Variable load control method for turbine of generator set

Abstract

The invention relates to the technical field of generator set control, and discloses a generator set steam turbine variable load control method, which specifically comprises the following steps: acquiring operation data of a steam turbine under a normal working condition; establishing a load transfer function model according to a generator set steam turbine operation mechanism; performing model identification on the acquired steam turbine operation data; proportional and integral terms are introduced into basic sliding mode control, and a PI integral terminal sliding mode controller is constructed according to a model obtained through identification; improving a basic particle swarm optimization algorithm by adopting a dynamic parameter method; and determining an objective function according to the system tracking error and the control quantity, optimizing the sliding mode surface parameters by using an improved particle swarm optimization algorithm, and determining an optimal sliding mode controller. Compared with a traditional control mode, the variable load control method for the steam turbine of the generator set has the advantages that the steam turbine runs at a more reasonable working condition point through accurate load control, and equipment abrasion and energy waste caused by load fluctuation are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of generator set control, and particularly relates to a method for controlling the variable load of a steam turbine of a generator set. Background Art

[0002] In recent years, with the implementation of the national policy of off-peak power consumption, the metallurgical industry has actively responded to the national power consumption policy. Combining the production needs of enterprises with the production environment of off-peak power consumption has caused the generator sets of metallurgical self-provided power plants to adjust their loads more frequently. As the direct energy conversion device of the generator set, the steam turbine thus puts forward higher requirements for the output control ability of the generator set.

[0003] The steam turbine is a 150MW subcritical, double-cylinder double-exhaust, combined high-pressure and intermediate-pressure cylinder, once-through reheat extraction condensing steam turbine. The DCS system adopts the HB600 series control system of Hebang Energy. The hardware of this system consists of an engineer station, an operator station, a field control station (including a main control unit device and an I / O unit device), a communication control station, a system server, a monitoring network, etc.

[0004] In summary, currently, for the problem of variable load control of generator sets, semi-artificial semi-automatic control or fully automatic control methods based on PID are mainly adopted. Semi-artificial semi-automatic control often relies on the work experience of staff, resulting in poor real-time performance and accuracy of the control effect of this method; although the control method based on PID can ensure the stability of the control system, due to the production needs of enterprises and the production environment of off-peak power consumption, the steam turbine of the generator set needs to adjust its load more frequently, resulting in poor dynamic performance of the system. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for controlling the variable load of a steam turbine of a generator set, so as to solve the problem that due to the production needs of enterprises and the production environment of off-peak power consumption, the steam turbine of the generator set needs to adjust its load more frequently, and the current control method can no longer better adapt to the operating conditions of the unit.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for controlling the variable load of a steam turbine of a generator set includes the following steps:

[0008] Step 1: Data acquisition and preprocessing: Obtain the operation data of the steam turbine under normal conditions. The operation data includes main steam temperature, main steam pressure, regulating stage pressure, opening commands of each valve, main steam flow rate, and active power of the unit, and perform outlier rejection processing on the obtained data;

[0009] Step 2: Model establishment and identification: Based on the operating mechanism of the steam turbine, considering the work process of steam in each cylinder, the extraction process, and the steam volume characteristics, establish a load transfer function model, and then use the least squares identification method to identify the model; among them, the intermediate and low-pressure cylinders are combined, and at the same time, let T N = F HP ·(1 + λ)·T RH , and the following model calculation formula is obtained:

[0010]

[0011] In the formula: F HP is the high-pressure cylinder power ratio; λ is the natural overshoot coefficient of the high-pressure cylinder power; T CH is the high-pressure steam chamber volume time constant; T RH is the reheated steam volume time constant, and s is the complex variable in the Laplace transform;

[0012] Step 3: Controller design: Introduce proportional and integral terms on the basis of the basic sliding mode control to construct a PI integral terminal sliding mode controller; specifically as follows:

[0013] b. First, transform the transfer function into the following state equation, set the given input value r, and obtain the system error e1 and the error change rate e2;

[0014]

