CN106873395A - A kind of quick and various amount forecast Control Algorithm of oriented vinylalcohol pyrolysis furnace - Google Patents

Abstract

The present invention relates to a kind of quick and various amount forecast Control Algorithm of oriented vinylalcohol pyrolysis furnace, with offline and online two links.Offline link obtains the dynamic mathematical models and mathematics model of stable state of ethane cracking furnace production process by analysis or discrimination method, and sets control parameter.Measured value of the online link first according to cracking of ethylene furnace control system DCS calculates output predicted value, and the difference according to predicted value and measured value calculates predicated error so as to be corrected to stable state output predicted value；Then stable state control input predicted value is calculated by steady-state optimization or direct inversion technique；The control input value at current time is calculated finally according to given control parameter, DCS is passed to after being corrected for the dynamic property of control input value by lead-lag link.This invention simplifies the computation complexity of umlti-variable finite elements, the controlling cycle of umlti-variable finite elements is significantly decreased, can effectively solve the control problem of this kind of fast reaction process of ethane cracking furnace.

Description

A kind of quick and various amount forecast Control Algorithm of oriented vinylalcohol pyrolysis furnace

Technical field

The present invention relates to industrial process control field, more particularly to a kind of control method of oriented vinylalcohol pyrolysis furnace, specifically a kind of quick multi-variant control method.

Background technology

Ethylene unit is one of device of energy consumption maximum in petrochemical industry.Pyrolysis furnace is the key equipment of ethylene unit, is also the energy consumption rich and influential family of ethylene unit, and its energy consumption accounts for the 50%~60% of ethylene unit total energy consumption.The energy consumption for reducing pyrolysis furnace is one of important channel of reduction production cost of ethylene.Using Advanced Control Techniques, ethane cracking furnace is operated, it is possible to increase ethene, propylene receive solution stove and optimize control operation, be beneficial to reduce the energy consumption of ethylene unit.

Ethylene unit units such as cracking, compression, separation by being mainly made up of, and brief description of the process is as follows：

1st, operation is cracked

Receive from light hydrocarbon out-of-bounds, naphtha, send into pyrolysis furnace, plus dilution steam generation (DS) is cracked, the cracking gas for obtaining is (i.e.：The mixture of the components such as hydrogen, methane, ethene, ethane, propylene, propane, butadiene, drippolene, Pyrolysis fuel oil PFO), through waste heat boiler chilling, oil cooling, water-cooled to normal temperature, recovery section heat, and wherein most oils product separate after send into subsequent handling.Receive from the HP boiler water for out-of-bounds coming simultaneously and be translated into extra high pressure steam.

2nd, compression section

Future autothermic cracking operation cracking gas, compressed pressure-raising, for cryogenic separation provides condition.In compression process, cooling and separation, remove heavy hydrocarbon and water to cracking gas paragraph by paragraph, and are provided with alkali cleaning, the sour gas in cracking gas are removed, for piece-rate system provides qualified cracking gas.Refrigeration system is made up of propylene refrigeration system and ethene, methane binary refrigerating system, and the cryogen of different size is provided for cryogenic separation.Propylene, binary refrigerating system are multi-stage compression, the closed cycle system of multi-level throttle.

3rd, separation circuit

The cracking gas that compression section is come, through processes such as dehydration, deep cooling, hydrogenation and rectifying, obtains ethene, the propylene of high-purity, while obtaining byproduct H2, CH4, C3, LPG, mixing c4 fraction and drippolene.

Wherein, the technological process of cracking operation is as follows：

The hydrocarbon charging for coming from out-of-bounds has flow control to every boiler tube of pyrolysis furnace, hydrocarbon is preheated first in the topmost coil pipe of convection section, reinject dilution steam generation, the mixture of hydrocarbon and dilution steam generation is after convection section is further preheated, cracked into radiant coil, pyrolysis product is sent to waste heat boiler cooling, and waste heat boiler produces steam by being connected to the thermosiphon system of public drum.This enters drum after boiler feedwater is preheated in convection section boiler feedwater boiler tube.Saturated vapor is produced in waste heat boiler, is then overheated in two sections of superheater boiler tubes of convection section.Superheater outlet temperature is controlled by the way that boiler feedwater is injected into part superheated steam.Upper, in, pressure and temperature reducing is completed between the superheater boiler tube of bottom, steam returns to bottom superheater boiler tube, is finally superheated to design setting temperature.

Pyrolysis furnace uses ambient air as the source of oxygen of fuel combustion, and burner on sidewall (fixed combustion) is set according to basic load, then Bottom Nozzle Used is adjusted (with afterburning amount), to control coil outlet temperature.By the air mass flow for controlling the ventilation pressure of pyrolysis furnace to adjust Bottom Nozzle Used.Furthermore, it is possible to adjust the air plenum of single burner be run in its optimal ventilation pressure limit ensuring pyrolysis furnace.Burner hearth (ventilation) negative pressure is controlled by adjusting the rotating speed of air-introduced machine, and air-introduced machine can overcome the friction loss of flue gas flow in convection section to ensure the safe operation of pyrolysis furnace.

