CN115573819A - Gas turbine T-S model optimization method based on multi-model predictive control algorithm - Google Patents
Gas turbine T-S model optimization method based on multi-model predictive control algorithm Download PDFInfo
- Publication number
- CN115573819A CN115573819A CN202211301967.4A CN202211301967A CN115573819A CN 115573819 A CN115573819 A CN 115573819A CN 202211301967 A CN202211301967 A CN 202211301967A CN 115573819 A CN115573819 A CN 115573819A
- Authority
- CN
- China
- Prior art keywords
- model
- gas turbine
- data
- representing
- particle
- 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.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000005457 optimization Methods 0.000 title claims abstract description 20
- 239000013598 vector Substances 0.000 claims abstract description 25
- 230000004044 response Effects 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 239000002245 particle Substances 0.000 claims description 50
- 230000006870 function Effects 0.000 claims description 29
- 238000012546 transfer Methods 0.000 claims description 11
- 239000000446 fuel Substances 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 7
- 238000002485 combustion reaction Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 4
- 239000007789 gas Substances 0.000 description 41
- 230000000694 effects Effects 0.000 description 16
- 239000011159 matrix material Substances 0.000 description 12
- 230000033228 biological regulation Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 8
- 238000002474 experimental method Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000004088 simulation Methods 0.000 description 6
- 238000010248 power generation Methods 0.000 description 5
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 4
- 238000011217 control strategy Methods 0.000 description 3
- 238000012937 correction Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 239000003245 coal Substances 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000000630 rising effect Effects 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 241001212789 Dynamis Species 0.000 description 1
- 238000012369 In process control Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000005314 correlation function Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010965 in-process control Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02C—GAS-TURBINE PLANTS; AIR INTAKES FOR JET-PROPULSION PLANTS; CONTROLLING FUEL SUPPLY IN AIR-BREATHING JET-PROPULSION PLANTS
- F02C9/00—Controlling gas-turbine plants; Controlling fuel supply in air- breathing jet-propulsion plants
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F05—INDEXING SCHEMES RELATING TO ENGINES OR PUMPS IN VARIOUS SUBCLASSES OF CLASSES F01-F04
- F05D—INDEXING SCHEME FOR ASPECTS RELATING TO NON-POSITIVE-DISPLACEMENT MACHINES OR ENGINES, GAS-TURBINES OR JET-PROPULSION PLANTS
- F05D2270/00—Control
- F05D2270/70—Type of control algorithm
- F05D2270/707—Type of control algorithm fuzzy logic
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a gas turbine T-S model optimization method based on a multi-model predictive control algorithm, which comprises the steps of preprocessing data, performing cubic polynomial fitting by using five adjacent sampling points, preprocessing the data and improving the data quality; establishing a feature vector, namely establishing a feature vector combination of input and output variables; identifying the front part structure of the feature vector; carrying out posterior part parameter identification on the data obtained in the step S3; obtaining a T-S fuzzy model; and step response vector models corresponding to the working conditions are obtained from the T-S fuzzy models obtained in the step S5, and the optimal input quantity is obtained by utilizing a DMC algorithm to carry out optimal control on the output of the gas turbine.
Description
Technical Field
The invention relates to the technical field of dynamic working conditions of gas turbines, in particular to a gas turbine T-S model optimization method based on a multi-model predictive control algorithm.
Background
China as an energy resource big country actively participates in coping with global climate change problems, the energy and power industry will remodel the pattern, coal and electricity which are main energy resources of China are being changed to adjusting performance sources for a long time, and the trend of driving energy production and cleaning by renewable energy sources becomes an inevitable trend [1]. The wind and light power generation is limited by a plurality of factors, the uncertainty of output can bring serious impact to the stability adjustment and the flexible scheduling of the power system, and the new energy power generation as the main power generation energy can be finished at all times and still needs reasonable planning and time transition. In the period, the gas-steam combined cycle unit can support system transformation and can be matched with new energy development, and the gas-steam combined cycle unit is the best option under the existing conditions.
