CN101328836A - Multi-model self-adapting generalized forecast control method of gas turbine rotary speed system - Google Patents

Abstract

The invention discloses a method for controlling a multi-model self-adapting generalized prediction of a rotating speed system of a gas turbine, which is a control strategy for multi-model self-adapting generalized prediction, which is a method for adjusting a rotating speed system of a rotor of the gas turbine to make the rotating speed of the rotor fast and stable and have an agonic tracking set value. The method adopts a recursive least square identification method to obtain a model of a typical working condition point to form a multi-model collection of the rotating speed system of the gas turbine, and further designs each controller for sub-prediction in connection with each model; control quantity output by each sub-controller is weighted as actual control quantity of the rotating speed of the gas turbine to control the rotating speed of the gas turbine to realize optimality at a global scope, thereby overcoming the defect that the conventional control makes the system respond fast when a system loading point of the gas turbine has an abrupt change; and the controller of the control method has simple design, the switches among the sub-controllers are easy for engineering realization, thereby solving the control problems of nonlinearity and nonstationarity of the object parameter of the rotating speed of the rotor, and parameter jumping under a large-scale working condition.

Description

The multi-model self-adapting generalized forecast control method of gas turbine rotary speed system

Technical field

The present invention is a kind of multi-model self-adapting generalized predictive control strategy, and the gas turbine rotor rotary speed system is regulated, make rotor speed fast, a kind of method of stable, agonic tracking setting value, belong to thermal technology's automation field.

Background technique

Series of advantages such as combustion engine power station is efficient with it, environmental protection, energy-conservation, water saving have played the effect that becomes more and more important in power field.Rotor speed is most important control parameters in the gas turbine control system, and rotor speed is directly determining the power quality of production, keeps combustion machine stabilization of speed extremely important at rating value.The combustion machine is a complicated nonlinear systems, especially the dynamo-electric factory of most combustions is bearing the peak load regulation network task, the frequent wide variation of its operating mode, the unit nonlinear characteristics highlights more, especially add/slow down and and network process in, the stability requirement of rotating speed control and the rapid response of variation require this more obvious to contradiction, and conventional pid control algorithm often is difficult to satisfy not only fast but also steady requirement, so research and design advanced person's gas turbine rotary speed control algorithm has the meaning of particular importance.

Generalized predictive control is one of advanced control strategy of tool application and popularization value in process control industries, but conventional predictive control strategy often can not be adapted to overall operating mode.And multi-model process is the important development direction of self adaptive control, is one of research focus of current advanced control theory, is to solve a kind of effective method of dynamic characteristic with the complex industrial process Control of Nonlinear Systems of working conditions change.

For improving the gas turbine rotary speed controlling performance, the present invention is directed to the closely-related actual features of gas turbine rotary speed system nonlinear characteristics and operating conditions, gas turbine rotary speed control strategy based on multi-model self-adapting generalized predictive control has been proposed, multi-model self-adapting generalized predictive control is introduced the gas turbine rotor rotary speed system first, utilize the multi-model modeling method to approach the dynamic performance of the machine of operating mode process combustion on a large scale rotary speed system, design overall adaptive controller based on multi-model again, thereby combustion machine rotating speed is effectively controlled.

Summary of the invention

Technical problem: the multi-model self-adapting generalized forecast control method that the purpose of this invention is to provide a kind of gas turbine rotary speed system, be used for the gas turbine rotor rotary speed system, promptly solve gas turbine rotary speed system parametrical nonlinearity, time variation, on a large scale under the operating mode parameter saltus step etc. the method for problem.

Technological scheme: in order to overcome the problems referred to above, by adopting multi-model self-adapting generalized predictive control, remedy the deficiency of single model predictive control, be used for gas turbine rotary speed system, make system response time fast, performance of dynamic tracking is good, no regulating error.