Ethylene unit employs many new technologies in technique and equipment aspect, and domestically leading level is reached at the aspect such as yield and plant energy consumption of target product.From in terms of the design and configuring condition of current classical control system, ethylene unit has used DCS centralized Controls, facilitates process operation, improves the reliability of plant running.In actual motion, these control systems can give play to good control effect, for the quiet run of device provides strong guarantee.Simultaneously, the pressure of the energy-saving and emission-reduction of enterprise, quality and efficiency to technological operation and process control propose requirement higher, and device inside complicated Matter Transfer and recycle heat, route of transmission and the mode of internal system disturbance are then increased, production steady for device and optimization operation propose more challenges.DCS control systems control Various Complex control strategy using such as serials control, Ratio control, selection, to improve device production stability.But, it is mainly the angle from single-input single-output object due to classical control system to consider a problem, it is difficult to process that there is multivariable, multiple constraint, the device of close coupling characteristic as pyrolysis furnace reactor etc., also advanced technologies thought and operating experience moment cannot be acted on process units, classical control system requires that high yield, low consumed multiobjective optimal control have larger difficulty for both maintaining device quiet run, again.Additionally, ethane cracking furnace is than conventional chemical process, its dynamic responding speed is very fast, and the calculated load of common Advanced process control method is all than larger, it is difficult to adapt to so quick dynamic response process.Therefore, further excavating gear potentiality, it is necessary to use higher level Optimal Control Strategy and quick algorithm, to meet the process control needs that steady control and edge optimize simultaneously.

The content of the invention

The multivariable existed for above-mentioned ethane cracking furnace production process, the problem for requiring quick response, the present invention proposes a kind of quick and various amount control method, for solving the control problem of ethane cracking furnace.

The technical scheme that is used to achieve the above object of the present invention is：A kind of quick and various amount forecast Control Algorithm of oriented vinylalcohol pyrolysis furnace, including offline link and online link；

Offline link：The open-loop test data of disturbance input variable, control input variable and controlled output variable according to ethane cracking furnace obtain the dynamic step response Mathematical Modeling and mathematics model of stable state of cracking of ethylene process；

Online link：Disturbance input variable, control input variable and the controlled output variable of ethane cracking furnace are read in real time, the stable state increment size of control input is obtained using dynamic step response Mathematical Modeling and mathematics model of stable state, then control input value is obtained by lead and lag correction, is sent to the control system of ethane cracking furnace.

The offline link is comprised the following steps：

(1) it is controlled output variable with the total treating capacity of pyrolysis furnace, the branch road 1COT temperature difference, the branch road 3COT temperature difference, the flow deviation of branch road 1, the flow deviation of branch road 3,1 and 3 liang of total COT, A room thinner ratio of branch road, the branch road 2COT temperature difference, the branch road 4COT temperature difference, the flow deviation of branch road 2, the flow deviation of branch road 4,2 and 4 liang of total COT, B room thinner ratios of branch road, oxygen content of smoke gas；

(2) it is control input variable with the hydrocarbon charging flow of branch road 1, the hydrocarbon charging flow of branch road 3, the dilution steam generation inlet amount of branch road 1, the dilution steam generation inlet amount of branch road 3, A rooms fuel atmospheric pressure, the hydrocarbon charging flow of branch road 2, the dilution steam generation inlet amount of 4 hydrocarbon charging flow branch road of branch road 2, the dilution steam generation inlet amount of branch road 4, B rooms fuel atmospheric pressure and combustion chamber draft；

(3) it is disturbance input variable with cracking furnace tube outlet pressure, calorific value of fuel gas/fuel air tightness, fuel atmospheric pressure, thinner ratio, feedstock property；

(4) the open-loop test data of above-mentioned disturbance input variable, control input variable and controlled output are read, the dynamic step response Mathematical Modeling and mathematics model of stable state for obtaining cracking of ethylene process is recognized by least square method；

(5) control parameter is set, including：Steady state time TTSS, control umber of beats M, time constant T₁And T₂。

The dynamic step response Mathematical Modeling：

Wherein,

a_ij=[a_ij(1) … a_ij(N)]^T, i=1 ..., p, that is, the quantity for being controlled output variable are p；J=1 ..., m, the i.e. quantity of control input variable are m；N is modeling time domain；a_ijDimension be N × 1；

b_il=[b_il(1) … b_il(N)]^T, l=1 ..., n, b_ilDimension be N × 1；

Represent in k moment whole controlled quentity controlled variable u₁,…u_mInitial prediction of the output variable at following N number of moment is controlled when keeping constant, is the vector that a dimension is (p × N) × 1,ElementRepresent to controlled output y_iIt is that a dimension is the vector of N × 1 in the initial prediction at following N number of moment,ElementRepresent y_iIn the initial prediction at following h-th moment；Represent in k moment whole controlled quentity controlled variable u₁,…u_mModel predication value of the output variable at following N number of moment is controlled when keeping constant, is the vector that a dimension is (p × N) × 1,ElementRepresent to controlled output y_iIt is that a dimension is the vector of N × 1 in the output predicted value at following N number of moment,ElementRepresent y_iIn the model predication value at following h-th moment, h=1 ... N；A is control input coefficient matrix, and B is disturbance input sytem matrix；Δ u (k) is k moment control input increment sizes, is the vector that a dimension is m × 1, its element Δ u_jK () represents control input u_jControl input increment size, u (k) represent k moment control input values, be a dimension be m × 1 vector；Δ d (k) is k moment disturbance input increment sizes, is the vector that a dimension is n × 1, element Δ d_lK () represents disturbance input d_lDisturbance input increment size, d (k) be k moment disturbance inputs, be a dimension be m × 1 vector.

