A PSO-ELM-based supercritical unit overheating predictive control method

Through the supercritical unit overheating prediction control method based on PSO-ELM, the particle swarm algorithm is used to optimize the limit learning machine model, and the large inertia and delay problems in the supercritical unit overheating steam temperature control are solved, achieving a more efficient control effect.

CN114895567BActive Publication Date: 2024-11-12GUODIAN SCI & TECH RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210571954.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-24
Publication Date
2024-11-12
Estimated Expiration
2042-05-24

AI Technical Summary

Technical Problem

The existing supercritical unit overheating steam temperature control technology has problems of large inertia and large delay, and conventional cascade controllers are difficult to achieve effective control effects, especially during rapid peak regulating and frequency regulation of the unit.

Method used

The supercritical unit overheating prediction control method based on PSO-ELM is adopted. By constructing a local prediction model, the particle swarm algorithm is used to optimize the connection weight and threshold of the limit learning machine model, the correction value of the steam outlet temperature is generated, and the feedback of the system output is compensated. Inertia and delay improve control stability.

Benefits of technology

It improves the accuracy and performance of overheating steam temperature control, simplifies on-site debugging, achieves better control effects, and can effectively improve the steam temperature control performance when the unit operating conditions change.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114895567B_ABST
    Figure CN114895567B_ABST
Patent Text Reader

Abstract

The present invention discloses a superheat prediction control method for a supercritical unit based on PSO-ELM, comprising the following steps: constructing a prediction controller, the prediction controller is used to realize superheated steam temperature control of a supercritical unit of a power plant; based on the prediction controller, by obtaining external disturbances on the flue gas side and a transfer function of a mixing process of desuperheating water and high-temperature steam, the superheated steam temperature of the supercritical unit is predicted and controlled through a local prediction model based on PSO-ELM after the desuperheating water and the high-temperature steam are mixed, and the superheated steam temperature outlet temperature is corrected according to the prediction result, thereby controlling the superheated steam temperature of the supercritical unit; the present invention effectively improves the control performance by proposing a local prediction steam temperature control strategy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of overheating control of supercritical units in power plants, and in particular to a supercritical unit overheating prediction and control method based on PSO-ELM. Background Art

[0002] For coal-fired units, main steam temperature and reheat steam temperature are one of the important indicators for evaluating the safety and economy of boiler operation. At present, coal-fired units are gradually developing in the direction of high parameters and large capacity, and supercritical and ultra-supercritical units have become the main force of coal-fired power generation. With the transformation of my country's energy structure, the large-scale grid connection of new energy has made conventional coal-fired units bear more and more peak-shaving and frequency-regulating tasks. Overtemperature pipe burst and other phenomena have generally become the core bottleneck restricting the ultra-supercritical units from increasing the variable load rate. The desuperheating water system is usually designed for abnormal steam temperature or accidents, but in the process of rapid peak-shaving and frequency-regulating response of the unit, desuperheating water control has gradually become a common means of main steam temperature control. Due to the nonlinear, large inertia and hysteresis characteristics of boiler steam temperature, desuperheating water control faces huge challenges. The current superheated steam temperature control is mainly based on conventional cascade controllers, which often cannot achieve good control effects. Therefore, there is an urgent need for a supercritical unit superheating predictive control method based on PSO-ELM to solve existing technical problems. Summary of the invention

[0003] In order to solve the above problems, the purpose of the present invention is to propose a superheat prediction control method for supercritical units based on PSO-ELM. Aiming at the characteristics of large inertia and large delay in the superheated steam temperature control system, the output of the controller is fed back to the input through accurate and effective prediction by the ELM model to compensate for the inertia and delay of the actual system output feedback, thereby improving the control stability of the system.

[0004] In order to achieve the above technical objectives, the present application provides a supercritical unit overheating prediction control method based on PSO-ELM, comprising the following steps:

[0005] Construct a predictive controller to control the superheated steam temperature of a supercritical unit in a power plant.

