CN1815225A - Compound-fertilizer nutrient-content soft-measuring modeling method beased on hybrid multiple models - Google Patents

Abstract

Present invention discloses a modeling method for nutrient content soft measuring based on mix multiple mold type compound fertilizer. Modeling method features 1, whole pattern is mixed model consisting of based on data driven 3 submodel and based on simplify mechanical models 1 submodel, wherein the outputting of data driven submodel used as inputting of based on simplify mechanical models submodel, 2, simplifying mechanical models related model parameter determining method, determining potassium chloride and concentrated sulfuric acid extent of reaction parameter according to chloride ion content and material balance in finished product, 3, method determining (NH4)2HPO4 quality, according to neutralization degree value and measuring method, combining concerned chemical equation to calculate said quality. Said invention has advantages of Said invention has advantages of simple model building, strong explicable ability, high reliability, and fine extrapolate ability.

Description

Based on the composite fertilizer's nutrient content soft-measuring modeling method that mixes multi-model

Technical field

The present invention relates to a kind of based on the composite fertilizer's nutrient content soft-measuring modeling method that mixes multi-model.

Background technology

Compound (mixing) fertilizer is the inexorable trend of chemical fertilizer industrial development in modern age, and in the country of chemical fertilizer industry prosperity, its output has accounted for 70%～80% of total fertilizer application amount.The production of compound in the Tenth Five-year plan of China (mixing) fertilizer is as the direction of Fertilizer Industry structural adjustment.In recent years, the production development of China's compound (mixing) fertilizer is very fast, has successively built up the fertile process units of thousands of cover granulating methods compound (mixing).Sulfenyl nitrogen-phosphorus-potassium (S-NPK) composite fertilizer is fit to all kinds of cereal crops and industrial crops with it and especially avoids the welcome that the advantage of chlorine industrial crops is subjected to wide farmers deeply.In recent years, S-NPK composite fertilizer developed rapidly in China, and output rises significantly.

Domestic composite fertilizer production new technology-potassium chloride low temperature conversion and slurry-spraying pelletizing drying process are produced sulfenyl nitrogen-phosphorus-potassium compound fertilizer, promptly organically combine producing phosphoric acid, phosphorus ammonium and glazier's salt, Phosphoric Acid Concentration or slip enrichment facility have not only been cancelled, and reduced the difficulty that high temperature method is produced the glazier's salt process, can produce high-quality and efficient sulfenyl nitrogen-phosphorus-potassium compound fertilizer.

Fig. 1 has shown the typical process flow of potassium chloride low temperature conversion and slurry-spraying pelletizing method production S-NPK composite fertilizer, potassium chloride continuously adds in the 1# reactive tank according to a certain ratio with the concentrated sulphuric acid after metering and reacts, when reacting, material overflow to 2# reactive tank continuously continues reaction.Resultant of reaction potassium acid sulfate slip continuously overflow to hydrogen potassium nitration mixture groove and the phosphoric acid,diluted that contains phosphorus pentoxide 19%～23% of certain proportioning be prepared into nitration mixture, continuously be sent to multiple fertile nitration mixture groove.Nitration mixture and the ammonia that flows out A, B line from multiple fertile nitration mixture groove enters A respectively, B spool formula reactor carries out neutralization reaction; reacted slip enters flash tank and removes moisture; overflow is to the groove of whitewashing behind the tankful; then by spray pump will in and the pressurized air sent here of slip and compressed air station spray into granulating machine by granulating machine head spray gun with spray pattern; contact with the returning charge of falling with the form of expecting curtain; under the rotation of hot blast that heat generator is sent here and granulating machine, form compound fertilizer granules, send from the granulating machine tail at last.

This device is produced is nitrogen, phosphorus pentoxide, kali nutrient content be 13%, 17%, 15% sulfenyl nitrogen-phosphorus-potassium compound fertilizer of (allow respectively ± 1.5% error).Sulfenyl nitrogen-phosphorus-potassium compound fertilizer is a kind of technology content high product, if the content of nutrients of nitrogen, phosphorus pentoxide, kali is lower than These parameters in the complex fertilizer product, then the complex fertilizer product of Sheng Chaning is defective, the processing again of need doing over again; And if nutrients of nitrogen, phosphorus pentoxide, potassium oxide content are too high in the complex fertilizer product, then can increase cost.In order to guarantee that complex fertilizer product is qualified and energy-saving and cost-reducing, nutrient content must remain in the certain limit in the general industry requirement composite fertilizer finished product, so the control in real time of the nutrient content of complex fertilizer product is extremely important.

In actual production, the content of the nutrients of nitrogen in the complex fertilizer product, phosphorus pentoxide, kali all can't on-line measurement, but adopts the off-line lab analysis to obtain, but off-line analysis lags behind a few hours, and analytical sampling number of times few (inferior/8 hour) can not satisfy the requirement of control far away.There are two kinds of approach to solve this problem, first development in-line analyzer; It two is to use soft-measuring technique.Because the former is long the construction cycle, expense costliness, and easy care not, therefore best solution is to use second kind of approach, promptly set up the soft-sensing model of composite fertilizer's nutrient content, the value of each nutrient content of On-line Estimation composite fertilizer under the prerequisite that does not increase hardware investment.