The mathematics model of stable state is Δ y_ss(k)=K_uΔu_ss(k)+K_dΔd(k)；

Wherein, Δ y_ss(k)=y_ss(k)-y_ss(k-1), y_ssK () represents that the k moment is controlled the steady-state value of output, Δ y_ssK () represents that the k moment is controlled the stable state increment size of output, Δ u_ss(k)=u_ss(k)-u_ss(k-1), u_ssK () represents the steady-state value of k moment control inputs, Δ u_ssK () represents the stable state increment size of k moment control inputs, K_uIt is the steady state gain matrix of m × p, K_dIt is the steady state gain matrix of m × n：

And, steady state gain matrix K_uEach element and corresponding dynamic step-response coefficients model vector a_ijThere is k_u,ij=a_ij(N)；Steady state gain matrix K_dEach element and corresponding dynamic step-response coefficients model vector b_ilThere is k_d,il=b_il(N)。

The online link is comprised the following steps：

1) controlled output valve y (k), exogenous disturbances value d (k), control input value u (k-1) of ethane cracking furnace are obtained from the DCS of ethane cracking furnace；

2) by the output predicted value after controlled output valve y (k) to current time and correction of the last moment to current timeMake the difference dynamic error e (k) for obtaining current time：

Wherein,ElementRepresent to controlled output y_iLast moment to the output predicted value after the correction at current time；

3) the output predicted value at current time is calculated by dynamic step response Mathematical ModelingCurrent time output predicted value is corrected using dynamic error e (k), obtains the stable state output predicted value after current time correction

4) the stable state increment value Δ u of control input is calculated_ss(k)；

5) according to control input stable state increment value Δ u_ssK () and control umber of beats M calculate control input increment value Δ u before current time correction_T(k)=Δ u_ss(k)/M；

6) by lead and lag correction link, current time control input increment value Δ u (k) is obtained：

Δ u (k)=Transform (Δ u_T(k))

Wherein, Transform is lead and lag correction conversion, and specific continuous domain transformation model is

7) current time control input value u (k)=u (k-1)+Δ u (k) is calculated；

8) control input value u (k) is communicated the DCS for passing to ethane cracking furnace by OPC.

Output predicted value after correctionIt is the output predicted value after the correction at k-1 moment, is obtained from 1 moment to k-1 moment recursion by following recurrence formula：

Wherein, It is the output predicted value after current time correction；It is the output predicted value at current time, H is the unit matrix that a dimension is (p × N) × (p × N), initialization e (0), Afterwards each moment the output predicted value after the correction of subsequent time can be obtained with recursion, then be can obtain when k takes k-1 in above formula (k-1) (i-1)+1 element be

The stable state increment value Δ u for calculating control input_ss(k), when input and output number is equal and provides output set point y_TWhen, the stable state increment size of the control input

The stable state increment value Δ u for calculating control input_ss(k), when input and output number is unequal and provides output set point y_TWhen, the stable state increment value Δ u of the control input_ssK () is obtained by solving following steady-state optimization

Wherein, u_min、u_maxIt is the lower and upper limit of control input variable；Wherein, y_Tmin、y_TmaxTo be controlled the expectation target lower limit and the expectation target upper limit of output variable；Q is weighting matrix.

The stable state increment value Δ u for calculating control input_ss(k), when input and output number is unequal and provides output set point y_TAnd provide input set point u_TWhen, the stable state increment value Δ u of the control input_ssK () is obtained by solving following steady-state optimization

Wherein, u_Tmin、u_TmaxIt is the expectation target lower limit and the expectation target upper limit of control input variable, R is weighting matrix.

The stable state increment value Δ u for calculating control input_ss(k), when the requirement of no input and output set point only has input variable cost coefficient c requirements, the stable state increment value Δ u of the control input_ssK () is obtained by solving following steady-state optimization

Wherein, y_min、y_maxIt is the lower and upper limit of control input variable, c is the cost coefficient of control input variable.

The present invention has advantages below and beneficial effect：

1. the present invention proposes a kind of quick and various amount control method, on-line calculation and governing equation quantity can be greatly reduced, the solution of control law just becomes very simple and quick, smaller to time constant or very little production process can make quickly response, can effectively solve the problems, such as the multivariable Control of ethane cracking furnace.

2. the present invention can be to solve the stable state controlling increment of control input by the change of multivariable mathematics model of stable state and output set point, then can be solving corresponding dynamic control increment according to the stable state controlling increment.

3. the output stable state of Linear System process of the present invention is only related with final control input stable state to initial steady state operating point, so as to avoid other states of pilot process, avoiding each moment goes to calculate the complex matrix of big dimension, so as to greatly simplify process and save the calculating time.

4. this invention simplifies the computation complexity of umlti-variable finite elements, the controlling cycle of umlti-variable finite elements is significantly decreased, can effectively solve the control problem of this kind of fast reaction process of ethane cracking furnace.

Brief description of the drawings

Fig. 1 is umlti-variable finite elements flow chart.

Specific embodiment

Below in conjunction with the accompanying drawings and embodiment the present invention is described in further detail.

The present invention has offline and online two links.Offline link obtains the dynamic mathematical models and mathematics model of stable state of ethane cracking furnace production process by analysis or discrimination method, and sets control parameter.Measured value of the online link first according to cracking of ethylene furnace control system (DCS) calculates output predicted value, and the difference according to predicted value and measured value calculates predicated error so as to be corrected to stable state output predicted value；Then stable state control input predicted value is calculated by steady-state optimization or direct inversion technique；The control input value at current time is calculated finally according to given control parameter, DCS is passed to after being corrected for the dynamic property of control input value by lead-lag link.

As shown in figure 1, a kind of quick and various amount forecast Control Algorithm of oriented vinylalcohol pyrolysis furnace, including offline link and online link.

