CN110887038B - Combustion self-adaptive control system and method for circulating fluidized bed boiler - Google Patents

Combustion self-adaptive control system and method for circulating fluidized bed boiler Download PDF

Info

Publication number
CN110887038B
CN110887038B CN201911369743.5A CN201911369743A CN110887038B CN 110887038 B CN110887038 B CN 110887038B CN 201911369743 A CN201911369743 A CN 201911369743A CN 110887038 B CN110887038 B CN 110887038B
Authority
CN
China
Prior art keywords
control
output
self
vector
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911369743.5A
Other languages
Chinese (zh)
Other versions
CN110887038A (en
Inventor
柴庆宣
商孟尧
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Cosco Control Engineering Co ltd
Original Assignee
Harbin Cosco Control Engineering Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Harbin Cosco Control Engineering Co ltd filed Critical Harbin Cosco Control Engineering Co ltd
Priority to CN201911369743.5A priority Critical patent/CN110887038B/en
Publication of CN110887038A publication Critical patent/CN110887038A/en
Application granted granted Critical
Publication of CN110887038B publication Critical patent/CN110887038B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
    • F23C—METHODS OR APPARATUS FOR COMBUSTION USING FLUID FUEL OR SOLID FUEL SUSPENDED IN  A CARRIER GAS OR AIR 
    • F23C10/00—Fluidised bed combustion apparatus
    • F23C10/18—Details; Accessories
    • F23C10/28—Control devices specially adapted for fluidised bed, combustion apparatus
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
    • F23N—REGULATING OR CONTROLLING COMBUSTION
    • F23N1/00—Regulating fuel supply
    • F23N1/02—Regulating fuel supply conjointly with air supply
    • F23N1/022—Regulating fuel supply conjointly with air supply using electronic means

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • General Engineering & Computer Science (AREA)
  • Fluidized-Bed Combustion And Resonant Combustion (AREA)
  • Feedback Control In General (AREA)

Abstract

A self-adaptive control system and method for the combustion of a circulating fluidized bed boiler, the control system comprises a self-correcting controller and an RBF neural network model estimator; the RBF neural network model estimator performs model estimation; the self-correcting controller performs setting, control and decoupling compensation on model control characteristic parameters given by the RBF neural network, and outputs a control quantity vector, and the control method comprises the following steps: setting given quantity vectors and collecting control quantity vectors; collecting as output vector; taking the control quantity vector and the output quantity vector as input, acting on an RBF neural network model estimator to carry out model estimation, and outputting a model control characteristic parameter to a self-correcting controller; and taking the given quantity vector and the output quantity vector as feedback, inputting the control gain matrix into a self-correcting controller, and completing parameter setting, decoupling compensation and self-adaptive control of the circulating fluidized bed boiler combustion system by the self-correcting controller. The invention has safe control and stable operation.

