Vibration reduction method and device for steam turbine of thermal power plant

Abstract

The invention relates to a thermal power plant steam turbine vibration reduction method and a vibration reduction device, and aims to overcome the inherent defects that a traditional differential evolution algorithm is slow in convergence speed and easy to fall into local optimum when approaching a globally optimal solution, an L-BFGS algorithm is innovatively introduced for improvement, and a DE-L-BFGS hybrid optimization algorithm is formed. According to the hybrid algorithm, the defects are effectively overcome, and the calculation efficiency and the global optimization capacity of parameter optimization of the TMD are greatly improved. A DE-L-BFGS hybrid algorithm is applied, collaborative optimization screening is specially carried out on two core parameters of rigidity and damping of the TMD, after the optimized parameter combination is applied to the TMD, rigorous vibration reduction effect verification and actual engineering application show that the vibration reduction performance of the TMD is remarkably superior to that of a state before optimization, and the TMD has a good vibration reduction effect. The effectiveness of the hybrid algorithm in the aspect of improving the vibration reduction efficiency of the TMD is fully verified.

Description

Technical Field

[0001] The present invention relates to the technical field of steam turbine vibration reduction, in particular to a method and a device for reducing vibration of a steam turbine in a thermal power plant. Background Art

[0002] Steam turbines, the heart of thermal power plants, are subject to stringent safety and stability requirements, with vibration being a key performance indicator. Severe vibration can cause unexpected turbine shutdowns at best, or even equipment damage or even casualties. Therefore, diagnosing and addressing vibration faults in thermal power plants is crucial. However, due to the influence of operating parameters, structure, and characteristics, vibration faults can vary in frequency. If these faults are not effectively addressed during operation, long-term operation can lead to unexpected equipment shutdowns or damage.

[0003] A tuned mass damper (TMD) is a vibration reduction system consisting of a mass, spring, and damper. When equipment experiences significant vibration, the TMD generates a reaction force that significantly reduces the vibration. Common TMD vibration reduction devices require a customized design based on a specific vibration fault frequency and the equipment structure. Their mass, stiffness, and damping are not adjustable, and they can only address faults at a single turbine vibration frequency, failing to address faults at other vibration frequencies. Furthermore, traditional thermal power plant turbine vibration protection is divided into two directions: horizontal and vertical. Exceeding the standard in either direction can cause equipment to unexpectedly stop and potentially be damaged. However, typical TMD vibration reduction devices can only address vibration in a specific direction. Therefore, improvement and innovation are imperative. Summary of the Invention

[0004] In view of the above situation and to overcome the defects of the prior art, the purpose of the present invention is to provide a vibration reduction method and vibration reduction device for a thermal power plant steam turbine, which can realize manual adjustment of TMD mass and self-adjustment of damping and stiffness. However, there is a mutual influence between the damping and stiffness parameters, and an optimization algorithm needs to be adopted. Optimization should be carried out according to the actual vibration conditions of the steam turbine to screen out the optimal parameter combination.

[0005] The technical solution provided by the present invention is a method for reducing vibration of a steam turbine in a thermal power plant, comprising the following steps:

[0006] Step 1: Obtain the main structure vibration characteristic information and set the TMD mass m

[0007] According to the design parameters of the steam turbine bearings in thermal power plants, the mass M, stiffness K, and damping C of the main structure are determined, and the mass m of the TMD is set so that m = μM (the mass ratio μ is generally 0.01-0.05);

[0008] Clip leads are used to extract signals from turbine bearing position sensors (speed sensors, eddy current sensors, etc.), and the obtained vibration signal F0(t) is used as external excitation for the main structure. Sensors are installed at each turbine bearing position, so this extraction method eliminates the need for additional sensors.

