product hardness offline inspection result out after, after its synthesis, form new continuous annealing production data sample, be stored in enterprise database.After new samples collection when nearest one month completes, the re-training process of belt steel product hardness forecasting model can be started, thus ensure that forecasting model can follow the tracks of the up-to-date operating mode of continuous annealing unit.
Present embodiment is based on the continuous annealing production technology shown in Fig. 1, producing temper is the belt steel product of T5, and according to continuous annealing product hardness Online integration study forecasting procedure principle, by continuous annealing product hardness Online integration proposed by the invention study method and application in present embodiment, as shown in Figure 2, comprise the steps:
S1: the band steel for temper being T5, sample is produced in the continuous annealing gathering nearest 3 months from enterprise database, amounts to acquisition 350 samples, wherein each sample is made up of following 52 data item, comprising 14 band steel information data items, be respectively 9 strip steel specification information data items and inlet thickness, throat width, carbon content, sulfur content, phosphorus content, Fe content, nitrogen content, silicone content and total aluminium content, and 5 hot rolling information data items and tapping temperature, Average curl temperature, average final rolling temperature, average finishing temperature and continuous acid-washing-rolling extensibility (CDCM extensibility), 37 continuous annealing associated production status data items, are respectively central sections speed, HF stove 1 district furnace temperature, HF stove 2 district furnace temperature, HF stove 3 district furnace temperature, HF stove 4 district furnace temperature, HF stove 5 district furnace temperature, HF outlet of still belt steel temperature, SF stove 1 district furnace temperature, SF stove 2 district furnace temperature, SF outlet of still belt steel temperature, SCF stove 1 district furnace temperature, SCF stove 2 district furnace temperature, SCF outlet of still belt steel temperature, 1C stove 1 district furnace temperature, 1C stove 2 district furnace temperature, 1C stove 3 district furnace temperature, 1C stove refrigerating gas temperature, 1C outlet of still belt steel temperature, 1OA stove 1 district furnace temperature, 1OA stove 2 district furnace temperature, 1OA outlet of still belt steel temperature, 2OA stove 1-1 district furnace temperature, 2OA stove 1-2 district furnace temperature, 2OA stove 2-1 district furnace temperature, 2OA stove 2-2 district furnace temperature, 2OA outlet of still belt steel temperature, 2C stove furnace temperature, 2C refrigerating gas temperature, 2C outlet of still belt steel temperature, WQ stove 1 district water temperature, WQ stove 2 district water temperature, planisher extensibility, planisher strip steel at entry tension force, strip tension in the middle of planisher, planisher outlet strip tension, planisher 1# frame roll-force and 2# frame roll-force, 1 belt steel product hardness actual value data item.It is as shown in table 1 that sample set is produced in the continuous annealing gathered, wherein every a line represents a sample, each is classified as a data item, and arranges to the order of last belt steel product hardness actual value data item according to from strip steel specification information data item, hot rolling information data item, continuous annealing associated production status data item successively.With S
irepresent sample, S
i=(s
i1, s
i2..., s
ik..., s
i52), wherein i=1,2 ..., 350, s
ikrepresent a kth data item of i-th sample.
In nearest 3 months of table 1., temper is that sample set is produced in the continuous annealing of the belt steel product of T5
S2: the continuous annealing shown in his-and-hers watches 1 is produced sample set and carried out data processing;
S2.1: each the column data item respectively in his-and-hers watches 1 is normalized, to eliminate the impact that between different pieces of information item, dimension is different;
S2.2: adopt the gross error detection method based on cluster for the sample set after normalized, rejects the sample comprising human error, comprises the following steps:
S2.2.1: calculate each sample S respectively
iand the minimum Eustachian distance between other 349 samples
Wherein i=1,2 ..., 350;
S2.2.2: the mean value calculating the minimum Eustachian distance of all samples
by d
avg3 times, i.e. 3d
avgas the first threshold value (3d
avgthreshold value for obtaining through experiment);
S2.2.3: will be d with the minimum Eustachian distance of other 349 samples in sample set
i>3d
avgsample reject, (be d with the minimum Eustachian distance of other 349 samples by remaining sample simultaneously
i≤ 3d
avgsample) carry out cluster according to nearby principle, obtain 36 clusters;
S2.2.4: calculate each cluster C respectively
u(u=1,2 ..., 36) and other 35 clusters between minimum Eustachian distance
Wherein
Represent cluster C
ucenter (c
u1..., c
u52) in the value of a kth data item, the i.e. arithmetic mean of a kth data item in all samples of comprising of this cluster, wherein | C
u| represent cluster C
uthe quantity of middle sample;
S2.2.5: the mean value D calculating the minimum Eustachian distance of 36 clusters
avg, and by D
avg3 times, i.e. 3D
avgas the second threshold;
S2.2.6: the minimum Eustachian distance between other 35 clusters is greater than the second threshold value and the cluster that in cluster, sample size is less than 10 and the sample that comprises thereof are rejected; Wherein 3D
avgwith the threshold value that the sample size 10 in cluster is through experiment acquisition;
Continuous annealing shown in table 1 produces sample set after the data processing of S2.1 and S2.2, has 5 samples comprising human error disallowable, as shown in table 2.