Description

Combustion self-adaptive control system and method for circulating fluidized bed boiler
Technical Field
The invention relates to a boiler combustion control system and method, in particular to a circulating fluidized bed boiler combustion self-adaptive control system and method.
Background
The circulating fluidized bed boiler has the characteristics of wide fuel applicability, low combustion temperature, large load adjustment range, meeting the environmental protection requirement and the like, and is widely applied to industries such as electric power, thermoelectricity and the like. The circulating fluidized bed boiler has large combustion particles, long combustion process and large thermal inertia of the boiler, so that the combustion process of the fluidized bed boiler has large hysteresis characteristic. The main steam pressure, the hearth temperature, the oxygen content of the flue gas, the hearth pressure and other working condition state quantities interact, and the main steam pressure and the hearth temperature are particularly seriously coupled; the adjustment quantities of the boiler fuel quantity, the limestone quantity, the primary air quantity, the secondary air quantity, the slag discharge quantity and the like also influence each other, so that the combustion process has strong coupling characteristics. The combustion system of the circulating fluidized bed boiler is a large-lag, strong-coupling and time-varying nonlinear system, and an accurate mathematical model is difficult to establish to describe the dynamic process of the system. The automatic control target of the circulating fluidized bed boiler combustion system is as follows: and the load is quickly adjusted to respond to the load change demand of the power grid, and meanwhile, the main steam pressure is kept stable. The hearth temperature is kept in the range of 850-900 ℃, and the desulfurization effect is the best in the range. Too low a hearth temperature can cause furnace flameout, and too high a hearth temperature can cause coking. Maintaining reasonable oxygen content in the flue gas, and burning insufficiently when the oxygen content is too low; too high oxygen content results in great heat loss and reduced heat efficiency.
At present, the automatic control scheme of the circulating fluidized bed boiler combustion system at home and abroad mostly adopts the conventional proportional-integral-derivative (PID) control, and the effect is not ideal in practical application: the change from the coal feeding amount to the change of the main steam pressure of the boiler needs more than 10 minutes, and the large inertia and large hysteresis characteristics of the boiler object cause the main steam pressure of the boiler to swing greatly, so that the overshoot is serious. The main task of the primary air is to establish a stable circulating fluidization state, with a limited regulation range. The temperature of the hearth is frequently changed due to factors such as coal quality, load and the like, and the primary air is used for controlling the temperature of the hearth to be frequently changed, so that the fluidization state of bed materials is seriously influenced, and the safe operation of the boiler is threatened. The automatic control can only automatically operate in a local range under a stable environment, and cannot automatically operate stably for a long time when the load of a boiler or the type of coal is greatly changed.
Disclosure of Invention
The invention provides a self-adaptive control system and a method for combustion of a circulating fluidized bed boiler, which are used for overcoming the defects of the prior art and controlling the safe and stable operation. The self-adaptive control system and the method take the coal feeding quantity, the primary air quantity and the secondary air quantity as control quantities, and automatically control the main steam pressure, the hearth temperature and the flue gas oxygen quantity which are strongly coupled in the circulating fluidized bed combustion system. And estimating combustion control characteristic model parameters of the circulating fluidized bed combustion system by using a Radial Basis Function (RBF) neural network, and performing decoupling compensation control based on a characteristic vector space of a dynamic gain matrix of a control object to realize adaptive regulation control of the control quantity.
The technical scheme of the invention is as follows:
the first scheme is as follows: a self-adaptive control system for the combustion of a circulating fluidized bed boiler comprises a self-correcting controller and an RBF neural network model estimator;
the RBF neural network model estimator is used for carrying out model estimation on coal feeding quantity, primary air quantity and secondary air quantity which serve as control quantity vectors, and main steam pressure, hearth temperature and flue gas oxygen quantity which serve as output quantity vectors, and estimating model control characteristic parameters; calculating by using the model control parameters to obtain a control gain matrix of the CFB boiler combustion system;
the self-correcting controller is used for calculating and outputting a control quantity vector to a CFB boiler combustion system by using main steam pressure given quantity, hearth temperature given quantity, flue gas oxygen given quantity as given quantity vectors, main steam pressure, hearth temperature and flue gas oxygen as output quantity vectors and the control gain matrix as input quantities through parameter setting, control and decoupling compensation.
Scheme II: a self-adaptive control method for combustion of a circulating fluidized bed boiler comprises the following steps:
setting main steam pressure, hearth temperature and flue gas oxygen quantity as given quantity vectors; collecting coal feeding quantity, primary air quantity and secondary air quantity which are used as control quantity vectors; collecting main steam pressure, hearth temperature and flue gas oxygen amount as output quantity vectors;
secondly, the coal feeding quantity, the primary air quantity and the secondary air quantity which are used as control quantity vectors, and the main steam pressure, the hearth temperature and the flue gas oxygen quantity which are used as output quantity vectors are used as input quantities and are applied to an RBF neural network model estimator for model estimation, control characteristic parameters of an output model are calculated, and a control gain matrix obtained by calculation of the control characteristic parameters of the model is input to a self-correcting controller;
and thirdly, the main steam pressure, the hearth temperature and the flue gas oxygen quantity of the given quantity vector are used as feedback, the main steam pressure, the hearth temperature and the flue gas oxygen quantity of the output quantity vector are input into a self-correcting controller, the parameter setting control and decoupling compensation are completed, the control quantity vector is output to the CFB boiler combustion system, and the steps are repeated so as to realize the self-adaptive control of the circulating fluidized bed boiler combustion system.
Compared with the prior art, the invention has the beneficial effects that:
the invention adopts a Radial Basis Function (RBF) neural network estimator and a self-correcting controller to take the coal feeding quantity, the primary air quantity and the secondary air quantity as control quantity vectors as adjusting targets, controls the main steam pressure, the hearth temperature and the flue gas oxygen quantity existing in the output quantity vectors, and combines the given quantity to carry out model estimation, parameter real-time update, control characteristic model parameter control decoupling and compensation, thereby realizing the self-adaptive control of the CFB boiler combustion system.
The self-adaptive control method is applied to the CFB boiler combustion system, and can effectively solve the problems of large hysteresis, strong coupling and time-varying nonlinearity existing in the automatic control of the CFB boiler combustion system. And performing parameter estimation on the fluidized bed boiler combustion system by utilizing the arbitrary approximation characteristic of a Radial Basis Function (RBF) neural network to obtain a control gain matrix of the combustion control system. The self-correcting controller performs decoupling control based on the eigenvector space of the dynamic gain matrix of the control object. And a control parameter updating algorithm is designed to realize self-adaptive adjustment control, so that the circulating fluidized bed boiler combustion system can quickly respond to the large-range adjustment of the boiler load. The prediction output estimated by the RBF neural network model can effectively overcome the hysteresis characteristic of a controlled object, real-time online optimization is carried out on RBF neural network parameters, the time-varying characteristic of the controlled object is overcome, long-time safe and stable operation is realized, operators are liberated from heavy manual operation, and a good foundation is laid for further improving the operation efficiency and improving the economic benefit of enterprises.
Drawings
FIG. 1 is a block diagram of an adaptive control system for a combustion system of a circulating fluidized bed boiler;
FIG. 2 is a block diagram of an RBF neural network model estimator;
FIG. 3 is a block diagram of a self-calibrating controller;
FIG. 4 is a flow diagram of adaptive control of a circulating fluidized bed boiler combustion system;
FIG. 5 is a flow chart of RBF neural network model estimation;
FIG. 6 is a diagram of the PID control parameter auto-tuning process.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-3, one embodiment provides a circulating fluidized bed boiler combustion adaptive control system, which includes a self-correcting controller and an RBF neural network model estimator;
the RBF neural network model estimator is used for carrying out model estimation on coal feeding quantity, primary air quantity and secondary air quantity which serve as control quantity vectors U, and main steam pressure, hearth temperature and flue gas oxygen quantity which serve as output quantity vectors Y, and estimating model control characteristic parameters; calculating by using the model control parameters to obtain a control gain matrix of the CFB boiler combustion system;
the self-correcting controller is used for calculating and outputting a control quantity vector U to a CFB boiler combustion system by using main steam pressure given quantity, hearth temperature given quantity and flue gas oxygen given quantity as given quantity vectors R, and main steam pressure, hearth temperature and flue gas oxygen as output quantity vectors Y and the control gain matrix as input quantities through parameter setting, control and decoupling compensation. In fig. 1, the operation protection subfunction realizes an automatic control overrun release function and an operation safety protection function. Yp represents an estimated value of the output quantity, and Ep represents an estimation error of the output quantity.
Furthermore, the RBF neural network is a three-layer feedforward network with a single hidden layer, simulates a neural network structure of local adjustment in the human brain and mutually covering receiving domains, is a local approximation neural network, and can approximate any continuous function with any precision. Fig. 2 is a structural diagram of an RBF neural network estimator, and the RBF neural network model estimator is composed of an input layer, a hidden layer and an output layer. In fig. 2, 6 input layer neurons, 7 hidden layer neurons, and 3 output layer neurons, the network optimization parameters include: radial basis function base width B, center C, weight W.
Further, FIG. 3 shows a self-correcting controller structure, which includes a PID controller and a decoupling compensator; the decoupling compensator and the control object form a generalized control object, the variables of the generalized control object are decoupled to realize compensation of the decoupled control quantity, and the PID controller receives main steam pressure and hearth temperature serving as set quantity vectorsAnd a given signal of the oxygen content of the flue gas, and measurement signals of the main steam pressure, the hearth temperature and the oxygen content of the flue gas which are used as output quantity vectors, set and control are carried out on the given control characteristic model parameters, the output end of the PID controller is connected with the input end of the decoupling compensator, and the decoupling compensator carries out output of a control quantity vector U. The control characteristic model parameters form a control dynamic gain matrix (Jac matrix in the figure), UdRepresenting the PID controller output without decoupling compensation.
Referring to fig. 1, 4 and 5, in another embodiment, there is provided a circulating fluidized bed boiler combustion adaptive control method including:
setting main steam pressure, hearth temperature and flue gas oxygen quantity as given quantity vectors R; collecting coal feeding quantity, primary air quantity and secondary air quantity which are used as control quantity vectors U; collecting main steam pressure, hearth temperature and flue gas oxygen quantity as an output quantity vector Y;
secondly, taking the coal feeding quantity, the primary air quantity and the secondary air quantity as a control quantity vector U, and the main steam pressure, the hearth temperature and the flue gas oxygen quantity as an output quantity vector Y as input quantities, acting on an RBF neural network model estimator to carry out model estimation, calculating control characteristic parameters of an output model, and inputting a control gain matrix obtained by calculating the control characteristic parameters of the model to a self-correcting controller;
and thirdly, the main steam pressure, the hearth temperature, the flue gas oxygen quantity and the main steam pressure, the hearth temperature and the flue gas oxygen quantity of the given quantity vector R and the output quantity vector Y are used as feedback, the control gain matrix is input into the self-correcting controller to complete parameter setting control and decoupling compensation, and the control quantity vector U is output to the CFB boiler combustion system, and the steps are repeated so as to realize the self-adaptive control of the circulating fluidized bed boiler combustion system.
Further, fig. 5 shows an estimation process of the RBF neural network model, where epsilon in fig. 5 is a given threshold of the prediction error, and J is a sum of squares of the estimation errors of the neural network model; in the second step, the RBF neural network model estimation and gain matrix control algorithm is as follows:
6.1, determining an estimation target of the RBF model:
the used RBF neural network estimator consists of an input layer, a hidden layer and an output layer;
input layer input vector:
X＝[x1,x2,x3,x4,x5,x6]T＝[u1,u2,u3,y1,y2,y3]T
wherein u is1,u2,u3Respectively representing coal feeding quantity, primary air quantity and secondary air quantity as control quantity vector U; y is1,y2,y3Respectively representing main steam pressure, hearth temperature and flue gas oxygen quantity as output quantity vectors Y; the network optimization parameters include: a radial basis function base width B, a center position C and a weight W;
output layer output vector:
Yp＝[yp1,yp2,yp3]T
wherein yp1，yp2，yp3Respectively representing a main steam pressure predicted value, a hearth temperature predicted value and a flue gas oxygen amount predicted value of output quantity;
hidden layer activation vector:
H＝[h1,h2,h3,h4,h5,h6,h7]T
the neuron activation function adopts a Gaussian base function:
Figure BDA0002339347250000051
wherein, bjThe base width of the radial basis function of the jth neuron of the hidden layer;
hidden layer jth neuron radial basis function center point central vector:
Cj＝[cj1,cj2,cj3,cj4,cj5,cj6]T，j＝1,2,...,7
radial basis function basis width vector:
B＝[b1,b2,b3,b4,b5,b6,b7]T
hidden layer and output layer weight matrix:
Figure BDA0002339347250000052
output layer output vector:
Yp＝W·H
Figure BDA0002339347250000053
the estimated model error is:
E(t)＝Y(t)-Yp(t)
E＝Y-W·H
the RBF model estimation target is:
Figure BDA0002339347250000054
minimizing the error between the estimated output and the real output, so that the estimated error of the RBF neural network model meets a given error threshold;
6.2 parameter updating algorithm meeting the estimation target of the RBF model:
optimizing the form of the target matrix:
Figure BDA0002339347250000061
work-up from 6.1 gave:
Figure BDA0002339347250000062
weight W updating algorithm:
Figure BDA0002339347250000063
Figure BDA0002339347250000064
estimation of parameter learning rate:
η∈[0,1]
estimating a parameter inertial damping coefficient:
α∈[0,1]
the radial basis function base width B updating algorithm:
Figure BDA0002339347250000065
wherein,
Figure BDA0002339347250000066
Figure BDA0002339347250000067
Figure BDA0002339347250000068
center position C update algorithm:
Figure BDA0002339347250000069
Figure BDA00023393472500000610
Figure BDA00023393472500000611
RBF neural network parameter updating formula:
Figure BDA0002339347250000071
wherein j is 1,2
Performing a parameter updating algorithm by adopting the model estimation, predicting and outputting Yp, calculating a prediction error Ep, and updating and adjusting the neural network parameters;
6.3 determining an estimation model based on RBF neural network parameters
The control gain is calculated, and since the model is unknown, approximated using an estimation model:
Figure BDA0002339347250000072
Figure BDA0002339347250000073
Figure BDA0002339347250000074
wherein k is 1,2,3, i is 1,2, 3;
obtaining a control gain matrix of a control object:
Figure BDA0002339347250000075
the eigenvectors of the control gain matrix form an eigenspace matrix V which satisfies:
Jac·V＝Λ·V
wherein,
Figure BDA0002339347250000076
the self-correcting controller carries out characteristic vector space decomposition according to the control gain matrix, decoupling compensation is carried out on the control quantity of the PID controller, control decoupling is achieved, and a control quantity vector U is output:
U＝VT·Ud
and completing the self-adaptive control of the circulating fluidized bed combustion system.
Further, adjusting the control characteristic parameters estimated by the model to realize compensation of the decoupled control quantity, further outputting a control quantity vector to finish self-adaptive control of the circulating fluidized bed combustion system, wherein a PID controller and a decoupling compensator in a self-correcting controller are adopted to finish the self-adaptive control; and automatically adjusting the parameters of the PID controller according to the estimated parameters and the control error. The method comprises the following specific steps:
given vector of quantities:
R＝[r1,r2,r3]T
output quantity vector:
Y＝[y1,y2,y3]T
controlling the error:
Ec＝(R-Y)
eck(t)＝rk(t)-yk(t)，k＝1,2,3
an incremental PID controller is adopted:
ΔUd＝[Δu1,Δu2,Δu3]T
Δuk＝kpk(eck(t)-eck(t-1))+kik(eck(t))+kdk(eck(t)-2eck(t-1)+eck(t-2))
definition of
xck1(t)＝eck(t)-eck(t-1)
xck2(t)＝eck(t)
xck3(t)＝eck(t)-2*eck(t-1)+eck(t-2)
Then
Δuk＝kpk*xck1(t)+kik*xck2(t)+kdk*xck3(t)
PID controller parameter matrix
Figure BDA0002339347250000081
Order to
Pk＝[kpk,kik,kdk]
Dk＝[xck1,xck2,xck3]T
Then
ΔUdk＝Pk·Dk
Udk(t)＝Udk(t-1)+ΔUdk
The decoupling compensation is obtained
U＝VT·Ud
Completing decoupling compensation of the PID controller output control quantity to obtain a control quantity vector of the self-adaptive controller;
the decoupling compensator is formed by a feature space matrix V transpose formed by feature vectors of a control gain matrix Jac, wherein Jac.V.. Lambda.V.
Based on the control gain matrix, the parameter setting process of the PID controller comprises the following steps:
control error Ec＝(R-Y)
Closed-loop control error optimization objective:
Figure BDA0002339347250000091
namely minimizing the mean square value of the closed-loop control error;
Figure BDA0002339347250000092
Figure BDA0002339347250000093
PID closed-loop control parameter updating algorithm:
Figure BDA0002339347250000094
wherein the closed loop optimization target pair outputs a variance:
Figure BDA0002339347250000095
control gain matrix of the control object:
Figure BDA0002339347250000096
the decoupling compensation by the control quantity can be obtained:
Figure BDA0002339347250000101
control amount versus control parameter variation:
Figure BDA0002339347250000102
finishing to obtain:
Figure BDA0002339347250000103
Figure BDA0002339347250000104
and (3) a control parameter updating algorithm:
Figure BDA0002339347250000105
Pk(t)＝Pk(t-1)+ηc·Eck·Λk·Vk·Dk T+αc(Pk(t-1)-Pk(t-2))
control parameter learning rate:
ηc∈[0,1]
control parameter inertial damping coefficient:
αc∈[0,1]
the control parameter updating algorithm is adopted to realize the automatic adjustment of the PID controller parameters, and the model estimation parameter updating algorithm and the control parameter updating algorithm are adopted to realize the self-adaptive control of the circulating fluidized bed boiler combustion system.
The automatic parameter adjustment process of the circulating fluidized bed boiler combustion system is shown in fig. 6, and the control parameters are automatically adjusted according to the model estimated parameters and the control errors. As shown in fig. 4, the control gain characteristic of the combustion system is estimated by combining the historical track offline training and online learning modes, the dynamic control gain matrix obtained by calculation according to the estimation model is subjected to adaptive PID parameter adjustment, the characteristic vector space decomposition is performed on the control gain matrix, and the decoupling compensation control is performed, so that the automatic control problem of the circulating fluidized bed combustion system can be effectively solved.
The present invention is not limited to the above embodiments, and those skilled in the art can make various changes and modifications without departing from the scope of the invention.