The technological scheme of multi-model self-adapting generalized Predictive Control System can adopt following steps to realize:

Step 1: when the gas turbine rated load and under the open loop stabilization operating mode, control system is provided with the typical condition point, apply low level pseudo-random signal excitation at the speed regulator output terminal, record the rotation speed change data, pick out the speed dynamic model with the recursion generalized least square method, constitute the multi-model set of combustion machine rotary speed system;

Step 2: at the speed dynamic model of step 1, adopt the sub-predictive controller of generalized predictive control constructing tactics, and the parameter of corresponding sub-predictive controller is adjusted;

Step 3: if the multi-model that can not satisfy step 1 based on the sub-predictive controller of step 2 is gathered the overlapping of Satisfactory Control scope between adjacent model, set up the partial model family of gas turbine, the multi-model set of step 1 is replenished; Otherwise enter step 4;

Step 4: control system is checked each sampling instant duty parameter situation of change, and the output of sub-predictive controller control increment is weighted working control increment as gas turbine rotary speed.

Step 1 control system is provided with 100%, 80%, 50% load typical condition point, and the model of these three typical condition points approaches the dynamic performance of gas turbine rotary speed system under overall operating mode.

The working control increment of step 4 gas turbine rotary speed is obtained by following weighted strategy by the output control increment of sub-predictive controller:

1. σ _i≤ σ＜σ _jThe time, Δ u=(1-x _j) Δ u _i+ x _jΔ u _j, $x_j=\frac{\mathrm{\σ}-{\mathrm{\σ}}_i}{{\mathrm{\σ}}_j-{\mathrm{\σ}}_i};$

2. 0＜σ＜σ ₁The time, Δ u=Δ u ₁

3. σ ₃During≤σ, Δ u=Δ u ₃

Wherein, σ ₁=50%, σ ₂=80%, σ ₃=100%, i=1,2, j=i+1, Δ u _i, Δ u _jBe respectively the output control increment of the sub-predictive controller of step 4.

Beneficial effect: utilize the multi-model modeling method to approach the dynamic performance of operating mode procedures system on a large scale, design overall adaptive controller based on multi-model again, act on the gas turbine rotor rotary speed system, make system responses rapid, no dynamic deviation, effectively overcome single model predictive control insurmountable gas turbine rotary speed system parametrical nonlinearity, time variation, on a large scale under the operating mode parameter saltus step etc. problem.

Description of drawings

Fig. 1 is based on the multi-model predictive controller of weighting scheme,

Fig. 2 is a gas turbine rotary speed control system schematic representation.

Embodiment

The present invention a kind ofly adopts multi-model self-adapting generalized predictive control at gas turbine rotor rotary speed system parametrical nonlinearity, time variation, parameter saltus step under the operating mode on a large scale, makes the control system response rapidly, the method for no dynamic deviation.Specific implementation method is,

Step 1: when the gas turbine rated load and under the open loop stabilization operating mode, apply low level pseudo-random signal excitation at the speed regulator output terminal, record many group rotation speed change data, pick out speed dynamic model under declared working condition point with the recursion generalized least square method, with the speed dynamic model that obtains other typical condition points with quadrat method, the rotating speed model of a plurality of typical condition points constitutes the multi-model set of combustion machine rotary speed system.

The model that the present invention sets up 100%, 80%, 50% these three typical condition points of load approaches the dynamic performance of gas turbine rotary speed system under overall operating mode, set up the partial model family of gas turbine, if can not satisfy the overlapping of Satisfactory Control scope between adjacent model based on many group controllers of model collection design, then the model collection be replenished.

Step 2: at the rotating speed model of above each typical condition point, adopt each sub-predictive controller of generalized predictive control constructing tactics, and respectively at these typical condition points, parameter to corresponding sub-predictive controller is adjusted, and reaches optimum up to the gas turbine rotary speed at each typical condition point;

Step 3: in each sampling instant, control system is checked current working parameter situation of change, and the controlled quentity controlled variable of each sub-controller output is weighted working control amount as gas turbine rotary speed, and the control gas turbine rotary speed is realized optimum in global scope.