At present, the report of the relevant composite fertilizer of Shang Weijian nutrient content soft-measuring modeling method.The method that present factory is adopted is that these nutrient contents samplings are manually chemically examined, and obtains by the off-line analysis mode, and the operative employee adjusts production run according to the laboratory assay value that these hysteresis reach a few hours.The shortcoming of this method is:

Artificial chemical examination lags behind big, reaches a few hours; Because chemical examination cost problem makes the sampling period all longer, therefore, these laboratory values are difficult to be directly used in quality control in addition.The operative employee mainly relies on experience separately to regulate, thereby makes the first-time qualification rate of product lower.The unacceptable product reprocessing of must doing over again has increased cost of products greatly.

Summary of the invention

The purpose of this invention is to provide a kind of based on the composite fertilizer's nutrient content soft-measuring modeling method that mixes multi-model.

Composite fertilizer's nutrient content soft-measuring modeling method based on the mixing multi-model: it mixes multi-model structure, promptly this model is by based on the A of data-driven, B line degree of neutralization soft-sensing model, based on the chloride ion content soft-sensing model of data-driven with import more/export more soft-sensing model based on composite fertilizer's nutrient content of simplifying mechanism model and constitute, and wherein more the output of A, B line degree of neutralization soft-sensing model and chloride ion content soft-sensing model imports/export more the input of soft-sensing model as composite fertilizer's nutrient content.

Described modeling method comprises the steps:

1), potassium chloride inventory that selection is relevant with the composite fertilizer nutrient content, concentrated sulphuric acid flow, hydrogen potassium 1# reactive tank temperature, hydrogen potassium 2# reactive tank temperature, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, sulfate radical density, A line nitration mixture flow, A line gas ammonia flow, B line nitration mixture flow, B line gas ammonia flow, gas ammonia pressure, A line degree of neutralization, B line degree of neutralization, chloride ion content totally 16 process variable, as the auxiliary variable of soft-sensing model;

2), the crucial auxiliary variable A line degree of neutralization and the B line degree of neutralization of composite fertilizer's nutrient content be the manual analysis value, rather than collection value in real time; At first use restricted memory partial least square algorithm to set up the dynamic soft sensor model of A, B line degree of neutralization; Wherein, the auxiliary variable of A line degree of neutralization soft-sensing model is: potassium chloride feeds intake, concentrated sulphuric acid flow, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, A line nitration mixture flow, A line gas ammonia flow, gas ammonia pressure.The auxiliary variable of B line degree of neutralization soft-sensing model is: potassium chloride inventory, concentrated sulphuric acid flow, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, B line nitration mixture flow, B line gas ammonia flow, gas ammonia pressure;

3), the key parameter chloride ion content simplified in the mechanism model is the manual analysis value, uses restricted memory partial least square algorithm to set up the dynamic soft sensor model of chloride ion content; The auxiliary variable of this model is potassium chloride inventory, concentrated sulphuric acid flow, hydrogen potassium 1# reactive tank temperature, hydrogen potassium 2# reactive tank temperature;

4), the output of A, B line degree of neutralization soft-sensing model and chloride ion content soft-sensing model is as the input of importing more/export more soft-sensing model based on composite fertilizer's nutrient content of simplifying mechanism model.

More described composite fertilizer's nutrient content based on the simplification mechanism model is imported/is exported more the soft-sensing model establishment step and be:

1) utilizes the result of chloride ion content soft-sensing model,, calculate the reaction rate of the potassium chloride and the concentrated sulphuric acid according to material balance.Produce one ton of composite fertilizer and need consume 235 kilograms in potassium chloride, suppose that chloride ion content is Cl_c (a chloride ion content soft-sensing model calculated value) in the composite fertilizer, then the computing formula of reaction rate is:

According to this reaction rate, in conjunction with chemical equation , can calculate KCl and H ₂SO ₄Reaction remaining KCl in back and H ₂SO ₄Amount and the KHSO of generation ₄Amount.

2) in the flow process of back, a neutralization reaction has only taken place, do not have other subsidiary reaction and exist.The chemical equation of neutralization reaction is:

According to known KHSO ₄, H ₂SO ₄, H ₃PO ₄Quality, can calculate corresponding product KNH in conjunction with first three equation ₄SO ₄, (NH ₄) ₂SO ₄, (NH ₄) H ₂PO ₄Quality and the NH of their combinations ₃Quality.

3) according to the result of calculation of A, B line degree of neutralization soft-sensing model and A, B line mixed flow flow, utilize weighted average calculation to go out the degree of neutralization of whole production line.The degree of neutralization of supposing the A line is count_a, and the degree of neutralization of B line is count_b, and A line nitration mixture flow is flow_a, and B line nitration mixture flow is flow_b, and then the degree of neutralization of whole production line is $\mathrm{count}\_a\×\frac{\mathrm{flow}\_a}{\mathrm{flow}\_a+\mathrm{flow}\_b}+\mathrm{count}\_b\×\frac{\mathrm{flow}\_b}{\mathrm{flow}\_a+\mathrm{flow}\_b}.$

Because degree of neutralization reflection just in and (NH in the slip ₄) ₂HPO ₄With (NH ₄) H ₂PO ₄Ratio, therefore, according to the assay method of degree of neutralization and known (NH ₄) H ₂PO ₄Quality, can calculate (NH in the last composite fertilizer ₄) ₂HPO ₄Quality, and calculate remaining (NH thus ₄) H ₂PO ₄Quality and reaction in the NH that consumes ₃Quality.(NH ₄) ₂HPO ₄The computing formula of quality is as follows:

$\frac{{\left({\mathrm{NH}}_4\right)}_2{\mathrm{HPO}}_4}{\left({\mathrm{NH}}_4\right)H_{2}P{O}_4}=\frac{0.2\×V_1}{0.1\×V_2-0.2\×V_1}=\frac{0.2\×\frac{V_1}{V_2}}{0.1-0.2\×\frac{V_1}{V_2}}$

V wherein ₁Sulfuric acid standard solution consumption when being titration, V ₂Standard solution of sodium hydroxide consumption when being titration,

Promptly

4) calculate the summation of last product,, can obtain the degree of nitrogen, phosphorus pentoxide, kali with the quality of the quality of the quality of the nitrogen that has calculated, phosphorus pentoxide and kali summation divided by last product.

Advantage of the present invention:

1) import more/export more soft-sensing model based on the composite fertilizer's nutrient content that mixes multi-model and combine based on simplifying modelling by mechanism and based on two kinds of modeling methods of data-driven modeling, the integrated advantage of these two kinds of modeling methods, model is simple, good reliability;

2) the data-driven modeling method in the model adopts the line modeling method based on restricted memory partial least square algorithm, has adaptation function, slow time-varying characteristics that can adaptive system;

3) use restricted memory partial least square algorithm, can overcome the saturability of sample;

4) to import more/export more soft-sensing model be dynamic model to composite fertilizer's nutrient content, the estimated accuracy height of model.

Description of drawings

Fig. 1 is the typical process flow that potassium chloride low temperature conversion and slurry-spraying pelletizing method generate sulfenyl nitrogen-phosphorus-potassium compound fertilizer;

Fig. 2 is based on composite fertilizer's nutrient content soft-sensing model structure of mixing multi-model;

Fig. 3 is nitrogen content soft-sensing model calculated value and laboratory values curve;

Fig. 4 is phosphorus pentoxide soft-sensing model calculated value and laboratory values curve;

Fig. 5 is kali soft-sensing model calculated value and laboratory values curve.

Embodiment

Terminological interpretation of the present invention:

Soft measurement: in the modern industry production run, the variable that has a big class reactor product quality, they are for technology or reason economically, can not directly measure with conventional sensors, but come out by in-line analyzer detection or lab analysis, analytical cycle is long, and big (just analyze once every a few hours such as composite fertilizer's nutrient content content, analyzing lags behind reaches several hours) lags behind.But the real-time detection of these variablees for the continuous quiet run that guarantees product quality and process units and the productive capacity etc. of giving full play to device all tool play a very important role.If use the laboratory values that lags behind can't correctly reflect real-time operating mode, the operative employee leaves the economic benefit that excessive leeway also can reduce process units to be on the safe side usually simultaneously.In order to address this problem, soft-measuring technique has been obtained very big progress in recent years.Soft-measuring technique just is to use the process variable and the corresponding product quality analysis value of easy survey, adopt certain modeling method (modelling by mechanism, regression modeling etc.) set up a quality estimation model, under the prerequisite that does not increase hardware investment, estimate product quality in real time by this model.

Restricted memory: in the line modeling process, As time goes on number of samples increases gradually, if all samples that at this moment will collect in the past all are used for modeling, can cause old sample information to flood new sample information and causes the precise decreasing of model.The restricted memory method is that new sample of every increase just removes an old sample, is used for the data length of modeling to remain constantly, takes into account the time variation of the system of considering and the saturability of sample.

The partial least square algorithm: partial least square algorithm (PLS) is a kind of with the multivariate regression method of higher dimensional space information projection to the low-dimensional information space of being made up of several implicit variablees, comprised the important information in the raw data in the implicit variable, and be mutually independently between implicit variable.The PLS method not only can be eliminated the collinearity influence between raw data, can also handle the data that contain noise, and when the independent variable number is bigger than observation frequency, still can obtain result preferably with the PLS homing method.Since the data of some the very important reflection product qualities in large complicated petrochemical industry and the chemical plant installations obtain very difficult and relevant detection equipment very expensive, usually the sampled value of these data is seldom, and the dimension of the conventional variable relevant with these qualitative datas is very big, usually have tens of even up to a hundred, so the PLS method is particularly suitable for the soft sensor modeling of these complex appts.

Mixture model: soft-measuring modeling method can be divided into pure modelling by mechanism method and usually based on the modeling method of data-driven.These two kinds of methods respectively have its relative merits: the former can be from reflecting the rule of process in essence, the reliability height, and extrapolation is good, has interpretation; Shortcoming is that modeling process is more loaded down with trivial details, and for some complex processes, getable mechanism model generally also is through the model after some simplification.The latter is directly according to the direct modeling of I/O data of process, almost need not the priori of process object, but shortcoming also is significantly, is example with the neural network, as a kind of modeling method of black box, pace of learning is slow, promote poor performance, and model do not have interpretation, be difficult to determine that appropriate network architecture and study stop index, cause the over-fitting phenomenon easily, even noise also match may be come in.Because pure modelling by mechanism and caused hybrid modeling thought based on the limitation of these two kinds of methods of data-driven modeling, promptly for there being the process of simplifying mechanism model, can will simplify mechanism model and combine based on the model of data-driven, complement one another, simplify the priori that mechanism model provides, can save training sample for model based on data-driven; Model based on data-driven can compensate the not modeling characteristic of simplifying mechanism model again simultaneously.