Claims (7)

1. A self-adaptive control system for combustion of a circulating fluidized bed boiler is characterized in that: the self-correcting controller comprises a self-correcting controller and an RBF neural network model estimator;
the RBF neural network model estimator is used for carrying out model estimation on coal feeding quantity, primary air quantity and secondary air quantity which serve as control quantity vectors, and main steam pressure, hearth temperature and flue gas oxygen quantity which serve as output quantity vectors, and estimating model control characteristic parameters; calculating by using the model control parameters to obtain a control gain matrix of the CFB boiler combustion system;
the RBF neural network is a three-layer feedforward network with a single hidden layer, and the RBF neural network model estimator consists of an input layer, a hidden layer and an output layer;
the self-correcting controller is used for calculating and outputting a control quantity vector to a CFB boiler combustion system by taking main steam pressure given quantity, hearth temperature given quantity, flue gas oxygen given quantity as given quantity vectors, main steam pressure, hearth temperature and flue gas oxygen as output quantity vectors and the control gain matrix as input quantities through parameter setting, control and decoupling compensation;
the self-correcting controller comprises a PID controller and a decoupling compensator; the PID controller takes the set parameters, the main steam pressure, the hearth temperature and the flue gas oxygen quantity as set quantity vectors and the main steam pressure, the hearth temperature and the flue gas oxygen quantity as output quantity vectors as input quantities, executes control and outputs the PID controller output quantity which is not subjected to decoupling compensation, the output end of the PID controller is connected with the input end of the decoupling compensator, the decoupling compensator performs characteristic vector space decomposition on a control gain matrix to realize control decoupling, compensates the output quantity of the PID controller and outputs a control quantity vector after decoupling compensation.
2. A self-adaptive control method for combustion of a circulating fluidized bed boiler is characterized by comprising the following steps: it includes:
setting main steam pressure, hearth temperature and flue gas oxygen quantity as given quantity vectors R; collecting coal feeding quantity, primary air quantity and secondary air quantity which are used as control quantity vectors; collecting main steam pressure, hearth temperature and flue gas oxygen amount as output quantity vectors;
secondly, taking the coal feeding quantity, the primary air quantity and the secondary air quantity as control quantity vectors, and the main steam pressure, the hearth temperature and the flue gas oxygen quantity as output quantity vectors as input quantities, acting on an RBF neural network model estimator to carry out model estimation, calculating control characteristic parameters of an output model, and inputting a control gain matrix obtained by calculating the control characteristic parameters of the model to a self-correcting controller; the RBF neural network is a three-layer feedforward network with a single hidden layer, and the RBF neural network model estimator consists of an input layer, a hidden layer and an output layer;
thirdly, the main steam pressure, the hearth temperature and the flue gas oxygen quantity of the given quantity vector R, the main steam pressure, the hearth temperature and the flue gas oxygen quantity of the output quantity vector R are used as feedback, and the control gain matrix is input into the self-correcting controller,
the self-correcting controller comprises a PID controller and a decoupling compensator; the PID controller receives given signals of main steam pressure, hearth temperature and flue gas oxygen quantity as set quantity vectors and measurement signals of main steam pressure, hearth temperature and flue gas oxygen quantity as output quantity vectors, performs setting and control on given control characteristic model parameters, the output end of the PID controller is connected with the input end of the decoupling compensator, the decoupling compensator performs output control quantity vectors, completes parameter setting control and decoupling compensation, and outputs the control quantity vectors to the CFB boiler combustion system, and the steps are repeated so as to realize self-adaptive control of the circulating fluidized bed boiler combustion system.
3. The adaptive control method for combustion of a circulating fluidized bed boiler according to claim 2, wherein: the control gain matrix determined by the model control characteristic parameters is:
Figure FDA0002931744280000021
the eigenvectors of the control gain matrix form an eigenspace matrix V which satisfies:
Jac·V＝Λ·V
wherein,
Figure FDA0002931744280000022
the self-correcting controller carries out characteristic vector space decomposition and decoupling compensation according to the control gain matrix, realizes control decoupling, and outputs a control quantity vector U:
U＝VT·Ud
and completing the self-adaptive control of the circulating fluidized bed combustion system.
4. The adaptive control method for combustion of a circulating fluidized bed boiler according to claim 2 or 3, wherein: the RBF neural network model estimation and control gain matrix algorithm is as follows:
6.1, determining an estimation target of the RBF model:
the used RBF neural network estimator consists of an input layer, a hidden layer and an output layer;
input layer input vector:
X＝[x1,x2,x3,x4,x5,x6]T＝[u1,u2,u3,y1,y2,y3]T