Offline link is comprised the following steps：

(1) it is controlled output variable with the total treating capacity of pyrolysis furnace, the branch road 1COT temperature difference, the branch road 3COT temperature difference, the flow deviation of branch road 1, the flow deviation of branch road 3,1/3 liang of total COT, A room thinner ratio (hydrocarbon/steam) of branch road, the branch road 2COT temperature difference, the branch road 4COT temperature difference, the flow deviation of branch road 2, the flow deviation of branch road 4,2/4 liang of total COT, B room thinner ratio (hydrocarbon/steam) of branch road, oxygen content of smoke gas；

(2) it is control input variable with the hydrocarbon charging flow of branch road 1, the hydrocarbon charging flow of branch road 3, the dilution steam generation inlet amount of branch road 1, the dilution steam generation inlet amount of branch road 3, A rooms fuel atmospheric pressure, the hydrocarbon charging flow of branch road 2, the dilution steam generation inlet amount of 4 hydrocarbon charging flow branch road of branch road 2, the dilution steam generation inlet amount of branch road 4, B rooms fuel atmospheric pressure and combustion chamber draft；

(3) it is exogenous disturbances variable with cracking furnace tube outlet pressure, calorific value of fuel gas/fuel air tightness, fuel atmospheric pressure, thinner ratio, feedstock property；

(4) by OPC (OLE for Process Control, for the OLE of process control) above-mentioned disturbance input variable, control input variable and the controlled open-loop test data for exporting are read, the dynamic step response Mathematical Modeling and mathematics model of stable state for obtaining cracking of ethylene process are recognized by least square method.

(5) control parameter of quick and various amount PREDICTIVE CONTROL is set, including：Steady state time (TTSS), control umber of beats (M), time constant T₁And T₂, wherein,

Steady state time (TTSS)：System reaches the time that stable state is experienced again in the presence of Stepped Impedance Resonators；

Control umber of beats (M)：One control input is decomposed into some umber of beats to go to implement, signified umber of beats is exactly to control umber of beats；

Time constant T₁And T₂：Lead-lag linkMiddle parameter.

It is described that the dynamic step response Mathematical Modeling for obtaining cracking of ethylene process is recognized by least square method, specially：

By carrying out process Open loop step response test to ethane cracking furnace, the data for measuring obtain the output of process y via least square method_iTo control input variable u_jStep-response coefficients a_ijValue composition model vector on sampled point：

a_ij=[a_ij(1) … a_ij(N)]^T

Wherein, i=1, L, p, that is, the quantity for being controlled output variable are p；J=1, L, m, the i.e. quantity of control input variable are m；N is modeling time domain；a_ijDimension be N × 1；

The step-response coefficients model vector that exogenous disturbances are controlled output to process can equally be measured：

b_il=[b_il(1) … b_il(N)]^T

Wherein, l=1, L, n, b_ilDimension be N × 1；

Comprehensive Control is input into and exogenous disturbances, and the dynamic step response Mathematical Modeling of cracking of ethylene process can be expressed as：

Wherein,

Represent in k moment whole controlled quentity controlled variable u₁,…u_mInitial prediction of the output variable at following N number of moment is controlled when keeping constant, is the vector that a dimension is (p × N) × 1,ElementRepresent to controlled output y_iIt is that a dimension is the vector of N × 1 in the initial prediction at following N number of moment,ElementRepresent y_iIn the initial prediction at following h-th moment；Represent in k moment whole controlled quentity controlled variable u₁,…u_mModel predication value of the output variable at following N number of moment is controlled when keeping constant, is the vector that a dimension is (p × N) × 1,ElementRepresent to controlled output y_iIt is that a dimension is the vector of N × 1 in the model predication value at following N number of moment,ElementRepresent y_iIn the model predication value at following h-th moment；A is control input coefficient matrix, and B is disturbance input sytem matrix；Δ u (k) is k moment control input increment sizes, is the vector that a dimension is m × 1, its element Δ u_jK () represents control input u_jControl input increment size, u (k) represent k moment control input values, be a dimension be m × 1 vector；Δ d (k) is k moment disturbance input increment sizes, is the vector that a dimension is n × 1, element Δ d_lK () represents disturbance input d_lDisturbance input increment size, d (k) be k moment disturbance inputs, be a dimension be m × 1 vector.

The mathematics model of stable state is

Δy_ss(k)=K_uΔu_ss(k)+K_dΔd(k)

Wherein, Δ y_ss(k)=y_ss(k)-y_ss(k-1), y_ssK () represents that the k moment is controlled the steady-state value of output, Δ y_ssK () represents that the k moment is controlled the stable state increment size of output, Δ u_ss(k)=u_ss(k)-u_ss(k-1), u_ssK () represents the steady-state value of k moment control inputs, Δ u_ssK () represents the stable state increment size of k moment control inputs, K_uIt is the steady state gain matrix of m × p, K_dIt is the steady state gain matrix of m × n：

And, steady state gain matrix K_uEach element and corresponding dynamic step-response coefficients model vector a_ijThere is k_u,ij=a_ij(N)；Steady state gain matrix K_dEach element and corresponding dynamic step-response coefficients model vector b_ilThere is k_d,il=b_il(N)。

The online link is comprised the following steps：

(1) controlled output valve y (k), exogenous disturbances value d (k), control input value u (k-1) that ethane cracking furnace is obtained from the DCS of ethane cracking furnace are communicated by OPC；

(2) dynamic error for obtaining current time is made the difference by the output predicted value after the controlled output valve to current time and correction of the last moment to current time：

Wherein, e (k) is current time dynamic error, and y (k) is the controlled output valve at current time,Output predicted value after correction for last moment to current time,ElementRepresent to controlled output y_iLast moment to the output predicted value after the correction at current time；

(3) the output predicted value at current time is calculated by dynamic step response Mathematical ModelingCurrent time output predicted value is corrected using dynamic error e (k), obtains the stable state output predicted value after current time correction

(4) the stable state increment value Δ u of control input is calculated_ss(k)；

(5) according to control input stable state increment value Δ u_ssK () and control parameter M calculate control input increment value Δ u before current time correction_T(k)：

Δu_T(k)=Δ u_ss(k)/M

M is the parameter that user gives according to the requirement of control input argument action speed, will control input increment Delta u_ssK () is equally divided into M parts, obtain control input value increment before the correction at current time；

(6) by lead and lag correction link, current time control input increment value Δ u (k) is obtained：

Δ u (k)=Transform (Δ u_T(k))

Wherein, Transform is lead and lag correction conversion, and specific continuous domain transformation model isNeed its discretization herein.Different T₁、T₂Selection can be realized to Δ u_TThe advanced or correction or lag of (k).

