A kind of cement rotary kiln sintering process twice optimum method
Technical field
The invention belongs to automatic industrial production process control fields, and it is secondary to be related to a kind of cement rotary kiln sintering process
Optimal control method.
Background technique
In actual industrial production, industrial processes modeling and control strategy are just become more and more important, while by
It is more stringent in the operations specification that the high request to product quality and operational safety produces industrial processes, and then lead
The probability of control system failure is caused to increase.For example, the kiln hood of cement rotary kiln sprays in actual cement rotary kiln sintering process
The failure of coal actuator is especially common, is mainly reflected in kiln hood coal powder injection failsafe valve, this will make final clinker
Quality decline, at the same in view of actuator power-off, blocking in the case where, cement rotary kiln sintering process will be no longer controllable, according to this
Designing controller will be meaningless.Therefore, it is necessary to propose a kind of control method come the case where handling part actuator failures, to protect
The stability for hindering cement rotary kiln sintering process, ensures the high standard of final clinker, the steady production of high quality.
Summary of the invention
Object of the present invention is to the influences to improve cement rotary kiln sintering process control system part actuator failures, propose
A kind of cement rotary kiln sintering process twice optimum method.This method is established by acquisition inputoutput data first
Then process status is changed and exports tracking error group being combined into new process status amount, further established with this by process model
New process model finally designs controller with quadratic objective function, designs optimal more new law.
The technical scheme is that establishing a kind of water by means such as data acquisition, model foundation, controller designs
Mud rotary kiln sintering process twice optimum method, can effectively ensure that the stabilization of cement rotary kiln sintering process using this method
With the faults-tolerant control of optimal control performance and realization system.
Method and step of the invention includes:
Step 1 establishes industrial processes single-input single-output model, comprises the concrete steps that:
1-1. acquires the real-time running data of industrial processes first, establishes an industrial processes system model.
Industrial processes under indefinite interference are described as following form:
Wherein k indicates process operation moment, y (z), u (z), and e (z) respectively indicates the output of process y (k) at k moment, process
It inputs u (k), the form after the z-transform of indefinite interference w (k).Δ is difference operator, A (z-1), B (z-1), C (z-1) it is suitable respectively
When the corresponding multinomial of dimension.
The part 1-2. actuator failures are described as follows:
uF(k)=α u (k)
Wherein uF(k) be the k moment actuator practical control action, α indicate actuator failures influence degree.
1-3. is according to step 1-1,1-2, then the industrial processes model with part actuator failures is described as follows:
Wherein uF(z) u is indicatedF(k) differentiated form.
1-4. obtains following discrete transfer function form industrial processes model according to 1-3:
y(k+1)+F1y(k)+…+Fpy(k-p+1)
=H1u(k)+H2u(k-1)+…+Hqu(k-q+1)
Wherein y (k+1), y (k) ..., y (k-p+1) indicate k+1, k ..., the output of process at k-p+1 moment, u (k+1), u
(k) ..., u (k-q+1) indicates k+1, k ..., the process input at k-q+1 moment, F1,…,Fp, H1,…,HqRespectively indicate correspondence
The model coefficient of the output of process and input, p, q are the order of corresponding process output and input respectively.
Difference operator is added to above-mentioned process model and defines a new vector as follows:
Δxm(k)T=[Δ y (k) ... Δ y (k-p+1) Δ u (k-1) ... Δ u (k-q+1)]
Wherein m representation dimension, m=dim (Δ xm)=p+q-1, Δ xm(k)TIndicate turning for the state increment of k moment m dimension
It sets, Δ y (k+1), Δ y (k) ..., Δ y (k-p+1) indicates k+1, k ..., the output of process increment at k-p+1 moment, Δ u (k+
1), Δ u (k) ..., Δ u (k-q+1) indicate k+1, and the process of k ..., k-q+1 moment input increment.
1-5. further obtains differentiated process model according to step 1-4 are as follows:
Δxm(k+1)=AmΔxm(k)+BmΔu(k)
Δ y (k+1)=CmΔxm(k+1)
Wherein Δ xm(k),Δxm(k+1) k, the state increment of k+1 moment m dimension, when Δ y (k+1) indicates k+1 are respectively indicated
The output of process increment at quarter.H2,…,Hm-1,HmIndicate differentiated model coefficient.