wherein u is1,u2,u3Respectively representing coal feeding quantity, primary air quantity and secondary air quantity as control quantity vector U; y is1,y2,y3Respectively representing main steam pressure, hearth temperature and flue gas oxygen quantity as output quantity vectors Y; the network optimization parameters include: a radial basis function base width B, a center position C and a weight W;
output layer output vector:
Yp＝[yp1,yp2,yp3]T
wherein yp1，yp2，yp3Respectively showing the predicted value of the output main steam pressure, the predicted value of the hearth temperature and the predicted value of the smoke oxygen amount;
hidden layer activation vector:
H＝[h1,h2,h3,h4,h5,h6,h7]T
the neuron activation function adopts a Gaussian base function:
Figure FDA0002931744280000031
wherein, bjThe base width of the radial basis function of the jth neuron of the hidden layer;
hidden layer jth neuron radial basis function center point central vector:
Cj＝[cj1,cj2,cj3,cj4,cj5,cj6]T，j＝1,2,...,7
radial basis function basis width vector:
B＝[b1,b2,b3,b4,b5,b6,b7]T
hidden layer and output layer weight matrix:
Figure FDA0002931744280000032
output layer output vector:
Yp＝W·H
Figure FDA0002931744280000033
the estimated model error is:
E(t)＝Y(t)-Yp(t)
E＝Y-W·H
the RBF model estimation target is:
Figure FDA0002931744280000034
minimizing the error between the estimated output and the real output, and taking the error as the condition for completing the estimation of the RBF neural network model;
6.2 parameter updating algorithm meeting the estimation target of the RBF model:
optimizing the form of the target matrix:
Figure FDA0002931744280000041
work-up from 6.1 gave:
Figure FDA0002931744280000042
weight W updating algorithm:
Figure FDA0002931744280000043
Figure FDA0002931744280000044
estimation of parameter learning rate:
η∈[0,1]
estimating a parameter inertial damping coefficient:
α∈[0,1]
the radial basis function base width B updating algorithm:
Figure FDA0002931744280000045
wherein,
Figure FDA0002931744280000046
Figure FDA0002931744280000047
Figure FDA0002931744280000048
center position C update algorithm:
Figure FDA0002931744280000049
Figure FDA00029317442800000410
Figure FDA00029317442800000411
RBF neural network parameter updating formula:
Figure FDA0002931744280000051
wherein j is 1,2
Performing a parameter updating algorithm by adopting the model estimation, performing prediction output, calculating prediction errors, and updating and adjusting neural network parameters;
6.3 determining an estimation model based on RBF neural network parameters
Estimation model
Figure FDA0002931744280000052
Figure FDA0002931744280000053
Figure FDA0002931744280000054
Wherein k is 1,2,3, i is 1,2, 3;
obtaining a control gain matrix of a control object:
Figure FDA0002931744280000055
and finishing the CFB combustion system control characteristic estimation.
5. The adaptive control method for combustion of a circulating fluidized bed boiler according to claim 4, wherein: the process of outputting the control quantity vector U is completed by a PID controller and a decoupling compensator, and specifically comprises the following steps:
given vector of quantities:
R＝[r1,r2,r3]T
output quantity vector:
Y＝[y1,y2,y3]T
controlling the error:
Ec＝(R-Y)
eck(t)＝rk(t)-yk(t)，k＝1,2,3
an incremental PID controller is adopted:
ΔUd＝[Δu1,Δu2,Δu3]T
Δuk＝kpk(eck(t)-eck(t-1))+kik(eck(t))+kdk(eck(t)-2eck(t-1)+eck(t-2))
definition of
xck1(t)＝eck(t)-eck(t-1)
xck2(t)＝eck(t)
xck3(t)＝eck(t)-2*eck(t-1)+eck(t-2)
Then
Δuk＝kpk*xck1(t)+kik*xck2(t)+kdk*xck3(t)
PID controller parameter matrix
Figure FDA0002931744280000061
Order to
Pk＝[kpk,kik,kdk]
Dk＝[xck1,xck2,xck3]T
Then
ΔUdk＝Pk·Dk
Udk(t)＝Udk(t-1)+ΔUdk
The decoupling compensation is obtained
U＝VT·Ud
Decoupling compensation is carried out on the control quantity of the PID controller, control decoupling is realized, and a control quantity vector U of the self-adaptive controller is obtained;
the decoupling compensator is formed by a feature space matrix V transpose formed by feature vectors of a control gain matrix Jac, wherein Jac.V.. Lambda.V.
6. The adaptive control method for combustion of a circulating fluidized bed boiler according to claim 5, wherein: the parameter setting process of the PID controller comprises the following steps:
control error Ec＝(R-Y)
Closed-loop control error optimization objective:
Figure FDA0002931744280000071
namely minimizing the mean square value of the closed-loop control error;
Figure FDA0002931744280000072
Figure FDA0002931744280000073
PID closed-loop control parameter updating algorithm:
Figure FDA0002931744280000074
wherein the closed loop optimization target pair outputs a variance:
Figure FDA0002931744280000075
control gain matrix of the control object:
Figure FDA0002931744280000076
the decoupling compensation by the control quantity can be obtained:
Figure FDA0002931744280000077
control amount versus control parameter variation:
Figure FDA0002931744280000078
finishing to obtain:
Figure FDA0002931744280000079
Figure FDA0002931744280000081
and (3) a control parameter updating algorithm:
Figure FDA0002931744280000082
Pk(t)＝Pk(t-1)+ηc·Eck·Λk·Vk·Dk T+αc(Pk(t-1)-Pk(t-2))
control parameter learning rate:
ηc∈[0,1]
control parameter inertial damping coefficient:
αc∈[0,1]
by adopting the control parameter updating algorithm, the automatic adjustment of the PID controller parameters is realized, and the self-adaptive control of the CFB boiler combustion system is completed.
7. The adaptive control method for combustion of a circulating fluidized bed boiler according to claim 2,3, 5 or 6, wherein: the CFB boiler combustion system model estimation is further completed by combining an RBF neural network offline training mode and an online rolling optimization mode.
CN201911369743.5A 2019-12-26 2019-12-26 Combustion self-adaptive control system and method for circulating fluidized bed boiler Active CN110887038B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911369743.5A CN110887038B (en) 2019-12-26 2019-12-26 Combustion self-adaptive control system and method for circulating fluidized bed boiler