[0006] Based on the prediction controller, by obtaining the external disturbance on the flue gas side and the transfer function of the mixing process of desuperheating water and high-temperature steam, the correction value of the steam outlet temperature is generated through the constructed local prediction model based on PSO-ELM after the desuperheating water and high-temperature steam are mixed. The local prediction model is used to represent the process of desuperheating water and high-temperature steam mixing and flowing through the superheater / reheater for heat exchange;

[0007] The supercritical unit is predicted according to the prediction model, and the superheated steam outlet temperature is corrected according to the prediction result, so as to control the superheated steam temperature of the supercritical unit.

[0008] Preferably, in the process of constructing the PSO-ELM-based local estimation model after the desuperheated water is mixed with the high-temperature steam, the PSO-ELM-based local estimation model is constructed according to the extreme learning machine model and the particle swarm algorithm.

[0009] Preferably, in the process of constructing a local estimation model based on PSO-ELM, the extreme learning machine model consists of three layers of neurons, including an input layer, a hidden layer and an output layer, wherein the input layer, the hidden layer and the output layer are fully connected.

[0010] Preferably, in the process of constructing the local estimation model based on PSO-ELM, a first connection weight between the input layer and the hidden layer, a second connection weight between the output layer and the hidden layer, and a threshold of the hidden layer are obtained;

[0011] Obtaining an expected value of a local estimation model according to the first connection weight, the threshold, and the second connection weight;

[0012] Based on the expected value, an output sample is generated by acquiring an input sample according to a local estimation model, wherein the output sample is used to generate an outlet correction value of the mixed steam temperature.

[0013] Preferably, in the process of obtaining the expected value of the local estimation model, the matrix expression of the first connection weight is:

[0014]

[0015] Among them, w ln Represents the connection weight between the lth hidden layer neuron and the nth input layer neuron.

[0016] Preferably, in the process of obtaining the expected value of the local estimation model, the matrix expression of the threshold is:

[0017]

[0018] The matrix form of the second connection weight is:

[0019]

[0020] The matrix expression of the expected value is:

[0021]

[0022] Preferably, in the process of generating the output sample, the expression for generating the output sample is:

[0023] Hβ=T

[0024] Among them, H is the output of the hidden layer node, β is the output weight, and T is the expected output.

[0025] Preferably, in the process of generating output samples, the output of the hidden layer node is obtained through the activation function, and the output is expressed as:

[0026]

[0027] Preferably, in the process of constructing the local estimation model, the first connection weight and threshold of the extreme learning machine model are optimized by a particle swarm algorithm to obtain an optimal network structure, wherein:

[0028] The particle swarm algorithm is expressed as:

[0029]

[0030] in, represents the position of the d-th dimension of the k-th generation of the i-th particle, represents the velocity of the d-dimensional i-th particle in the k-th generation, i∈[1,2,…N],d∈[1,2,…,D], r 1 and r 2 Generated by a random number between 0 and 1, Pbest represents the individual extreme value, gbest represents the group extreme value, c 1 and c 2 It is the acceleration factor, which is used to adjust the particle to move towards the individual optimal point or the global optimal point.

[0031] Preferably, in the process of obtaining the optimal network structure, the algorithm optimization process of obtaining the optimal network structure is as follows:

[0032] (1) Randomly initialize the initial velocity and initial position of particles in the particle population, set the evolutionary generation, population size, and population range;

[0033] (2) Set the corresponding fitness function according to the target problem and calculate the initial fitness;

[0034] (3) Particles update their positions and velocities based on individual extreme values ​​and group extreme values;

[0035] (4) Particle swarm velocity update and particle swarm position update;

[0036] (6) Calculate the updated fitness;

[0037] (7) Compare fitness, select particles with better fitness and update extreme points;

[0038] (8) If the maximum number of iterations is reached or the fitness value reaches the set value, the optimization is completed; otherwise, repeat steps 3-7.

[0039] The present invention discloses the following technical effects:

[0040] 1. The particle swarm algorithm is used to optimize parameter selection and improve the prediction accuracy of the model;

[0041] 2. The method proposed in this paper is simpler and easier to debug on site. It can effectively improve the superheated steam temperature control performance and achieve good control effect.