5 samples of human error that what table 2. was disallowable comprise
S3: use PCA method to carry out dimensionality reduction to front 14 band steel information data items of 345 samples remaining in sample set and 37 the associated production status data items matrix that totally 51 data item are formed, keeping characteristics root is more than or equal to 11 pivots of 1, is stored in pivot transition matrix M; Simultaneously, also according to 345 samples remaining in sample set, adopt the integrated study modeling method based on LSSVM, set up its product hardness forecasting model for this kind of temper band steel, and this kind of set up temper belt steel product hardness forecasting model is stored in off-line continuously annealing steel strip product hardness forecasting model storehouse; The described method setting up its product hardness forecasting model for this kind of temper band steel, comprises the steps:
S3.1: be that 345 sample standard deviations give identical weighted value, i.e. the weight w of each sample
i=1/345, and learning machine index l=1 is set;
S3.2: based on current sample weights, use LSSVM method is trained and obtains the sub-learning machine model of a belt steel product hardness forecast; Method is: in LSSVM, use Radial basis kernel function, namely
wherein x
irepresent front 51 input data item of i-th training sample, σ
2for kernel functional parameter, concrete step is as follows:
S3.2.1: in span, the LSSVM parameter combinations that random generation 20 is initial.Remember that each parameter combinations is P
j=(γ
j, σ
2 j), wherein γ
jwith σ
2 jrepresent parameter combinations P respectively
jin to the penalty coefficient of error and kernel functional parameter.γ
jwith σ
2 jspan be respectively [5,50] and [0.1,2.0];
S3.2.2: for each parameter combinations P
j=(γ
j, σ
2 j), solve a system of linear equations as follows:
Wherein n=345 is total sample number amount, y
i(i=1 ..., 345) and represent the output of i-th sample, be namely with steel hardness.Thus obtain the model parameter α of the LSSVM corresponding to this parameter combinations
i(i=1 ..., 345) and b, then use the LSSVM model set up to predict its hardness, i.e. Hardness Prediction value according to front 51 input data item of each sample data
and then calculate this parameter combinations P
jevaluation index, the root-mean-square error namely predicted the outcome
S3.2.3: using LSSVM model minimum for current RMSE value as best LSSVM model, be designated as LSSVM
best.
S3.2.4: based on current 20 parameter combinations, obtains 20 new parameter combinations: first in accordance with the following methods, Stochastic choice two parameter combinations P
f=(γ
f, σ
2 f) and P
g=(γ
g, σ
2 g); Secondly, a new parameter combinations P is produced
j=(γ
j, σ
2 j), γ
jat scope [γ
min– 0.5I
1, γ
max+ 0.5I
1] interior random generation, wherein γ
min=min{ γ
f, γ
g, γ
max=max{ γ
f, γ
g, I
1=γ
max– γ
min; Similarly, σ
2 jat scope [σ
2 min– 0.5I
2, σ
2 max+ 0.5I
2] interior random generation, wherein σ
2 min=min{ σ
2 f, σ
2 g, σ
2 max=max{ σ
2 f, σ
2 g, I
2=σ
2 max– σ
2 min.
S3.2.5: for new 20 parameter combinations produced, the method using S3.2.2 to propose calculates its evaluation index.From current 40 parameter combinations, get front 20 parameter combinations that evaluation index is less, if obtain better LSSVM model, then upgrade LSSVM
best, delete remaining 20 second-rate parameter combinations.