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911369743.5A CN110887038B (en) 2019-12-26 2019-12-26 Combustion self-adaptive control system and method for circulating fluidized bed boiler

Publications (2)

Publication Number Publication Date
CN110887038A CN110887038A (en) 2020-03-17
CN110887038B true CN110887038B (en) 2021-05-28

Family

ID=69753249

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911369743.5A Active CN110887038B (en) 2019-12-26 2019-12-26 Combustion self-adaptive control system and method for circulating fluidized bed boiler

Country Status (1)

Country Link
CN (1) CN110887038B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023036924A1 (en) * 2021-09-09 2023-03-16 Sumitomo SHI FW Energia Oy Method for determining a local temperature anomaly in a fluidized bed of a reactor, method for calibrating a numerical model of a fluidized bed of a reactor, method for estimating risk of fluidized bed reactor bed sintering, method of controlling a fluidized bed reactor, as well as a reactor

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111561694B (en) * 2020-06-03 2022-01-25 国网浙江省电力有限公司电力科学研究院 Method and system for improving low-load SCR inlet smoke temperature of coal-fired boiler
CN112212322B (en) * 2020-09-22 2022-08-26 河北国超热力工程有限公司 Intelligent control method for optimizing combustion of thermodynamic circulating fluidized bed boiler
CN113091088B (en) * 2021-04-14 2022-10-25 南京邮电大学 Boiler combustion generalized predictive control method based on two-stage neural network
CN113217908A (en) * 2021-05-06 2021-08-06 江西江右净达热能科技有限公司 Biomass briquette fuel combustion system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106224948A (en) * 2016-09-23 2016-12-14 凤阳海泰科能源环境管理服务有限公司 A kind of self adaptation CFBB control method
CN110296436A (en) * 2019-07-09 2019-10-01 威立雅（哈尔滨）热电有限公司 Computer improves boiler pressuring fire and restarts success rate and starting stationarity method