The combined gas-steam circulating unit belongs to the field of thermal power generation, and is different from traditional coal powder thermal power unit, and its combustion system is implemented by gas turbine, and the fuel is natural gas (methane CH) 4 ) And after full combustion, higher heat value and zero pollution emission are obtained. The gas-steam combined cycle unit has obvious emission reduction effect and can provide high-occupation-ratio effective capacity. In recent years, many accidents of power shortage and power failure occur in our country, and the problems that the supply of a power generation side and the supply of a power grid side in a power system are unbalanced and the power source capacity is not matched with the maximum load are reflected. The reason is that the residential consumption is increased and the social electricity utilization structure is changed, especially in winter and summer, the electricity utilization load is increased, the peak valley difference of a power grid is continuously enlarged, the frequency modulation and voltage regulation capacity is insufficient, and the severity of the problem of power guarantee supply is highlighted. Therefore, the capacity of the newly added gas-steam combined cycle unit can effectively deal with and relieve the problem of power grid peak regulation.
However, the special operating characteristics of the gas turbine also bring difficulty to understand the dynamic characteristics of the gas turbine, and in addition, the nonlinearity and the coupling of the gas turbine itself are not good choices for directly researching the gas turbine by using a mechanism method, in recent years, the trend and the hot point for researching the nonlinear system are to decompose the nonlinear system into a plurality of linear system models and then fit each linear system to a global nonlinear system by a fusion mode, and the most typical of the multi-model strategy is a T-S clustering fuzzy model. A model of a nonlinear system is not required to be given, only the relation between input and output data of the gas turbine needs to be researched, a fuzzy clustering and identification method is used for dividing a data space into a plurality of subspaces, and each group of simple linear models is expressed in a segmented mode. Establishing an accurate and universal gas turbine model is the basis for developing optimization and control, which is of great practical significance.
Disclosure of Invention
In view of the shortcomings of the prior art, the invention aims to provide a gas turbine T-S model optimization method based on a multi-model predictive control algorithm.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a gas turbine T-S model optimization method based on a multi-model predictive control algorithm comprises the following steps:
s1, data preprocessing, namely performing cubic polynomial fitting by using five adjacent sampling points, preprocessing the data and improving the data quality;
s2, establishing a feature vector, namely establishing a feature vector combination of input and output variables;
s3, identifying the structure of the front part of the feature vector;
s4, performing post-part parameter identification on the data obtained in the step S3;
s5, obtaining a T-S fuzzy model;
s6, obtaining a step response vector model of the corresponding working condition from the T-S fuzzy model obtained in the step S5;
and S7, the optimal input quantity is obtained by utilizing a DMC algorithm to carry out optimal control on the output of the gas turbine.
It should be noted that, in step S1, a five-point cubic smoothing formula is adopted to perform data preprocessing, where:
let the original data be recorded as x, the filtered data be recorded as y, and the specific filtering algorithm rule is expressed as formula:
it should be noted that, the combination of the feature vectors in step S2 is as follows:
wherein, X 1 A combination of eigenvectors representing power, k representing the kth sampling instant, and also in discrete systems, understood as orders, representing X 2 Combination of eigenvectors of exhaust temperature, u 1 Representing the input variable fuel quantity, u 2 Representing the inlet guide vane opening, y, of the input variable 1 Representing the output variable combustion engine power, y 2 Representing the output variable exhaust temperature.
It should be noted that, the step S3 adopts an improved FCM clustering algorithm, which specifically includes:
It should be noted that, the method further includes improving a particle swarm algorithm, specifically:
the method comprises the following steps of firstly, defining variables to be identified, setting an optimization objective function and a stopping condition, randomly generating a plurality of particle populations, and initializing positions and speeds of particles;
secondly, according to the value preliminarily identified by Matlab as a reference, restricting the interval of the optimal solution of the particles and accelerating the speed of searching the optimal solution;
thirdly, calculating the fitness value of each particle, and recording the current historical optimal solution and the global optimal solution of each particle;
fourthly, determining the distance between each particle and the global optimal particle according to a formula, and respectively calculating the inertia weight and the learning factor of the particles;
fifthly, updating and recording the speed and the position of the particles, covering the original historical optimal solution and the original global optimal solution, substituting into an objective function to solve, judging whether a stopping condition is met, if so, performing the next step, otherwise, performing the third step;
sixthly, outputting the globally optimal particle individual, namely the solution of the identified parameter;
obtaining a field accurate transfer function of the gas turbine by improving a particle swarm optimization, wherein the improved transfer formula is as follows:
the invention has the beneficial effects that:
the invention simplifies the gas turbine into a two-input and two-output model based on a 'decomposition-fusion' strategy and a data-driven T-S fuzzy identification modeling method. Further, model structure identification is carried out through an improved fuzzy C-means clustering (FCM) algorithm, model parameters are identified through a least square method with a result correction mode and a recursion mode, a mathematical model with linear rule expression of fuzzy statements is obtained, and accuracy and universality of the model are verified through regression model indexes.
