CN104765347A - Yield real-time prediction method in residual oil delayed coking process - Google Patents

Abstract

The invention discloses a yield real-time prediction method in the residual oil delayed coking process. The method combines features of a delayed coking reactor, and a mathematical model of an adiabatic reactor in the axial direction is established. Based on the material balance principle and the thermal balance principle, processing is conducted on collected device real-time data, an intelligent optimization algorithm is combined, real-time correction is conducted on delayed coking reaction kinetic parameters, and a mechanism model can accurately describe device actual operation conditions. Based on the accurate mechanism model, the influence of key operation and technological conditions on the yield of delayed coking products is analyzed, the neural network modeling technology is combined, correlation is conducted between the operation conditions and the yield data, a yield surrogate model is established, and the calculation speed of the yield data is increased. Piecewise linearization is conducted according to the yield influencing trend, a linear equation is solved, corresponding Delta-Base yield data of the PIMS planning model is obtained, the yield real-time prediction of a delayed coking device is achieved, and the theoretical support for establishing the accurate planning and optimizing PIMS model is provided.

Description

Yield real-time predicting method in a kind of residual oil delayed coking

Technical field

The present invention relates to the delayed coking yield real-time predicting method merging slag oil cracking reaction mechanism and device real time execution characteristic, the method may be used for delayed coking modeling and simulating, model real time correction and and PIMS production planning optimization model build in real time.

Background technology

Delayed coking is the main machining method of modern oil refining process for heavy oil.Delayed Coking Technology is under uniform temperature and pressure condition, by heating furnace Fast Heating, slag oil crack is become the process of dry gas, liquefied gas, gasoline, diesel oil and wax oil.

The delayed coking unit process flow diagram of general refinery as shown in Figure 1.Residual oil from vacuum distillation tower enters delayed coking raw material surge tank after flowmeter, fractionator bottom is entered after pump pressurization, with high-temperature oil gas (430 ~ 440 DEG C) heat exchange from coke drum, lightweight oil in feedstock oil is evaporated, again overheated coking oil gas is down to the temperature can carrying out fractionation simultaneously.Feedstock oil is extracted out together with recycle oil at the bottom of the tower of fractionator, delivers to heating furnace and is heated to about 500 DEG C, then enter coke drum bottom through four-way valve.The feedstock oil of heat carries out the reaction such as cracking, condensation in coke drum, finally generates coke.Coke is gathered in coke drum, and reaction oil gas is overflowed from coke drum top, enters fractionator, obtains coking gas, gasoline, diesel oil, wax oil and recycle oil.

In actual production process, the yield of the products such as coking gas, gasoline, diesel oil and wax oil is subject to the impact of raw material and operating conditions, causes actual output and production schedule desired value often to there is certain deviation.Trace it to its cause, key is that current production schedule PIMS model all adopts fixing yield, is difficult to accurate description device actual moving process.Therefore, how to set up an energy accurately, the yield model of quantitative description delayed coking unit is the key improving current oil refining apparatus Production planning model accuracy.

Delayed coking raw material is mostly heavy, residual oil inferior, and its composition is very complicated, and thus delayed coking reaction mechanism is also very complicated, therefore has certain difficulty to its modelling.But generally speaking delayed coking belongs to hydro carbons thermal conversion reaction, be a kind of parallel sequence reaction of complexity, reaction is carried out according to cracking and condensation both direction substantially.Cracking reaction is thermonegative reaction, is produced the Small molecular such as carburet hydrogen by cracking reaction; Condensation reaction is themopositive reaction, and produced the large molecules such as condensed-nuclei aromatics by condensation reaction, the share shared by the former is obviously greater than the latter, and thus delayed coking belongs to thermonegative reaction generally, needs at high temperature to carry out.