Bm=[H1 0 0 … 0 10 0]T,Cm=[1 00 ... 000 0]
1-6. definition output tracking error are as follows:
E (k)=y (k)-r (k)
Wherein e (k) indicates that the output tracking error at k moment, r (k) indicate the setting value at k moment.
The dynamic process for further obtaining output tracking error is expressed as follows:
E (k+1)=e (k)+CmAmΔxm(k)+CmBmΔu(k)
Wherein e (k+1) indicates the output tracking error at k+1 moment.
It is as follows to finally obtain new process status spatial model according to above step by 1-7.:
Z (k+1)=Az (k)+B Δ u (k)
Wherein z (k)=[Δ xm(t) e(t)]T, z (k), z (k+1) respectively indicate k, the new process status at k+1 moment.
Step 2, design industrial processes controller, specifically:
2-1. is based on step 1, and the objective function for selecting industrial processes twice optimum is following form:
Wherein J is objective function, and Q, R are the weighting matrix of corresponding new process status and input increment respectively, and Q is pair
Angle battle array, Q=diag { qj1,qj2,…,qjp+q-1,qje, qj1,qj2,…,qjp+q-1It is the output of process variation and input variation respectively
Weighting parameters, qjeIt is the weighting parameters for exporting tracking error.
The minimization of object function of 2-2. solution procedure 2-1 obtains the optimal of industrial processes twice optimum device
Controlling increment are as follows:
Δ u (k)=- R-1BT[I+K$BR-1BT]-1K$Az(k)
Wherein K$=AT[I+K$BR-1BT]-1K$A+Q=ATK$A-ATK$B(R+BTK$B)-1BTK$A+Q,
K$For the solution for meeting above-mentioned Riccati equation.
2-3. repeats step 1.6 to 2.2 and continues to solve new optimal more new law, obtain optimum control increasing in subsequent time
It measures Δ u (k), acts on control object, and circuit sequentially.
The invention has the advantages that: be different from traditional control method, the present invention by interprocedual introduce state change and
Tracking error is exported, the adjusting of controller is more flexible, so that Fault Tolerance is promoted, part actuator failures influence
To improve.
Specific embodiment
By taking cement rotary kiln sintering process as an example:
In cement process production process, cement rotary kiln sintering process is the important ring in manufacture of cement.Cement is raw
After the completion of material preparation, cement slurry goes successively to cement rotary kiln, and the coal powder injection kiln hood of rotary kiln starts to rotary kiln coal powder injection at this time,
Rotary kiln is heated, and clinker is reacted, and as the temperature of rotary kiln clinkering zone rises to a certain extent, cement slurry is gradually
It is transformed into clinker.
Step 1 establishes cement rotary kiln sintering process single-input single-output model, comprises the concrete steps that:
1-1. acquires the real-time running data of cement rotary kiln sintering process first, establishes a cement rotary kiln and was burnt into
Journey system model.Cement rotary kiln sintering process under indefinite interference is described as following form:
Wherein k indicates the time of running of cement rotary kiln sintering process, y (z), u (z), and e (z) respectively indicates the k moment and turns round
The temperature y (k) of kiln, kiln hood coal powder injection input valve opening u (k), the form after the z-transform of indefinite interference w (k).Δ is that difference is calculated
Son, A (z-1), B (z-1), C (z-1) be appropriate dimension corresponding multinomial.
1-2. kiln hood coal powder injection input valve opening partial fault is described as follows:
uF(k)=α u (k)
Wherein uFIt (k) is that the kiln hood coal powder injection at k moment inputs the practical aperture of valve, α indicates that kiln hood coal powder injection input valve is opened
Spend the influence degree of failure.
1-3. is according to step 1-1,1-2, then the cement rotary kiln with part kiln hood coal powder injection input valve opening failure
Sintering process is described as follows:
Wherein uF(z) u is indicatedF(k) differentiated form.