[0015] In the formula, x is the system state vector; u is the control quantity; y is the output quantity; A and b are parameters;

[0016] b. Introduce proportional and integral terms on the basis of the basic sliding mode surface to construct the PI integral terminal sliding mode surface, as follows:

[0017] Among them:

[0018] In the formula: k p , k i are the proportional coefficient and the integral coefficient respectively, and both are non-zero positive constants, which are used to eliminate the state deviation, control the convergence speed, and eliminate the steady-state error respectively;

[0019] c. The control law design adopts the exponential reaching law and is improved by replacing the sign function sgn(s) with the saturation function sat(s) as:

[0020]

[0021] In the formula, ε and k are used to adjust the system reaching the sliding mode surface time and weaken the chattering, and σ is the variable critical value;

[0022] d. Finally obtain

[0023] Step 4: Algorithm improvement and parameter optimization. The basic particle swarm optimization algorithm is improved using the dynamic parameter method. According to the objective function value of the particle, the inertia weight ω, acceleration constants c1, c2 are changed. When the current objective function value of the particle is greater than the average objective function value of the particle, the parameters are adjusted according to the following calculation formula to expand the search range:

[0024]

[0025] In the formula: ω min , ω max are the minimum and maximum weight values respectively; f, f min , f avg are the current objective function value of the particle, the minimum objective function value of the particle, and the average objective function value of the particle respectively;

[0026] Then, the improved particle swarm optimization algorithm is used to optimize the sliding mode surface parameters and the exponential reaching law coefficient to determine the optimal sliding mode controller.

[0027] Preferably, the source of the operation data obtained in Step 1 is the operation history data of the steam turbine unit stored in the DCS system historical database.

[0028] Preferably, when establishing the load transfer function model in Step 2, the description of the work process of steam in each cylinder should follow the first and second laws of thermodynamics, accurately calculate the power share coefficient of each cylinder. For the extraction steam link, consider the influence of the extraction steam volume, extraction steam pressure, and extraction steam temperature on the steam flow and energy distribution. In the model identification process, the calculation process of the least squares identification method is carried out according to the matrix operation rules.

[0029] Preferably, when the dynamic parameter improves the basic particle swarm optimization algorithm in Step 4, the update formulas of the inertia weight ω and the acceleration constants c1, c2 are calculated based on the comparison results of the particle objective function value with the minimum, average, and maximum objective function values, and a specific parameter selection strategy is adopted for the 10% particles with the worst fitness.

[0030] Preferably, when the current objective function value of the particle is less than the average objective function value of the particle, the search range is reduced for more accurate search. At the same time, for the 10% particle individuals with the worst fitness after each round of iteration,

[0031] the search range maximization parameter selection strategy is adopted;

[0032] Then, according to the system tracking error e(t) and the control quantity u(t), the objective function is determined:

[0033]

[0034] In the formula: a and b are the weights of the system tracking error and the control quantity respectively, representing the degree of emphasis on the tracking error and energy consumption.

[0035] Preferably, the optimization process is specifically as follows: Generate an initial particle swarm in the solution space, including random positions and velocities, construct a sliding mode control system to calculate the objective function value, evaluate the particle fitness, update the individual and global best positions, update the particle parameters layer by layer according to the fitness, and iterate cyclically until the algorithm termination condition is met, and finally determine the optimal sliding mode controller.

[0036] Preferably, the termination condition is that the algorithm reaches the maximum number of iterations or the increment of the best fitness value is less than a given threshold.

[0037] Advantages of the present invention:

[0038] 1. The present invention constructs a PI integral terminal sliding mode controller by introducing proportional and integral terms into the basic sliding mode control, and uses an improved particle swarm algorithm to optimize the sliding mode surface parameters. The proportional term can quickly eliminate the state deviation and control the convergence speed, the integral term can eliminate the system steady-state error, and the improved particle swarm algorithm can find a better parameter combination, so that the control system can respond quickly when facing load change commands and effectively reduce the adjustment time;

[0039] 2. In the present invention, a saturation function is used to replace the sign function by improving the sliding mode reaching law, effectively weakening the system chattering. At the same time, the parameters optimized by the improved particle swarm algorithm make the system have stronger robustness to interference. Even when affected by external interferences such as steam parameter fluctuations and grid voltage changes, the control system can still work stably, ensure the safe and reliable operation of the unit, reduce the risk of operation failures caused by interference, and maintain the stability of the output power of the generator set.