The core of accurate yield model is reaction mechanism dynamics accurately.As everyone knows, delayed coking reaction comprises alkane thermal transition, naphthenic hydrocarbon thermal transition, aromatic hydrocarbons thermal transition and the multiple reaction such as sulfur-bearing, nitrogenous non-hydrocarbons thermal transition, and mechanism is comparatively complicated.At present, for similar complex reaction system, the general method of lump that adopts carries out dynamic analysis.So, reaction core mechanism can be retained, can also reaction network be simplified simultaneously, reduce Chemical kinetic parameter estimation difficulty.In recent years, external many engineering corporatioies and scientific research institution, as Uop Inc., Mobil oil company, Sinopec Research Institute of Petro-Chemical Engineering, University of Virginia, East China University of Science, all delayed coking reaction is conducted in-depth research, propose respective lumped watershed hydrologic model model.This patent considers model accuracy and counting yield, and the basis of 11 lumped models is set up delayed coking reaction device mechanism model, for real-time estimate delayed coking unit critical product yield.

Summary of the invention

The invention provides and a kind ofly merge yield real-time predicting method in the delayed coking of slag oil cracking reaction mechanism and device real time execution characteristic.

Concrete technical scheme is as follows:

Yield real-time predicting method in a kind of residual oil delayed coking, comprises the following steps:

(1) utilize the data communication interface between VB.net Development of Software Platform refinery PHD real-time data base and mechanism model, realize the real-time data capture of residual oil delayed coking; Based on material balance, energy equilibrium and field instrument situation, data are in harmonious proportion, reject useless and wrong real time data;

(2) according to the device data of Real-time Collection, minimum as optimization aim using the difference of two squares of the model predication value of delayed coking reaction device outlet and actual value, intelligent optimization algorithm is utilized to solve, matching reactive kinetics parameters, realization mechanism model real time correction;

(3) on accurate model basis, analyze the critical product yield under different material quality, processing capacity and operating procedure condition, set up product yield database;

(4) the neural network agent model of product yield database training delayed coking reaction process is utilized, piece-wise linearization is carried out according to yield change trend curve, calculating Delta-Base data, providing theory support for setting up accurate planning optimization PIMS model.

The real time data gathered in step (1) is residual oil raw material feed loading, density, sulfur content, carbon residue content, reactor inlet temperature, reaction pressure and product flow information.

Described in step (2), device data refer to feed loading, temperature of reaction, pressure and product information.

The optimized algorithm adopted in step (2) is the differential evolution algorithm with trigonometric mutation, and defines optimization aim and be:

$\min f\left(x\right)=\underset{i}{\mathrm{\Σ}}{\left(\frac{x_{\mathrm{actual}}-x_{\mathrm{calculate}}}{x_{\mathrm{actual}}}\right)}^2$

In formula: decision variable x comprises the Dynamics Factors of each reaction, x _actualand x _calculaterepresent product oil quality yield that is actual and that calculate respectively.

Step (3) Raw quality refers to material density, sulfur content and carbon residue content; Operating procedure condition refers to reactor inlet temperature and reaction pressure, and critical product then refers to rich gas, gasoline, diesel oil and wax oil.

Employ neural network model in step (4) to carry out alternative mechanism model and carry out prediction and calculation; Wherein neural network model adopts backpropagation (Back propagation, BP) neural network, chooses 6 neural network input variables: feed loading, density, sulfur content, carbon residue content, reactor inlet temperature, reaction pressure; 5 output variables: rich gas, gasoline, diesel oil, wax oil, coke; Hidden layer is set to 7 layers; Utilize agent model to carry out real-time estimate to delayed coking product yield, and obtain corresponding Delta-Base value.

Beneficial effect of the present invention is as follows: propose a kind of delayed coking critical product yield real-time predicting method merging slag oil cracking reaction mechanism and device real time execution characteristic.The method relies on mechanism model, integrated application imitation technology, real time correction technology and nerual network technique, Delta-Base data accurately can be obtained according to device practical operation situation, provide Theory and technology to support for oil refining process PIMS production planning optimization model builds in real time.

Accompanying drawing explanation

Fig. 1 is delayed coking simplified flow chart;

Fig. 2 is 11 lumped watershed hydrologic model model reaction networks;

Fig. 3 is model real time correction simplified flow chart;

Fig. 4 is multilayer feedforward neural network structural representation;

Fig. 5 is neural network device agent model structural drawing.

Embodiment

Below by embodiment, the present invention is specifically described.What be necessary to herein means out is; following examples are only for the invention will be further described; can not limiting the scope of the invention be interpreted as, some nonessential improvement and adjustment that professional and technical personnel's content according to the present invention in this field is made, still belong to protection scope of the present invention.