[0009] Step 2: Taking the main structure displacement time history x(t) as the target, determine the fitness function f

[0010] 2.1 After numerical calculation, the complex displacement time history x(t) of the main structure is obtained:

[0011]

[0012] Where: x m , are the displacement, velocity and acceleration of the main structure respectively;

[0013] x t , are the displacement, velocity, and acceleration of TMD, respectively;

[0014] m, k, and c are the mass, stiffness, and damping of the TMD, respectively;

[0015] 2.2 The fitness function f is determined with the goal of minimizing the peak displacement of the main structure min|x(t)|, that is:

[0016] f(k,c)=min|x(t)|

[0017] 2.3 To ensure that the optimized parameters do not cause the TMD to exceed the mechanical limit, a penalty function is introduced:

[0018]

[0019] Therefore, the fitness function is:

[0020] f(k,c)=min|x(t)|+λ·Penalty(k,c)

[0021] Among them, λ is the penalty factor, ranging from 0 to 1, and the constraint is strengthened by increasing λ;

[0022] Step 3: DE-L-BFGS hybrid algorithm optimizes TMD parameters

[0023] 3.1 Population initialization G=0,G=1,2,3,…,MaxGen,where G is the number of iterations, and the maximum number of iterations MaxGen=100; the population size is selected as NP=30; according to the stiffness and damping size of the main structure, the optimization boundary conditions of k and c of TMD can be manually set to ensure that k∈[k min ,k max],c∈[c min , c max ], where k min ,k max are the lower and upper limits of k, c min , c max are the lower and upper limits of c respectively;

[0024] 3.2 Randomly generate initial individual X i =(k i ,c i ), that is, the stiffness and damping parameters corresponding to TMD:

[0025] k i =k min +rand(0,1)·(k max -k min )

[0026] c i =c min +rand(0,1)·(c max -c min )

[0027] Among them, rand(0,1) is a uniformly distributed random number, k i 、c i are randomly generated individual X i Stiffness and damping, i = 1, 2, 3, ..., NP;

[0028] 3.3 For each individual X i =(k i , c i ), calculate x i (t);

[0029] 3.4 Calculate the fitness function value: f(k i , c i )=min|x(t)|+0.5·Penalty(k i , c i ).

[0030] 3.5 Start Differential Evolution (DE) iteration

[0031] 1) Mutation: for each generation of individual X i Use DE / rand / 1 strategy to generate mutation vector V i :

[0032] V i =X r1 +F·(X r2 -X r3 )

[0033] Among them, r1, r2, and r3 are different random indexes; the scaling factor F is 0.8 (generally between 0.5 and 1.0). The larger F is, the stronger the exploratory nature (global search); the smaller F is, the stronger the exploitative nature (local search).

[0034] 2) Crossover: In order to increase the diversity of the population, the mutation vector V i and the target vector X i Perform mixing and generate the test vector U through binomial crossover i =(u i，k ,u i,c ):

[0035]

[0036] Among them, the crossover probability CR is 0.9 (generally between 0 and 1.0). The larger the CR is, the more components the test vector inherits from the mutation vector, the higher the population diversity, the faster the convergence, but it may also skip the optimal solution; the smaller the CR is, the more components the test vector inherits from the target vector, the stronger the development, but the convergence may become slower; u i,k 、u i,c are the i-th trial vector U i stiffness and damping.

[0037] 3) Boundary processing, truncating out-of-range parameters:

[0038] u i,k =max(k minn ,min(k max ,u i,k ))

[0039] u i,c =max(c min ,min(c max ,u i，c ))

[0040] 4) Selection, comparison test vector U i With the original individual X i Adaptability:

[0041]

[0042] 3.6 Determine if conditions are met

[0043] Let t = 1, Tc=20, where: Tc is the trigger condition for executing the L-BFGS local optimization algorithm, which is executed every 20 generations, that is: Tc=20; t is the number of triggers; is the initial optimal fitness function value.

[0044] Determine whether i>NP is true

[0045] 1) If established,

[0046] Let G=G+1, and determine whether G>MaxGen holds.

[0047] If it is true, end the loop, output the results, and select the optimal parameters k and c;

[0048] If not, determine whether G / (Tc×t)=1.

[0049] If so, proceed to step 3.7. After completion, set t = t + 1 and return to step 3.5.

[0050] If not, return to step 3.5;

[0051] 2) If not established,

[0052] Set i = i + 1 and return to step 3.5.