Secondly, the established gas turbine model is combined with dynamic matrix predictive control and applied to an output predictive control strategy. According to results, the T-S fuzzy model established by the invention is combined with a hybrid algorithm of predictive control, and the effect is better compared with the PID control.
Drawings
FIG. 1 shows the result of five-point cubic smoothing of input data according to the present invention;
FIG. 2 is a five-point three-pass smoothing result of the output data of the present invention;
FIG. 3 is a graph of the error function for different gamma factor improvements in the present invention;
FIG. 4 is a diagram of the identification results of a T-S model of a gas turbine according to the present invention;
FIG. 5 is a graphical representation comparing the excitation and actual response of the gas turbine engine according to the present invention;
FIG. 6 is a schematic diagram of a gas turbine output prediction control according to the present invention;
FIG. 7 is a power setpoint disturbance control strategy diagram of the present invention;
FIG. 8 is a schematic diagram of dynamic response of a power target value tracking experiment in the present invention;
FIG. 9 is a control strategy structure under external disturbance in the present invention;
FIG. 10 is a schematic diagram of a step response of an exhaust temperature disturbance signal in a simulation test according to the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings, and it should be noted that the present embodiment is based on the technical solution, and the detailed implementation and the specific operation process are provided, but the protection scope of the present invention is not limited to the present embodiment.
The invention relates to a gas turbine T-S model optimization method based on a multi-model predictive control algorithm, which comprises the following steps:
s1, preprocessing data, namely performing cubic polynomial fitting by using five adjacent sampling points, preprocessing the data and improving the data quality;
s2, establishing a feature vector, namely establishing a feature vector combination of input and output variables;
s3, performing front part structure identification on the feature vectors;
s4, performing postcursor parameter identification on the data obtained in the step S3;
s5, obtaining a T-S fuzzy model;
s6, obtaining a step response vector model of the corresponding working condition from the T-S fuzzy model obtained in the step S5;
and S7, utilizing a DMC algorithm to obtain the optimal input quantity to carry out optimal control on the output of the gas turbine.
Further, in step S1 of the present invention, a five-point cubic smoothing formula is adopted to perform data preprocessing, wherein:
let the original data be recorded as x, the filtered data be recorded as y, and the specific filtering algorithm rule is expressed as formula:
further, the feature vector combination in step S2 of the present invention is:
wherein, X 1 A combination of eigenvectors representing power, k representing the kth sampling instant, and also in discrete systems, understood as orders, representing X 2 Combination of eigenvectors of exhaust temperature, u 1 Representing the input variable fuel quantity, u 2 Representing the inlet guide vane opening, y, of the input variable 1 Representing the output variable combustion engine power, y 2 Representing the output variable exhaust temperature.