Embodiment 1

Implementation method of the present invention is specifically introduced below in conjunction with chart:

1, delayed coking reaction modelling by mechanism

Raw material is divided into stable hydrocarbon, light aromatic hydrocarbons, heavy aromatics, maltha, hard colloid and bituminous matter six component and be divided into by thermal transition product on the basis of gas, gasoline, stable hydrocarbon cracking intermediate oil, unsaturation hydrocarbon pyrolysis intermediate oil and coke, set up 11 lumped watershed hydrologic model models, as shown in Table 1 and Table 2.Reaction network is as Fig. 2, and reaction network comprises 28 reaction rate constants.

Table 1 delayed coking dynamics research achievement

Dynamics	Stoichiometric number	Time	Researcher
				Thermogravimetry (TG)	1	1992	Yang Jitao
Parallel reactor dynamics	2	1978	Rich winning chance of a specified duration
				Parallel-consecutive reaction dynamics	3	1989	Yang Jiamo
Eight lumped watershed hydrologic model	11	1987	Toru Takatsuka
				11 lumped watershed hydrologic model	28	1999	Zhou Xiaolong

The detailed component of table 211 lumped watershed hydrologic model model

First-order kinetics equation group:

$\frac{{\mathrm{dC}}_s}{\mathrm{dt}}=-(k_{\mathrm{SG}}+k_{\mathrm{SL}}+k_{\mathrm{SV}1})C_s---\left(1\right)$

$\frac{{\mathrm{dC}}_{V1}}{\mathrm{dt}}=k_{\mathrm{SV}1}{C}_S-(k_{V1G}+k_{V1L})C_{V1}---\left(2\right)$

$\frac{{\mathrm{dC}}_L}{\mathrm{dt}}=k_{\mathrm{SL}}{C}_S+k_{V1L}{C}_{V1}---\left(3\right)$

$\frac{{\mathrm{dC}}_G}{\mathrm{dt}}=k_{\mathrm{SG}}{C}_S+k_{V1G}{C}_{V1}---\left(4\right)$

$\frac{{\mathrm{dC}}_{A1}}{\mathrm{dt}}=-(k_{A1G}+k_{A1L}+k_{A1V2}+k_{A1C})C_{A1}---\left(5\right)$

$\frac{{\mathrm{dC}}_{V2}}{\mathrm{dt}}=k_{A1V2}{C}_{A1}-(k_{V2G}+k_{V2L}+k_{V2G})C_{V2}---\left(6\right)$

$\frac{{\mathrm{dC}}_G}{\mathrm{dt}}=k_{A1G}{C}_{A1}+k_{V2G}{C}_{V2}---\left(7\right)$

$\frac{{\mathrm{dC}}_L}{\mathrm{dt}}=k_{A1L}{C}_{A1}+k_{V2L}{C}_{V2}---\left(8\right)$

$\frac{{\mathrm{dC}}_c}{\mathrm{dt}}=k_{A1C}{C}_{A1}+k_{V2C}{C}_{V2}---\left(9\right)$

S: stable hydrocarbon, Al: light aromatic hydrocarbons, Ah: heavy aromatics, R1: maltha, Rh: hard colloid, B: bituminous matter,

G: gasoline, V1: stable hydrocarbon intermediate oil, V2: unsaturation hydrocarbon intermediate oil, C: coke

Model hypothesis:

Do not react to each other between a six reaction lumped components that () hypothesis asphaltum oil raw material comprises.

B () responds and is one-level non-reversible reaction.

When c middle distillate that () stable hydrocarbon lump cracking generates and the middle distillate further cracking that all the other five lumped component cracking are produced, reactivity worth is different.

D () stable hydrocarbon lump cracking does not generate coke.

2, data acquisition and mediation

Because industry spot situation is complicated, production run is subject to the impact of many factors, the complete delayed coking model set up according to the mechanism reported in experiment document cannot accurate simulation commercial plant, therefore needs in conjunction with field device actual motion characteristic to correct kinetic model.For ensureing the accuracy of basic data, reconciliation process need be carried out to the data of on-site collection.