(7) current time control input value u (k) is calculated：

U (k)=u (k-1)+Δ u (k)

(8) control input value u (k) is communicated the DCS (Distributed Control System, dcs) for passing to ethane cracking furnace by OPC.

Output predicted value after the correction is：

Wherein, It is the output predicted value after current time correction；It is the output predicted value at current time, H is the unit matrix that a dimension is (p × N) × (p × N), initialization e (0), Afterwards each moment the output predicted value after the correction of subsequent time can be obtained with recursion.

When input and output number is equal and provides output set point y_TWhen, described stable state control input increment is

When input and output number is unequal and provides output set point y_TWhen, described stable state control input increment is obtained by solving following steady-state optimization

Wherein, u_min、u_maxIt is the lower and upper limit of control input variable；Wherein, y_Tmin、y_TmaxTo be controlled the expectation target lower limit and the expectation target upper limit of output variable；Q is weighting matrix.

When input and output number is unequal and provides output set point y_TAnd provide input set point u_TThe stable state control input increment of Shi Suoshu is obtained by solving following steady-state optimization

Wherein, u_Tmin、u_TmaxIt is the expectation target lower limit and the expectation target upper limit of control input variable.R is weighting matrix.

When the requirement of no input and output set point only has input variable cost coefficient c requirements, described stable state control input increment is obtained by solving following steady-state optimization

Wherein, y_min、y_maxIt is the lower and upper limit of control input variable, c is the cost coefficient of control input variable.

Present invention specific implementation is as follows, and workflow is divided into offline and online two links：

Wherein, offline link is comprised the following steps：

It is controlled output variable with the total treating capacity of ethane cracking furnace (entering the cracked gas of ethane cracking furnace), the branch road 1COT temperature difference, branch road 3COT (control of temprature) temperature difference, the flow deviation of branch road 1, the flow deviation of branch road 3,1/3 liang of total COT, A room thinner ratio (hydrocarbon/steam) of branch road, the branch road 2COT temperature difference, the branch road 4COT temperature difference, the flow deviation of branch road 2, the flow deviation of branch road 4,2/4 liang of total COT, B room thinner ratio (hydrocarbon/steam) of branch road, oxygen content of smoke gas；The quantity for controlling output variable is 14, i.e. parameter p=14.

It is control input variable with the hydrocarbon charging flow of branch road 1, the hydrocarbon charging flow of branch road 3, the dilution steam generation inlet amount of branch road 1, the dilution steam generation inlet amount of branch road 3, A rooms fuel atmospheric pressure, the hydrocarbon charging flow of branch road 2, the dilution steam generation inlet amount of 4 hydrocarbon charging flow branch road of branch road 2, the dilution steam generation inlet amount of branch road 4, B rooms fuel atmospheric pressure and combustion chamber draft；The quantity of control input variable is 10, i.e. m=10.

With cracking furnace tube outlet pressure, calorific value of fuel gas/fuel air tightness, fuel atmospheric pressure, thinner ratio, feedstock property (ratio shared by each constituent of alkane, cycloalkane, aromatic hydrocarbon) for exogenous disturbances variable；Disturbance input variable is 5, i.e. n=5.

The open-loop test data of above-mentioned disturbance input variable, control input variable and controlled output are read by OPC, the dynamic step response Mathematical Modeling and mathematics model of stable state for obtaining cracking of ethylene process are recognized by least square method.Here, suppose that the coefficient number of step response Mathematical Modeling is 30, Sampling period is one minute, then model time domain N=30, steady state time TTSS=30 minutes.Control umber of beats M=5, T₁=10 minutes, T₂=20 minutes.

By carrying out process step response test to ethane cracking furnace, the data for measuring obtain the output of process y via least square method_iTo control input variable u_jStep response a_ijValue composition model vector on sampled point：

a_ij=[a_ij(1) … a_ij(30)]^T

Wherein, i=1 ..., 14, j=1 ..., 10, a_ijDimension be 30 × 1；

The step-response coefficients model vector that exogenous disturbances are controlled output to process can equally be measured：

b_il=[b_il(1) … b_il(N)]^T

Wherein, l=1 ..., 5, b_ilDimension be 30 × 1.

Comprehensive Control is input into and exogenous disturbances, and the dynamic step response Mathematical Modeling of cracking of ethylene process can be expressed as：

Wherein,

Represent in k moment whole controlled quentity controlled variable u₁,…u_mOutput variable is controlled when keeping constant in following 30 initial predictions at moment, is the vector that a dimension is 140 × 1,ElementRepresent to controlled output y_iIt is vector that a dimension is 30 × 1 in following 30 initial predictions at moment,ElementRepresent_yiIn the initial prediction at following h-th moment；Represent in k moment whole controlled quentity controlled variable u₁,…u_mOutput variable is controlled when keeping constant in following 30 model predication values at moment, is the vector that a dimension is 140 × 1,ElementRepresent to controlled output y_iIt is vector that a dimension is 30 × 1 in following 30 model predication values at moment,ElementRepresent y_iIn the model predication value at following h-th moment；A is control input coefficient matrix, and B is disturbance input sytem matrix；Δ u (k) is k moment control input increment sizes, is the vector that a dimension is 10 × 1, its element Δ u_jK () represents control input u_jControl input increment size, u (k) represent k moment control input values, be a dimension be 10 × 1 vector；Δ d (k) is k moment disturbance input increment sizes, is the vector that a dimension is 5 × 1, element Δ d_lK () represents disturbance input d_lDisturbance input increment size, d (k) be k moment disturbance inputs, be a dimension be 10 × 1 vector.