The present invention is further illustrated below in conjunction with accompanying drawing and by embodiment:

Adopt certain sulfuric acid plant's sulfenyl nitrogen-phosphorus-potassium compound fertilizer process units that the validity of concrete implementation step of the present invention and institute's established model is described.The typical process flow of potassium chloride low temperature conversion and slurry-spraying pelletizing method generation sulfenyl nitrogen-phosphorus-potassium compound fertilizer as shown in Figure 1.

The foundation of soft-sensing model is imported/exported more to composite fertilizer's nutrient content more

1, the coupling of the sequential between leading variable and the process variable

The promptly definite leading variable of sequential coupling is with respect to the retardation time of each process variable.Because the existence of various disturbing factors, actual industrial process is difficult to reach desirable stable state, therefore, in order to make soft-sensing model under the situation that various disturbing factors exist, still can estimate the variation tendency of leading variable more exactly, when gathering training sample, must specify the retardation time of each model input variable correspondence.

Table 1 is depicted as each auxiliary variable and corresponding value retardation time thereof in composite fertilizer's nutrient content soft-sensing model.

Table 1 auxiliary variable and corresponding retardation time thereof are shown

Chloride ion content soft-sensing model auxiliary variable	Retardation time (minute)
		x 1: the potassium chloride inventory	0
x 2: concentrated sulphuric acid flow	0
		x 3: hydrogen potassium 1# transformation pot temperature	0
x 4: hydrogen potassium 2# transformation pot temperature	0
		A line degree of neutralization soft-sensing model auxiliary variable	Retardation time (minute)
x 1: the potassium chloride inventory	30
		x 2: concentrated sulphuric acid flow	30
x 3: the phosphoric acid,diluted flow	30
		x 4: phosphoric acid,diluted concentration	30
x 5: phosphoric acid,diluted density	30
		x 6: the sulfate radical density in the phosphoric acid,diluted	30
x 7: A line nitration mixture flow	0

x 8: A line gas ammonia flow	0
		x 9: gas ammonia pressure	0
B line degree of neutralization soft-sensing model auxiliary variable	Retardation time (minute)
		x 1: the potassium chloride inventory	30
x 2: concentrated sulphuric acid flow	30
		x 3: the phosphoric acid,diluted flow	30
x 4: phosphoric acid,diluted concentration	30
		x 5: phosphoric acid,diluted density	30
x 6: the sulfate radical density in the phosphoric acid,diluted	30
		x 7: B line nitration mixture flow	0
x 8: B line gas ammonia flow	0
		x 9: gas ammonia pressure	0
Multiple fertilizer nutrient content soft-sensing model auxiliary variable
		x 1: the potassium chloride inventory	90
x 2: concentrated sulphuric acid flow	90
		x 3: chloride ion content	90
x 4: the phosphoric acid,diluted flow	90
		x 5: phosphoric acid,diluted concentration	90
x 6: phosphoric acid,diluted density	90
		x 7: the sulfate radical density in the phosphoric acid,diluted	90
x 8: A line nitration mixture flow	60
		x 9: B line nitration mixture flow	60
x 10: A line degree of neutralization	60
		x 11: B line degree of neutralization	60

2, adopt restricted memory PLS algorithm computation chlorion percentage composition and A, B line degree of neutralization

Because the model algorithm of chlorion percentage composition and A line, B line degree of neutralization is identical, be introduced in this example that is calculated as with the chlorion percentage composition.Restricted memory PLS algorithm steps is as follows:

1) (X Y), carries out standardization with sample, and data window length is 40, and the training sample number during promptly each modeling is 40 to gather training sample set.The step of standardization is as follows:

Computation of mean values and variance

The number of input variable X is M, then X=[x ₁x ₂X _M], the number of Y is 1, and number of samples is N, and then the computing formula of the average of sample and variance is as follows:

The mean value computation formula: $\stackrel{\‾}{X}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{X}_i,\stackrel{\‾}{Y}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{Y}_i---\left(1\right)$

The variance computing formula: ${\mathrm{\σ}}_x^2=\frac{1}{N-1}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(X_i-\stackrel{\‾}{X}),{\mathrm{\σ}}_y^2=\frac{1}{N-1}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(Y_i-\stackrel{\‾}{Y})---\left(2\right)$

Wherein, X, σ _xBe the vector of M dimension, Y, σ _yBe scalar.

Data normalization is handled: $X_1=\frac{X-\stackrel{\‾}{X}}{{\mathrm{\σ}}_x},Y_1=\frac{Y-\stackrel{\‾}{Y}}{{\mathrm{\σ}}_y}---\left(3\right)$

Formula (3) carries out standardization to data, and wherein the vector of the standardization formulate same dimension of X removes, after raw data is handled through formula (3), and new data X ₁, Y ₁Average be 0, variance is 1.