[0040] In summary, the variable load control method for the steam turbine of the generator set provided by the present invention, compared with the traditional control method, the accurate load control enables the steam turbine to operate at a more reasonable working condition point, reducing equipment wear and energy waste caused by load fluctuations. The improvement of the dynamic response performance ensures that the unit can quickly adapt to load changes and avoids excessive energy consumption caused by adjustment lags. The improved control method optimizes the power distribution of each cylinder of the steam turbine, improves the energy conversion efficiency, thereby enhancing the stability of the unit operation and the overall economic benefits, contributing to the efficient and stable operation of the generator set, achieving significant optimizations at multiple key levels, and being of great significance to improving the overall operation efficiency and stability of the generator set.

[0041] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. Description of the Drawings

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0043] Figure 1 It is a flowchart of the steam turbine variable load control method for the generator set of the present invention;

[0044] Figure 2 It is a schematic structural diagram of the steam turbine in the present invention;

[0045] Figure 3 It is a flowchart of optimizing the sliding mode surface coefficient by the improved particle swarm algorithm of the present invention;

[0046] Figure 4 It is a block diagram of the control system built in the embodiment of the present invention. Specific embodiments

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0048] The present invention adopts a steam turbine variable load control method for a generator set, and the method includes:

[0049] Step S1: Obtain operation data such as the main steam flow rate, main steam temperature, main steam pressure, and active power of the unit under normal operating conditions;

[0050] Step S2: Establish its load transfer function model according to the operating mechanism of the steam turbine of the generator set;

[0051] Step S3: Use the steam turbine operation data obtained in step S1 for model identification;

[0052] Step S4: Introduce proportional and integral terms into the basic sliding mode control, and construct a PI integral terminal sliding mode controller according to the model identified in step S2;

[0053] Step S5: Improve the basic particle swarm optimization algorithm by using the dynamic parameter method;

[0054] Step S6: Determine the objective function according to the system tracking error and the control quantity, and use the improved particle swarm optimization algorithm to optimize the sliding mode surface parameters to determine the optimal sliding mode controller.

[0055] Among them, the operating data of the steam turbine under normal operating conditions is obtained, and the data acquisition source is the operating historical data of the steam turbine unit stored in the DCS system historical database. The operating data of the steam turbine to be obtained includes main steam temperature, main steam pressure, regulating stage pressure, opening commands of each valve, main steam flow rate, and active power of the unit. Obvious abnormal data is removed, and a statistical-based outlier detection method, such as the three-sigma principle or box plot method, is used to ensure the accuracy and stability of the data, providing a reliable basis for subsequent model establishment and analysis.

[0056] As an implementation manner of the present invention, specifically, please refer to Figure 1 :

[0057] The simplified schematic diagram of the steam turbine structure, such as Figure 1 shown, the main steam enters the high-pressure cylinder, reheater, intermediate-pressure cylinder, and low-pressure cylinder in sequence through the valve, drives the blades to drive the rotor to do work, and at the same time, there are multiple extraction steam links to provide heating steam for the high-temperature heaters.

[0058] The output power of the steam turbine is the sum of the powers of the high-pressure, intermediate-pressure, and low-pressure cylinders, and the power of each cylinder is calculated according to the power share coefficient. When establishing the load transfer function model, it is necessary to fully consider the energy conversion process of the main steam in the high-pressure cylinder, reheater, intermediate-pressure cylinder, and low-pressure cylinder and the influence of the multiple extraction steam links on the steam flow rate and energy distribution. According to the power calculation method of each cylinder and the characteristics of the steam volume link, the model is simplified and constructed. Relatively speaking, the high-pressure cylinder and the reheater steam system are the main steam volume links of the steam turbine. Therefore, to simplify the steam turbine model, the intermediate-pressure and low-pressure cylinders can be combined, and the transfer function of the combined steam turbine model is shown in Equation (1).

[0059]

[0060] In the formula: F HP is the power ratio of the high-pressure cylinder; λ is the natural overshoot coefficient of the high-pressure cylinder power; T CH is the volume time constant of the high-pressure steam chamber; T RH is the volume time constant of the reheater steam, and s is the complex variable in the Laplace transform. Specifically, it is a variable introduced adaptively according to the influence of actual environmental, equipment, and site factors.

[0061] Let T N = F HP ·(1 + λ)·T RH , then the transfer function of the steam turbine model in Equation (1) can be recorded as shown in Equation (2).

[0062]

[0063] As an implementation manner of the present invention, specifically, first, model identification is performed on the obtained steam turbine operation data, and the identification method is the least squares identification method:

[0064] The model to be identified is shown in Equation (2), and its z-transform can obtain the form shown in Equation (3):

[0065]

[0066] In the formula: u(z) is the system input, and y(z) is the system output.

[0067] The difference equation corresponding to Equation (3) is:

[0068]

[0069] Since z(k) = y(k) + v(k), where z(k) is the actually measurable output and v(k) is the system noise, Equation (4) can be written in the following form:

[0070]

[0071] Define the system input-output matrix and the parameter matrix to be identified as shown in Equation (6) and Equation (7) respectively:

[0072] h(k) = [-y(k - 1), -y(k - 2), u(k - 1), u(k - 2)], (6);

[0073] θ = [a1, a2, b1, b2] T , (7);

[0074] The operation data of each item of the generator set steam turbine collected in step S1 is used as the input and output and imported into the input-output matrix shown in Equation (6).

[0075] Then the matrix form of the model to be identified is shown in Equation (8):

[0076] Z m = H m θ + V m , (8);

[0077] Among them,

[0078] θ = [a1, a2, b1, b2] T , V m = [v(1) v(2)... v(m)] T ;

[0079] The loss function is as shown in Equation (9):

[0080]

[0081] To achieve the best identification effect, the loss function should be minimized. According to the extreme value theorem, we have:

[0082]

[0083] Then the least squares estimate is:

[0084]

[0085] As an implementation manner of the present invention, specifically:

[0086] Introduce proportional and integral terms into the basic sliding mode control, and construct a PI integral terminal sliding mode controller according to the model identified in step S2, specifically as follows:

[0087] The basic design of the sliding mode control includes: sliding mode surface design and control law design. Among them, the sliding mode surface determines the dynamic characteristics of the system to a certain extent, and the control law determines the convergence speed of the system to a certain extent.

[0088] For the transfer function shown in Equation (2), convert it into a state equation as:

[0089]

[0090] In the formula: x is the system state vector; u is the control quantity; y is the output quantity; A and b are parameters.

[0091] Since the object of study in this paper is a second-order system, and the set input value of the system is r, then the system error e1 and error change rate e2 are obtained as shown in Equation (13):

[0092]

[0093] This method introduces proportional and integral terms on the basis of the basic sliding mode surface, and constructs the PI integral terminal sliding mode surface as follows:

[0094]

[0095] In the formula: k p 、k i are the proportional coefficient and integral coefficient respectively, and both are non-zero positive constants. Among them, the proportional term is used to eliminate the state deviation and control the convergence speed, and the integral term is used to eliminate the steady-state error of the system.

[0096] The design of the control law requires the selection of an approaching law. Since the constant-speed approaching law is difficult to balance a faster approaching rate and a smaller system chattering, this method adopts the exponential approaching law, as shown in Equations (15) and (16):

[0097]

[0098] where ε > 0 is used to adjust the time for the system to reach the sliding surface; k > 0 can also be used to adjust the time for the system to reach the sliding surface and can weaken chattering and improve the dynamic characteristics of the system.

[0099] Since there is a discontinuous sign function sgn(s) in the exponential reaching law shown in Equation (15), this will cause the problem of system chattering, and the larger ε is, the stronger the system chattering. To further eliminate the chattering problem, the saturation function sat(s) is used to replace the sign function sgn(s), and the improved sliding mode reaching law is shown in Equations (17) and (18):

[0100]

[0101] where σ is the variable critical value.

[0102] The control law can be obtained from Equations (12), (13), (14), and (17) as follows:

[0103]

[0104] The value ranges of the proportional coefficient and integral coefficient of its sliding surface should be reasonably set according to the specific operating characteristics and control requirements of the steam turbine. In the design of the control law, the parameters ε, k of the exponential reaching law and the critical value of the saturation function should be determined by experiment or simulation optimization to achieve the rapid convergence and stable control of the system, and at the same time effectively reduce the chattering phenomenon.

[0105] As an implementation manner of the present invention, specifically:

[0106] The particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence. The algorithm randomly initializes a group of particles, and each particle contains a set of velocity and position information. Each set of position information is regarded as a feasible solution, and the quality of the particle is determined by the fitness. The particles move in the feasible solution space according to the iterative rules, record the individual best position and the group best position after each iteration, and obtain the optimal solution after multiple iterative searches. In this embodiment, the basic particle swarm optimization algorithm is improved by the dynamic parameter method.

[0107] The iterative formula of the basic particle swarm optimization algorithm is as follows:

[0108] Velocity update formula:

[0109] v i,j (t + 1) = ωv i,j (t) + c1r1[pbest i,j -x i,j (t)] + c2r2[gbest j -x i,j (t)], (20);

[0110] Position update formula:

[0111] x i,j (t+1)=x i,j (t)+v i,j (t+1),j=1,2,···,d, (21);

[0112] Where: v i,j (t) is the j-th component of the flight velocity vector of particle i of the t-th generation; x i,j (t) is the j-th component of the position vector of particle i of the tth generation; c1, c2 are acceleration constants, which are used to adjust the maximum step size of learning; r1, r2 are random functions with a value range of [0, 1], which are used to increase randomness; ω is the inertia weight, which is a non-negative number and is used to adjust the search range of the solution space; pbest is the individual best position; gbest is the global best position.

[0113] For the basic PSO algorithm, the inertia weight ω and acceleration parameters c1, c2 in equation (20) are generally constants. Therefore, during the algorithm iteration process, the particles lack diversity and tend to converge to local extreme values ​​prematurely. To address this problem, this paper adopts a dynamic parameter strategy to change the inertia weight ω and acceleration parameters c1, c2 according to the particle's objective function value to obtain more diverse and high-quality particle individuals and get rid of local extreme value interference as much as possible. The objective function value of the system in this paper reflects the system tracking error and energy consumption, so the smaller the value, the closer to the optimization target. Therefore, this method adopts the following inertia weight and acceleration update strategy: when the particle's current objective function value is greater than the particle's average objective function value, it indicates that the current particle position is far from the optimal position, and the parameters should be adjusted to expand the search range; when the particle's current objective function value is less than the particle's average objective function value, it indicates that the current particle position is close to the optimal position, so adjust the parameters to narrow the search range for a more accurate search.

[0114] Based on the above analysis, the inertia weight and acceleration parameter update formula adopted by this method is as follows:

[0115]

[0116]

[0117] Where: min ,ω max are the minimum and maximum weight values ​​respectively; f,f min ,f avg They are respectively the current objective function value of the particle, the minimum objective function value of the particle, and the average objective function value of the particle.

[0118] The value range of inertia weight is (ω min ,ωmax ) The value ranges of the acceleration parameters c1 and c2 are their corresponding maximum and minimum values respectively. When updating the particle parameters, the accuracy and stability of the calculation process should be ensured to avoid numerical overflow or calculation errors. During the optimization process using the improved particle swarm optimization algorithm, a suitable random number generation method should be adopted to ensure the randomness and uniformity of particle initialization, and improve the search efficiency and optimization quality of the algorithm.

[0119] Meanwhile, to ensure that the algorithm always maintains the ability to jump out of local traps during the iteration process, for the 10% of particle individuals with the worst fitness after each round of iteration, a parameter selection strategy of maximizing the search range is adopted, and the update formula is shown in Equation (25):

[0120]

[0121] As an implementation manner of the present invention, specifically, refer to Figure 2 :

[0122] Use the improved particle swarm optimization algorithm to optimize the sliding mode surface parameters, specifically as follows:

[0123] This method uses the improved PSO algorithm to find a set of optimal values for the sliding mode surface coefficient and the exponential reaching law coefficient, so as to minimize the system tracking error and energy consumption.

[0124] According to the above requirements, the objective function of the system is obtained as follows:

[0125]

[0126] Where: e(t) is the system tracking error; u(t) is the control quantity; a and b are the weights of the system tracking error and the control quantity respectively, representing the degree of emphasis on the tracking error and energy consumption.

[0127] The weights a and b should be reasonably adjusted according to the degree of emphasis on the system tracking error and energy consumption in the actual application scenario. During the algorithm iteration process, information such as the objective function value, particle position, and velocity of each iteration should be recorded to analyze and optimize the algorithm performance. Meanwhile, when judging the algorithm termination condition, the increment of the best fitness value should be accurately calculated to ensure that the algorithm converges to a satisfactory solution.

[0128] The specific process of optimizing the sliding mode parameters based on the improved PSO algorithm is as follows:

[0129] A. Generate an initial particle swarm in the solution space, including random positions and velocities;

[0130] B. Construct a sliding mode control system based on the particle information, and calculate the objective function value of the system according to the tracking error and control quantity of the control system;

[0131] C. Evaluate the fitness of the particles. In this paper, the fitness is selected as the objective function value;

[0132] D. For each particle, compare its current fitness value with its historical best fitness value. If the fitness value of the particle at the current position is lower, update the current position as the individual historical best position;

[0133] E. For each particle, compare its current fitness value with the fitness value of the historical global best position. If the fitness value of the particle at the current position is lower, update the current position as the historical global best position;

[0134] F. Arrange the particles in ascending order of fitness value. The first 90% update the inertia weight and acceleration parameters of each particle according to Equations (22), (23) and (24) respectively, and the last 10% update the inertia weight and acceleration parameters of each particle according to Equation (25);

[0135] G. Update the velocity and position of each particle according to Equations (20) and (21);

[0136] H. Determine whether the algorithm meets the end condition. If the end condition is met, output the global best position information; if the end condition is not met, jump to step (B). The end condition is that the algorithm reaches the maximum number of iterations or the increment of the best fitness value is less than the given threshold.

[0137] The final control system block diagram is as Figure 3 shown.

[0138] The above content is only an example and explanation of the concept of the present invention. Those skilled in the art of this technology can make various modifications, supplements or use similar methods to replace the specific embodiments described, as long as they do not deviate from the concept of the invention or exceed the scope defined by this claim book, they should fall within the protection scope of the present invention.

Claims

1. A method for controlling the variable load of a steam turbine in a generator set, characterized in that, It includes the following steps: Step 1: Data acquisition and preprocessing: Obtain the operation data of the steam turbine under normal conditions. The operation data includes main steam temperature, main steam pressure, regulating stage pressure, opening commands of each valve, main steam flow rate, and unit active power, and perform outlier rejection processing on the obtained data. Step 2: Model establishment and identification: Based on the operating mechanism of the steam turbine, considering the work done by steam in each cylinder, the extraction process, and the steam volume characteristics, a load transfer function model is established, and then the least squares identification method is used to identify the model; among them, the intermediate and low-pressure cylinders are combined, and at the same time, let T N = F HP ·(1 + λ)·T RH , and the following model calculation formula is obtained: Where: F HP is the high-pressure cylinder power ratio; λ is the natural overshoot coefficient of the high-pressure cylinder power; T CH is the volume time constant of the high-pressure steam chamber; T RH is the volume time constant of the reheated steam, and s is the complex variable in the Laplace transform; Step 3: Controller design: Introduce proportional and integral terms on the basis of basic sliding mode control to construct a PI integral terminal sliding mode controller. Specifically as follows: a. First, transform the transfer function into the following state equation, set the given input value r, and obtain the system error e1 and error change rate e2. In the formula, x is the system state vector; u is the control quantity; y is the output quantity; A and b are parameters. b. Introduce proportional and integral terms on the basis of the basic sliding mode surface to construct the PI integral terminal sliding mode surface as follows: Wherein: where: k p , k i are the proportional coefficient and the integral coefficient respectively, and both are non-zero positive constants, which are used to eliminate the state deviation, control the convergence rate, and eliminate the steady-state error respectively; c. The exponential reaching law is adopted for the control law design and it is improved by replacing the sign function sgn(s) with the saturation function sat(s) as follows: In the formula, ε and k are used to adjust the system reaching the sliding mode surface time and weaken chattering, and σ is the variable critical value. d. Finally obtained Step 4: Algorithm improvement and parameter optimization. Use the dynamic parameter method to improve the basic particle swarm optimization algorithm. According to the objective function value of the particle, change the inertia weight ω, acceleration constants c1, c2. When the current objective function value of the particle is greater than the average objective function value of the particle, adjust the parameters according to the following calculation formula to expand the search range: where: ω min , ω max are the minimum and maximum weight values respectively; f, f min , f avg are the current objective function value of the particle, the minimum objective function value of the particle, and the average objective function value of the particle respectively; Then use the improved particle swarm optimization algorithm to optimize the sliding mode surface parameters and the exponential reaching law coefficient to determine the optimal sliding mode controller.

2. A method for controlling the variable load of a steam turbine in a generator set according to claim 1, characterized in that, The source of the operation data in Step 1 is the operation history data of the steam turbine unit stored in the DCS system historical database.

3. A variable load control method for a steam turbine of a generator set according to claim 1, characterized in that, When establishing the load transfer function model in Step 2, the description of the work process of steam in each cylinder should follow the first and second laws of thermodynamics, accurately calculate the power share coefficient of each cylinder. For the extraction steam link, consider the influence of the extraction steam volume, extraction steam pressure, and extraction steam temperature on the steam flow rate and energy distribution. In the model identification process, the calculation process of the least squares identification method is carried out according to the matrix operation rules.

4. A method for controlling the variable load of a steam turbine in a generator set according to claim 1, characterized in that, When using the dynamic parameter to improve the basic particle swarm optimization algorithm in Step 4, the update formulas of the inertia weight ω and acceleration constants c1, c2 are calculated according to the comparison results of the particle objective function value with the minimum, average, and maximum objective function values, and for the 10% particles with the worst fitness, a specific parameter selection strategy is adopted.

5. A method for controlling the variable load of a steam turbine in a generator set according to claim 4, characterized in that, When the current objective function value of the particle is less than the average objective function value of the particle, narrow the search range for more accurate search. At the same time, for the 10% particle individuals with the worst fitness after each round of iteration, take... Search range maximization parameter selection strategy; Then, determine the objective function according to the system tracking error e(t) and control quantity u(t): In the formula: a and b are the weights of the system tracking error and control quantity respectively, representing the degree of emphasis on the tracking error and energy consumption.

6. A variable load control method for a steam turbine of a generator set according to claim 1, characterized in that, The optimization process is specifically as follows: Generate an initial particle population in the solution space, including random positions and velocities, construct a sliding mode control system to calculate the objective function value, evaluate the particle fitness, update the individual and global best positions, update the particle parameters layer by layer according to the fitness, and perform iterative loops until the algorithm end condition is met, and finally determine the optimal sliding mode controller.

7. A method for controlling the variable load of a steam turbine in a generator set according to claim 6, characterized in that, The end condition is that the algorithm reaches the maximum number of iterations or the increment of the best fitness value is less than the given threshold.

Images