Control input is to the steady-state model of controlled output：

Δy_ss=K_uΔu_ss+K_dΔd

Wherein, Δ y_ss(k)=y_ss(k)-y_ss(k-1), y_ssK () represents that the k moment is controlled the steady-state value of output, Δ y_ssK () represents that the k moment is controlled the stable state increment size of output, Δ u_ss(k)=u_ss(k)-u_ss(k-1), u_ssK () represents the steady-state value of k moment control inputs, Δ u_ssK () represents the stable state increment size of k moment control inputs, K_uIt is 14 × 10 matrix, K_dIt is 14 × 5 matrix.

Steady state gain matrix K_uIt is described as follows

Steady state gain matrix K_uEach element and corresponding dynamic step-response coefficients model vector a_ijThere is relation：k_u,ij=a_ij(30)。

Steady state gain matrix K_dIt is described as follows

Steady state gain matrix K_dEach element and corresponding dynamic step-response coefficients model vector b_ilThere is relation：k_d,il=b_il(30)。

Online link is comprised the following steps：

(1) controlled output valve y (k), exogenous disturbances value d (k), control input value u (k-1) that ethane cracking furnace is obtained from the DCS of ethane cracking furnace are communicated by OPC；

(2) dynamic error for obtaining current time is made the difference by the output predicted value after the controlled output valve to current time and correction of the last moment to current time：

Wherein, e (k) is current time dynamic error, and y (k) is the controlled output valve at current time,Output predicted value after correction for last moment to current time,ElementRepresent to controlled output y_iLast moment to the output predicted value after the correction at current time；

Output predicted value after wherein correctingCalculated by following formula：

Wherein, It is the output predicted value after current time correction；It is the output predicted value at current time, H is the unit matrix that a dimension is (p × N) × (p × N), initialization e (0), Afterwards each moment the output predicted value after the correction of subsequent time can be obtained with recursion；

(3) the output predicted value at current time is calculated by dynamic step response Mathematical ModelingCurrent time output predicted value is corrected using dynamic error e (k), obtains the stable state output predicted value after current time correction

(4) the stable state increment value Δ u of control input is calculated_ss(k), specially：

Because of control input variable number in the present embodiment and controlled output variable number, control system does not have input and output set point, the cost coefficient of control input variable is only gived, then the stable state increment size of control input can be obtained by solving following steady-state optimization：

(5) according to control input stable state increment value Δ u_ssK () and control parameter M calculate control input increment value Δ u before current time correction_T(k)：

Δu_T(k)=Δ u_ss(k)/M

M is the parameter that user gives according to the requirement of control input argument action speed, will control input increment Delta u_ssK () is equally divided into M parts, obtain control input value increment before the correction at current time；

(6) by lead and lag correction link, current time control input increment value Δ u (k) is obtained：

Δ u (k)=Transform (Δ u_T(k))

Wherein, Transform is lead and lag correction conversion, and specific continuous domain transformation model isNeed its discretization herein.

(7) current time control input value u (k) is calculated：

U (k)=u (k-1)+Δ u (k)

(8) control input value u (k) is communicated the DCS (Distributed Control System, dcs) for passing to ethane cracking furnace by OPC.

Claims (10)

1. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace, it is characterised in that including offline link With online link；

Offline link：Disturbance input variable, control input variable and controlled output according to ethane cracking furnace become The open-loop test data of amount obtain the dynamic step response Mathematical Modeling and stable state mathematical modulo of cracking of ethylene process Type；

Online link：The disturbance input variable of ethane cracking furnace, control input variable and controlled defeated are read in real time Go out variable, the stable state increment of control input is obtained using dynamic step response Mathematical Modeling and mathematics model of stable state Value, then obtains control input value by lead and lag correction, sends to the control system of ethane cracking furnace.

2. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 1, it is special Levy and be, the offline link is comprised the following steps：

(1) with the total treating capacity of pyrolysis furnace, the branch road 1COT temperature difference, the branch road 3COT temperature difference, the flow of branch road 1 Deviation, the flow deviation of branch road 3,1 and 3 liang of total COT, A room thinner ratio of branch road, the branch road 2COT temperature difference, The branch road 4COT temperature difference, the flow deviation of branch road 2, the flow deviation of branch road 4,2 and 4 liang of branch roads total COT, B Room thinner ratio, oxygen content of smoke gas are controlled output variable；

(2) with the hydrocarbon charging flow of branch road 1, the hydrocarbon charging flow of branch road 3, the dilution steam generation inlet amount of branch road 1, The dilution steam generation inlet amount of branch road 3, A rooms fuel atmospheric pressure, the hydrocarbon charging flow of branch road 2, the hydrocarbon charging stream of branch road 4 Amount branch road 2 dilution steam generation inlet amount, the dilution steam generation inlet amount of branch road 4, B rooms fuel atmospheric pressure and combustion chamber draft It is control input variable；

(3) with cracking furnace tube outlet pressure, calorific value of fuel gas/fuel air tightness, fuel atmospheric pressure, dilute Ratio, feedstock property are released for disturbance input variable；

(4) the open-loop test data of above-mentioned disturbance input variable, control input variable and controlled output are read, The dynamic step response Mathematical Modeling and stable state mathematics for obtaining cracking of ethylene process are recognized by least square method Model；