Further, the step S3 of the present invention adopts an improved FCM clustering algorithm, specifically
Comprises the following steps:
Further, the invention also includes an improved particle swarm algorithm, specifically:
the method comprises the following steps of firstly, defining variables to be identified, setting an optimization objective function and a stopping condition, randomly generating a plurality of particle populations, and initializing positions and speeds of particles;
secondly, according to the value preliminarily identified by Matlab as a reference, restricting the interval of the optimal solution of the particles and accelerating the speed of searching the optimal solution;
thirdly, calculating the fitness value of each particle, and recording the current historical optimal solution and the global optimal solution of each particle;
fourthly, determining the distance between each particle and the global optimal particle according to a formula, and respectively calculating the inertia weight and the learning factor of the particles;
fifthly, updating and recording the speed and the position of the particles, covering the original historical optimal solution and the original global optimal solution, substituting into an objective function to solve, judging whether a stopping condition is met, if so, performing the next step, otherwise, performing the third step;
sixthly, outputting the globally optimal particle individual, namely the solution of the identified parameter;
obtaining a field accurate transfer function of the gas turbine by improving a particle swarm optimization, wherein the improved transfer formula is as follows:
examples
It should be noted that, aiming at the nonlinear characteristics of multiple input/output variables, strong coupling between variables and the like of the gas turbine, a system identification method based on data driving is adopted to establish a mathematical model capable of accurately expressing the gas turbine under the condition of only input/output data, and the method is one of the methods for solving the problem of nonlinear system modeling at present. The method adopts a T-S fuzzy model identification algorithm to model the gas turbine, uses a linear fuzzy model to describe the dynamic characteristic of a nonlinear system, applies an optimization theory to fuzzy model identification, establishes a gas turbine identification strategy flow, and further obtains the T-S fuzzy model of the gas turbine.
1. Data pre-processing
The method uses a five-point cubic smoothing formula, namely, five adjacent sampling points are used for cubic polynomial fitting, data are preprocessed, the algorithm is simple, and the processing effect is good. Let the original data be x, and the filtered data be xThe specific filtering algorithm rule is expressed as follows:
the data of this paper are taken from the DCS emulation machine, have carried out the lift load experiment to gas steam combined cycle unit, have gathered the unit and have fallen earlier to 700MW from full load 800MW and then to 600 MW's experimental data, and the standby group stabilizes after a period, rises the unit from 600MW to full load 800MW again, and the centre still passes through at 700MW stage, and the rate of change of unit lift load all is 1, and data sampling cycle is 1 second. The processing formulas for the last two data points are similar to the processing of the first two points, specifically, the processing of the first two sample points in practice is replaced by the third point, the processing of the last two sample points is performed by applying a five-point cubic smoothing formula to perform data preprocessing on two input variables and two output variables of the gas turbine, and the processing results are respectively shown in fig. 1 and fig. 2.
2. Establishing feature vectors
For a multi-input multi-output variable system, data have coupling and nonlinearity, in order to know the influence of variables on the system at different moments, the input and output variables need to be analyzed by means of a cross-correlation function, the analysis index is to determine the correlation coefficient of the input variables at which moment to the system output, and the closer the value of the correlation coefficient is to 1, the greater the influence on the output is. And the correlation coefficient is analyzed, overfitting can be avoided, and the modeling precision of the model is improved, so that the characteristic vector combination of input and output variables needs to be constructed.
Through repeated experiments, the following formula is determined as a feature vector combination:
wherein, X 1 A combination of eigenvectors representing power, k representing the kth sampling instant, and also in discrete systems, understood as orders, representing X 2 Combination of eigenvectors of exhaust temperature, u 1 Representing the input variable fuel quantity, u 2 Representing the inlet guide vane opening, y, of the input variable 1 Representing the output variable combustion engine power, y 2 Representing the output variable exhaust temperature.
3. Front piece identification-improved FCM clustering algorithm
The fuzzy C-means clustering algorithm (FCM) is a flexible clustering, and the core idea is to determine the degree of each data point belonging to a certain cluster by taking the data membership as a standard, and divide the data sample points into intervals[0,1]. Firstly, randomly selecting a plurality of clustering centers, searching an optimal target weighting point according to a target function, repeatedly adjusting a fuzzy membership matrix and a clustering center matrix to obtain the minimum iteration number, selecting the optimal clustering center, and clustering the result by a fuzzy membership matrix U = [ mu ], [ mu ] ij ]E, R represents, U is a sample set, and finally all linear sections are connected through a membership function. Solving the optimal solution problem of the objective function through the Euclidean distance from the sample point to the clustering center, wherein the objective function of the FCM algorithm can be expressed as the following formula:
because the FCM algorithm has strong dependence on the distance of the sample points, the target function is disturbed by outliers and noise points in the sample cluster, and the target function is easily trapped in a local minimum value in the iteration process, so that the final model precision is influenced. Therefore, a regularization factor gamma for controlling and improving an error function is introduced to reduce the influence of outliers and noise points and improve the effect of effective sample points.