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1203274C (en) * 2003-07-29 2005-05-25 厦门厦大海通自控有限公司 Optimizing control system for combustion process of circulating fluid bed in boiler
CN101551103B (en) * 2009-04-30 2010-11-10 东莞德永佳纺织制衣有限公司 Automatic boiler combustion control system of circulating fluid bed
CN101556038B (en) * 2009-05-27 2010-09-15 北京和隆优化控制技术有限公司 Optimization control system for stable operation and economical combustion of circulating fluidized-bed boiler
CN101788809B (en) * 2009-08-17 2013-03-06 杭州和利时自动化有限公司 Coordinated control system (CCS) of large-size circulating fluidized bed boiler (CFBB) unit
CN101713536B (en) * 2009-12-03 2011-06-29 太原理工大学 Control method of combustion system of circulating fluidized bed boiler
CN102494336B (en) * 2011-12-16 2013-09-25 浙江大学 Combustion process multivariable control method for CFBB (circulating fluidized bed boiler)
TW201544766A (en) * 2014-05-16 2015-12-01 中原大學 Fluidized bed boiler and method for enhancing operational efficiency of the same
CN104296131B (en) * 2014-10-23 2015-09-30 东南大学 A kind of multivariable cooperative control method of twin furnace Properties of CFB
US10253974B1 (en) * 2015-02-27 2019-04-09 Morgan State University System and method for biomass combustion
CN105020705B (en) * 2015-03-04 2017-06-09 内蒙古瑞特优化科技股份有限公司 Burning in circulating fluid bed boiler performance method for real-time optimization control and system
EP3106747A1 (en) * 2015-06-15 2016-12-21 Improbed AB Control method for the operation of a combustion boiler
CN104932274B (en) * 2015-07-06 2017-08-29 东南大学 One kind coordinates control controlled device transfer function model discrimination method
CN105240846B (en) * 2015-10-09 2017-06-16 南京信息工程大学 The Process of Circulating Fluidized Bed Boiler control method of multivariable GPC optimization
CN107023825B (en) * 2016-08-31 2019-01-22 西安艾贝尔科技发展有限公司 Fluidized-bed combustion boiler control and combustion optimizing system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106224948A (en) * 2016-09-23 2016-12-14 凤阳海泰科能源环境管理服务有限公司 A kind of self adaptation CFBB control method
CN110296436A (en) * 2019-07-09 2019-10-01 威立雅（哈尔滨）热电有限公司 Computer improves boiler pressuring fire and restarts success rate and starting stationarity method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023036924A1 (en) * 2021-09-09 2023-03-16 Sumitomo SHI FW Energia Oy Method for determining a local temperature anomaly in a fluidized bed of a reactor, method for calibrating a numerical model of a fluidized bed of a reactor, method for estimating risk of fluidized bed reactor bed sintering, method of controlling a fluidized bed reactor, as well as a reactor
WO2023036427A1 (en) * 2021-09-09 2023-03-16 Sumitomo SHI FW Energia Oy Method for determining a local temperature anomaly in a fluidized bed of a combustion boiler, method for calibrating a numerical model of a fluidized bed of a combustion boiler, method for estimating risk of fluidized bed combustion boiler bed sintering, method of controlling a fluidized bed boiler, as well as a combustion boiler

Also Published As

Publication number Publication date
CN110887038A (en) 2020-03-17

Similar Documents

Publication Publication Date Title
CN110887038B (en) Combustion self-adaptive control system and method for circulating fluidized bed boiler
CN111413872B (en) Air cavity pressure rapid active disturbance rejection method based on extended state observer
CN109901403A (en) A kind of face autonomous underwater robot neural network S control method
CN102841540A (en) MMPC-based supercritical unit coordination and control method
CN102494336A (en) Combustion process multivariable control method for CFBB (circulating fluidized bed boiler)
CN107270283B (en) Multivariable constraint predictive control method based on circulating fluidized bed unit
CN110181510B (en) Mechanical arm trajectory tracking control method based on time delay estimation and fuzzy logic
CN107450326A (en) Contragradience finite time bilateral teleoperation control method and computer-readable recording medium
CN111176115A (en) Valve position control method based on fuzzy neural network and humanoid intelligent control
CN112650169B (en) Generator set main parameter control system based on enthalpy value and fuel online heat value calculation
CN111290282B (en) Predictive control method for thermal power generating unit coordination system
CN111562744A (en) Boiler combustion implicit generalized predictive control method based on PSO algorithm
CN110579968A (en) Prediction control strategy for ultra-supercritical unit depth peak regulation coordination system
Meng et al. NN-based asymptotic tracking control for a class of strict-feedback uncertain nonlinear systems with output constraints
CN111259525A (en) Model prediction control method for nonlinear unstable wind power engine room suspension system
Zhang et al. Adaptive self-regulation PID control of course-keeping for ships
CN111413865B (en) Disturbance compensation single-loop superheated steam temperature active disturbance rejection control method
CN113835342A (en) Disturbance rejection prediction control method of superheated steam temperature system
CN109062030A (en) Thermal power unit plant load prediction PID control method based on laguerre function model
CN117032209A (en) Robust self-adaptive neural network control method for under-actuated ship
CN116239022A (en) Bridge crane positioning anti-swing model-free self-adaptive control method
CN112947606A (en) Boiler liquid level control system and method based on BP neural network PID predictive control
Yin et al. An Adaptive Control Method for Robot Constant Force Polishing Device
Huang et al. Neural sliding-mode control of engine torque
CN107065538B (en) Dum boiler-Steam Turbine fuzzy tracking control method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
CB02 Change of applicant information

Address after: 150001 5th floor, complex building, No.9 Donghu Road, Harbin Development Zone, Heilongjiang Province

Applicant after: Harbin COSCO Control Engineering Co.,Ltd.

Address before: 150070 5th floor, complex building, No.9 Donghu Road, Harbin Development Zone, Heilongjiang Province

Applicant before: Harbin Institute of Technology COSCO Industrial Control Co.,Ltd.

CB02 Change of applicant information
GR01 Patent grant
GR01 Patent grant