2) call the partial least square algorithm, obtain soft-sensing model ${\hat{Y}}_1=X_1\mathrm{\β}.$ The partial least square algorithm can be realized with following NIPALS recursion formula:

(1)E ₀＝X ₁，F ₀＝Y ₁，h＝0

(2)h＝h+1，u _h＝F _h-1

(3)w _h ^T＝u _h ^TE _h-1/(u _h ^Tu _h)

(4) to w _hMake normalized: w _h=w _h/ ‖ w _h‖

(5)t _h＝E _h-1w _h/(w _h ^Tw _h)

(6)q _h ^T＝t _h ^TF _h-1/(t _h ^Tt _h)

(7) to q _hMake normalized: q _h=q _h/ ‖ q _h‖

(8)u _h＝F _h-1q _h/(q _h ^Tq _h)

(9) check u _hConvergence, if convergence then changes (10), otherwise change (3)

(10)p _h ^T＝t _h ^TE _h-1/(t _h ^Tt _h)

(11)b _h＝t _h ^Tu _h/(t _h ^Tt _h)

(12)E _h＝E _h-1-t _hp _h ^T，F _h＝F _h-1-b _h·t _h·q _h ^T

(13) check whether h arrives set-point a,, otherwise change (14) if do not change for (2) step

(14) finish

Wherein W is that M * a ties up matrix, W=[w ₁w ₂W _a]; P is that M * a ties up matrix, P=[p ₁p ₂P _a]; B is that a * a ties up diagonal matrix, B=diag (b ₁b ₂B _a); Q is 1 * a dimension matrix; A is the number of implicit variable, can rule of thumb determine with test.Above NIPALS algorithm can calculate matrix W, P, and B, Q, it is as follows to calculate linear regression model (LRM) according to these matrixes:

${\hat{Y}}_1=X_1\mathrm{\β},\mathrm{\β}=\stackrel{\‾}{W}{\mathrm{BQ}}^T---\left(4\right)$

Wherein The output valve of the soft-sensing model of setting up for standardized data; Wherein W is that M * a ties up matrix, w ₁=w ₁, ${\stackrel{\‾}{w}}_h=\underset{i=1}{\overset{h-1}{\mathrm{\Π}}}\left(1{-{\stackrel{\‾}{w}}_{i}p}_i^T\right),h=2,\·\·\·a.$

3) model that obtains in the step 2 is the soft-sensing model that obtains after the standardization, and this model can not directly be used for the original procedure variable is predicted, need carry out anti-standardization, obtain after the processing about the soft-sensing model formula of original sample variable being,

$\hat{Y}=\mathrm{XC}+b---\left(5\right)$

Wherein

C＝αβ (6)

b＝ Y- XC (7)

α＝σ _y/diag(σ _x) (8)

4) estimate the chloride ion content value according to soft-sensing model and the process variable value that newly collects.

5) check whether new chloride ion content assay value is arranged,, otherwise change step 4 if the step 6 of commentaries on classics is arranged.

6) according to the retardation time in the table 1 from the DCS process database collect new training sample and with the chloride ion content assay value new training sample that partners.With process variable x _iBe example, the correction formula of average, variance is as follows:

${\stackrel{\‾}{x}}_{i,N+1}\frac{N}{N+1}{\stackrel{\‾}{x}}_{i,N}+\frac{1}{N+1}{x}_{i,N+1}---\left(9\right)$

${\mathrm{\σ}}_{i,N+1}^2=\frac{N-1}{N}{\mathrm{\σ}}_{i,N}^2+\frac{1}{N+1}{(x_{i,N+1}-{\stackrel{\‾}{x}}_{i,N+1})}^2---\left(10\right)$

Formula (9) is the online correction formula of average, and formula (10) is the online correction formula of variance.X wherein _{I, N}And x _{I, N+1}Represent variable x respectively _iThe average of correspondence when having N sampled value and N+1 sampled value; σ _{I, N} ²And σ _{I, N+1} ²Represent variable x respectively _iCorresponding variance when having N sampled value and N+1 sampled value.

7) a oldest sample is removed, new sample is added, keep new sample data length of window constant, the training sample under the new window is carried out standardization with revised average and variance.Change step 2.

The model and the computation process of A line degree of neutralization and B line degree of neutralization are the same, its corresponding auxiliary variable and being shown in Table 1 retardation time.

3, adopt the simplification mechanism model to calculate composite fertilizer's nutrients of nitrogen, phosphorus pentoxide, potassium oxide content, step is as follows:

1) according to K among the inventory of KCl and the KCl ₂The percentage composition of O, and in conjunction with K ₂O is scaled the equation of KCl, can calculate the wherein content of KCl.

The chemical molecular formula of potassium chloride is KCl, is white cube crystallization, proportion 1.99,768 ℃ of fusing points.Contain 63.17% kali (K in the pure product ₂O).Can dissolve 27.6 gram potassium chloride in per 100 gram water in the time of 0 ℃; In the time of 100 ℃ 56.7 grams.In most countries, as the potassium chloride of fertilizer wherein the content of kali generally be 58～60% (promptly containing 92～95% potassium chloride).Because the index of potash fertilizer is to represent with the content of kali in the world, potassium chloride is as a kind of potash fertilizer, and its potassium content also is to represent with the content of kali.The content of being calculated potassium chloride by the content of kali can calculate by following formula:

If potassium oxide content is x%, then potassium chloride purity y is:

$y=(\frac{39\×2}{39\×2+16}\·x\%)\×100\%/\left(\frac{39}{39+35.5}\right)---\left(11\right)$

After obtaining the percentage composition of potassium chloride, just can calculate input potassium chloride in the quality and the content of impurity wherein of the potassium chloride that contains.