Will be (D) in the target function of the original FCM ij ) 2 The part is modified into the formula:
as shown in FIG. 3, the left and right regions of the contour are observed whenIn time, the larger the value of gamma is, the more obvious the inhibition effect on noise points and outliers is; when in useThe effect of valid data points in the sample set can also reduce the effect of the error function.
4. Back piece identification-recursive least square method parameter identification
Obtained from actual systemsInput variable matrix of (X) = [ X ] kj ](k =1,2., nj =1,2., d) and the sum output vector y = [ y ] 1 ,y 2 ,...,y n ] T Where n is the number of samples and d is the dimension. Thus, the back-piece parameters of the model can be obtained from the following matrix equation:
whereinIs an error vector, phi = [ phi ] 1 ,φ 2 ,...,φ r ](r = c × (d + 1)) is a coefficient matrix, and the formula can be seen from the foregoing description:
φ i =[φ i (x 1 ),φ i (x 2 ),...,φ i (x n )],(i=1,2,...,r)
in order to solve the above problem, if the least square method is used, a singular matrix may be encountered and the solution cannot be performed, and the recursive least square method (RLS) is generated to avoid the inversion operation of a high-order matrix, and the accuracy of parameter identification can be improved. Iterative formula for estimating parameter values:
5. Model testing
In order to test the universality and generalization capability of the model, the established model is tested, the specific steps are consistent with the training set, only the selected test set data groups are different, the test data is extracted from one section of the unit load-increasing stage, the total number of 800 groups of data is obtained, and the related data indexes are distributed as shown in the following table:
the identification result of the T-S fuzzy model of the gas turbine under the test set is shown in FIG. 4.
Under the test set, the output variable evaluation indexes of the T-S fuzzy model of the gas turbine are shown in the following table:
according to the identification result and the evaluation index of the test set model, the fitting effect of the model output and the actual output is good, the numerical values of the mean square error and the decision coefficient are also in an acceptable range, and the tested T-S fuzzy model has better model generalization capability.
6. Dynamic matrix predictive control algorithm (DMC)
Dynamic matrix control (dynami, DMC) is one of the predictive control representative algorithms which are generated in the industrial process at the earliest in model predictive control, is a predictive control algorithm based on a system step response curve, is widely applied to the industrial fields of chemical industry, energy production and the like, combines the linear programming and constraint control problems in process control, solves the problem of static optimization, and is suitable for predictive control of a controlled linear object with self balance. According to the prior art, the DMC algorithm comprises three parts, namely a predictive model, rolling optimization and feedback correction, and the invention considers that the related formula, principle and derivation process thereof are known to those skilled in the art as belonging to the prior art.
7. Transfer function
The improved particle swarm algorithm specifically comprises the following steps:
the method comprises the following steps of firstly, defining variables to be identified, setting an optimization objective function and a stopping condition, randomly generating a plurality of particle populations, and initializing positions and speeds of particles;
secondly, according to the value preliminarily identified by Matlab as a reference, restricting the interval of the optimal solution of the particles and accelerating the speed of searching the optimal solution;
thirdly, calculating the fitness value of each particle, and recording the current historical optimal solution and the global optimal solution of each particle;
fourthly, determining the distance between each particle and the global optimal particle according to a formula, and respectively calculating the inertia weight and the learning factor of the particles;
fifthly, updating and recording the speed and the position of the particles, covering the original historical optimal solution and the original global optimal solution, substituting into an objective function to solve, judging whether a stopping condition is met, if so, performing the next step, otherwise, performing the third step;
and sixthly, outputting the globally optimal particle individual, namely the solution of the identified parameter.