(5) control parameter is set, including：Steady state time TTSS, control umber of beats M, time constant T₁And T₂。

3. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 2, it is special Levy and be, the dynamic step response Mathematical Modeling：

${\tilde{y}}_{N1}\left(k\right)={\tilde{y}}_{N0}\left(k\right)+A\mathrm{\Δ}u\left(k\right)+B\mathrm{\Δ}d\left(k\right)$

Wherein,

${\tilde{y}}_{N1}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{1,N1}\left(k\right)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,N1}\left(k\right)\end{array}\right],{\tilde{y}}_{i,N1}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{i,1}(k+1|k)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{i,1}(k+N|k)\end{array}\right],{\tilde{y}}_{N0}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{1,N0}\left(k\right)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,N0}\left(k\right)\end{array}\right],{\tilde{y}}_{i,N0}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{i,0}(k+1|k)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{i,0}(k+N|k)\end{array}\right],$

$A=\left[\begin{array}{ccc}{a}_{11}& ...& a_{1m}\\ \·& & \·\\ \·& & \·\\ \·& & \·\\ a_{p1}& ...& a_{pm}\end{array}\right],\mathrm{\Δ}u\left(k\right)=\left[\begin{array}{c}{\mathrm{\Δu}}_1\left(k\right)\\ \·\\ \·\\ \·\\ {\mathrm{\Δu}}_m\left(k\right)\end{array}\right]=u\left(k\right)-u(k-1),u\left(k\right)=\left[\begin{array}{c}{u}_1\left(k\right)\\ \·\\ \·\\ \·\\ u_m\left(k\right)\end{array}\right],$

$B=\left[\begin{array}{ccc}{b}_{11}& ...& b_{1n}\\ \·& & \·\\ \·& & \·\\ \·& & \·\\ b_{p1}& ...& b_{pn}\end{array}\right],\mathrm{\Δ}d\left(k\right)=\left[\begin{array}{c}{\mathrm{\Δd}}_1\left(k\right)\\ \·\\ \·\\ \·\\ {\mathrm{\Δd}}_n\left(k\right)\end{array}\right]=d\left(k\right)-d(k-1),d\left(k\right)=\left[\begin{array}{c}{d}_1\left(k\right)\\ \·\\ \·\\ \·\\ d_n\left(k\right)\end{array}\right]$

a_ij=[a_ij(1) … a_ij(N)]^T, i=1 ..., p, that is, the quantity for being controlled output variable are p； J=1 ..., m, the i.e. quantity of control input variable are m；N is modeling time domain；a_ijDimension be N × 1；

b_il=[b_il(1) … b_il(N)]^T, l=1 ..., n, b_ilDimension be N × 1；

Represent in k moment whole controlled quentity controlled variable u₁,…u_mOutput variable is controlled when keeping constant N number of in future The initial prediction at moment, is that a dimension is the vector of (p × N) × 1,ElementIt is right to represent Controlled output y_iIt is that a dimension is the vector of N × 1 in the initial prediction at following N number of moment,'s ElementRepresent_yiIn the initial prediction at following h-th moment；Represent complete at the k moment Portion controlled quentity controlled variable u₁,…u_mModel predication value of the output variable at following N number of moment is controlled when keeping constant, is one Dimension is the vector of (p × N) × 1,ElementRepresent to controlled output y_iAt following N number of moment Output predicted value, be a dimension be N × 1 vector,ElementRepresent y_iNot Come h-th model predication value at moment, h=1 ... N；A is control input coefficient matrix, and B is disturbance input system System matrix；Δ u (k) is k moment control input increment sizes, is the vector that a dimension is m × 1, its element Δ u_j(k) Represent control input u_jControl input increment size, u (k) represent k moment control input values, be that a dimension is The vector of m × 1；Δ d (k) is k moment disturbance input increment sizes, is the vector that a dimension is n × 1, element Δd_lK () represents disturbance input d_lDisturbance input increment size, d (k) be k moment disturbance inputs, be a dimension It is the vector of m × 1.

4. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 2, it is special Levy is that the mathematics model of stable state is

Δy_ss(k)=K_uΔu_ss(k)+K_dΔd(k)

Wherein, Δ y_ss(k)=y_ss(k)-y_ss(k-1), y_ssK () represents that the k moment is controlled the steady-state value of output, Δ y_ss(k) Represent that the k moment is controlled the stable state increment size of output, Δ u_ss(k)=u_ss(k)-u_ss(k-1), u_ssK () represents that the k moment is controlled Make the steady-state value of input, Δ u_ssK () represents the stable state increment size of k moment control inputs, K_uIt is the stable state of m × p Gain matrix, K_dIt is the steady state gain matrix of m × n：

And, steady state gain matrix K_uEach element and corresponding dynamic step-response coefficients model vector a_ijHave k_u,ij=a_ij(N)；Steady state gain matrix K_dEach element and corresponding dynamic step-response coefficients model vector b_il There is k_d,il=b_il(N)。

5. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 1, its It is characterised by, the online link is comprised the following steps：

1) from the DCS of ethane cracking furnace obtain controlled output valve y (k) of ethane cracking furnace, exogenous disturbances value d (k), Control input value u (k-1)；

2) it is defeated after the correction by controlled output valve y (k) to current time with last moment to current time Go out predicted valueMake the difference dynamic error e (k) for obtaining current time：

$e\left(k\right)=\left[\begin{array}{c}{e}_1\left(k\right)\\ \·\\ \·\\ \·\\ e_p\left(k\right)\end{array}\right]=y\left(k\right)-{\tilde{y}}_{Cor}\left(k\right|k-1),{\tilde{y}}_{Cor}\left(k\right|k-1)=\left[\begin{array}{c}{\tilde{y}}_{1,Cor}\left(k\right|k-1)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,Cor}\left(k\right|k-1)\end{array}\right]$