The working control increment of gas turbine rotary speed is obtained by following weighted strategy by the control increment of sub-controller:

1. σ _i≤ σ＜σ _jThe time, Δ u=(1-x _j) Δ u _i+ x _jΔ u _j, $x_j=\frac{\mathrm{\σ}-{\mathrm{\σ}}_i}{{\mathrm{\σ}}_j-{\mathrm{\σ}}_i};$

2. 0＜σ＜σ ₁The time, Δ u=Δ u ₁

3. σ ₃During＜σ, Δ u=Δ u ₃

Wherein, σ ₁=50%, σ ₂=80%, σ ₃=100%, i=1,2, j=i+1, Δ u _i, Δ u _jBe respectively each sub-controller output control increment.

Under gas turbine rated load open loop stabilization operating mode, apply low level pseudo-random signal (PRBS) excitation at the speed regulator output terminal, record many group rotation speed change data, according to the inputoutput data of experiment gained, pick out speed dynamic model under declared working condition point with the recursion generalized least square method.But with the speed dynamic model that obtains two other typical condition point with the quadrat method identification.At the rotating speed model under above three kinds of typical conditions, on industrial control software WINCC 6.0, design sub-GPC controller respectively, set up the multi-model control system according to Fig. 1, replace PID rotational speed governor output control fuel control valve, finally regulate rotating speed with the multi-model global controller.In the implementation process, adjust, realize optimal control based on the combustion machine rotating speed of multi-model self-adapting generalized predictive control by antithetical phrase GPC controller important parameter.Conventional PID and multi-model self-adapting generalized predictive control rotating speed response are compared, prove that multi-model self-adapting generalized predictive control has combustion machine rotating speed to control effect preferably.

Claims (3)

1, a kind of multi-model self-adapting generalized forecast control method of gas turbine rotary speed system, this method comprises the steps it is characterized in that:

Step 1: when the gas turbine rated load and under the open loop stabilization operating mode, control system is provided with the typical condition point, apply low level pseudo-random signal excitation at the speed regulator output terminal, record the rotation speed change data, pick out the speed dynamic model with the recursion generalized least square method, constitute the multi-model set of combustion machine rotary speed system;

Step 2: at the speed dynamic model of step 1, adopt the sub-predictive controller of generalized predictive control constructing tactics, and the parameter of corresponding sub-predictive controller is adjusted;

Step 3: if the multi-model that can not satisfy step 1 based on the sub-predictive controller of step 2 is gathered the overlapping of Satisfactory Control scope between adjacent model, set up the partial model family of gas turbine, the multi-model set of step 1 is replenished; Otherwise enter step 4;

Step 4: control system is checked each sampling instant duty parameter situation of change, and the output of sub-predictive controller control increment is weighted working control increment as gas turbine rotary speed.

2, gas turbine rotary speed system multi-model self-adapting generalized forecast control method according to claim 1, it is characterized in that step 1 control system is provided with 100%, 80%, 50% load typical condition point, the model of these three typical condition points approaches the dynamic performance of gas turbine rotary speed system under overall operating mode.

3, gas turbine rotary speed system multi-model self-adapting generalized forecast control method according to claim 1, the working control increment that it is characterized in that step 4 gas turbine rotary speed is obtained by following weighted strategy by the output control increment of sub-predictive controller:

1. σ _i≤ σ＜σ _jThe time, Δ u=(1-x _j) Δ u _i+ x _jΔ u _j, $x_j=\frac{\mathrm{\σ}-{\mathrm{\σ}}_i}{{\mathrm{\σ}}_j-{\mathrm{\σ}}_i};$

2. 0＜σ＜σ ₁The time, Δ u=Δ u ₁

3. σ ₃During≤σ, Δ u=Δ u ₃

Wherein, σ ₁=50%, σ ₂=80%, σ ₃=100%, i=1,2, j=i+ ₁, Δ u _i, Δ u _jBe respectively the output control increment of the sub-predictive controller of step 4.

Images