1) on-site data gathering: Large-scale Refinery all adopts MES system to carry out daily Production&Operations Management at present, wherein comprises the real-time dataBase system for pen recorder real-time operating conditions.Field data obtains the item needing building database interface and provide respective counts strong point.The present invention utilizes VB.net interfacing development delay coker field data Real-time Collection technology, can realize reading in on-the-spot real time data and storing.The data gathered are needed mainly to comprise feedstock property data (density, sulfur content, carbon residue content), the data such as feed loading, reactor inlet temperatures, pressure and critical product output.

2) data reconciliation process: by the limitation of Site Detection instrument reliability, often there is the problem such as material imbalance, heat imbalance in the data directly got from DCS, therefore can not be directly used in apparatus for establishing model.In order to ensure the accuracy of model sample data, be necessary to set up mediation standard to the data of collection in worksite, the following several method of concrete use:

(1) material balance means for correcting turnover logistics flux is adopted;

(2) energy equilibrium means for correcting turnover energy meter data are adopted;

(3) according to the codomain of statistics and knowhow determination data, judge the accuracy of data according to this, misdata is deleted from local data base;

(4) for the data that cannot gather in given period, setting up the computing formula of redundancy, deriving this point by gathering other data.

3, mechanism model real time correction

Based on the various balance equations in the kinetic rate form chosen and reactor, choose the predicted value of reactor outlet critical product yield and the variance of experimental data and industry spot image data and minimum be target, the optimization problem that the deterministic process of parameter of reaction kinetics model is converted into function is solved, that is:

$\min f\left(x\right)=\underset{i}{\mathrm{\Σ}}{\left(\frac{x_{\mathrm{actual}}-x_{\mathrm{calculate}}}{x_{\mathrm{actual}}}\right)}^2---\left(10\right)$

In formula, decision variable x comprises the Dynamics Factors of each reaction, x _actualand x _calculaterepresent product oil quality yield that is actual and that calculate respectively.For such optimization aim, the present invention uses the difference algorithm with trigonometric mutation to solve problem.

Difference algorithm (differential evolution, DE) be a kind of based on population random search algorithm, it has, and structure is simple, fast convergence rate, robustness high.The Variation mechanism of algorithm, the method namely generating filial generation is:

r′＝r ₁+F*(r ₂-r ₃) (11)

Wherein, r ' is newly-generated offspring individual, r ₁, r ₂, r ₃be that the parent that three of random selecting in population are different is individual, F is differential evolution operator, is generally a constant.

Because this objective decision variables number is numerous, when causing Algorithm for Solving, calculated amount is very large, therefore needs to improve algorithm, accelerates its speed of convergence.The present invention have selected the improvement difference algorithm with trigonometric mutation, and the method is proved to be has remarkable effect in raising algorithm the convergence speed, and its Mutation Strategy improved can be expressed as:

r′＝(r ₁+r ₂+r ₃)/3+(p ₂-p ₁)(r ₁-r ₂)+(p ₃-p ₂)(r ₂-r ₃)+(p ₁-p ₃)(r ₃-r ₁) (12)

Wherein

p ₁＝|f(r ₁)|/p′

p ₂＝|f(r ₂)|/p′

p ₃＝|f(r ₃)|/p′

(13)

p′＝|f(r ₁)|+|f(r ₂)|+|f(r ₃)| (14)

The simplified flow chart of model real time correction as shown in Figure 3.

This example chooses the actual average data in certain refinery residual oil delayed coking unit week as explanation, and the operation operating mode of device is: reactor pressure 0.16MPag, reactor inlet temperature 497 DEG C, feed rate 175870kg/h, internal circulating load 39480kg/h.According to above-mentioned condition, obtain the model output data after correcting by Algorithm for Solving as shown in table 3, obtain the kinetic parameter of calibration model in table 4 simultaneously.

The prediction of table 3 calibration model exports the contrast (%) with real data

Lumped model reactive kinetics parameters after table 4 corrects

Sequence number	Reaction	Value
			1	Thermal load deviation [kJ/h]	0
2	Cracking generates naphtha	4.61E-02
			3	Cracking generates diesel oil	8.00E-02
4	Cracking generates wax oil	1.00E-02
			5	Lytic activity	0.6
6	Coking is active	23.29
			7	Cracking generates sulfuretted hydrogen	1.400437

4, delayed coking product yield is analyzed

By the realization of above step, mechanism model accurately can reflect the practical operation situation of delayed coking unit by the correction of device real-time running data.Bonding mechanism model is the main contents of this step to device crucial product yield analysis under different operating condition and material condition.