[0053] 3.7 L-BFGS local optimization

[0054] The DE algorithm suffers from a drawback: when approaching the global optimum, its convergence rate slows due to a fixed step size, making it difficult to achieve a high-precision solution. The quasi-Newton method (L-BFGS) is the most commonly used method for solving unconstrained nonlinear programming problems, offering advantages such as fast convergence and low memory usage. Therefore, using the L-BFGS algorithm to locally optimize the DE algorithm can effectively overcome the DE algorithm's drawbacks of slow convergence and low-precision solutions.

[0055] A method of optimizing the DE algorithm using the L-BFGS algorithm is proposed, which effectively improves the speed and accuracy of optimizing the TMD parameters.

[0056] 1) Set L-BFGS parameters: Convergence tolerance is set to ∈ = 10 -6 , the maximum number of iterations is set to N = 20;

[0057] 2) Sort the fitness values ​​of all current individuals in ascending order, select the first m fitness functions and the corresponding X i , take m = 5, and take the first m X i The individuals constitute the initial vector Y0.

[0058] 3) Calculate the initial vector Y0 using the L-BFGS algorithm:

[0059]

[0060] in: is the gradient of the objective function; H nis the approximate Hessia inverse matrix; α n is the step size, satisfying the Wolfe condition; n = 0, 1, 2, ..., N;

[0061] The projection method is used to ensure that the iteration point is always within the feasible region:

[0062] Y n+1 =clip(Y n+1 ,[k min ,c min ],[k max , c max ])

[0063] That is: k=max(k min ,min(k max ,u i,k )), c=max(c min ,min(c max ,u i,c ))

[0064] when Or when n>N, the optimal individual Y is obtained n ,make

[0065] 4) Individuals after L-BFGS algorithm optimization Calculate the optimal fitness

[0066] 5) If (Original individual fitness), then Replace the original individual Right now:

[0067]

[0068] 3.8 Screening the optimal TMD parameters

[0069] After optimization by DE-L-BFGS hybrid algorithm, the best individual As the optimal parameters of TMD stiffness and damping, the optimal individual is input into the TMD stiffness and damping adjustment execution controller;

[0070] Step 4: TMD parameter adjustment

[0071] 1) TMD mass adjustment: The mass of the TMD central mass is changed by adding or removing the counterweight;

[0072] 2) TMD stiffness adjustment: The optimal stiffness k obtained by optimizing the DE-L-BFGS hybrid algorithm is input into the integrated controller and converted into a corresponding current to control the operation of the servo motor (first motor). The ball screw (first ball screw) converts the servo motor's rotational work into linear motion, thereby adjusting the compression or extension of the spring to achieve the purpose of controlling the spring preload and changing the spring stiffness. The spring preload is detected in real time by a strain gauge sensor, and the output electrical signal is fed back to the integrated controller to calculate the deviation (Δ = ±1%) to complete the closed-loop control of the system.

[0073] 3) TMD damping adjustment: The optimal damping c, obtained by optimizing the DE-L-BFGS hybrid algorithm, is input into the integrated controller and converted into a corresponding current to control the operation of the servo motor (second motor). The ball screw (second ball screw) converts the servo motor's rotational work into linear motion, thereby adjusting the distance between the permanent magnet and the conductor plate, thereby controlling the magnetic force between the conductor and the permanent magnet and changing the damping. The distance between the permanent magnet and the conductor is detected in real time by a distance sensor, and the output electrical signal is fed back to the integrated controller. The deviation (Δ = ±1%) is calculated to complete the closed-loop control of the system.

[0074] Preferably, the value of λ in step 2 is 0.5.

[0075] The differential evolution (DE) algorithm is based on the evolutionary philosophy of genetic algorithms. Its operational principle is that individuals in a population continuously compete and improve to generate temporary individuals, which are then randomly recombined to achieve population evolution, ultimately achieving the optimal solution to the target problem. This algorithm primarily includes initialization, mutation, crossover, and selection steps. Due to its strong search capabilities in optimization problems with significant uncertainty, it is widely used in industrial production. However, the DE algorithm suffers from the drawback that when approaching the global optimum, the fixed step size causes the convergence rate to decrease, making it difficult to achieve a high-precision solution. Consequently, in practical applications, the stiffness and damping of the TMD cannot be quickly adjusted to the optimal values, significantly reducing the vibration reduction effect and posing the risk of turbine failure or equipment damage due to excessive vibration. The quasi-Newton method (L-BFGS) is a method for solving unconstrained nonlinear programming problems. It has the advantages of fast convergence and low memory usage, and can be used to improve the DE algorithm.