The improved particle swarm algorithm parameter settings used herein are respectively that the inertia weight is 0.9, both learning factors are set to 2, considering the possibility that each particle can be used as an optimal solution, the number of particles and the iteration times are set to be the same and are both 100, and the optimal objective function is set to be the sum of squares of the difference between the model identification output and the actual output. And according to the parameters of the preliminarily determined transfer function to be identified, defining a number domain of the parameter to be identified according to the boundary of [ -5,5 ]. The identified transfer function excites the original input through the lsim function in Matlab, and then is compared with the actual output processed by the zero initial value, as shown in FIG. 5, respectively, the curves have the same basic trend, and the error is in the precision requirement. Therefore, the new transfer function obtained by the above method is used as PID control of the model as shown in the following formula.
8. Dynamic matrix predictive control based on TTT fuzzy model
The fuzzy model is the first step in implementing the non-linear predictive control. The multi-variable linear system with dimensional output can be decomposed into r multi-input single-output systems according to the output number, for the first multi-input single-output system, cl clustering is obtained through fuzzy clustering, sub-model description shown in the form formula is obtained, the sub-model description is further converted into an increment form, the requirement of a DMC algorithm on control increment as a variable is met, meanwhile, the influence of constants in each linear model is eliminated, and the fuzzy rule after deformation is expressed as:
wherein,representing input vectors of the l-th multiple-input single-output system, d l As its dimension, y l For its output, a first parameter in the first submodel, an output parameter representing a pair of the first submodel, a cluster center cl and a linear model parameterAll determined by off-line model, and corresponding membership degree of each sub-modelThe system can be fused into a global model of the multi-input single-output system, and the expression is as follows:
specifically, in the research on the output control of the gas turbine, two outputs of the gas turbine can be considered separately, and a power output identification model and an exhaust temperature output identification model are respectively established and combined with a DMC algorithm. Firstly, a step response vector model corresponding to working conditions is obtained through the established T-S fuzzy model of the gas turbine, and then optimal input quantity is obtained by combining a DMC algorithm to carry out optimal control on the output of the gas turbine. The key point of the method lies in the dynamic performance characteristics of the gas turbine, and the DMC algorithm carries out feedback correction through the identification error of the model. In summary, a schematic diagram 6 of the structure of the predictive control of the gas turbine output is established.
9. Control system simulation under given value disturbance
The given value disturbance is to observe the regulation function of the prediction controller by changing the target value of the output quantity, specifically, by changing the target value of the exhaust gas temperature, the prediction controller will regulate the main control input quantities such as the fuel quantity, the IGV opening degree and the like according to a new target set value, and act the main control input quantities into a T-S fuzzy model of the gas turbine, and the actual output of the gas turbine is sent into the prediction controller, so that a closed loop feedback link is formed, and the stability of the system is increased. The schematic structural diagram of the control system under the given value disturbance is shown in FIG. 7.
For the tracking experiment of the given power value, under the stable working condition of 700MW of the unit selected herein, the corresponding power of the gas turbine is 227MW, the power is set to be 237MW, 217MW, 228MW and 235MW, respectively, each change is executed after the power output is stable, the tracking response capability of the system under the T-S combined DMC algorithm is tested by continuously setting the constant power value of one rise, one fall and two rises, the parameter setting selected by the DMC algorithm is that the control time domain is 60, the control time domain is 1, and the result combined with the PID control effect is shown in fig. 8.
As can be seen from the figure, the control effect of the T-S combined DMC algorithm is better than that of the conventional PID control, the control has good response tracking capability and basically has no steady-state error, and the corresponding regulation quality is shown in the following table:
as can be seen from the table, for the change of the power set value, the DMC algorithm based on the T-S fuzzy model can quickly respond to the power output, and has very small overshoot and short adjustment time; and the response of the conventional PID control effect is slightly slower than that of a hybrid algorithm, the overshoot is large, the conventional PID control effect is easy to sink into jitter, and the adjustment time is prolonged.