Wherein,ElementRepresent to controlled output y_iLast moment to it is current when Output predicted value after the correction at quarter；

3) the output predicted value at current time is calculated by dynamic step response Mathematical ModelingUsing dynamic State error e (k) is corrected to current time output predicted value, obtains the stable state output after current time correction pre- Measured value

${\tilde{y}}_{Cor}(k+N|k)=\left[\begin{array}{c}{\tilde{y}}_{1,Cor}(k+N|k)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,Cor}(k+N|k)\end{array}\right]=\left[\begin{array}{c}{\tilde{y}}_{1,N1}(k+N|k)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,N1}(k+N|k)\end{array}\right]+\left[\begin{array}{c}{e}_1\left(k\right)\\ \·\\ \·\\ \·\\ e_p\left(k\right)\end{array}\right].$

4) the stable state increment value Δ u of control input is calculated_ss(k)；

5) according to control input stable state increment value Δ u_ssK () and control umber of beats M calculate control before current time correction Input increment value Δ u_T(k)=Δ u_ss(k)/M；

6) by lead and lag correction link, current time control input increment value Δ u (k) is obtained：

Δ u (k)=Transform (Δ u_T(k))

Wherein, Transform is lead and lag correction conversion, and specific continuous domain transformation model is

7) current time control input value u (k)=u (k-1)+Δ u (k) is calculated；

8) control input value u (k) is communicated the DCS for passing to ethane cracking furnace by OPC.

6. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 5, its It is characterised by, the output predicted value after the correctionFor the output after the correction at k-1 moment is pre- Measured value, is obtained by following recurrence formula from 1 moment to k-1 moment recursion：

${\tilde{y}}_{Cor}\left(k\right)={\tilde{y}}_{N1}\left(k\right)+He\left(k\right)$

${\tilde{y}}_{Cor}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{1,Cor}\left(k\right)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,Cor}\left(k\right)\end{array}\right],{\tilde{y}}_{i,Cor}\left(k\right)=\left[\begin{array}{c}{\tilde{y}}_{1,Cor}(k+1|k)\\ \·\\ \·\\ \·\\ {\tilde{y}}_{p,Cor}(k+N|k)\end{array}\right]$

Wherein, It is the output predicted value after current time correction； It is the output predicted value at current time, H is the unit matrix that a dimension is (p × N) × (p × N), initialization e (0),Afterwards each moment the output predicted value after the correction of subsequent time can be obtained with recursion Arrive, then be can obtain when k takes k-1 in above formula(k-1) (i-1)+1 element As

7. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 5, its It is characterised by the stable state increment value Δ u of the calculating control input_ss(k), when input and output number is equal and provides defeated Go out set point y_TWhen, the stable state increment size of the control input

8. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 5, its It is characterised by the stable state increment value Δ u of the calculating control input_ss(k), when input and output number is unequal and is given Output set point y_TWhen, the stable state increment value Δ u of the control input_ssK () is obtained by solving following steady-state optimization

$\left\{\begin{array}{c}\underset{{\mathrm{\Δu}}_{ss}\left(k\right)}{min}J=\left|\right|y_{ss}\left(k\right)-y_T|{|}_Q^2\\ s.t.\\ y_{ss}\left(k\right)={\tilde{y}}_{Cor}(k+N|k)+K_u\mathrm{\Δ}{u}_{ss}(k)+K_d\mathrm{\Δ}d(k)\\ u_{\min }\≤u_{ss}(k-1)+\mathrm{\Δ}{u}_{ss}(k-1)\≤u_{max}\\ y_{T\min }\≤{\mathrm{\Δy}}_{ss}\left(k\right)\≤y_{T\max }\end{array}\right.$

Wherein, u_min、u_maxIt is the lower and upper limit of control input variable；Wherein, y_Tmin、y_TmaxIt is controlled The expectation target lower limit and the expectation target upper limit of output variable；Q is weighting matrix.

9. the quick and various amount forecast Control Algorithm of a kind of oriented vinylalcohol pyrolysis furnace according to claim 5, its It is characterised by the stable state increment value Δ u of the calculating control input_ss(k), when input and output number is unequal and is given Output set point y_TAnd provide input set point u_TWhen, the stable state increment value Δ u of the control input_ssK () passes through Following steady-state optimization is solved to obtain

$\left\{\begin{array}{c}\underset{{\mathrm{\Δu}}_{ss}\left(k\right)}{min}J=\left|\right|y_{ss}\left(k\right)-y_T|{|}_Q^2+|\left|u_{ss}\right(k)-u_T|{|}_R^2\\ s.t.\\ y_{ss}\left(k\right)={\tilde{y}}_{Cor}(k+N|k)+K_u\mathrm{\Δ}{u}_{ss}(k)+K_d\mathrm{\Δ}d(k)\\ u_{T\min }\≤u_{ss}(k-1)+\mathrm{\Δ}{u}_{ss}(k-1)\≤u_{Tmax}\\ y_{T\min }\≤y_{ss}\left(k\right)\≤y_{T\max }\end{array}\right.$

Wherein, u_Tmin、u_TmaxIt is the expectation target lower limit and the expectation target upper limit of control input variable, R is to add Weight matrix.

10. a kind of quick and various amount control method according to claim 5, it is characterised in that the calculating control Make the stable state increment value Δ u of input_ss(k), when the requirement of no input and output set point only has input variable cost coefficient During c requirements, the stable state increment value Δ u of the control input_ssK () is obtained by solving following steady-state optimization

Wherein, y_min、y_maxIt is the lower and upper limit of control input variable, c is the cost system of control input variable Number.