[0042] 3. The method of the present invention uses computer language to simulate the dynamic characteristics of the object, which has better accuracy;

[0043] 4. The present invention starts from the process characteristics of the water spray cooling object system, adopts the estimation model method, and proposes a local estimation steam temperature control strategy, which effectively improves the control performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 It is a conventional superheated steam temperature cascade PID control block diagram according to an embodiment of the present invention;

[0046] Figure 2 It is a logic block diagram of the control method of the present invention;

[0047] Figure 3 is a schematic diagram of the prediction results of the present invention;

[0048] Figure 4 This is a diagram showing the application effect of the present invention on a 1000MW ultra-supercritical boiler. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0050] like Figure 1-4 As shown, the present invention discloses a supercritical unit overheating prediction control method based on PSO-ELM, comprising the following steps:

[0051] Construct a predictive controller to control the superheated steam temperature of a supercritical unit in a power plant.

[0052] Based on the prediction controller, by obtaining the external disturbance on the flue gas side and the transfer function of the mixing process of desuperheating water and high-temperature steam, the correction value of the steam outlet temperature is generated through the constructed local prediction model based on PSO-ELM after the desuperheating water and high-temperature steam are mixed. The local prediction model is used to represent the process of desuperheating water and high-temperature steam mixing and flowing through the superheater / reheater for heat exchange;

[0053] The supercritical unit is predicted according to the prediction model, and the superheated steam outlet temperature is corrected according to the prediction result, so as to control the superheated steam temperature of the supercritical unit.

[0054] Further preferably, in the process of constructing a local estimation model based on PSO-ELM after the cooling water and the high-temperature steam are mixed, the present invention constructs a local estimation model based on PSO-ELM according to an extreme learning machine model and a particle swarm algorithm.

[0055] Further preferably, in the process of constructing a local estimation model based on PSO-ELM, the extreme learning machine model mentioned in the present invention is composed of three layers of neurons, including an input layer, a hidden layer and an output layer, wherein a full connection is adopted between the input layer, the hidden layer and the output layer.

[0056] Further preferably, in the process of constructing the local estimation model based on PSO-ELM, the present invention obtains the first connection weight between the input layer and the hidden layer, the second connection weight between the output layer and the hidden layer, and the threshold of the hidden layer;

[0057] Obtaining an expected value of a local estimation model according to the first connection weight, the threshold, and the second connection weight;

[0058] Based on the expected value, an output sample is generated by acquiring an input sample according to a local estimation model, wherein the output sample is used to generate an outlet correction value of the mixed steam temperature.

[0059] Further preferably, in the process of obtaining the expected value of the local estimation model, the matrix expression of the first connection weight mentioned in the present invention is:

[0060]

[0061] Among them, w ln Represents the connection weight between the lth hidden layer neuron and the nth input layer neuron.

[0062] Further preferably, in the process of obtaining the expected value of the local estimation model, the matrix expression of the threshold mentioned in the present invention is:

[0063]

[0064] The matrix form of the second connection weight mentioned in the present invention is:

[0065]

[0066] The matrix expression of the expected value mentioned in the present invention is:

[0067]

[0068] Further preferably, in the process of generating output samples of the present invention, the expression for generating output samples mentioned in the present invention is:

[0069] Hβ=T

[0070] Among them, H is the output of the hidden layer node, β is the output weight, and T is the expected output.

[0071] Further preferably, in the process of generating output samples, the output of the hidden layer node is obtained by the activation function mentioned in the present invention, and the output is expressed as:

[0072]

[0073] Further preferably, in the process of constructing the local estimation model, the present invention optimizes the first connection weight and threshold of the extreme learning machine model through a particle swarm algorithm to obtain an optimal network structure, wherein:

[0074] The particle swarm algorithm is expressed as:

[0075]

[0076] in, represents the position of the d-th dimension of the k-th generation of the i-th particle, represents the velocity of the d-th dimension of the ith particle in the k-th generation, i∈[1,2,…N], d∈[1,2,…,D], r1 and r2 are generated by random numbers between 0 and 1, Pbest represents the individual extreme value, gbest represents the group extreme value, c1 and c2 are acceleration factors, and the acceleration factor is used to adjust the particle to move toward the individual optimal point or the global optimal point.