10. Control system simulation under external disturbance
The external disturbance is to add a step signal between the controller and the controlled object and observe the regulation effect of the predictive controller, and the external disturbance can reflect the disturbance processing capacity of the controlled object. In the section, an irrelevant variable such as rotating speed is used as a step signal of disturbance, two different step signals of a rising edge and a falling edge are set to simulate external disturbance when the exhaust temperature output is stable, and the adjustment and recovery capacity of the exhaust temperature under the action of a predictive controller is observed. The schematic diagram of the control system under the step signal disturbance is shown in fig. 9.
In the experiment, the exhaust temperature is increased from 622 ℃ to 632 ℃, the response index under the primary exhaust temperature transfer function is checked, for the setting of the change of the step signal in the experiment, the step response of a rising edge and a falling edge is respectively carried out when the simulation time is 500S and 1000S, the adjusted controller parameters are still kept, the processing capacity of a DMC algorithm and a conventional PID (proportion integration differentiation) on the disturbance based on a T-S fuzzy model is observed under the disturbance of the same fuel quantity, and the dynamic characteristic of the exhaust temperature under the disturbance of the step signal is shown in FIG. 10.
It is seen that the exhaust temperature fluctuates due to the change of the step signal, but the control effect of the DMC algorithm based on the T-S fuzzy model is significantly superior to that of the conventional PID control, the former can be quickly restored to the set value of the exhaust temperature, and the latter requires a certain time for restoration.
The following table shows that under the same step signal disturbance, the DMC algorithm based on the T-S fuzzy model has smaller overshoot, shows that the DMC algorithm has certain suppression capability on the disturbance, and is better than the recovery capability after the conventional PID disturbance in each time performance.
The simulation results and the experimental data are integrated, the simulation effect of the DMC algorithm based on the T-S fuzzy model is superior to the control effect of the traditional PID on target value tracking and step signal disturbance experiments, the result values on the regulation quality such as overshoot, peak time, regulation time and the like are basically smaller than the result value of the PID, and the DMC algorithm based on the T-S fuzzy model has the advantages of high regulation speed, good stability and strong disturbance suppression capability.
Various other changes and modifications to the above-described embodiments and concepts will become apparent to those skilled in the art from the above description, and all such changes and modifications are intended to be included within the scope of the present invention as defined in the appended claims.
Claims (5)
1. The method for optimizing the T-S model of the gas turbine based on the multi-model predictive control algorithm is characterized by comprising the following steps of:
s1, preprocessing data, namely performing cubic polynomial fitting by using five adjacent sampling points, preprocessing the data and improving the data quality;
s2, establishing a feature vector, namely establishing a feature vector combination of input and output variables;
s3, identifying the structure of the front part of the feature vector;
s4, performing postcursor parameter identification on the data obtained in the step S3;
s5, obtaining a T-S fuzzy model;
s6, obtaining a step response vector model of the corresponding working condition from the T-S fuzzy model obtained in the step S5;
and S7, utilizing a DMC algorithm to obtain the optimal input quantity to carry out optimal control on the output of the gas turbine.
2. The method for optimizing the T-S model of the gas turbine based on the multi-model predictive control algorithm according to claim 1, wherein in the step S1, a five-point cubic smoothing formula is adopted for data preprocessing, wherein:
let the original data be recorded as x, the filtered data be recorded as y, and the specific filtering algorithm rule is expressed as formula:
3. the method for optimizing the T-S model of the gas turbine based on the multi-model predictive control algorithm according to claim 1, wherein the combination of the eigenvectors in the step S2 is as follows:
wherein, X 1 A combination of eigenvectors representing power, k representing the kth sampling instant, and also in discrete systems, understood as orders, representing X 2 Combination of eigenvectors of exhaust temperature, u 1 Representing the input variable fuel quantity, u 2 Representing the inlet guide vane opening, y, of the input variable 1 Representing the output variable combustion engine power, y 2 Representing the output variable exhaust temperature.