The device critical product yield utilizing the mechanism model corrected to obtain for different material characteristic (being mainly density, sulfur content and carbon residue content), feed loading, reactor inlet temperature and pressure is analyzed, and wherein critical product mainly comprises: rich tail gas of coking plant, coker gasoline, coker gas oil, wax oil and coke.Set up local data library storage yield data, and the read-write of data can be realized.

Under table 5 and table 6 are respectively reactor inlet temperature and material quality (density, sulfur content and carbon residue content) change, crucial production yield rate analysis example.

The yield of the lower critical product of table 5 reactor inlet temperature change

The yield of the lower critical product of table 6 material quality change

Note: a, density, sulfur content, carbon residue content are respectively 0.973, and 1.8387%, 12.5647

B, density, sulfur content, carbon residue content are respectively 0.987, and 2.5186%, 14.2285

C, density, sulfur content, carbon residue content are respectively 0.979, and 2.2485%, 13.2864

D, density, sulfur content, carbon residue content are respectively 0.978, and 1.7800%, 12.2515

E, density, sulfur content, carbon residue content are respectively 0.968, and 1.8072%, 11.7737

As shown in table 5, in the process that temperature of reactor raises gradually, the dry gas in delayed coking product, gasoline, diesel yield constantly rise, and this contributes to residual oil owing to improving temperature C-C cleavage reaction occurs, can more lightweight oil and fuel gas.But also can find out from table, temperature is higher, coking yield also raises wax oil yield simultaneously gradually to be reduced, and this is conducive to the polycondensation of condensed ring hydro carbons owing to heating up to generate coke, is conducive to wax oil cracking simultaneously and generates vapour, diesel oil.

As shown in table 6, along with material quality is deteriorated, oil density becomes large, sulfur content and carbon residue content all increase, and cause the raising corresponding to coking yield of wax oil in delayed coking product, simultaneously because light constituent in raw material reduces, light-end products in product, dry gas, gasoline and diesel oil, yield declines.As can be seen here, material quality (density, sulfur content and carbon residue content) has important impact for device product yield.

5, product yield real-time estimate

Also very high requirement is proposed to the counting yield of model while delayed coking unit yield real-time estimate technical requirement mechanism model has high accuracy.The calibration model mechanism introduced above is very complicated, and solving speed is very slow, and does not easily restrain, and obviously cannot meet the needs of real-time estimate, therefore needs utilize nerual network technique to set up real-time estimate that agent model carrys out implement device yield badly.

The present invention adopts backpropagation (Back propagation, BP) neural network, and this is one most widely used neural network structure in process control.BP neural network structure as shown in Figure 4.Total is made up of L layer neuron, and ground floor is input layer, and last one deck is output layer, and other layer is hidden layer, can obtain each neuronic model to be:

$y_{\mathrm{pjk}}=\underset{i=0}{\overset{N_{j=1}}{\mathrm{\Σ}}}{w}_{\mathrm{jki}}{x}_{p,j-1,j}---\left(15\right)$

$\left\{\begin{array}{c}{x}_{\mathrm{pjk}}=f\left(y_{\mathrm{pjk}}\right)\\ x_{\mathrm{pjo}}=1\end{array}\right.---\left(16\right)$

In formula:

Y _pjk: in j layer each neuron of kth under p group sample state and export;

X _pjk: in j layer, the sum functions of a kth neuron under p group sample state exports;

W _pjk: p neuronic link weights in i-th neuron to j layer in j-1 layer.Wherein w _pjobe defined as a j layer kth neuronic threshold values.

F (y): neuronic nonlinear activation function.

The object of neural network learning finds out a series of weights, after making the often group input vector of sample act on network, the actual output vector of its network is consistent with the desired output vector of sample, whole learning process is the connection weights in adjustment network between each neuron, makes the error energy function of following network reach minimum:

$E=\frac{1}{2}\underset{p}{\mathrm{\Σ}}\underset{k=1}{\overset{H_L}{\mathrm{\Σ}}}{(o_{\mathrm{pk}}-x_{\mathrm{plk}})}^2---\left(17\right)$

Wherein:

O _pk: after p group sample input vector acts on network, a kth expectation value that neuronic function exports in network output layer.

The BP algorithm problem concerning study solving above-mentioned Multilayer Feedforward Neural Networks, this learning algorithm is made up of signal forward-propagating and error back propagation.Traditional BP algorithm can briefly be summarized as follows.

$w_{\mathrm{jki}}(t+1)=w_{\mathrm{jki}}\left(t\right)+\mathrm{\η}\underset{p}{\mathrm{\Σ}}{\mathrm{\δ}}_{\mathrm{pjk}}{x}_{p,j-1,i}+\mathrm{\α}[w_{\mathrm{jki}}\left(t\right)-w_{\mathrm{kji}}(t-1)]---\left(18\right)$

Wherein:

$\left\{\begin{array}{c}{\mathrm{\δ}}_{\mathrm{plk}}=f^{\′}\left(y_{\mathrm{pjk}}\right)(o_{\mathrm{pk}}-x_{\mathrm{plk}})\\ {\mathrm{\δ}}_{\mathrm{pjk}}=f^{\′}\left(y_{\mathrm{pjk}}\right)\underset{i=1}{\overset{N_{j+1}}{\mathrm{\Σ}}}{\mathrm{\δ}}_{p,j+1,i}{w}_{j+1,i,k}\end{array}\right.j=L-1,...,1---\left(19\right)$

In formula:

δ _pjk: after p group sample input vector acts on network, the control information that in j layer, a kth neuron function exports;

T: learning time;

F ' (y): the single order derived function of neuron activation functions.

Formula (15)-(19) are the learning rules of weights in BP network, formula learning speed and situation term coefficient are generally determined by experience, in traditional BP algorithm they can not with network structure, network state and external learning environment Auto-matching, need people for adjusting during network training.

According to above-mentioned neural network structure, choose neural network and input 6: Residual cracking amount, fed feed density, sulfur content, carbon residue content, temperature of reaction, reaction pressure; Export 5: rich gas, gasoline, diesel oil, wax oil, coke; Hidden layer is set to 7 layers.With this neural network training agent model, see Fig. 5.

According to analytical data is bright above, the delayed coking unit critical product yield under performance variable impact presents nonlinear trend.The present invention utilizes neural network agent predicts model to calculate, and non-linear object is carried out piece-wise linearization process, produces corresponding Delta-Base data, realizes yield real-time estimate.Result of calculation can be used for the production schedule PIMS model setting up delayed coking, improves the precision of model.

Utilize the Delta-Base data of neural network agent model calculation element critical product yield as shown in table 7, its computation process is divided into two steps:

The first step: calculate the prophetic yields of the every critical product of device as Base data according to given residual oil quality (density, sulfur content and carbon residue content) and operating conditions;

Second step: according to the free variable of setting, automatically calculate this variable change in set-point neighborhood under product yield predicted data, and calculate the variable quantity that unit variance changes lower yield, this variable quantity is namely as the Delta data of plant yield prediction, and computing formula is as follows:

${\mathrm{Delta}}_{\mathrm{val}}=\frac{Y_{+}^{\mathrm{keyproduct}}-Y_{-}^{\mathrm{keyproduct}}}{{\mathrm{Val}}_{+}-{\mathrm{Val}}_{-}}=\frac{\mathrm{\Δ}{Y}^{\mathrm{keyproduct}}}{\mathrm{\ΔVal}}---\left(20\right)$

Wherein, Val represents the free variable (as material density, sulfur content and carbon residue content etc.) of specifying, Y ^keyproductrepresent critical product prophetic yields.

Table 7 device real-time yield prediction Delta-base value

Text	BA1 a	SP1 b	SU1 c	CN1 d	R1F e
						Free variable		1	1	1
Coker feedstock					1
						Coking dry gas	-0.0623	0.0000	0.0024	0.0007
Coker gasoline	-0.1630	0.0004	0.0352	0.0110
						Coker gas oil	-0.2810	0.0007	0.0587	0.0183
Wax oil	-0.2114	-0.002	-0.1595	-0.0497
						Coke	-0.2818	-0.0008	-0.0632	-0.0197
Loss	-0.0005	0.0000	0.0000	0.0000
						Material balance	-1.0000	0.0000

Note: a, BA1: benchmark

B, SP1: density

C, SU1: sulfur content

D, CN1: carbon residue

E, R1F: controlled condition

Table 7 is for free variable with material density, sulfur content and carbon residue content, respectively for the Delta_Base data that gasoline and aromatic hydrocarbons two schemes obtain, wherein BA represents the Base data of corresponding product yield, free variable 1 represents material density change 0.01, free variable 2 represents feed sulphur content change 1%, free variable 3 represents raw material carbon residue content change 1%, and under each variable change, the situation of change of corresponding critical product yield, is the Delta data of product yield.The computing formula of critical product yield is as follows:

Y _product＝C _product/C _Inlet(21)

Wherein, Y _productrepresent product yield, C _productrepresent the mass rate of product, C _inletrepresent the mass rate of feedstock.

By the carrying out of above step, the present invention can realize the delayed coking yield real-time estimate merging slag oil cracking reaction mechanism and device real time execution characteristic.The method with 11 lumped watershed hydrologic model models for theoretical foundation, actual industrial data are utilized to carry out real time correction to mechanism model, and the analysis data obtained by calculating calibration model carry out neural network training agent model, overcome the limitation that mechanism model computing velocity is slow.Utilize the yield of neural network agent model computing relay coker critical product, realize associating of operating conditions and Delta-Base data, reaching the effect of the real-time estimate of delayed coking unit yield, providing theory support for setting up accurately planning optimization PIMS model.

Claims (6)

1. a yield real-time predicting method in residual oil delayed coking, is characterized in that, comprise the following steps:

(1) utilize the data communication interface between VB.net Development of Software Platform refinery PHD real-time data base and mechanism model, realize the real-time data capture of residual oil delayed coking; Based on material balance, energy equilibrium and field instrument situation, data are in harmonious proportion, reject useless and wrong real time data;

(2) according to the device data of Real-time Collection, minimum as optimization aim using the difference of two squares of the model predication value of delayed coking reaction device outlet and actual value, intelligent optimization algorithm is utilized to solve, matching reactive kinetics parameters, realization mechanism model real time correction;

(3) on accurate model basis, analyze the critical product yield under different material quality, processing capacity and operating procedure condition, set up product yield database;

(4) the neural network agent model of product yield database training delayed coking reaction process is utilized, piece-wise linearization is carried out according to yield change trend curve, calculating Delta-Base data, providing theory support for setting up accurate planning optimization PIMS model.

2. yield real-time predicting method according to claim 1, it is characterized in that, the real time data gathered in step (1) is residual oil raw material feed loading, density, sulfur content, carbon residue content, reactor inlet temperature, reaction pressure and product flow information.

3. yield real-time predicting method according to claim 1, is characterized in that, described in step (2), device data refer to feed loading, temperature of reaction, pressure and product information.

4. yield real-time predicting method according to claim 1, is characterized in that, the optimized algorithm adopted in step (2) is the differential evolution algorithm with trigonometric mutation, and defines optimization aim and be:

$\min f\left(x\right)=\underset{i}{\mathrm{\Σ}}{\left(\frac{x_{\mathrm{actual}}-x_{\mathrm{calculate}}}{x_{\mathrm{actual}}}\right)}^2$

In formula: decision variable x comprises the Dynamics Factors of each reaction, x _actualand x _calculaterepresent product oil quality yield that is actual and that calculate respectively.

5. yield real-time predicting method according to claim 1, is characterized in that, step (3) Raw quality refers to material density, sulfur content and carbon residue content; Operating procedure condition refers to reactor inlet temperature and reaction pressure, and critical product then refers to rich gas, gasoline, diesel oil and wax oil.

6. yield real-time predicting method according to claim 1, is characterized in that, employs neural network model and carry out alternative mechanism model and carry out prediction and calculation in step (4); Wherein neural network model adopts reverse transmittance nerve network, chooses 6 neural network input variables: feed loading, density, sulfur content, carbon residue content, reactor inlet temperature, reaction pressure; 5 output variables: rich gas, gasoline, diesel oil, wax oil, coke; Hidden layer is set to 7 layers; Utilize agent model to carry out real-time estimate to delayed coking product yield, and obtain corresponding Delta-Base value.