S3.2.6: repeat S3.2.4 and S3.2.5, amounts to 500 times (repeating to be for 500 times to obtain the good LSSVM model of quality within given working time), by obtained LSSVM
beststore, as l sub-learning machine, and calculate each sample x corresponding to this model
itraining error
S3.3: according to each sample x
itraining error ζ
iadjustment sample weights, method is: first, arranges ε=0, for all samples, if ζ
i>0.01, then ε=ε+w
i; Secondly, β is set
l=ε × ε; Again, if sample x
itraining error ζ
i>0.01, then the weight adjusting this sample is w
i=w
i× β
l, otherwise its weight adjusting is w
i=1; Thus the weight of the sample making training error larger increases, and the less sample weights of training error diminishes; Finally, to all weight w
ibe normalized.
S3.4: reuse the method that above-mentioned S3.2 and S3.3 proposes, sets up the sub-learning machine model of 5 belt steel product hardness forecasts, the best parameter group of each sub-learning machine model and corresponding β thereof
lvalue, average training error are as shown in table 3.
The correlation parameter of table 3. belt steel product hardness pre-man who brings news of appointment's learning machine model
| Numbering l | γ | σ 2 | β l | Average relative error (%) | RMSE |
| 1 | 23.408 | 0.806 | 0.0125 | 0.50059 | 4.12081 |
| 2 | 20.9353 | 0.845 | 0.778 | 0.539676 | 4.34022 |
| 3 | 49.6097 | 0.731 | 0.570 | 0.533521 | 4.30655 |
| 4 | 25.0794 | 0.702 | 0.563 | 0.604213 | 4.95296 |
| 5 | 21.8188 | 0.712 | 0.181 | 0.480041 | 3.86035 |
S3.5: based on set up 5 sub-learning machine models, the belt steel product Hardness Prediction result according to following formula determination integrated study machine model:
Wherein
front 51 input items for sample x, the predicted value calculated by l belt steel product hardness forecasting model.The training result of the T5 temper belt steel product hardness forecasting model after 5 sub-learning machine model integrated as shown in Figure 3.
Except 345 training samples of used nearest 3 months, from enterprise database, acquisition time 41 belt steel product hardness actual values of 1 month are more early as the test sample book of T5 temper belt steel product hardness forecasting model, the validity of the continuous annealing product hardness forecasting model based on integrated study proposed with inspection institute.The belt steel product Hardness Prediction result obtained by T5 temper belt steel product hardness forecasting model and the comparative result of these 41 belt steel product hardness actual values are as shown in table 4 and Fig. 4, can find out, method proposed by the invention is in the estimated performance of continuous annealing product hardness, the consensus forecast deviation of the test sample book relatively do not learnt is 0.585%, and the variation tendency of belt steel product actual hardness can be followed, there is good generalization ability and robustness.
Table 4. test sample book contrasts with the predicting the outcome of continuous annealing product hardness forecasting model based on integrated study
S3.6: the integrated study model this temper belt steel product hardness forecast is stored in off-line continuously annealing steel strip product hardness forecasting model storehouse;
S4: when actual continuous annealing production temper is the band steel of T5, selects above-mentioned T5 temper belt steel product hardness forecasting model from continuously annealing steel strip product hardness forecasting model storehouse, start continuous annealing and produce;
S5: be 1 second with the maximum sampling period of continuous annealing production line up-sampling point, get its 2 times sampling periods as T5 temper belt steel product hardness forecasting model, the associated production status data reading each sampled point produce when every coiled strip steel product head passes every 2 seconds from the sampling of continuous annealing production run and control system, and itself and band steel information are synthesized continuous annealing production process data, namely continuous annealing produces first 51 of sample, be called process input vector, simultaneously, record is carried out to the associated production status data that sampled point each on continuous annealing production line produces when every coiled strip steel product head passes,
S6: use clustering method to carry out gross error detection to process input vector, method is: the minimum Eustachian distance between first 51 that first in computation process input vector and current database, sample is produced in each continuous annealing of T5 temper band steel, calculate minimum Eustachian distance mean value again, if this mean value is greater than the first threshold value 3d calculated in S2.2.2
avg, then think that it is normal sample, then enter into next step S7; Otherwise think that this sample packages is containing human error, deletes this process input vector, abandon the hardness forecast in this cycle, and wait for the process input vector in next sampling period;
S7: utilize the pivot transition matrix M obtained in step 3, dimensionality reduction is carried out to current process input vector, obtain the process input vector represented with pivot;
S8: based on the process input vector after dimensionality reduction, use the band steel hardness number that the forecast of T5 temper belt steel product hardness forecasting model is corresponding, and shown by operating platform display device, as shown in Figure 5, this upper left corner, interface can show the band steel hardness predicted value corresponding to current production status in real time, facilitate site operation personnel to be monitored production run by this predicted value, thus ensure that product quality is in acceptability limit.