5. The gas turbine T-S model optimization method based on the multi-model predictive control algorithm according to claim 1, further comprising improving a particle swarm algorithm, specifically:
the method comprises the following steps of firstly, defining variables to be identified, setting an optimization objective function and a stopping condition, randomly generating a plurality of particle populations, and initializing positions and speeds of particles;
secondly, according to the value preliminarily identified by Matlab as a reference, restricting the interval of the optimal solution of the particles and accelerating the speed of searching the optimal solution;
thirdly, calculating the fitness value of each particle, and recording the current historical optimal solution and the global optimal solution of each particle;
fourthly, determining the distance between each particle and the global optimal particle according to a formula, and respectively calculating the inertia weight and the learning factor of the particles;
fifthly, updating and recording the speed and the position of the particles, covering the original historical optimal solution and the original global optimal solution, substituting into an objective function to solve, judging whether a stopping condition is met, if so, performing the next step, otherwise, performing the third step;
sixthly, outputting the globally optimal particle individual, namely the solution of the identified parameter;
obtaining a field accurate transfer function of the gas turbine by improving a particle swarm optimization, wherein the improved transfer formula is as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211301967.4A CN115573819A (en) | 2022-10-24 | 2022-10-24 | Gas turbine T-S model optimization method based on multi-model predictive control algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211301967.4A CN115573819A (en) | 2022-10-24 | 2022-10-24 | Gas turbine T-S model optimization method based on multi-model predictive control algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115573819A true CN115573819A (en) | 2023-01-06 |
Family
ID=84586319
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211301967.4A Pending CN115573819A (en) | 2022-10-24 | 2022-10-24 | Gas turbine T-S model optimization method based on multi-model predictive control algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115573819A (en) |
-
2022
- 2022-10-24 CN CN202211301967.4A patent/CN115573819A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Heidary et al. | Survey on microgrids frequency regulation: Modeling and control systems | |
Wang et al. | Self-adaptive discrete grey model based on a novel fractional order reverse accumulation sequence and its application in forecasting clean energy power generation in China | |
Huang et al. | LSTM-MPC: A deep learning based predictive control method for multimode process control | |
CN107870564B (en) | Anti-interference control method for fuel cell with decoupling performance | |
CN112947672B (en) | Maximum power point tracking method and device for photovoltaic cell | |
CN116070769A (en) | Ultra-short-term wind power plant power multi-step interval prediction modularization method and device thereof | |
Tian et al. | Generalized predictive PID control for main steam temperature based on improved PSO algorithm | |
CN114722693A (en) | Optimization method of two-type fuzzy control parameter of water turbine regulating system | |
CN115986839A (en) | Intelligent scheduling method and system for wind-water-fire comprehensive energy system | |
Xu et al. | Design of type-2 fuzzy fractional-order proportional-integral-derivative controller and multi-objective parameter optimization under load reduction condition of the pumped storage unit | |
CN112015082A (en) | Machine furnace coordination system control method based on fuzzy gain scheduling prediction control | |
CN116865343B (en) | Model-free self-adaptive control method, device and medium for distributed photovoltaic power distribution network | |
CN106681133A (en) | Method for identifying hydroelectric generating set model improved type subspace closed loop | |
CN113393119A (en) | Stepped hydropower short-term scheduling decision method based on scene reduction-deep learning | |
CN108803342B (en) | Unit unit load quick response prediction control method | |
CN115573819A (en) | Gas turbine T-S model optimization method based on multi-model predictive control algorithm | |
CN116430726A (en) | Method for controlling turboset based on subtractive clustering and fuzzy neural network | |
CN116154875A (en) | Photovoltaic power station active power optimization distribution method based on TCN and error function | |
CN113189871B (en) | Thermal power generating unit flexible operation-oriented coordination control strategy | |
Li et al. | A gray rbf model improved by genetic algorithm for electrical power forecasting | |
CN110850710A (en) | Hydroelectric generating set control optimization method based on model-free adaptive control | |
Ge et al. | An Improved Distributed Maximum Power Point Tracking Technique in Photovoltaic Systems Based on Reinforcement Learning Algorithm | |
CN112699600B (en) | Thermal power operating parameters and NOxMethod for analyzing partial return between emission concentrations | |
CN111273563B (en) | Prediction control method based on AGC (automatic gain control) comprehensive index of heat supply unit | |
Guo et al. | Neural Network Modeling and Predictive Control of Low Oxygen Combustion System in Petrochemical Heating Furnace |
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 |