2) according to the flow of the concentrated sulphuric acid with and percent concentration, can calculate the consumption of sulfuric acid.

In composite fertilizer produces used sulfuric acid be 98% and more than the concentrated sulphuric acid, its concentration 98% refers to massfraction, the density of 98% sulfuric acid is 1.84g/ml, i.e. 1840Kg/L.Because the computing formula of massfraction is:

And the computing formula of density is:

Therefore, the flow when sulfuric acid is Vm ³During/h, per hour the sulfuric acid quality of Liu Ruing is: V * 1 * 1840 * 98%.

3) because phosphoric acid,diluted is impure, wherein contain the sulfate of solubility, as magnesium sulphate, iron sulfate, aluminium sulphate etc.Therefore the density of the sulfate radical in the known phosphoric acid,diluted, can estimate the quality of the soluble sulphate in the phosphoric acid,diluted according to the flow of this density and phosphoric acid,diluted.

According to the flow of phosphoric acid,diluted with and P ₂O ₅Percent concentration and density, can calculate wherein P ₂O ₅Content and the quality of the phosphoric acid of generation.

If the flow of phosphoric acid,diluted is Vm ³/ h, its density is d Kg/L, the density of sulfate radical is d ₁Kg/L, P ₂O ₅Concentration be p%, the P of Liu Ruing per hour ₂O ₅Quality be: V * 1 * (d-d ₁) * p%.The pure phosphoric acid amount z that generates is: $z=\frac{V\·(d-d_1)\·p\%}{142}\×196.$ This equation is to obtain according to following chemical equation:

142 98*2

V·(d-d ₁)·p％ z (14)

4) extent of reaction of KCl and sulfuric acid is certainly less than 100%, and its extent of reaction can calculate an approximate value according to the degree of chlorion in the finished product.Calculation equation is as follows:

The percentage composition of supposing chlorion in the finished product is x%, and then the chlorion quality in one ton of composite fertilizer is x% * 1000.One ton of composite fertilizer of known again production approximately consumes 235 kilograms in potassium chloride, has calculated above the purity of used potassium chloride to be y%.The chlorion total amount of then producing one ton of composite fertilizer's consumption is: 235 * y% * 35.45/ (35.45+39), wherein 35.45 is atomic weight of chlorine, the 39th, and the atomic weight of potassium.So, when KCl and strong sulfuric acid response, the number percent of not participating in the chlorion of reaction is:

$\frac{x\%\×1000}{y\%\×235\×\frac{35.45}{35.45+39}}\×100\%---\left(15\right)$

Then the extent of reaction is:

$(1-\frac{x\%\×1000}{y\%\×235\×\frac{35.45}{35.45+39}})\×100\%---\left(16\right)$

According to this extent of reaction, can calculate KCl and H ₂SO ₄Reaction remaining KCl in back and H ₂SO ₄Amount and the KHSO of generation ₄Amount.Equation is as follows:

The assumed response degree is z%, and the inventory of potassium chloride is a, and then the potassium chloride that falls of reaction consumes is: a * y% * z%, remaining KCl is: a-a * y% * z%.According to chemical reaction equation once, the quality of the sulfuric acid b of consumption and the potassium acid sulfate c of generation is:

74.45 98 136

a×y％×z％ b c (17)

Wherein $b=\frac{a\×y\%\×z\%}{74.45}\×98,$ $c=\frac{a\×y\%\×z\%}{74.45}\×136.$

Because calculated the sulfuric acid quality of inflow, so can calculate remaining sulfuric acid quality be: V * 1 * 1840 * 98%-b.

Chloride ion content is the manual analysis value in the finished product, analysis frequency is time/8 hours, therefore, at first chloride ion content is set up the value of calculating chloride ion content based on the soft-sensing model of restricted memory partial least square algorithm in real time, and ask the dereaction degree with this Model Calculation value.

5) because in the operation of back, a neutralization reaction only having taken place, does not have other subsidiary reaction and exist.Chemical equation is:

Because known KHSO ₄Quality, according to equation , can calculate KHSO ₄Consume how many ammonias altogether, and generated how many KNH ₄SO ₄

6) according to equation Can calculate remaining sulfuric acid and consume how many ammonias altogether, and generate how many (NH ₄) ₂SO ₄

7) according to equation Can calculate phosphoric acid with equation and consume how many ammonias altogether, and generate how many (NH ₄) H ₂PO ₄Because some phosphorous acid (NH wherein ₄) H ₂PO ₄React with ammonia again and generated phosphoric acid diamino (NH ₄) ₂HPO ₄, therefore to calculate whole ammonias and last all quality of products of consumption, must know that having how many phosphorous acids and ammonia that reaction has taken place has generated how much phosphoric acid diamino again.Because therefore the ratio of phosphorous acid and phosphoric acid diamino, can utilize this index of degree of neutralization to calculate during this index of degree of neutralization has just in time reflected and in the slip.The detection method of degree of neutralization is as follows:

To in and the degree of neutralization of slip when detecting, the solution that is used for titration is:

The sulfuric acid standard solution: $c\left(\frac{1}{2}{H}_{2}S{O}_4\right)=0.1\mathrm{mol}/L;$

Standard solution of sodium hydroxide: c (NaOH)=0.1mol/L.