S9: after current band steel has been produced, intercept band steel taping head part and carry out offline inspection, obtaining this coiled strip steel product hardness actual value is 64, line item of going forward side by side;
S10: the continuous annealing production process data of this recorded coiled strip steel and this coiled strip steel product hardness actual value 64 are carried out mating and synthesize, forms a new continuous annealing production sample, line item of going forward side by side;
S11: upgrade the continuous annealing of T5 temper band steel and produce sample, method is: for T5 temper band steel, if after its continuous annealing production sample collection of nearest one month completes, then first sample being produced in the continuous annealing of this month joins in enterprise database, then sample standard deviation is produced in all continuous annealings of time that month the earliest in this database to delete
S12: according to the method for S2 to S4, sample is produced according to the T5 temper band steel continuous annealing after upgrading, re-training and renewal are carried out to T5 temper belt steel product hardness forecasting model, thus makes T5 temper belt steel product hardness forecasting model can follow current continuous annealing actual production operating mode to adjust accordingly.
S13: when beginning is produced in new T5 temper band steel continuous annealing, repeat S4 to S13, realize the online forecasting to T5 temper belt steel product hardness.
Present embodiment realizes the continuous annealing product hardness Online integration study forecasting procedure of present embodiment by continuous annealing product hardness Online integration study forecast system, and this system comprises with lower module: database maintenance module, continuous annealing production run MBM, data preprocessing module and product hardness online forecasting module.Database maintenance module has been mainly used to the operation such as deletion and new samples data importing to enterprise's historical data; Data preprocessing module has been mainly used to the operation such as gross error detection and PCA dimensionality reduction for production run sample.Continuous annealing production run MBM in order to complete the integrated study modeling function based on LSSVM, and shows the training and testing result of the belt steel product hardness forecasting model set up.Product hardness online forecasting module is mainly used to realize continuously annealing steel strip product hardness online forecasting function.The continuous annealing product hardness Online integration study forecast system of present embodiment, sample with the continuous annealing production run shown in Fig. 1 to match with control system to use, can attach it to the sampling of continuous annealing production run when using at the scene with the computing machine at control system place, this computing machine is as the hardware platform of the inventive method.The inventive method is by the continuous annealing production run of installing from this computing machine sampling and the real-time associated production status data reading each sampled point of continuous annealing production run in control system, after carrying out pre-service, form continuous annealing production process data, re-use integrated study model and real-time prediction is carried out to current belt steel product hardness, and result is shown on this computer screen, facilitate site operation personnel to grasp the quality information of band steel in real time, and take the regulating measures of being correlated with when quality fluctuation.
Claims (5)
1. a continuous annealing product hardness Online integration study forecasting procedure, is characterized in that: comprise the steps:
Step 1: for each temper band steel, sample is produced in the continuous annealing of reading from database in the nearest m time period, and sample set is produced in the continuous annealing obtained containing B sample;
Step 2: sample set is produced in the continuous annealing for each temper band steel, carries out data processing;
Step 3: according to the sample set after step 2 data processing, asks for pivot transition matrix M and adopts for often kind of temper band steel and set up its product hardness forecasting model based on the integrated study modeling method of LSSVM and form off-line continuously annealing steel strip product hardness forecasting model storehouse;
Step 4: according to the band steel temper treating that continuous annealing is produced, selects corresponding temper belt steel product hardness forecasting model from continuously annealing steel strip product hardness forecasting model storehouse, starts continuous annealing and produces;
Step 5: with the maximum sampling period T of continuous annealing production line up-sampling point
maxz doubly as the sampling period of belt steel product hardness forecasting model, Real-time Obtaining continuous annealing production process data, as the process input vector of this kind of temper belt steel product hardness forecasting model, record is carried out to the associated production status data that sampled point each on continuous annealing production line produces when every coiled strip steel product head passes simultaneously; Described production process data is each sampled point the associated production status data of generation and synthesis of corresponding band steel information when every coiled strip steel product head passes on continuous annealing production line; Described band steel information comprises himself specification information and hot rolling information;
Step 6: adopt clustering method to carry out gross error detection to process input vector, if this process input vector comprises human error, then delete this process input vector, abandon the belt steel product hardness forecast in this cycle, and wait for the process input vector in next sampling period; Otherwise, then step 7 is entered into;