[0077] Further preferably, in the process of obtaining the optimal network structure of the present invention, the algorithm optimization process of obtaining the optimal network structure mentioned in the present invention is as follows:

[0078] (1) Randomly initialize the initial velocity and initial position of particles in the particle population, set the evolutionary generation, population size, and population range;

[0079] (2) Set the corresponding fitness function according to the target problem and calculate the initial fitness;

[0080] (3) Particles update their positions and velocities based on individual extreme values ​​and group extreme values;

[0081] (4) Particle swarm velocity update and particle swarm position update;

[0082] (6) Calculate the updated fitness;

[0083] (7) Compare fitness, select particles with better fitness and update extreme points;

[0084] (8) If the maximum number of iterations is reached or the fitness value reaches the set value, the optimization is completed; otherwise, repeat steps 3-7.

[0085] Embodiment 1: The specific implementation process of the present invention includes:

[0086] 1. Improvement of control strategy

[0087] The traditional superheated steam temperature cascade PID control block diagram is as follows Figure 1 As shown: PID1 and PID2 are the main regulator and sub-regulator of cascade control respectively; G1(s) is the transfer function of the mixing process of desuperheating water and high-temperature steam; G2(s) is the transfer function of the heat transfer process of the heat exchanger; D1(s) is the external disturbance of superheated steam; D2(s) is the external disturbance on the flue gas side. This control method is widely used in steam temperature control.

[0088] In order to ensure the accuracy of model predictive control, the present invention separates the model of the mixing process G1(s) of desuperheated water and high-temperature steam, and constructs a local prediction model of the mixing of desuperheated water and high-temperature steam. On the basis of taking into account the reliability of cascade control, data modeling is performed based on the mixed temperature T1 to predict and correct the outlet temperature deviation, so as to solve the large inertia and large delay problems of the heat transfer part G2(s), thereby improving the original cascade control performance.

[0089] The control block diagram of the present invention is as follows: Figure 2 As shown:

[0090] After the desuperheated water is mixed with the high-temperature steam, it flows through the superheater / reheater for heat exchange. This heat exchange process can be expressed by a certain model. Ignoring the change in the heat transfer capacity of the boiler under steady-state conditions, the model between the mixed steam temperature T1 and the outlet temperature T2 can be predicted using data modeling.

[0091] 2. Extreme Learning Machine Theory

[0092] Extreme learning machine is a single hidden layer neural network, which is a machine learning algorithm widely used in the field of data modeling. The model consists of three layers of neurons, including input layer, hidden layer and output layer, and the input layer, hidden layer and output layer are fully connected.

[0093] According to the model structure, the connection weights between the input layer and the hidden layer are expressed in the form of a matrix:

[0094]

[0095] In the formula, w ln Represents the connection weight between the lth hidden layer neuron and the nth input layer neuron.

[0096] The threshold of the hidden layer neurons is expressed in vector form:

[0097]

[0098] The connection weights between the output layer and the hidden layer are expressed in matrix form:

[0099]

[0100] For a single hidden layer neural network, assume that there are Q samples, matrix X represents the sample input set, and matrix Y represents the sample output set. The sample matrix structures are as follows:

[0101]

[0102] The expected value of the model is T, which is expressed in matrix form as:

[0103] T=[t 1t 2 ...t Q ] 1×Q (5)

[0104]

[0105] Formula (1-6) can be simplified as:

[0106] Hβ=T' (7)

[0107] Among them, H is the output of the hidden layer node, β is the output weight, and T is the expected output.