If sulfuric acid standard solution consumption is V during titration ₁, the standard solution of sodium hydroxide consumption is V ₂, then the computing formula of degree of neutralization is: 1+V ₁/ V ₂Degree of neutralization value according to analysis obtains can obtain V ₁/ V ₂Value.According to the amount of substance concentration of reaction formula and standard solution, can calculate (the NH of generation then ₄) H ₂PO ₄With (NH ₄) ₂HPO ₄Ratio.The quality of the phosphoric acid that goes out according to previous calculations can calculate (the NH of generation at last ₄) H ₂PO ₄With (NH ₄) ₂HPO ₄Quality.Detailed process is as follows:

2 1 2

0.2×V ₁ 0.1×V ₁ 0.2×V ₁ (18)

2 2

0.1×V ₂0.1×V ₂ (19)

Therefore, in and slip in contained (NH ₄) ₂HPO ₄With (NH ₄) H ₂PO ₄Ratio be:

$\frac{{\left(N{H}_4\right)}_2\mathrm{HP}{O}_4}{\left(N{H}_4\right)H_{2}P{O}_4}=\frac{0.2\×V_1}{0.1\×V_2-0.2\×V_1}=\frac{0.2\×\frac{V_1}{V_2}}{0.1-0.2\×\frac{V_1}{V_2}}---\left(20\right)$

Again according to chemical equation: Can calculate H by known 4 quality ₃PO ₄Generate (NH ₄) H ₂PO ₄Quality, then according in obtaining above and the (NH of slip ₄) ₂HPO ₄With (NH ₄) H ₂PO ₄Ratio can calculate (the NH of generation ₄) ₂HPO ₄Quality.

Degree of neutralization is the manual analysis value, analysis frequency is time/2 hours, therefore, at first set up the degree of neutralization value of calculating A, B line based on the degree of neutralization soft-sensing model of restricted memory partial least square algorithm in real time, participate in top calculating in conjunction with the degree of neutralization that A, B line nitration mixture flow weighted average calculation go out whole production line with this Model Calculation value.

8) KNH that utilizes the front to calculate ₄SO ₄Amount, (NH ₄) ₂SO ₄Amount, (NH ₄) H ₂PO ₄Amount, (NH ₄) ₂HPO ₄Amount can calculate the quality of last all products.Add the amount of the remaining KCl in front, and the quality of the soluble sulphate that contains in the amount, phosphoric acid,diluted of impurities in the potassium chloride, can calculate last gross mass.The consumption of the ammonia that calculates according to above-mentioned three steps can calculate to contain how much nitrogen in the product altogether, promptly can calculate the number percent of nitrogen in the product divided by last gross mass.In like manner, K ₂The quality of O and P ₂O ₅The quality front calculate, also can calculate K divided by last gross mass ₂The number percent of O and P ₂O ₅Number percent.

9) according to analysis, the degree of the nitrogen that the degree of the nitrogen of the complex fertilizer product that aforementioned calculation obtains, kali, phosphorus pentoxide and lab analysis obtain, kali, phosphorus pentoxide should be consistent.Therefore, according to above-mentioned analytical calculation step, can calculate the degree of nitrogen in the product, kali, phosphorus pentoxide in real time.

Composite fertilizer's nutrient content soft-measuring modeling method effect:

Be used for certain sulfuric acid plant's sulfenyl nitrogen-phosphorus-potassium compound fertilizer generating apparatus according to soft-sensing model that this modeling method is built.For the validity of model is described, adopt on October 1st, 2,005 2 computational data and laboratory assay values to analyze comparison to the soft-sensing model in 10 of October 25, three of laboratory assay value every days, collect 57 valid data (during shutdown of short term is arranged) altogether.Nitrogen content, P ₂O ₅Content, K ₂The soft-sensing model output valve of O content and the curve ratio of assay value are seen accompanying drawing 3,4,5, and the relative accuracy of soft measuring instrument is as shown in table 2.

Table 2 composite fertilizer nutrient content soft measuring instrument relative accuracy (2005.10.1-10.25)

	Maximum relative deviation %	Minimum relative deviation %	Mean square deviation %
				Nitrogen	4.28	0	1.46
Phosphorus pentoxide	4.88	0.05	1.99
				Kali	6.62	0.02	2.57

From table 2, composite fertilizer's soft-sensing model precision is higher; From accompanying drawing 3,4,5, the calculated value of composite fertilizer's soft-sensing model and laboratory assay value trend comparison coincide.This shows to have good estimated performance for the sulfenyl nitrogen-phosphorus-potassium compound fertilizer nutrient content, be applicable to industrial online computing application fully by soft-sensing model that this modeling method is built.

Claims (3)

1, a kind of based on the composite fertilizer's nutrient content soft-measuring modeling method that mixes multi-model, it is characterized in that it mixes the multi-model structure, promptly this model is by based on the A of data-driven, B line degree of neutralization soft-sensing model, based on the chloride ion content soft-sensing model of data-driven with import more/export more soft-sensing model based on composite fertilizer's nutrient content of simplifying mechanism model and constitute, and wherein more the output of A, B line degree of neutralization soft-sensing model and chloride ion content soft-sensing model imports/export more the input of soft-sensing model as composite fertilizer's nutrient content.

2, according to claim 1 a kind of based on the composite fertilizer's nutrient content soft-measuring modeling method that mixes multi-model, it is characterized in that described modeling method comprises the steps:

1), potassium chloride inventory that selection is relevant with the composite fertilizer nutrient content, concentrated sulphuric acid flow, hydrogen potassium 1# reactive tank temperature, hydrogen potassium 2# reactive tank temperature, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, sulfate radical density, A line nitration mixture flow, A line gas ammonia flow, B line nitration mixture flow, B line gas ammonia flow, gas ammonia pressure, A line degree of neutralization, B line degree of neutralization, chloride ion content totally 16 process variable, as the auxiliary variable of soft-sensing model;

2), the crucial auxiliary variable A line degree of neutralization and the B line degree of neutralization of composite fertilizer's nutrient content be the manual analysis value, rather than collection value in real time; At first use restricted memory partial least square algorithm to set up the dynamic soft sensor model of A, B line degree of neutralization; Wherein, the auxiliary variable of A line degree of neutralization soft-sensing model is: potassium chloride feeds intake, concentrated sulphuric acid flow, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, A line nitration mixture flow, A line gas ammonia flow, gas ammonia pressure.The auxiliary variable of B line degree of neutralization soft-sensing model is: potassium chloride inventory, concentrated sulphuric acid flow, phosphoric acid,diluted flow, phosphoric acid,diluted concentration, phosphoric acid,diluted density, B line nitration mixture flow, B line gas ammonia flow, gas ammonia pressure;

3), the key parameter chloride ion content simplified in the mechanism model is the manual analysis value, uses restricted memory partial least square algorithm to set up the dynamic soft sensor model of chloride ion content; The auxiliary variable of this model is potassium chloride inventory, concentrated sulphuric acid flow, hydrogen potassium 1# reactive tank temperature, hydrogen potassium 2# reactive tank temperature;

4), the output of A, B line degree of neutralization soft-sensing model and chloride ion content soft-sensing model is as the input of importing more/export more soft-sensing model based on composite fertilizer's nutrient content of simplifying mechanism model.

3. a kind of composite fertilizer's nutrient content soft-measuring modeling method based on the mixing multi-model according to claim 2 is characterized in that, more described composite fertilizer's nutrient content based on the simplification mechanism model is imported/exported more the soft-sensing model establishment step and be:

1) utilizes the result of chloride ion content soft-sensing model,, calculate the reaction rate of the potassium chloride and the concentrated sulphuric acid according to material balance.Produce one ton of composite fertilizer and need consume 235 kilograms in potassium chloride, suppose that chloride ion content is Cl_c (a chloride ion content soft-sensing model calculated value) in the composite fertilizer, then the computing formula of reaction rate is:

According to this reaction rate, in conjunction with chemical equation , can calculate KCl and H ₂SO ₄Reaction remaining KCl in back and H ₂SO ₄Amount and the KHSO of generation ₄Amount;

2) in the flow process of back, a neutralization reaction has only taken place, do not have other subsidiary reaction and exist.The chemical equation of neutralization reaction is:

According to known KHSO ₄, H ₂SO ₄, H ₃PO ₄Quality, can calculate corresponding product KNH4SO in conjunction with first three equation ₄, (NH ₄) ₂SO ₄, (NH ₄) H ₂PO ₄Quality and the NH of their combinations ₃Quality;

3) according to the result of calculation of A, B line degree of neutralization soft-sensing model and A, B line mixed flow flow, utilize weighted average calculation to go out the degree of neutralization of whole production line.The degree of neutralization of supposing the A line is count_a, and the degree of neutralization of B line is count_b, and A line nitration mixture flow is flow_a, and B line nitration mixture flow is flow_b, and then the degree of neutralization of whole production line is $\mathrm{count}\_a\×\frac{\mathrm{flow}\_a}{\mathrm{flow}\_a+\mathrm{flow}\_b}+\mathrm{count}\_b\×\frac{\mathrm{flow}\_b}{\mathrm{flow}\_a+\mathrm{flow}\_b},$

Because degree of neutralization reflection just in and (NH in the slip ₄) ₂HPO ₄With (NH ₄) H ₂PO ₄Ratio, therefore, according to the assay method of degree of neutralization and known (NH ₄) H ₂PO ₄Quality, can calculate (NH in the last composite fertilizer ₄) ₂HPO ₄Quality, and calculate remaining (NH thus ₄) H ₂PO ₄Quality and reaction in the NH that consumes ₃Quality.(NH ₄) ₂HPO ₄The computing formula of quality is as follows:

$\frac{{\left({\mathrm{NH}}_4\right)}_2\mathrm{HP}{O}_4}{\left({\mathrm{NH}}_4\right)H_{2}P{O}_4}=\frac{0.2{\×V}_1}{0.1\×V_2-0.2{\×V}_1}=\frac{0.2\×\frac{V_1}{V_2}}{0.1-0.2\×\frac{V_1}{V_2}}$

Wherein ₁Sulfuric acid standard solution consumption when being titration, V ₂Standard solution of sodium hydroxide consumption when being titration,

Promptly

4) calculate the summation of last product,, can obtain the degree of nitrogen, phosphorus pentoxide, kali with the quality of the quality of the quality of the nitrogen that has calculated, phosphorus pentoxide and kali summation divided by last product.

Images