Step 7: utilize the pivot transition matrix M obtained in step 3, dimensionality reduction is carried out to current process input vector, obtain the process input vector represented with pivot;
Step 8: based on the process input vector after dimensionality reduction, uses this kind of temper belt steel product hardness forecasting model to forecast belt steel product hardness;
Step 9: after the continuous annealing of this coiled strip steel has been produced, intercepts band steel taping head part and carries out offline inspection, obtain this coiled strip steel product hardness actual value, line item of going forward side by side;
Step 10: product hardness actual value corresponding with it for the continuous annealing production process data of recorded every coiled strip steel carried out mating and synthesize, forms a new continuous annealing production sample; Each continuous annealing is produced sample and is comprised H data item; A described H data item comprises several relevant continuous annealing production status data item, several band steel information data item and belt steel product hardness actual values;
Step 11: upgrade continuous annealing and produce sample;
Step 12: according to step 1 to the method for step 3, produce sample according to the continuous annealing after upgrading, respectively each temper belt steel product hardness forecasting model is upgraded;
Step 13: when beginning is produced in new band steel continuous annealing, repeated execution of steps 4 to step 13, realizes the online forecasting of known temper belt steel product hardness.
2. continuous annealing product hardness Online integration study forecasting procedure according to claim 1, is characterized in that: produce sample set for the continuous annealing of each temper band steel in described step 2 and carry out data processing, comprise the steps:
Step 2.1: be normalized each data item in sample set respectively, to eliminate the impact that between different pieces of information item, dimension is different;
Step 2.2: adopt the gross error detection method based on cluster for the sample set after normalized, in Rejection of samples set, A comprises the sample of human error;
Step 2.2.1: calculate the minimum Eustachian distance d between each sample and other B-1 sample respectively
i(i=1,2 ..., B);
Step 2.2.2: the mean value d calculating the minimum Eustachian distance of all samples
avg, by d
avgq doubly as the first threshold;
Step 2.2.3: by the minimum Eustachian distance d in sample set and between other B-1 sample
ithe sample being greater than the first threshold value is rejected, and remaining sample is carried out cluster simultaneously, obtains P cluster;
Step 2.2.4: calculate the minimum Eustachian distance D between each cluster and other P-1 cluster respectively
u(u=1,2 ..., P);
Step 2.2.5: the mean value D calculating the minimum Eustachian distance of all clusters
avg, and by D
avgr doubly as the second threshold;
Step 2.2.6: the minimum Eustachian distance between other P-1 cluster is greater than the second threshold value and the cluster that in cluster, sample size is less than t and the sample that comprises thereof are rejected.
3. continuous annealing product hardness Online integration study forecasting procedure according to claim 1, it is characterized in that: the method asking for pivot transition matrix M according to the sample set after step 2 data processing in described step 3 is: utilize PCA method, dimensionality reduction is carried out to the matrix that production process data all in the sample set after step 2 data processing are formed, keeping characteristics root is more than or equal to the pivot of 1, and the pivot of reservation is stored in pivot transition matrix M.
4. continuous annealing product hardness Online integration study forecasting procedure according to claim 1, it is characterized in that: adopt the integrated study modeling method based on LSSVM to set up the method for its product hardness forecasting model according to the sample set after step 2 data processing for often kind of temper band steel in described step 3, comprise the steps:
Step 3.2.1: for sample standard deviations all in the sample set after step 2 data processing give identical weighted value, i.e. the weight w of each sample
i=1/ (B-A);
Step 3.2.2: based on current sample weights, uses the sub-learning machine model of a LSSVM method establishment belt steel product hardness forecast, calculates the root-mean-square error of this model and the training error of each sample;
Step 3.2.3: according to the training error adjustment sample weights of each sample, method is: the weight of the sample that weight increases and training error is less of the sample first making training error larger reduces, obtain new sample weights, then new sample weights is normalized;
Step 3.2.4: repeated execution of steps 3.2.2 is to step 3.2.3, set up the sub-learning machine model of N number of continuously annealing steel strip product hardness forecast, and calculate the training error of the root-mean-square error of each sub-learning machine model and each sample corresponding to each sub-learning machine model;
Step 3.2.5: according to the root-mean-square error of each sub-learning machine model that step 3.2.4 calculates, calculate the weight of each sub-learning machine model;
Step 3.2.6: according to N number of sub-learning machine model and respective weight thereof, be an integrated study machine model by N number of sub-learning machine model integrated, this integrated study machine model is this kind of temper belt steel product hardness forecasting model, and it exports the weighting composite value of the belt steel product hardness predicted value for each sub-learning machine model output;
Step 3.2.7: by this kind of temper belt steel product hardness forecasting model stored in off-line continuously annealing steel strip product hardness forecasting model storehouse.