1-4. obtains the cement rotary kiln sintering process model of following discrete transfer function form according to 1-3:
y(k+1)+F1y(k)+…+Fpy(k-p+1)
=H1u(k)+H2u(k-1)+…+Hqu(k-q+1)
Wherein m representation dimension, m=dim (Δ xm)=p+q-1, y (k+1), y (k) ..., y (k-p+1) indicate k+1,
The kiln temperature at k ..., k-p+1 moment, u (k+1), u (k) ..., u (k-q+1) indicate k+1, k ..., the kiln at k-q+1 moment
Head coal powder injection inputs valve actuator aperture, F1,…,Fp, H1,…,HqRespectively indicate corresponding kiln temperature and kiln hood coal powder injection valve
The model coefficient of aperture, p, q are the order of corresponding kiln temperature and kiln hood coal powder injection valve opening respectively.
Difference operator is added to above-mentioned cement rotary kiln sintering process model and defines a new vector as follows:
Δxm(k)T=[Δ y (k) ... Δ y (k-p+1) Δ u (k-1) ... Δ u (k-q+1)]
Wherein Δ xm(k)TIndicate the transposition of the cement rotary kiln sintering process state increment of k moment m dimension, m=dim (Δ
xm)=p+q-1, Δ y (k+1), Δ y (k) ..., Δ y (k-p+1) indicate that k+1, the kiln temperature of k ..., k-p+1 moment increase
Amount, Δ u (k+1), Δ u (k) ..., Δ u (k-q+1) indicate k+1, and the kiln hood coal powder injection of k ..., k-q+1 moment input valve opening
Increment.
1-5. further obtains differentiated cement rotary kiln sintering process model according to step 1-4 are as follows:
Δxm(k+1)=AmΔxm(k)+BmΔu(k)
Δ y (k+1)=CmΔxm(k+1)
Wherein Δ xm(k),Δxm(k+1) k is respectively indicated, the cement rotary kiln sintering process state increment of k+1 moment m dimension,
The kiln temperature increment at Δ y (k+1) expression k+1 moment.H2,…,Hm-1,HmIndicate differentiated model coefficient.
Bm=[H1 0 0 … 0 10 0]T,Cm=[1 00 ... 000 0]
1-6. defines kiln temperature tracking error are as follows:
E (k)=y (k)-r (k)
Wherein e (k) indicates that the kiln temperature tracking error at k moment, r (k) indicate the kiln temperature setting at k moment
Value.
The dynamic process for further obtaining kiln temperature tracking error is expressed as follows:
E (k+1)=e (k)+CmAmΔxm(k)+CmBmΔu(k)
Wherein e (k+1) indicates the kiln temperature tracking error at k+1 moment.
It is as follows to finally obtain new cement rotary kiln sintering process model according to above step by 1-7.:
Z (k+1)=Az (k)+B Δ u (k)
Wherein z (k)=[Δ xm(t) e(t)]T, z (k), z (k+1) respectively indicate k, the new cement rotary kiln at k+1 moment
Sintering process state.
Step 2, design cement rotary kiln sintering process controller, specifically:
2-1. is based on step 1, and the objective function for selecting cement rotary kiln sintering process twice optimum is following form:
Wherein J is objective function, and Q, R are corresponding new cement rotary kiln sintering process state and kiln hood coal powder injection input respectively
The weighting matrix of valve opening increment, and Q is diagonal matrix, Q=diag { qj1,qj2,…,qjp+q-1,qje, qj1,qj2,…,
qjp+q-1It is the weighting parameters of kiln temperature variation and kiln hood coal powder injection input valve variation, qjeIt is kiln temperature tracking error
Weighting parameters.
The minimization of object function of 2-2. solution procedure 2-1 obtains cement rotary kiln sintering process twice optimum device
Optimal kiln hood coal powder injection input valve opening increment are as follows:
Δ u (k)=- R-1BT[I+K$BR-1BT]-1K$Az(k)
Wherein K$=AT[I+K$BR-1BT]-1K$A+Q=ATK$A-ATK$B(R+BTK$B)-1BTK$A+Q, K$It is above-mentioned to meet
The solution of Riccati equation.
2-3. repeats step 1.6 to 2.2 and continues to solve new optimal more new law in subsequent time, obtains optimal kiln hood spray
Coal inputs valve opening increment Delta u (k), acts on kiln hood coal powder injection input valve, and circuit sequentially.