[0108] The output data matrix H of the hidden layer can be obtained by calculating the activation function. The specific structure is:

[0109]

[0110] 3. Particle Swarm (PSO) Theory

[0111] In PSO, particles are the basic elements of the particle swarm algorithm. Each particle has three characteristics, namely position x (the position of the particle in the solution space), velocity v (determines the direction and speed of the next movement) and fitness. In each iteration, the particles in the population can learn based on their own experience and the "experience" of the best particle in the population, that is, update their own positions based on the positions of the individual extreme value Pbest and the group extreme value Gbest. The individual extreme value Pbest refers to the position of the particle with the highest fitness value among the individuals, and the group extreme value Gbest refers to the position of the particle with the highest fitness value among all particles in the population. Each time a particle iterates, the fitness value is calculated, and the speed and direction of movement are determined by comparing the fitness of the new generation of particles. In this way, it is iterated step by step, and eventually tends to the optimal solution. The specific update formula is as follows:

[0112]

[0113] in, represents the position of the d-th dimension of the k-th generation of the i-th particle, represents the velocity of the d-dimensional particle of the i-th particle in the k-th generation, i∈[1,2,…N],d∈[1,2,…,D]; r 1 and r 2 Generated by a random number between 0 and 1; c 1 and c 2 It is an acceleration factor, which is used to adjust the particle to move towards the individual optimal point or the global optimal point.

[0114] The algorithm optimization process is as follows:

[0115] (1) Randomly initialize the initial velocity and initial position of particles in the particle population. Set the evolutionary generation, population size, and population range;

[0116] (2) Set the corresponding fitness function according to the target problem and calculate the initial fitness;

[0117] (3) Particles update their positions and velocities based on individual extreme values ​​and group extreme values.

[0118] (4) Particle swarm velocity update and particle swarm position update;

[0119] (6) Calculate the updated fitness;

[0120] (7) Compare fitness, select particles with better fitness and update extreme points;

[0121] (8) If the maximum number of iterations is reached or the fitness value reaches the set value, the optimization is completed; otherwise, repeat steps 3-7.

[0122] 4. PSO-ELM prediction model

[0123] In order to ensure that the ELM model maintains a high prediction accuracy when the hidden layer nodes are optimal, the present invention adopts the PSO optimization algorithm to optimize it. The PSO algorithm is used to optimize the connection weights and thresholds between the input layer and the hidden layer of the ELM to obtain the optimal network structure, thereby enhancing the stability of the ELM while ensuring the convergence speed of the ELM.

[0124] Using the above steps, 1250 sets of variable load data were selected for simulation prediction experiments, 1000 sets of data were selected as training sets to train the model, and the remaining 250 sets were used as prediction sets. PSO was used to optimize the connection weights and thresholds between the ELM input layer and the hidden layer. In the PSO algorithm, the maximum number of iterations was set to 1000, the population size was set to 300, and the acceleration factor c was set to 0. 1 =1.5,c 2 =3. The fitness function is the sum of the absolute values ​​of the difference between the predicted value and the actual value. Compare the prediction results of the estimation model with the actual data. The prediction results are as follows: Figure 3 shown.

[0125] The trained prediction model was applied to the actual site and was actually applied in the desuperheating water control of a 1000MW ultra-supercritical boiler. Compared with the classic cascade control, the combination of local prediction model correction can achieve better results for desuperheating water control. Figure 4The figure shows the application effect on a 1000MW ultra-supercritical boiler. Compared with the original cascade PID control, the hybrid model control based on cascade and local estimation model estimation correction proposed in the present invention can effectively improve the steam temperature control performance as the unit operating conditions change, and the maximum control deviation quickly drops to within ±5°C.