5. continuous annealing product hardness Online integration study forecasting procedure according to claim 1, it is characterized in that: the method upgrading continuous annealing production sample in described step 11 is: for a certain temper band steel, after the continuous annealing production sample collection in its nearest O time period completes, then sample being produced in the continuous annealing in this nearest O time period joins in enterprise database, sample is produced in the continuous annealing in the O time period the earliest in this database simultaneously and deletes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410843307.8A CN104517162B (en) | 2014-12-30 | 2014-12-30 | A kind of continuous annealing product hardness Online integration learns forecasting procedure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410843307.8A CN104517162B (en) | 2014-12-30 | 2014-12-30 | A kind of continuous annealing product hardness Online integration learns forecasting procedure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104517162A true CN104517162A (en) | 2015-04-15 |
| CN104517162B CN104517162B (en) | 2017-09-05 |
Family
ID=52792422
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410843307.8A Active CN104517162B (en) | 2014-12-30 | 2014-12-30 | A kind of continuous annealing product hardness Online integration learns forecasting procedure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104517162B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106345823A (en) * | 2016-08-31 | 2017-01-25 | 北京首钢自动化信息技术有限公司 | On-line real-time mechanical property prediction method based on hot rolled steel coil production processes |
| CN106845825A (en) * | 2017-01-18 | 2017-06-13 | 西安交通大学 | It is a kind of to be traced to the source and control method based on the cold rolling of strip steel quality problems for improving PCA |
| CN106886799A (en) * | 2017-03-17 | 2017-06-23 | 东北大学 | A kind of continuous annealing band steel quality online test method based on hybrid integrated study |
| CN107541597A (en) * | 2016-06-29 | 2018-01-05 | 宝山钢铁股份有限公司 | Strip running deviation monitoring and diagnostic method and the system of continuous annealing unit soaking pit |
| CN108034804A (en) * | 2017-12-08 | 2018-05-15 | 中国地质大学（武汉） | A kind of method and system of continuous annealing unit stove area energy consumption modeling |
| CN108229072A (en) * | 2018-02-10 | 2018-06-29 | 东北大学 | Continuous annealing soaking pit strip tension On-line Measuring Method based on data parsing |
| CN108281192A (en) * | 2017-12-29 | 2018-07-13 | 诺仪器（中国）有限公司 | Human body component prediction technique based on Ensemble Learning Algorithms and system |
| CN110321658A (en) * | 2019-07-10 | 2019-10-11 | 江苏金恒信息科技股份有限公司 | A kind of prediction technique and device of plate property |
| CN110989522A (en) * | 2019-12-06 | 2020-04-10 | 东北大学 | Multi-steel-coil-oriented optimal setting method for technological parameters in continuous annealing production process |
| CN111563686A (en) * | 2020-05-13 | 2020-08-21 | 中冶南方工程技术有限公司 | Cold-rolled silicon steel quality judgment method based on full-process data |
| TWI718928B (en) * | 2020-04-10 | 2021-02-11 | 中國鋼鐵股份有限公司 | Method and system for evaluating a width of a steel strip |
| CN112598026A (en) * | 2020-12-07 | 2021-04-02 | 东北大学 | Continuous annealing production process fault diagnosis method based on unbalanced data deep learning |
| CN114943364A (en) * | 2022-03-31 | 2022-08-26 | 苏州新凌电炉有限公司 | Intelligent visual management method and device for mesh belt furnace |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008099457A1 (en) * | 2007-02-09 | 2008-08-21 | Toshiba Mitsubishi-Electric Industrial Systems Corporation | Process line control unit and method of controlling the line |
| CN101899563A (en) * | 2009-06-01 | 2010-12-01 | 上海宝钢工业检测公司 | PCA (Principle Component Analysis) model based furnace temperature and tension monitoring and fault tracing method of continuous annealing unit |
| CN102033523A (en) * | 2009-09-25 | 2011-04-27 | 上海宝钢工业检测公司 | Strip steel quality forecasting, furnace condition early-warning and fault diagnosis method based on partial least square |