[0076] The present invention achieves significant beneficial technical effects through the combination of algorithm innovation and device structure design, which are specifically embodied in:

[0077] 1. Significantly Improved Algorithm Performance: Addressing the inherent drawbacks of traditional differential evolution (DE) algorithms, such as slow convergence and susceptibility to local optima when approaching the global optimal solution, the L-BFGS algorithm was innovatively introduced to improve upon this, resulting in a DE-L-BFGS hybrid optimization algorithm. This hybrid algorithm effectively overcomes these drawbacks, significantly improving the computational efficiency and global optimization capabilities of tuned mass damper (TMD) parameter optimization.

[0078] 2. Excellent TMD parameter optimization: The DE-L-BFGS hybrid algorithm was applied to collaboratively optimize and screen the two core parameters of the TMD: stiffness and damping. After applying the optimized parameter combination to the TMD, rigorous vibration reduction validation and actual engineering applications demonstrated significantly better vibration reduction performance than before optimization, fully demonstrating the effectiveness of this hybrid algorithm in improving the TMD's vibration reduction performance.

[0079] 3. Innovative and practical design of adaptive vibration damping device: A TMD vibration damping device specifically designed for steam turbines in thermal power plants has been designed. Its core advantages are:

[0080] (1) Intelligent parameter adaptation: Based on the mass and vibration characteristics of the main structure (turbine), the device can manually adjust the TMD mass; more importantly, the control device integrating the DE-L-BFGS hybrid algorithm can automatically adjust the stiffness and damping parameters of the TMD according to the operating conditions, achieving adaptive matching and dynamic optimization of stiffness and damping.

[0081] (2) Wide applicability: Through the above-mentioned adaptive adjustment mechanism, the device can effectively cope with various operating conditions and complex vibration problems under various fault frequencies that occur during turbine operation, demonstrating excellent engineering applicability and universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a schematic diagram of the overall concept of the present invention.

[0083] Figure 2 Flow chart of the method of the present invention.

[0084] Figure 3 This is the TMD parameter optimization flow chart of the present invention.

[0085] Figure 4 Schematic diagram (cross-sectional view) of the TMD structure of the present invention.

[0086] Figure 5 This is a vibration displacement trend diagram of the main structure of an application example of the present invention.

[0087] Figure 6 This is the frequency response characteristic diagram of the main structure of the application example of the present invention.

[0088] Figure 7 This is a schematic diagram of the vibration condition of an application example of the present invention before TMD is installed.

[0089] Figure 8 This is a vibration spectrum diagram of the application example of the present invention before TMD is installed.

[0090] Figure 9 This is a schematic diagram of the vibration condition after TMD is installed in the application example of the present invention.

[0091] Figure 10 This is the vibration spectrum diagram after TMD is installed in the application example of the present invention. DETAILED DESCRIPTION

[0092] The specific embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0093] This paper designs a vibration reduction device for steam turbines in thermal power plants. It uses an advanced integrated controller and uses the improved differential evolution (DE-L-BFGS) algorithm to screen out the optimal stiffness and damping combination parameters to achieve self-adjustment of TMD stiffness and damping. The device can be applied to vibration reduction in various operating conditions, various fault frequencies, and multiple directions of steam turbines. It has the advantages of easy installation, rapid transplantation, and no need for customized design. The overall concept diagram is shown as follows: Figure 1 The implementation process is as shown in Figure 2 As shown; the optimization flow chart is as follows Figure 3 As shown; the schematic diagram of TMD structure is shown Figure 4As shown, the present invention also discloses a vibration reduction device for a steam turbine in a thermal power plant, including a TMD (tuned mass damper), the TMD including a shell 1 and a mass block 2 arranged in the shell, the mass block 2 is placed in the middle of the internal space of the shell through an adjustment mechanism A arranged on the front, rear, left and right four sides thereof, the four sides of the mass block 2 are respectively covered with a conductor plate 3, the adjustment mechanism includes a first motor 4, a first ball screw, a first bearing seat 7, a second motor, a second ball screw, and a second bearing seat 15, the first motor 4 is fixed to the center of the side wall of the shell, the first ball screw includes a first ball screw located inside the shell and coaxially connected to the first motor The first screw 6 on the output shaft and the first sliding nut 5 provided on the first screw convert the rotary motion of the screw shaft into the linear motion of the first sliding nut by the rolling of the balls of the first ball screw between the first screw (screw shaft) and the first sliding nut, thereby driving the permanent magnet plate to slide and adjusting the distance between the permanent magnet plate and the conductor plate, so as to achieve the purpose of controlling the magnetic force between the conductor plate and the permanent magnet plate and changing the damping; a permanent magnet plate 8 is fixed on the side of the first sliding nut 5 facing the mass block, facing the conductor plate on the corresponding side of the mass block, and the permanent magnet plate 8 is fixed to the first sliding nut and slides along the axis of the first screw with it, and a hole is opened in the center for The first screw passes through the center hole, and the end of the first screw away from the first motor is rotatably set on the first bearing seat 7. The first bearing seat is fixed in the housing. The side of the first bearing seat 7 away from the first motor is fixedly connected to a spring guide tube 9 coaxially arranged with the first screw. A stiffness adjustment spring 10 is sleeved on the spring guide tube. The second motor 11 is fixed to the housing on one side of the first motor. The second ball screw includes a second screw 13 located inside the housing and coaxially connected to the output shaft of the second motor 11 and a second sliding nut 12 provided on the second screw. A spring compression plate 14 is fixed to the side of the second sliding nut 12 facing the mass block. The spring compression plate 14 is fixed to the side of the second sliding nut 12 facing the mass block. One end of the plate 14 away from the second sliding nut is slidingly connected to the spring guide tube, one end of the stiffness adjustment spring 10 is connected to the conductor plate corresponding to the mass block, and the other end is connected to the spring compression plate 14. The ball of the second ball screw rolls between the second screw (screw shaft) and the second sliding nut to realize the conversion of the rotational motion of the screw shaft into the linear motion of the second sliding nut. The second sliding nut drives the spring compression plate to slide, thereby adjusting the compression or extension of the spring to achieve the purpose of controlling the spring preload and changing the spring stiffness. The end of the second screw away from the second motor is rotatably set on the second bearing seat 15, and the second bearing seat is fixed in the housing.

[0094] In order to ensure the use effect, the stiffness adjustment spring 10 is provided with a strain gauge sensor for real-time detection of its stress condition, which is used to detect the preload force of the stiffness adjustment spring in real time;

[0095] A counterweight loading bolt (not shown) can be provided on the top of the mass block 2, and a counterweight block can be provided. The counterweight block can be tightened and fixed on the mass block by a nut, thereby changing the weight of the mass block;

[0096] The conductor plate is provided with a distance measuring sensor for detecting the distance between the permanent magnet plate and the conductor plate in real time.

[0097] The vibration reduction device also includes a controller, the input end of the controller is respectively connected to the signal output end of each bearing position and speed sensor of the turbine, the output end of the strain gauge sensor, and the output end of the distance measuring sensor, and the output end of the controller is respectively connected to the input end of the first motor and the second motor.

[0098] Verification of vibration reduction effect of DE-L-BFGS hybrid algorithm optimizing TMD parameters:

[0099] To verify the effectiveness of DE-L-BFGS in optimizing the two parameters of TMD, an external force with an excitation frequency of f = 3 Hz was applied to the main structure, the mass ratio μ = 0.03 was selected, and the main structure parameters were set as shown in Table 1.

[0100] Table 1 Main structural parameters

[0101]

[0102] After DE-L-BFGS optimization of the two TMD parameters, the optimal stiffness and damping were 57691N / m and 100N·s / m, respectively.

[0103] Taking the main structure vibration displacement time history as the target, the vibration reduction effects and frequency response characteristics of the main structure without TMD, with unoptimized TMD and with optimized TMD were compared. Figure 4 、 5 shown.

[0104] Depend on Figure 5 、 6 As can be seen, the main structure without the TMD has the largest vibration displacement, followed by the main structure with the unoptimized TMD. The main structure with the optimized TMD has the smallest vibration displacement, indicating the best vibration reduction effect. Furthermore, the installation of the TMD did not change the frequency characteristics of the main structure, only reducing the vibration amplitude. Verification results show that the DE-L-BFGS hybrid algorithm can optimize both TMD parameters and effectively suppress the main structure's vibration. Therefore, the DE-L-BFGS hybrid algorithm achieves a balance between global search and local optimization, significantly improving the optimization efficiency of TMD parameters and ensuring the feasibility and high performance of the engineering design.

[0105] The present invention has achieved the same or similar technical effects through practical application, and the application examples are as follows:

[0106] At present, the capacity of large domestic steam turbine generator sets is 300MW, 600MW, and 1000MW. The overall quality of their bearings varies. Taking the low-pressure cylinder support bearing box of Harbin Steam Turbine's 600MW capacity unit as an example, the vibration displacement of the turbine bearing seat, the relative vibration displacement of the rotating shaft in the X and Y directions, and the key phase signal are obtained from the turbine instrument monitoring system TSI. The vibration test analyzer used is the SK9172 model produced by Shanghai Shuke Measurement and Control Instrument Co., Ltd. During the operation of the turbine, the vibration of the No. 4 bearing continued to increase, with the maximum increase reaching 55μm (such as Figure 7 8). The steam turbine had no opportunity to be shut down for processing. To ensure equipment safety, the site decided to install the TMD device of the present application to reduce the vibration of the No. 4 bearing. The structural parameters of the No. 4 bearing box are shown in Table 2.

[0107] Table 2 Main structural parameters of No. 4 bearing box

[0108]

[0109] After the TMD was installed on site, the vibration of bearing No. 4 decreased significantly and was basically stable at around 17μm (e.g. Figure 9 、 10 This effectively ensures the safe operation of the equipment.

Claims

1. A method for reducing vibration of a steam turbine in a thermal power plant, characterized in that: The following steps are involved: Step 1: Obtain the main structure vibration characteristic information and set the TMD mass m According to the design parameters of the steam turbine bearing of the thermal power plant, the mass M, stiffness K and damping C of the main structure are determined, and the mass m of the TMD is set so that m = μM; The signal of the turbine bearing position sensor is led out using a clip lead, and the obtained vibration signal F0(t) is used as the external excitation of the main structure; Step 2: Taking the main structure displacement time history x(t) as the target, determine the fitness function f 2.1 After numerical calculation, the complex displacement time history x(t) of the main structure is obtained: Where: x m , are the displacement, velocity and acceleration of the main structure respectively; x t , are the displacement, velocity, and acceleration of TMD, respectively; m, k, and c are the mass, stiffness, and damping of the TMD, respectively; 2.2 The fitness function f is determined with the goal of minimizing the peak displacement of the main structure min|x(t)|, that is: f(k,c)=min|x(t)| 2.3 To ensure that the optimized parameters do not cause the TMD to exceed the mechanical limit, a penalty function is introduced: Therefore, the fitness function is: f(k,c)=min|x(t)|+λ·Penalty(k,c) Among them, λ is the penalty factor, ranging from 0 to 1, and the constraint is strengthened by increasing λ; Step 3: DE-L-BFGS hybrid algorithm optimizes TMD parameters 3.1 Population initialization G=0,G=1,2,3,…,MaxGen,where G is the number of iterations, and the maximum number of iterations MaxGen=100; the population size is selected as NP=30; according to the main structure stiffness and damping size, the optimization boundary conditions of k and c of TMD can be manually set to ensure k∈[k min ,k max ],c∈[c min ,c max ], where k min ,k max are the lower and upper limits of k, c min ,c max are the lower and upper limits of c respectively; 3.2 Randomly generate initial individual X i =(k i ,c i ), that is, the stiffness and damping parameters corresponding to TMD: k i =k min +rand(0,1)·(k max -k min ) c i =c min +rand(0,1)·(c max -c min ) Among them, rand(0,1) is a uniformly distributed random number, k i 、c i are randomly generated individual X i Stiffness and damping, i = 1, 2, 3, ..., NP; 3.3 For each individual X i =(k i ,c i ), calculate x i (t); 3.4 Calculate the fitness function value: f(k i ,c i )=min|x(t)|+0.5·Penalty(k i ,c i ). 3.5 Start Differential Evolution (DE) iteration 1) Mutation, for each generation of individual X i Use DE / rand / 1 strategy to generate mutation vector V i : V i =X r1 +F·(X r2 -X r3 ) Among them, r1, r2, and r3 are different random indexes; the scaling factor F is 0.

8. The larger F is, the stronger the exploratory nature (global search); the smaller F is, the stronger the exploitative nature (local search). 2) Crossover: In order to increase the diversity of the population, the mutation vector V i and the target vector X i Perform mixing and generate the test vector U through binomial crossover i =(u i,k ,u i,c ): Among them, the crossover probability CR is 0.9 (generally between 0 and 1.0). The larger the CR is, the more components the test vector inherits from the mutation vector, the higher the population diversity, the faster the convergence, but it may also skip the optimal solution; the smaller the CR is, the more components the test vector inherits from the target vector, the stronger the development, but the convergence may become slower; u i,k 、u i,c The i-th trial vector U i stiffness and damping. 3) Boundary processing, truncating out-of-range parameters: u i,k =max(k min ,min(k max ,u i,k )) u i,c =max(c min ,min(c max ,u i,c )) 4) Select and compare the test vector U i With the original individual X i Adaptability: 3.6 Determine if conditions are met Let t = 1, Tc=20, where: Tc is the trigger condition for executing the L-BFGS local optimization algorithm, which is executed every 20 generations, that is: Tc=20; t is the number of triggers; is the initial optimal fitness function value. Determine whether i>NP is true 1) If established, Let G=G+1, and determine whether G>MaxGen holds. If it is true, end the loop, output the results, and select the optimal parameters k and c; If not, determine whether G / (Tc×t)=1. If so, proceed to step 3.

7. After completion, set t = t + 1 and return to step 3.

5. If not, return to step 3.5; 2) If not established, Set i = i + 1 and return to step 3.

5. 3.7 L-BFGS Local Optimization 1) Set L-BFGS parameters: Convergence tolerance is set to ∈ = 10 -6 , the maximum number of iterations is set to N = 20; 2) Sort the fitness values ​​of all current individuals in ascending order, select the first m fitness functions and the corresponding X i , take m = 5, and take the first m X i The individuals constitute the initial vector Y0. 3) Perform L-BFGS algorithm calculation on the initial vector Y0: in: is the gradient of the objective function; H n is the approximate Hessia inverse matrix; α n is the step size, satisfying the Wolfe condition; n = 0, 1, 2, ..., N; The projection method is used to ensure that the iteration point is always within the feasible region: Y n+1 =clip(Y n+1 ,[k min ,c min ],[k max ,c max ]) That is: k = max(k min , min(k max , u i,k )), c = max(c min , min(c max , u i,c )) when Or when n>N, the optimal individual Y is obtained n ,make 4) Individuals after L-BFGS algorithm optimization Calculate the optimal fitness 5) If but Replace the original individual Right now: 3.8 Screening the optimal TMD parameters After optimization by DE-L-BFGS hybrid algorithm, the best individual As the optimal parameters of TMD stiffness and damping, the optimal individual is input into the TMD stiffness and damping adjustment execution controller; Step 4: TMD parameter adjustment 1) TMD stiffness adjustment: The optimal stiffness k obtained by the DE-L-BFGS hybrid algorithm is input into the integrated controller, converted into a corresponding current to control the operation of the servo motor. The ball screw converts the servo motor's rotational work into linear motion, thereby adjusting the compression or extension of the spring to achieve the purpose of controlling the spring preload and changing the spring stiffness. The spring preload is detected in real time by a strain gauge sensor, and the output electrical signal is fed back to the integrated controller. The deviation is calculated to complete the closed-loop control of the system. 2) TMD damping adjustment: The optimal damping c obtained by optimizing the DE-L-BFGS hybrid algorithm is input into the integrated controller and converted into a corresponding current to control the operation of the servo motor. The ball screw converts the servo motor's rotational work into linear motion, thereby adjusting the distance between the permanent magnet and the conductor plate, thereby controlling the magnetic force between the conductor and the permanent magnet and changing the damping. The distance between the permanent magnet and the conductor is detected in real time by a distance sensor, and the output electrical signal is fed back to the integrated controller to calculate the deviation and complete the closed-loop control of the system.

2. The vibration reduction device and vibration reduction method for a steam turbine in a thermal power plant according to claim 1, characterized in that: The value of λ in step 2 is 0.

5.

3. A steam turbine vibration reduction device for a thermal power plant, comprising a TMD, characterized in that: The TMD comprises a shell (1) and a mass block (2) arranged in the shell. The mass block (2) is placed in the middle of the shell internal space through an adjustment mechanism (A) arranged on the front, rear, left and right sides of the mass block (2). The four sides of the mass block (2) are respectively covered with a conductor plate (3). The adjustment mechanism comprises a first motor (4), a first ball screw, a first bearing seat (7), a second motor, a second ball screw, and a second bearing seat (15). The first motor (4) is fixed at the center of the shell side wall. The first ball screw comprises a shaft located inside the shell and coaxially connected to the output shaft of the first motor (4). The first screw (6) on the first screw and the first sliding nut (5) arranged on the first screw, the ball of the first ball screw rolls between the first screw and the first sliding nut to realize the conversion of the rotary motion of the screw shaft into the linear motion of the first sliding nut, thereby driving the permanent magnet plate to slide, adjusting the distance between the permanent magnet plate and the conductor plate, and achieving the purpose of controlling the magnetic force between the conductor plate and the permanent magnet plate and changing the damping; the first sliding nut (5) is fixed with a permanent magnet plate (8) facing the conductor plate corresponding to the mass block on the side facing the mass block, and the end of the first screw away from the first motor is rotated and arranged on the first bearing On the seat (7), a spring guide tube (9) coaxially arranged with the first screw is fixedly connected to the side of the first bearing seat (7) away from the first motor, and a stiffness adjustment spring (10) is sleeved on the spring guide tube. The second motor (11) is fixed to the housing on one side of the first motor. The second ball screw includes a second screw (13) located inside the housing and coaxially connected to the output shaft of the second motor (11) and a second sliding nut (12) arranged on the second screw. A spring compression plate (14) is fixed to the side of the second sliding nut (12) facing the mass block. The spring compression plate (14) is away from the second sliding nut. One end of the nut is slidably connected to the spring guide tube, one end of the stiffness adjustment spring (10) is connected to the conductor plate corresponding to the mass block, and the other end is connected to the spring compression plate (14). The balls of the second ball screw roll between the second screw (screw shaft) and the second sliding nut to realize the conversion of the rotary motion of the screw shaft into the linear motion of the second sliding nut. The second sliding nut drives the spring compression plate to slide, thereby adjusting the compression amount or extension amount of the spring, achieving the purpose of controlling the spring preload and changing the spring stiffness. The end of the second screw away from the second motor is rotatably arranged on the second bearing seat (15).

4. The steam turbine vibration reduction device for a thermal power plant according to claim 3, characterized in that: The stiffness adjusting spring (10) is provided with a strain gauge sensor for detecting its stress condition in real time, and is used for detecting the preload force of the stiffness adjusting spring in real time.

5. The steam turbine vibration reduction device for a thermal power plant according to claim 3, characterized in that: The conductor plate is provided with a distance measuring sensor for detecting the distance between the permanent magnet plate and the conductor plate in real time.

6. The steam turbine vibration reduction device for a thermal power plant according to claim 3, characterized in that: The vibration reduction device also includes a controller, the input end of the controller is respectively connected to the signal output end of each bearing position and speed sensor of the turbine, the output end of the strain gauge sensor, and the output end of the distance measuring sensor, and the output end of the controller is respectively connected to the input end of the first motor and the second motor.

Images