[0126] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0127] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0128] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A supercritical unit overheating prediction control method based on PSO-ELM, characterized in that: The following steps are involved: The cascade control includes a main regulator, a sub-regulator, G1(s) and G2(s) which are cascaded in sequence. D1(s) is input between the sub-regulator and G1(s), and D2(s) is input between G1(s) and G2(s). The output feedback of G1(s) is the input of the sub-regulator, and the output feedback of G2(s) is the input of the main regulator. G1(s) is the transfer function of the mixing process of desuperheating water and high-temperature steam, G2(s) is the transfer function of the heat transfer process of the heat exchanger, D1(s) is the external disturbance of the superheated steam, and D2(s) is the external disturbance on the flue gas side. Constructing a predictive controller for implementing superheated steam temperature control of a supercritical unit of a power plant; Based on the prediction controller, the model of the mixing process G1(s) of the cooling water and high-temperature steam is separated by obtaining the external disturbance on the flue gas side and the transfer function of the mixing process of the cooling water and the high-temperature steam; the cooling water and the high-temperature steam are mixed and flow through the superheater / reheater for heat exchange, and the change of the heat transfer capacity under the steady-state condition of the boiler is ignored to construct a local prediction model of the mixing of the cooling water and the high-temperature steam; the correction value of the steam outlet temperature is generated by the constructed local prediction model based on PSO-ELM after the cooling water and the high-temperature steam are mixed, and the obtained prediction model correction value and the outlet temperature T2 are fed back as the input of the main regulator of the cascade control; wherein the local prediction model is used to represent the process of the cooling water and the high-temperature steam being mixed and flowing through the superheater / reheater for heat exchange; In the process of constructing the local estimation model based on PSO-ELM after the desuperheated water and the high-temperature steam are mixed, the local estimation model based on PSO-ELM is constructed according to the extreme learning machine model and the particle swarm algorithm; The supercritical unit is predicted according to the local estimation model, and the superheated steam outlet temperature is corrected according to the prediction result, so as to control the superheated steam temperature of the supercritical unit.

2. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 1, characterized in that: In the process of constructing a local estimation model based on PSO-ELM, the extreme learning machine model is composed of three layers of neurons, including an input layer, a hidden layer and an output layer, wherein the input layer, the hidden layer and the output layer are fully connected.

3. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 2 is characterized by: In the process of constructing the local estimation model based on PSO-ELM, obtaining a first connection weight between the input layer and the hidden layer, a second connection weight between the output layer and the hidden layer, and a threshold of the hidden layer; Acquire an expected value of the local estimation model according to the first connection weight, the threshold, and the second connection weight; Based on the expected value, an output sample is generated by acquiring an input sample according to the local estimation model, wherein the output sample is used to generate a mixed steam temperature outlet correction value.

4. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 3 is characterized by: In the process of obtaining the expected value of the local estimation model, the matrix expression of the first connection weight is: Among them, w ln Represents the connection weight between the lth hidden layer neuron and the nth input layer neuron.

5. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 3 is characterized by: In the process of obtaining the expected value of the local estimation model, the matrix expression of the threshold is: The matrix form of the second connection weight is: The matrix expression of the expected value is: T=[t1 t2…t Q ] 1×Q 6. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 5, characterized in that: In the process of generating output samples, the expression for generating output samples is: Hβ=T Among them, H is the output of the hidden layer node, β is the output weight, and T is the expected output.

7. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 6, characterized in that: In the process of generating output samples, the output of the hidden layer node is obtained through the activation function, and the output is expressed as:

8. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 7, characterized in that: In the process of constructing the local estimation model, the first connection weight of the extreme learning machine model and the threshold are optimized by the particle swarm algorithm to obtain the optimal network structure, wherein: The particle swarm algorithm is expressed as: in, represents the position of the d-th dimension of the k-th generation of the i-th particle, represents the velocity of the d-th dimension of the ith particle in the k-th generation, i∈[1,2,…N], d∈[1,2,…,D], r1 and r2 are generated by random numbers between 0 and 1, Pbest represents the individual extreme value, gbest represents the group extreme value, c1 and c2 are acceleration factors, and the acceleration factors are used to adjust the particle to move toward the individual optimal point or the global optimal point.

9. The supercritical unit overheating prediction control method based on PSO-ELM according to claim 8 is characterized in that In the process of obtaining the optimal network structure, the algorithm optimization process of obtaining the optimal network structure is as follows: (1) Randomly initialize the initial velocity and initial position of particles in the particle population, set the evolutionary generation, population size, and population range; (2) Set the corresponding fitness function according to the target problem and calculate the initial fitness; (3) Particles update their positions and velocities based on individual extreme values ​​and group extreme values; (4) Particle swarm velocity update and particle swarm position update; (6) Calculate the updated fitness; (7) Compare fitness, select particles with better fitness and update extreme points; (8) If the maximum number of iterations is reached or the fitness value reaches the set value, the optimization is completed; otherwise, repeat steps (3)-(7).