| CN102564644A (en) * | 2011-12-20 | 2012-07-11 | 东北大学 | Temperature online measuring method for plate blank in production process of heating furnace |
-
2014
- 2014-12-30 CN CN201410843307.8A patent/CN104517162B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008099457A1 (en) * | 2007-02-09 | 2008-08-21 | Toshiba Mitsubishi-Electric Industrial Systems Corporation | Process line control unit and method of controlling the line |
| CN101899563A (en) * | 2009-06-01 | 2010-12-01 | 上海宝钢工业检测公司 | PCA (Principle Component Analysis) model based furnace temperature and tension monitoring and fault tracing method of continuous annealing unit |
| CN102033523A (en) * | 2009-09-25 | 2011-04-27 | 上海宝钢工业检测公司 | Strip steel quality forecasting, furnace condition early-warning and fault diagnosis method based on partial least square |
| CN102564644A (en) * | 2011-12-20 | 2012-07-11 | 东北大学 | Temperature online measuring method for plate blank in production process of heating furnace |
Non-Patent Citations (2)
| Title |
|---|
| 彭俊 等: "基于平均轨迹的连续退火过程带钢硬度预测方法", 《控制工程》 * |
| 王超 等: "基于正交信号校正的连续退火机组带钢质量预测方法", 《仪表技术》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107541597A (en) * | 2016-06-29 | 2018-01-05 | 宝山钢铁股份有限公司 | Strip running deviation monitoring and diagnostic method and the system of continuous annealing unit soaking pit |
| CN107541597B (en) * | 2016-06-29 | 2019-03-05 | 宝山钢铁股份有限公司 | Strip running deviation monitoring and diagnostic method and the system of continuous annealing unit soaking pit |
| CN106345823B (en) * | 2016-08-31 | 2018-02-27 | 北京首钢自动化信息技术有限公司 | The method of online real-time estimate mechanical performance based on coils of hot-rolled steel production procedure |
| CN106345823A (en) * | 2016-08-31 | 2017-01-25 | 北京首钢自动化信息技术有限公司 | On-line real-time mechanical property prediction method based on hot rolled steel coil production processes |
| CN106845825B (en) * | 2017-01-18 | 2020-04-28 | 西安交通大学 | Strip steel cold rolling quality problem tracing and control method based on improved PCA |
| CN106845825A (en) * | 2017-01-18 | 2017-06-13 | 西安交通大学 | It is a kind of to be traced to the source and control method based on the cold rolling of strip steel quality problems for improving PCA |
| CN106886799A (en) * | 2017-03-17 | 2017-06-23 | 东北大学 | A kind of continuous annealing band steel quality online test method based on hybrid integrated study |
| CN106886799B (en) * | 2017-03-17 | 2019-08-02 | 东北大学 | A kind of continuous annealing band steel quality online test method based on hybrid integrated study |
| CN108034804A (en) * | 2017-12-08 | 2018-05-15 | 中国地质大学（武汉） | A kind of method and system of continuous annealing unit stove area energy consumption modeling |
| CN108034804B (en) * | 2017-12-08 | 2019-04-23 | 中国地质大学（武汉） | A kind of method and system of continuous annealing unit furnace area's energy consumption modeling |
| CN108281192A (en) * | 2017-12-29 | 2018-07-13 | 诺仪器（中国）有限公司 | Human body component prediction technique based on Ensemble Learning Algorithms and system |
| CN108281192B (en) * | 2017-12-29 | 2022-03-22 | 一诺仪器（中国）有限公司 | Human body component prediction method and system based on ensemble learning algorithm |
| CN108229072A (en) * | 2018-02-10 | 2018-06-29 | 东北大学 | Continuous annealing soaking pit strip tension On-line Measuring Method based on data parsing |
| WO2021004198A1 (en) * | 2019-07-10 | 2021-01-14 | 江苏金恒信息科技股份有限公司 | Plate performance prediction method and apparatus |
| CN110321658A (en) * | 2019-07-10 | 2019-10-11 | 江苏金恒信息科技股份有限公司 | A kind of prediction technique and device of plate property |
| GB2600213A (en) * | 2019-07-10 | 2022-04-27 | Jiangsu Jinheng Information Tech Co Ltd | Plate performance prediction method and apparatus |
| GB2600213B (en) * | 2019-07-10 | 2023-06-14 | Jiangsu Jinheng Information Tech Co Ltd | Method and apparatus for predicitng plate performance |
| CN110321658B (en) * | 2019-07-10 | 2023-09-01 | 江苏金恒信息科技股份有限公司 | Method and device for predicting plate performance |
| CN110989522A (en) * | 2019-12-06 | 2020-04-10 | 东北大学 | Multi-steel-coil-oriented optimal setting method for technological parameters in continuous annealing production process |
| CN110989522B (en) * | 2019-12-06 | 2022-09-09 | 东北大学 | Multi-steel-coil-oriented optimal setting method for technological parameters in continuous annealing production process |
| TWI718928B (en) * | 2020-04-10 | 2021-02-11 | 中國鋼鐵股份有限公司 | Method and system for evaluating a width of a steel strip |
| CN111563686A (en) * | 2020-05-13 | 2020-08-21 | 中冶南方工程技术有限公司 | Cold-rolled silicon steel quality judgment method based on full-process data |
| CN112598026A (en) * | 2020-12-07 | 2021-04-02 | 东北大学 | Continuous annealing production process fault diagnosis method based on unbalanced data deep learning |
| CN112598026B (en) * | 2020-12-07 | 2023-09-22 | 东北大学 | Continuous annealing production process fault diagnosis method based on unbalanced data deep learning |
| CN114943364A (en) * | 2022-03-31 | 2022-08-26 | 苏州新凌电炉有限公司 | Intelligent visual management method and device for mesh belt furnace |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104517162B (en) | 2017-09-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104517162A (en) | On-line hardness forecasting method of continuous annealing product by means of integrated learning | |
| CN107377634B (en) | A kind of hot-strip outlet Crown Prediction of Media method | |
| CN106345823B (en) | The method of online real-time estimate mechanical performance based on coils of hot-rolled steel production procedure | |
| WO2021189739A1 (en) | Production plan-based iron and steel plant oxygen load prediction method | |
| WO2023130666A1 (en) | Strip steel plate convexity prediction method based on data-driving and mechanism model fusion | |
| CN108311544B (en) | A kind of roll-force Parameter Self-learning method and device | |
| CN104463344B (en) | Power network short-term load forecasting method and system | |
| CN101391268B (en) | Reverse optimization method of steel plate rolling and cooling controlling-process temperature institution | |
| CN104915518A (en) | Establishing method and application of two-dimensional prediction model of silicon content in hot metal in blast furnace | |
| CN103123483B (en) | Rolling deformation resistance prediction system of new steel based on clustering analysis | |
| CN104657602A (en) | Method and system for forecasting thickness of band steel in hot continuous rolling production process | |
| CN102564644B (en) | Temperature online measuring method for plate blank in production process of heating furnace | |
| CN108655186A (en) | Roll-force presetting method based on artificial neural network and mathematical model | |
| CN104899425A (en) | Variable selection and forecast method of silicon content in molten iron of blast furnace | |
| CN101782769B (en) | Quick prediction method of average flowing-through time on basis of index compensation | |
| CN109657878A (en) | A kind of Air-conditioning Load Prediction method and device | |
| CN102069095A (en) | Statistical learning-based method for predicting and controlling finish rolling temperature in fine rolling | |
| CN108334988A (en) | A kind of short-term Load Forecasting based on SVM | |
| CN107633316A (en) | The power network short-term load forecasting method of adaptive local nonlinear regression deviation compensation | |
| CN103853939A (en) | Combined forecasting method for monthly load of power system based on social economic factor influence | |
| CN104238498A (en) | Method for predicting hot continuous rolling pipeline steel structure and mechanical property | |
| CN109492335A (en) | A kind of annealing furnace furnace temperature prediction technique and system | |
| CN109013717A (en) | A kind of hot continuous rolling centre base center portion temperature computation method | |
| CN106886799A (en) | A kind of continuous annealing band steel quality online test method based on hybrid integrated study | |
| CN111178605A (en) | Distribution network engineering project construction period prediction method based on feature selection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |