CN117518823A - Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit - Google Patents

Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit Download PDF

Info

Publication number
CN117518823A
CN117518823A CN202311630148.9A CN202311630148A CN117518823A CN 117518823 A CN117518823 A CN 117518823A CN 202311630148 A CN202311630148 A CN 202311630148A CN 117518823 A CN117518823 A CN 117518823A
Authority
CN
China
Prior art keywords
pid controller
neural network
steam temperature
main steam
adaptive
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311630148.9A
Other languages
Chinese (zh)
Inventor
尚家华
唐亮
王焕敏
徐义军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guixi Power Generation Co ltd
Original Assignee
Guixi Power Generation Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guixi Power Generation Co ltd filed Critical Guixi Power Generation Co ltd
Priority to CN202311630148.9A priority Critical patent/CN117518823A/en
Publication of CN117518823A publication Critical patent/CN117518823A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention belongs to the field of coal-fired power generation, and discloses a main steam temperature self-adaptive control method and a related device of a supercritical thermal power unit, wherein the method comprises the steps of obtaining an expected signal and an output signal of the main steam temperature of the large supercritical thermal power unit, and calculating an error signal group of the expected signal and the output signal; obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on the dendritic neural network model; performing self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network, and applying the fuzzy control parameters and the control parameters of the neural network to the PID controller to obtain a self-adaptive PID controller; and obtaining a control signal of the main steam temperature of the large supercritical thermal power unit through the self-adaptive PID controller according to the error signal group. The method has higher precision and adaptability, and can meet the main steam temperature control requirements of large inertia, large delay and nonlinear time variation.

Description

Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit
Technical Field
The invention belongs to the field of coal-fired power generation, and relates to a main steam temperature self-adaptive control method and a related device of a supercritical thermal power unit.
Background
As a high-efficiency, high-temperature and high-pressure coal-fired power generating unit, the large supercritical thermal power generating unit has the advantages of large installed capacity, high degree of automation, high operation parameters, high thermal efficiency, low coal consumption and the like. The technology has become the dominant trend of the development of thermal power plants, and can effectively control harmful gases (such as SO2, NOx and the like) generated by the combustion of coal while improving the power generation efficiency, thereby reducing the influence on the environment. In the supercritical state, the working pressure, temperature and other parameters of the boiler are far beyond the common working condition. Although a unit operating in such an environment can significantly improve the power generation efficiency, the harsh environment of high temperature and high pressure can also increase the vulnerability and control difficulty of the entire system. When the supercritical thermal power generating unit regulates the peak value in a large range, the regulation and control of the main steam temperature are particularly important.
However, control of the main steam temperature of supercritical thermal power plants faces significant challenges. This is because once-through boiler plants are numerous and are connected to each other in series or parallel, and the small disturbance of each plant may cause the response of the adjacent plant to affect the overall boiler system. For example, in a fume system, changes in the operation of the induced draft fan may cause negative pressure fluctuations in the furnace and changes in the excess air system, which in turn may cause a main steam temperature bias. Therefore, the once-through boiler can be regarded as a complex multi-input multi-output control object, and due to the strong parameter coupling effect, one input change can cause a plurality of output quantity changes, so that the dynamic characteristic of the once-through boiler is complex, and the control object presents large inertia, time-varying and nonlinear characteristics, thereby increasing the control difficulty.
At present, in the supercritical thermal power generating unit, the water spray temperature reducing system plays a key role as an important component part of a main steam temperature control system of the supercritical thermal power generating unit, and can effectively reduce interference in most cases. Traditional steam temperature control methods rely mainly on PID (process-Integration-Differentiation) algorithms. Wherein the proportional controller P generates an output by multiplying a deviation (a difference between an actual value and an expected value) by a coefficient, thereby affecting a response speed of the system, the integral controller I generates an output by integrating the accumulated deviation, thereby correcting a steady-state error in the system, and the differential controller D generates an output by differentiating a change rate of the deviation, thereby effectively suppressing an excessive adjustment or oscillation in the system response. However, the basic principle of the main steam temperature control system is to use the slower response of the desuperheater disturbance to counteract the faster heat absorption disturbance of the superheater. The controlled object has nonlinear, high inertia and significant time delay characteristics, and conventional PID control methods often fail to achieve the desired control quality level. Although cascade PID strategies can pre-process disturbances and temperature variations in the inner loop to solve the response speed problem of high inertia systems to some extent, these conventional methods have difficulties in obtaining optimal control parameters and processing model parameter disturbances. The performance of the model parameter interference caused by the obvious time delay characteristic and random load fluctuation of the superheated steam temperature object in the thermal power plant is poor, and the realization of high-precision control quality under various working conditions is a great challenge.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a main steam temperature self-adaptive control method and a related device of a supercritical thermal power unit.
In order to achieve the purpose, the invention is realized by adopting the following technical scheme:
the invention provides a main steam temperature self-adaptive control method of a supercritical thermal power unit, which comprises the following steps: acquiring an expected signal and an output signal of main steam temperature of a large supercritical thermal power generating unit, and calculating an error signal group of the expected signal and the output signal; obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on a dendritic neural network model; performing self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network, and applying the fuzzy control parameters and the control parameters of the neural network to the PID controller to obtain a self-adaptive PID controller; and obtaining a control signal of the main steam temperature of the large supercritical thermal power generating unit through the self-adaptive PID controller according to the error signal group.
Optionally, the obtaining, according to the error signal set, the fuzzy control parameter of the PID controller by a fuzzy control method includes: solving the membership of the error signal group through a preset membership function to obtain the membership of the error signal group relative to each preset fuzzy subset; based on the membership degree of the error signal group relative to each preset fuzzy subset, fuzzy reasoning is carried out through a preset fuzzy control table, and fuzzy quantity of fuzzy control parameters of the PID controller is obtained; and performing anti-blurring on the blurring amount of the blurring control parameter of the PID controller by using a weighted average method to obtain the blurring control parameter of the PID controller.
Optionally, the pre-trained neural network control parameter model based on the dendritic neural network model is obtained by the following method: constructing a dendritic neural network model; acquiring an expected signal and a training output signal of the main steam temperature of the large supercritical thermal power generating unit at the current training time, and acquiring an error signal group at the current training time according to the expected signal and the training output signal at the current training time; inputting the error signal group at the current training time into a dendritic neural network model to obtain the neural network control parameters of the PID controller at the current training time; obtaining fuzzy control parameters of the current training time of the PID controller through a fuzzy control method according to the error signal group of the current training time, carrying out self-adaptive weighted summation on the fuzzy control parameters of the current training time of the PID controller and the neural network control parameters, and applying the fuzzy control parameters and the neural network control parameters to the PID controller to obtain a self-adaptive PID controller of the current training time; according to the error signal group at the current training time, a training control signal of the main steam temperature of the large supercritical thermal power unit is obtained through a self-adaptive PID controller at the current training time and is applied, and an output signal of the main steam temperature of the large supercritical thermal power unit at the next training time is obtained; acquiring an expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, obtaining an error signal group of the next training time according to the output signal and the expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, and optimizing a dendritic neural network model according to the error signal group of the next training time; repeating the steps until the error between the output signal and the expected signal at the next training moment meets a preset error threshold, and taking the final dendritic neural network model as a neural network control parameter model.
Optionally, the adaptively weighted summing the fuzzy control parameter of the PID controller and the neural network control parameter and applying the summed fuzzy control parameter and the neural network control parameter to the PID controller, to obtain the adaptive PID controller includes: and carrying out self-adaptive weighted summation on the fuzzy control parameter of the PID controller and the neural network control parameter by the following steps:
wherein alpha is 1 ，α 2 And alpha 3 All are self-adaptive weight parameters, and the value ranges are all between 0 and 1; k (k) P Is the proportional gain parameter k of the adaptive PID controller I Is the integral gain parameter k of the adaptive PID controller D Differential gain parameters for the adaptive PID controller;proportional gain parameter among fuzzy control parameters of PID controller,/for the control of the PID controller>Is an integral gain parameter among fuzzy control parameters of the PID controller,/and the like>A differential gain parameter among fuzzy control parameters of the PID controller; />Proportional gain parameter, among the neural network control parameters of the PID controller, +>Integral gain parameter of neural network control parameters for PID controller, +>Is a differential gain parameter among the neural network control parameters of the PID controller.
Optionally, the alpha 1 ，α 2 And alpha 3 And the fuzzy control parameters of the PID controller, the neural network control parameters of the PID controller and the error signal group are obtained through an Actor-Critic network model based on reinforcement learning.
Optionally, the obtaining, by the adaptive PID controller, the control signal of the main steam temperature of the large supercritical thermal power unit according to the error signal group includes: and according to the error signal group, obtaining a control signal of the main steam temperature of the large supercritical thermal power generating unit by the following formula:
wherein u is 0 (k) For the control signal of the main steam temperature of the large supercritical thermal power generating unit at the moment k, error (k) is an error signal group at the moment k, error (k-1) is an error signal group at the moment k-1, and k P Is the proportional gain parameter k of the adaptive PID controller I Is the integral gain parameter k of the adaptive PID controller D Is a differential gain parameter of the adaptive PID controller.
In a second aspect of the present invention, there is provided a main steam temperature control system for a large supercritical thermal power generating unit, comprising: the signal processing module is used for acquiring an expected signal and an output signal of the main steam temperature of the large supercritical thermal power generating unit and calculating an error signal group of the expected signal and the output signal; the parameter determining module is used for obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on a dendritic neural network model; the self-adaptive weighting module is used for carrying out self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network and applying the fuzzy control parameters and the control parameters of the neural network to the PID controller to obtain a self-adaptive PID controller; and the control module is used for obtaining a control signal of the main steam temperature of the large supercritical thermal power unit through the self-adaptive PID controller according to the error signal group.
In a third aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor implements the steps of the above-mentioned main steam temperature adaptive control method of a supercritical thermal power generating unit when executing the computer program.
In a fourth aspect of the present invention, a computer readable storage medium is provided, where a computer program is stored, where the computer program, when executed by a processor, implements the steps of the main steam temperature adaptive control method of a supercritical thermal power generating unit.
Compared with the prior art, the invention has the following beneficial effects:
according to the self-adaptive control method for the main steam temperature of the supercritical thermal power generating unit, an error signal group is calculated based on a desired signal and an output signal of the main steam temperature of the large supercritical thermal power generating unit, fuzzy control parameters of a PID controller are obtained through a fuzzy control method according to the error signal group, neural network control parameters of the PID controller are obtained through a pre-trained neural network control parameter model based on a dendritic neural network model, self-adaptive weighted summation is carried out on the fuzzy control parameters of the PID controller and the neural network control parameters and the self-adaptive weighted summation is applied to the PID controller, and then the self-adaptive PID controller is obtained through the self-adaptive PID controller, and the control signal of the main steam temperature of the large supercritical thermal power generating unit is obtained. By adaptively weighting the fuzzy control parameters and the neural network control parameters, the adaptive learning capacity of the method is stronger than that of the traditional artificial neural network. Aiming at the problems of obvious change of dynamic characteristics of main steam temperature, large change of parameter structure and the like when load changes, the dendritic neural network has higher precision and adaptability, can meet the main steam temperature control requirements of a complex boiler system with large inertia, large delay and nonlinear time variation, provides higher control precision and strong self-adaptive learning capability, can also effectively lower calculation complexity, and realizes the self-adaptive setting of control system parameters.
Drawings
Fig. 1 is a flowchart of a main steam temperature adaptive control method of a supercritical thermal power generating unit according to an embodiment of the invention.
Fig. 2 is a schematic diagram of a main steam temperature adaptive control method of a supercritical thermal power generating unit according to an embodiment of the invention.
Fig. 3 is a control schematic diagram of a conventional PID controller.
Fig. 4 is a schematic diagram of a fuzzy control method according to an embodiment of the present invention.
Fig. 5 is a block diagram of a single layer dendritic neuron network according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of forward propagation and backward update of a dendritic neural network model according to an embodiment of the present invention.
Fig. 7 is a schematic diagram of a mimo dendritic neural network including a hidden layer according to an embodiment of the present invention.
FIG. 8 is a schematic diagram of the structure principle of the reinforcement learning-based Actor-Critic model according to an embodiment of the present invention.
Fig. 9 is a block diagram of a main steam temperature control system of a large supercritical thermal power generating unit according to an embodiment of the invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
As described in the background art, when the supercritical thermal power generating unit is in peak shaving in a large range, the regulation and control on the temperature of the main steam are particularly important. If the main steam temperature is too high, thermal stresses on the unit steam lines, particularly on the heated surfaces of the boiler furnace, can be affected, which can accelerate the creep rate of the plant steel, resulting in plant damage. For example, when the main steam temperature reaches 593 ℃, the service life of the steel material is shortened from 10 ten thousand hours to 3 ten thousand hours, and when the temperature is seriously exceeded, the metal material may lose the strength, so that the pipeline is broken. On the other hand, if the main steam temperature is too low, the thermal efficiency of the generator set is reduced, which not only affects the power generation efficiency and increases the coal consumption; more importantly, in order to maintain unit power in the event of reduced main steam temperature, the opening of the turndown must be increased, which may lead to increased water and humidity in the final steam stage, corrosion to the turbine operating at high speed and possibly the initiation of water hammer events, affecting the life of the metal material. In addition, when the temperature fluctuation of the main steam is severe, the fatigue and creep of the metal materials can be accelerated, the expansion difference change of the rotor and the cylinder of the steam turbine can be caused, even severe vibration is generated, and the running safety of the generator set is seriously threatened.
The temperature of the superheated steam at the outlet of the boiler superheater of the large supercritical thermal power generating unit is highest (more than 500 ℃) in the whole boiler steam-water flow, and the control quality of the boiler steam temperature has an important influence on the safety and economic index of the thermal power generating unit, so that the method is one of the first-choice indexes for evaluating the unit operation quality. Therefore, the method has very important practical significance for accurately controlling the main steam temperature to be at the rated value and reducing the steam temperature fluctuation, and is important for ensuring the efficient, safe and stable operation of the unit. In general, the temporary deviation of the main steam temperature of the thermal power generating unit is required to be controlled to be within 10 ℃ of a set value, and the long-term deviation is required to be controlled to be within 5 ℃ of the set value, so that a very precise and reliable temperature control system is required.
At present, in the supercritical direct current thermal power generating unit, the regulation and control method of the main steam temperature has the function of adjusting the fuel-water ratio, namely, the main steam temperature coarse adjustment is realized by adjusting the proportional relation between the fuel supply and the water supply flow in the combustion system; the method is mainly used for maintaining the steam temperature stable for a long time, but has the problems of large inertia and large delay. In addition, water spray attemperators are currently commonly used to regulate the temperature of the primary steam, and such devices are widely used due to their simplicity, precise control, high temperature control capability, ease of automation, and the like. The working principle of the water spray desuperheater is that the water is atomized and guided into the superheater to absorb steam heat, so that the temperature of the water spray desuperheater is reduced; by adjusting the amount of water spray, the degree of decrease in the temperature of the main steam can be controlled. The main steam temperature is finely adjusted by using three-stage water spray attemperation equipment, so that after steam comes out of a superheater of a thermal power unit, the steam temperature can meet the design requirement of a system, and finally superheated steam with qualified temperature and pressure quality required by the steam turbine generator unit is obtained.
However, control of the main steam temperature of the supercritical thermal power generating unit faces a great challenge. This is because once-through boiler plants are numerous and are connected to each other in series or parallel, and the small disturbance of each plant may cause the response of the adjacent plant to affect the overall boiler system.
In addition, disturbances in other combustion conditions within the furnace, such as fluctuations in air or fuel levels, may cause the flue gas heat transfer to vary substantially simultaneously along the entire superheater length. Thus, the main steam temperature at the superheater outlet has a faster dynamic response characteristic, with a shorter characteristic time and a shorter characteristic lag time, which means that the main steam temperature control system needs to deal with this fast changing characteristic. Therefore, the key idea of designing the main steam temperature control system is to counteract fluctuation of the heat absorption capacity of the superheater with a higher response speed by utilizing the desuperheating water with a lower response speed; this is also the root cause of the difficulty in achieving good quality for the boiler steam temperature. In industrial processes, it is required that the primary steam temperature control system must have the capability to withstand various disturbances, so that the steam temperature fluctuates within allowable limits and ensure that the superheater temperature does not exceed allowable values.
Based on the traditional PID control method, in recent years, the quality of the main steam temperature control of the thermal power unit is improved to a certain extent by the aid of the fuzzy PID control algorithm, and the fuzzy PID control algorithm is successfully applied to actual situations. Thereafter, more scholars have proposed using genetic algorithms or particle swarm optimization algorithms to improve PID controllers. In recent years, with the rapid development of deep learning algorithms, a learner proposes a PID algorithm based on a fuzzy neural network, a learner proposes a main steam temperature PID cascade control system based on fuzzy reasoning of a radial basis neural network, and a learner also proposes a steam temperature control algorithm based on predictive control of a cyclic neural network. However, because the neural network is capable of approximating complex nonlinear functions with arbitrary precision, it has self-adaptation and self-learning capabilities, modeling uncertain objects; under the complex working condition environment, the neural network can automatically adjust system parameters to adapt to a new environment at the fastest speed, and the problem that the conventional PID control method is difficult to deal with can be effectively solved. However, when the traditional cell body neuron BPNN network is used for controlling the main steam temperature, the problems of high calculation complexity, difficulty in engineering practical application, poor control effect, overlong response time and the like caused by insufficient network expression capability and limited generalization capability are needed to be solved.
Therefore, the invention provides a supercritical thermal power unit main steam temperature self-adaptive control method based on a dendritic neural network and fuzzy PID, which is stronger in self-adaptive learning ability than the traditional artificial neural network by weighting the parameter output of the dendritic network module and the fuzzy module, aiming at the problems of insufficient expression ability, high computational complexity, limited generalization ability and the like of a conventional cell body neural network model on the basis of fuzzy neural network PID control. Aiming at the problems of obvious change of dynamic characteristics of main steam temperature, large change of parameter structure and the like when the load is changed, the dendritic neural network has higher precision and adaptability, and can meet the main steam temperature control requirements of a complex boiler system with large inertia, large delay and nonlinear time variation.
The invention is described in further detail below with reference to the attached drawing figures:
referring to fig. 1 and 2, in an embodiment of the present invention, a main steam temperature adaptive control method of a supercritical thermal power generating unit is provided, which can solve the problem that optimal control parameters in a PID control algorithm are difficult to obtain manually.
Specifically, the main steam temperature self-adaptive control method of the supercritical thermal power generating unit comprises the following steps:
S1: and acquiring an expected signal and an output signal of the main steam temperature of the large supercritical thermal power generating unit, and calculating an error signal group of the expected signal and the output signal.
S2: and obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on the dendritic neural network model.
S3: and carrying out self-adaptive weighted summation on the fuzzy control parameters of the PID controller and the neural network control parameters, and applying the self-adaptive weighted summation to the PID controller to obtain the self-adaptive PID controller.
S4: and obtaining a control signal of the main steam temperature of the large supercritical thermal power generating unit through the self-adaptive PID controller according to the error signal group.
Wherein the error signal group includes an error signal, an accumulated error signal, and a differential signal of the error signal.
According to the self-adaptive control method for the main steam temperature of the supercritical thermal power generating unit, an error signal group is calculated based on a desired signal and an output signal of the main steam temperature of the large supercritical thermal power generating unit, fuzzy control parameters of a PID controller are obtained through a fuzzy control method according to the error signal group, neural network control parameters of the PID controller are obtained through a pre-trained neural network control parameter model based on a dendritic neural network model, self-adaptive weighted summation is carried out on the fuzzy control parameters of the PID controller and the neural network control parameters and the self-adaptive weighted summation is applied to the PID controller, and then the self-adaptive PID controller is obtained through the self-adaptive PID controller, and the control signal of the main steam temperature of the large supercritical thermal power generating unit is obtained. By adaptively weighting the fuzzy control parameters and the neural network control parameters, the adaptive learning capacity of the method is stronger than that of the traditional artificial neural network. Aiming at the problems of obvious change of dynamic characteristics of main steam temperature, large change of parameter structure and the like when load changes, the dendritic neural network has higher precision and adaptability, can meet the main steam temperature control requirements of a complex boiler system with large inertia, large delay and nonlinear time variation, provides higher control precision and strong self-adaptive learning capability, can also effectively lower calculation complexity, and realizes the self-adaptive setting of control system parameters.
Specifically, referring to fig. 3, proportional-integral-differential control (PID) is currently the most widely used basic control algorithm, and the PID control principle is to implement control through the idea of feedback according to the deviation between a given input and an actual output.
Assuming that the error signal set is error (t), the proportional control considers the current error signal set value, the error signal set and the proportional control constant k p Multiplying to obtain a proportion control term; the integration control takes into account past error signal set values, which utilize the error signal set accumulated sumAnd integral control constant k i Multiplying to obtain an integral control term; the differential control is to consider future error signal group, calculate the first order derivative derror (t)/dt of error signal group, and then combine with differential control constant k d The multiplication results in a differential control term. Finally, the signal u (t) output by the PID controller, namely the input signal actually received by the main steam temperature regulating system (executing mechanism) is obtained by summing the proportional control item, the integral control item and the differential control item, and the input signal is shown as a formula (1):
wherein k is p ，k i And k d The three parameters are respectively proportional gain parameter, integral gain parameter and differential gain parameter of PID controller, which are parameters that can be adapted, and are determined by the main steam temperature characteristics of the controlled object, and the transfer function form can be written as error (t) is the current error signal group, and is obtained by calculating an error from an input signal and a system output signal, as shown in formula (2):
error(t)＝r(t)-y(t) (2)
in the practical use process, the input signal is generally required to be sampled with the period as T, so that the continuous time T is converted into discrete sampling time, the final differentiation uses first-order backward differential approximation, the integration uses summation approximation, and the discrete PID output signal expression of the kth sampling time is finally obtained:
in one possible implementation manner, obtaining the fuzzy control parameter of the PID controller by the fuzzy control method according to the error signal set includes: solving the membership of the error signal group through a preset membership function to obtain the membership of the error signal group relative to each preset fuzzy subset; based on the membership degree of the error signal group relative to each preset fuzzy subset, fuzzy reasoning is carried out through a preset fuzzy control table, and fuzzy quantity of fuzzy control parameters of the PID controller is obtained; and performing anti-blurring on the blurring amount of the blurring control parameter of the PID controller by using a weighted average method to obtain the blurring control parameter of the PID controller.
Specifically, the membership of the error signal group is solved through a preset membership function, and the membership of the error signal group relative to each preset fuzzy subset is obtained; based on the membership degree of the error signal group relative to each preset fuzzy subset, fuzzy reasoning is carried out through a preset fuzzy control table to obtain Fuzzy control parameter fuzzy quantity of PID controller; defuzzification is carried out on the fuzzy quantity of the fuzzy control parameter based on the PID controller, and the fuzzy quantity of the fuzzy control parameter of the PID controller is processed by using a weighted average method, namely delta k P ,Δk I ,Δk D And performing anti-blurring to obtain fuzzy control parameters of the PID controller. Wherein the output quantity Deltak P ,Δk I ,Δk D The fuzzy amount of each PID control parameter is calculated and output after fuzzy reasoning is carried out according to a fuzzy rule table, and the calculation process is shown in the following formulas (5), (6) and (7). The fuzzy amount of the fuzzy control parameter of the PID controller is a vector composed of membership degrees of a plurality of fuzzy subsets, which represents membership degrees of the variable in different fuzzy subsets.
Wherein the fuzzy subset comprises: negative Big (NB), negative Medium (NM), negative Small (NS), zero (ZO), positive Small (PS), median (PM), and Positive Big (PB).
In particular, PID controllers have become widely used control algorithms based on relatively simple principles. However, in practical application, the essence of the main steam temperature control system is to overcome the heat absorption disturbance of the superheater with a relatively slow reaction characteristic by using the relatively slow temperature reduction water disturbance, and the large inertia and the large time lag characteristic of the controlled object make the traditional PID controller difficult to realize a good control effect. In order to solve the problem, the fuzzy control method based on expert knowledge is integrated to optimize the traditional PID controller, so that the control of the main steam temperature of the large supercritical thermal power generating unit is realized.
The fuzzy control method is essentially a nonlinear control method in the intelligent control field, and the fuzzy control method carries out specific adjustment on the output of the PID control module according to expert knowledge stored in a fuzzy control table through the processes of fuzzification, fuzzy reasoning and anti-fuzzification. The method realizes smaller overshoot and shorter adjustment time, thereby realizing real-time optimization of gain parameters of the PID controller, and effectively overcoming the limitation that the traditional PID controller cannot adjust the parameters in real time.
Inputs to the fuzzy control method include the error E and the error rate of change of the system output from the actual valueE c Outputting a parameter k corresponding to the PID controller P ，k I And k D Associated correction value set u 1 (k)。
Referring to fig. 4, the algorithm flow and principle of the fuzzy control method. Wherein, fuzzification refers to solving membership (membership of a certain fuzzy subset) according to a membership function, and the fuzzy subset is generally defined as 7 classes, namely Negative Big (NB), negative Medium (NM), negative Small (NS), zero (ZO), positive Small (PS), medium (PM) and Positive Big (PB), which are used for describing input fuzzy variables, and then solving the membership by adopting a relevant membership function (such as Gaussian membership function) so as to determine a specific domain.
Where μ (x, a, b) is a gaussian membership function, x is an input variable, and a, b are parameters of different membership functions. The adjusting parameters a and b can be used for meeting the control requirements of different systems in different value ranges by changing the function shape. E and Ec fuzzy theory domains are { -n, …,0, …, n }, { -m, …,0, …, m }, respectively, where m, n are the discretized numbers of stages, respectively. In general, the range of fuzzy system input variables E and Ec is defined as the basic domain of arguments. So that the quantization factors from the basic domain to the fuzzy domain are respectively
After obtaining the membership degree, fuzzy reasoning needs to be performed, namely, the membership degree corresponding to the output of the fuzzy controller is solved according to a fuzzy control table, which is a decision process, and the fuzzy control table is generally derived from a fuzzy subset. The specific control rule is the key of the fuzzy controller, which has great influence on the control performance of the controller.
Aiming at the characteristics of a main steam temperature object of a thermal power unit, the embodiment designs a fuzzy control rule table, and the main principle is that the adjustment strength is increased when an error signal group is large; the PID controller is favored before the steady state phase and controls inertia during the transition phase.
TABLE 1 fuzzy control table
Wherein Δk is P ，Δk I ，Δk D The fuzzy amounts calculated and output by fuzzy reasoning are respectively represented, and can be used for correcting the PID control parameters. Δk P Plays a main regulation role, delta k I And Deltak D And setting according to actual requirements.
The following is Δk P Solving is an example, and a specific solving process is described. According to the fuzzy control table, the corresponding fuzzy relation is given as { R } 1 ，R 2 ，...，R n×m The calculation is as follows:
where m=7 and n=7, the total ambiguity R may be combined according to all ambiguity relations as described above:
then the output blur amount Deltak can be calculated from the blur amounts of the input error E and the error change rate Ec P ：
Δk P ＝(E×Ec)oR (7)
Δk I ，Δk D The calculation of (2) is the same as the above method.
Finally, defuzzifying is performed, the output quantity is defuzzified by using a weighted average method, and then the accurate value is scaled to the domain of the PID controller. Assuming that the exact value of the output is Out, there are:
wherein w is i Represents the fuzzy threshold value, mu of output quantity i Represents w i Corresponding membership in each fuzzy subset. The calculated exact value is then scaled to the domain of the PID controller. Under the main steam temperature control scene, after calculating an accurate value according to the control rule, assuming that the PID control quantity theory is [ -v, v]A scale factor g=v/L is defined, where l=7 corresponds to the number of quantizations in the basic domain. Finally, the corrected output of the fuzzy control to the PID control parameter can be obtained as follows:
In one possible embodiment, the pre-trained neural network control parameter model based on a dendritic neural network model is obtained by:
constructing a dendritic neural network model; acquiring an expected signal and a training output signal of the main steam temperature of the large supercritical thermal power generating unit at the current training time, and acquiring an error signal group at the current training time according to the expected signal and the training output signal at the current training time; inputting the error signal group at the current training time into a dendritic neural network model to obtain the neural network control parameters of the PID controller at the current training time; obtaining fuzzy control parameters of the current training time of the PID controller through a fuzzy control method according to the error signal group of the current training time, carrying out self-adaptive weighted summation on the fuzzy control parameters of the current training time of the PID controller and the neural network control parameters, and applying the fuzzy control parameters and the neural network control parameters to the PID controller to obtain a self-adaptive PID controller of the current training time; according to the error signal group at the current training time, a training control signal of the main steam temperature of the large supercritical thermal power unit is obtained through a self-adaptive PID controller at the current training time and is applied, and an output signal of the main steam temperature of the large supercritical thermal power unit at the next training time is obtained; acquiring an expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, obtaining an error signal group of the next training time according to the output signal and the expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, and optimizing a dendritic neural network model according to the error signal group of the next training time; repeating the steps until the error between the output signal and the expected signal at the next training moment meets a preset error threshold, and taking the final dendritic neural network model as a neural network control parameter model.
Specifically, in the actual working condition, the transfer function of the system is determined after the system is identified, and then the PID control parameters are adjusted on line when the system is subjected to PID control by using the dendritic neural network model, and meanwhile, the network parameters of the dendritic neural network model are updated. Collecting system output signals and expected signals in each sampling time interval, calculating to obtain a system error signal group, an accumulated error signal group and error differential signals, and taking the error signal group as the input of a dendritic neural network model; the PID parameter value is regulated through network output, a new output is obtained through a regulation and control system, a new error value is obtained through calculation, and gradient updating of the dendritic neural network model parameter is carried out through the new error value; finally, when the system output converges and reaches the expected signal, the parameters of the dendritic neural network model also reach the convergence, and a pre-trained neural network control parameter model is obtained.
Specifically, although the cascade PID control method and the fuzzy PID control strategy overcome the problem of too slow response speed caused by a large inertia controlled object to some extent, adjusting the gain parameters of the PID controller to adapt to the characteristics of an actual controlled system often faces serious challenges, and only a technician is relied on to manually adjust to obtain optimal control parameters. Therefore, to solve this problem, an improved PID control method using BP neural network can make controller adjustments for model changes according to a learning strategy of error feedback. Gain parameters of the related PID controller are obtained after the fuzzy controller is passed through, and are transmitted to the BP neural network for further updating and adjustment.
However, the conventional somatic neural network is limited in terms of insufficient expression capacity, high computational complexity and generalization capability. Aiming at the problems of obvious change of dynamic characteristics of main steam temperature, large change of parameter structure and the like when load changes, the traditional BP neural network has limited adaptive capacity and can not meet the main steam temperature control requirements of a complex boiler system with large inertia, large delay and nonlinear time variation.
Therefore, the invention realizes parameter control based on the dendritic neural network model, has stronger self-adaptive learning capacity compared with the traditional artificial neural network, and can automatically learn the optimal control parameters.
The dendritic neural network is different from the traditional machine learning algorithm in that the dendritic neural network is a white box model, which means that the model related parameters are controllable in accuracy, and meanwhile, the dendritic neural network has low computational complexity and has good application effects in aspects of classification, regression, system identification and the like. Dendritic neural networks are also divided into two processes, forward propagation and reverse gradient. The forward propagation calculation of the dendritic neural network is Hadamard Product (Hadamard Product) using a matrix, and referring to FIG. 5, the schematic diagram of the dendritic neurons, the forward propagation can be expressed as:
wherein A is l-1 And A l Respectively module input and output, and W l，l-1 Is the layer network parameter, symbolThe Hadamard product of the matrix is represented, the calculation process is to take two matrices with the same dimension, and the matrices with the same dimension are multiplied by each element correspondingly and are used for simulating the information transmission among multiple dendrites in a biological neuron network, X is the input of a dendrite neural network model, and the matrix is obtained by carrying out normalization data processing on an error signal group. The dendritic neural network under the three-layer network structure has the input and output expressed as:
wherein W is 10 The weight parameters between the input neuron and the first layer of dendritic network are calculated by matrix multiplication and then the output of the layer is obtained by Hadamard product calculationW 21 、W 32 The weight parameters between the first layer and the second layer and between the second layer and the third layer are respectively calculated in the same way as the first layer, the output of each layer is used as the input of the next layer, and finally the neural network control parameter Y of the PID controller is obtained by outputting.
As can be seen from equations (10) and (11), the computational complexity of the dendritic neural network is much lower than that of the conventional BP neural network, since the computational process only includes matrix multiplication and hadamard product.
Referring to fig. 6 and 7, the error back-gradient propagation of the dendritic neural network is substantially the same as the BP network, with the error calculation and gradient calculation being as follows:
/>
dA l-1 ＝(W l，l-1 ) T dZ l (14)
The parameters of the dendritic neural network are updated as follows:
W l，l-1 (new)＝W l，l-1(old) -αdW l，l-1 (16)
wherein,and Y represents the output and label of the dendritic neural network respectively, m represents the data sample size, and alpha is the learning rate. dA (dA) L Is the output error value of the dendritic neural network, dZ L Is the local gradient of the output layer, dA l Is the output error value of the intermediate layer, dZ l Is the local gradient of the intermediate layer. A is that l-1 Represents the output of layer 1, W l，l-1 Then the weight matrix between the first layer and the first layer is dW l，l-1 Is the updated value of the weight matrix between the first layer and the second layer, W l，i-1(old) And W is equal to l，l-1(new) Is the weight matrix between the first layer and the second layer before and after gradient updating.
In one possible implementation manner, the adaptively weighted summing the fuzzy control parameter of the PID controller and the neural network control parameter and applying the summed fuzzy control parameter and the neural network control parameter to the PID controller, so as to obtain an adaptive PID controller includes:
and carrying out self-adaptive weighted summation on the fuzzy control parameter of the PID controller and the neural network control parameter by the following steps:
wherein alpha is 1 ，α 2 And alpha 3 All are self-adaptive weight parameters, and the value ranges are all between 0 and 1; k (k) P Is the proportional gain parameter k of the adaptive PID controller I Is the integral gain parameter k of the adaptive PID controller D Differential gain parameters for the adaptive PID controller; Proportional gain parameter among fuzzy control parameters of PID controller,/for the control of the PID controller>Is an integral gain parameter among fuzzy control parameters of the PID controller,/and the like>A differential gain parameter among fuzzy control parameters of the PID controller; />Proportional gain parameter, among the neural network control parameters of the PID controller, +>Integral gain parameter of neural network control parameters for PID controller, +>Is a differential gain parameter among the neural network control parameters of the PID controller.
Wherein alpha is 1 ，α 2 And alpha 3 Are all determined by an Actor-Critic network model based on reinforcement learning.
Specifically, the determination is made by using a dual delay depth deterministic strategy gradient (Twin Delayed Deep Deterministic Policy Gradients, TD 3) algorithm, the inputs of which are the fuzzy control parameters of the PID controller, the neural network control parameters of the PID controller and the error signal set error (t).
The state space is set up hereinThe motion space is the adaptive parameter value a t ＝(α 1 ，α 2 ，α 3 ). Setting a bonus function toWherein u (t) is in state s t Take action a down t The adjustment amount of the rear control system and enter the next time state s t+1 。
Specifically, referring to fig. 8, the reinforcement learning-based Actor-Critic model used in the present invention includes two sets of neural networks: a set of policy function networks (Actor networks) which are agents for selecting actions according to the current state; another group is a cost function Network (Critical Network), using The motion value is estimated. Each group of neural networks comprises paired Target and Online networks, and the neural networks are updated in a soft update mode. Actor Online network Actor_o is defined as mu (s, phi), and the corresponding Target network Actor_t is defined as mu (s, phi) - ). Two pairs of Critic networks are defined in the same sense, critic_o1 being denoted Q (s, a; θ) 1 ) And critic_o2 is represented as Q (s, a; θ 2 ) The Target networks critic_t1 and critic_t2 corresponding to the two are denoted as Q (s, a; θ 1 ') and Q (s, a; θ 2 '), phi - ，θ 1 ，θ 2 ，θ 1 ′，θ 2 ' are weight parameters in each fully connected network component that make up the Actor-Critic network model.
The goal of the policy network is to represent the optimal policy μ(s) with respect to φ t The method comprises the steps of carrying out a first treatment on the surface of the Phi) to maximize jackpotThe gradient-up update parameter phi is used as shown in equation (18):
wherein α represents the learning rate, and the J (μ (Φ)) gradient is calculated as shown in the following formula (19):
the objective of the cost function network is to minimize the mean square error of the Q function to find the optimal parameter θ * As shown in the following formula (20), wherein θ ε { θ } 1 ，θ 2 The problem of Q value overestimation easily occurs in deep reinforcement learning can be solved by the design, so that the training process is more stable:
as shown in (21)Show, y t Is the target of the time difference (Temporal Difference, TD), Q(s) t ，a t The method comprises the steps of carrying out a first treatment on the surface of the θ) is the Critic network in state s t Action a taken at the time t Next, the expected sum of future postings from time t, γ represents the discount coefficient, θ' e { θ } 1 ′，θ 2 ′}：
y t ＝r t +γmin θ′ Q(s t+1 ，μ(s t+1 ；φ - )；θ′) (21)
After model training is finished, the output obtained by the fuzzy control methodOutput of neural network control parameter model>The error signal group error (t) is transmitted into the Actor network to output the corresponding action a t ＝(α 1 ,α 2 ,α 3 ). After the corresponding weighting coefficient is learned through the reinforcement learning network, the fuzzy control parameter of the PID controller and the neural network control parameter are subjected to weighted fusion according to a formula (17), so that the control parameter of the final PID controller of the system is obtained.
Specifically, the main steam temperature self-adaptive control method of the supercritical thermal power generating unit can realize the combined self-adaptive control of the main steam temperature through the combination of the fuzzy PID and the novel dendritic neural network, thereby improving the control effect, reducing the overshoot and reducing the adjustment time.
Aiming at the problem that a conventional PID control method is difficult to control a main steam temperature object with large inertia, uncertainty and nonlinearity, the invention integrates a fuzzy control algorithm, and a self-designed fuzzy control table corrects PID control parameters, so that the system is ensured to be greatly adjusted when the deviation is large, and the dynamic response of the system is more rapid; and inertia is effectively reduced near steady state, thereby achieving lower overshoot and shorter settling times.
Aiming at the problems that the optimal control parameters of the traditional PID algorithm and the fuzzy PID controller need to be tried manually in a large amount and are difficult to acquire automatically, the invention discloses a main steam temperature self-adaptive control method of a supercritical thermal power unit, which is used for self-adaptively learning the optimal control parameters through model training. Specifically, the error signal group is used as input, the fuzzy control parameters and the neural network control parameters of the corresponding PID controller are output through a fuzzy control method and a neural network control parameter model based on a dendritic neural network model respectively, the weight parameters of the neural network control parameter model are updated at the same time, finally, the controlled object (main steam temperature) is controlled after self-adaptive weighting adjustment according to the fuzzy control parameters and the neural network control parameters, and finally, the optimal control parameter self-adaptive learning is realized.
Compared with the existing PID algorithm based on the BP neural network, the main steam temperature self-adaptive control method of the supercritical thermal power generating unit has stronger learning characterization capability and generalization capability, and the Hadamard product operation effectively reduces the calculation complexity, so that the proposed algorithm ensures strong self-adaptive parameter learning capability, and meanwhile, has higher calculation efficiency and can better adapt to complex working conditions.
In addition, the main steam temperature self-adaptive control method of the supercritical thermal power generating unit benefits from the strong self-adaptive learning capacity of the dendritic neural network, and the dynamic characteristic of the main steam temperature when the main steam temperature changes along with the load can be well learned. The reason is that compared with the traditional BP neural network based on the traditional somatic neuron design, the dendritic neurons can learn complex logic operation better, have stronger generalization capability and lower computation complexity, and most importantly, overcome the problem of manually setting a large number of control parameters.
The following are device embodiments of the present invention that may be used to perform method embodiments of the present invention. For details not disclosed in the apparatus embodiments, please refer to the method embodiments of the present invention.
Referring to fig. 9, in still another embodiment of the present invention, a main steam temperature adaptive control system of a supercritical thermal power generating unit is provided, which can be used to implement the above-mentioned main steam temperature adaptive control method of a supercritical thermal power generating unit, and specifically, the main steam temperature adaptive control system of a supercritical thermal power generating unit includes a signal processing module, a parameter determining module, an adaptive weighting module, and a control module.
The signal processing module is used for acquiring an expected signal and an output signal of the main steam temperature of the large supercritical thermal power unit and calculating an error signal group of the expected signal and the output signal; the parameter determining module is used for obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on a dendritic neural network model; the self-adaptive weighting module is used for carrying out self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network and applying the fuzzy control parameters and the neural network to the PID controller to obtain a self-adaptive PID controller; and the control module is used for obtaining a control signal of the main steam temperature of the large supercritical thermal power unit through the self-adaptive PID controller according to the error signal group.
All relevant contents of each step related to the foregoing embodiment of the main steam temperature control method of the large supercritical thermal power unit may be cited to the functional description of the functional module corresponding to the main steam temperature control system of the large supercritical thermal power unit in the embodiment of the present invention, which is not described herein.
The division of the modules in the embodiments of the present invention is schematically only one logic function division, and there may be another division manner in actual implementation, and in addition, each functional module in each embodiment of the present invention may be integrated in one processor, or may exist separately and physically, or two or more modules may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules.
In yet another embodiment of the present invention, a computer device is provided that includes a processor and a memory for storing a computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate array (FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular to load and execute one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding functions; the processor provided by the embodiment of the invention can be used for the operation of the main steam temperature self-adaptive control method of the supercritical thermal power unit.
In yet another embodiment of the present invention, a storage medium, specifically a computer readable storage medium (Memory), is a Memory device in a computer device, for storing a program and data. It is understood that the computer readable storage medium herein may include both built-in storage media in a computer device and extended storage media supported by the computer device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the method for main steam temperature adaptive control of a supercritical thermal power generating unit in the above embodiments.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.

Claims (9)

1. A main steam temperature self-adaptive control method of a supercritical thermal power generating unit is characterized by comprising the following steps:
acquiring an expected signal and an output signal of main steam temperature of a large supercritical thermal power generating unit, and calculating an error signal group of the expected signal and the output signal;
obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on a dendritic neural network model;
performing self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network, and applying the fuzzy control parameters and the control parameters of the neural network to the PID controller to obtain a self-adaptive PID controller;
and obtaining a control signal of the main steam temperature of the large supercritical thermal power generating unit through the self-adaptive PID controller according to the error signal group.
2. The method for adaptively controlling the main steam temperature of the supercritical thermal power generating unit according to claim 1, wherein the obtaining the fuzzy control parameter of the PID controller by the fuzzy control method according to the error signal group comprises:
solving the membership of the error signal group through a preset membership function to obtain the membership of the error signal group relative to each preset fuzzy subset;
Based on the membership degree of the error signal group relative to each preset fuzzy subset, fuzzy reasoning is carried out through a preset fuzzy control table, and fuzzy quantity of fuzzy control parameters of the PID controller is obtained;
and performing anti-blurring on the blurring amount of the blurring control parameter of the PID controller by using a weighted average method to obtain the blurring control parameter of the PID controller.
3. The method for adaptively controlling the main steam temperature of the supercritical thermal power generating unit according to claim 1, wherein the pre-trained neural network control parameter model based on the dendritic neural network model is obtained by the following steps:
constructing a dendritic neural network model; acquiring an expected signal and a training output signal of the main steam temperature of the large supercritical thermal power generating unit at the current training time, and acquiring an error signal group at the current training time according to the expected signal and the training output signal at the current training time;
inputting the error signal group at the current training time into a dendritic neural network model to obtain the neural network control parameters of the PID controller at the current training time; obtaining fuzzy control parameters of the current training time of the PID controller through a fuzzy control method according to the error signal group of the current training time, carrying out self-adaptive weighted summation on the fuzzy control parameters of the current training time of the PID controller and the neural network control parameters, and applying the fuzzy control parameters and the neural network control parameters to the PID controller to obtain a self-adaptive PID controller of the current training time;
According to the error signal group at the current training time, a training control signal of the main steam temperature of the large supercritical thermal power unit is obtained through a self-adaptive PID controller at the current training time and is applied, and an output signal of the main steam temperature of the large supercritical thermal power unit at the next training time is obtained;
acquiring an expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, obtaining an error signal group of the next training time according to the output signal and the expected signal of the main steam temperature of the large supercritical thermal power generating unit at the next training time, and optimizing a dendritic neural network model according to the error signal group of the next training time;
repeating the steps until the error between the output signal and the expected signal at the next training moment meets a preset error threshold, and taking the final dendritic neural network model as a neural network control parameter model.
4. The method for adaptively controlling the main steam temperature of the supercritical thermal power generating unit according to claim 1, wherein the step of adaptively weighted summing the fuzzy control parameter of the PID controller and the neural network control parameter and applying the summed fuzzy control parameter and the neural network control parameter to the PID controller to obtain the adaptive PID controller comprises the steps of:
And carrying out self-adaptive weighted summation on the fuzzy control parameter of the PID controller and the neural network control parameter by the following steps:
wherein alpha is 1 ，α 2 And alpha 3 All are self-adaptive weight parameters, and the value ranges are all between 0 and 1; k (k) P Is the proportional gain parameter k of the adaptive PID controller I Is the integral gain parameter k of the adaptive PID controller D Differential gain parameters for the adaptive PID controller;proportional gain parameter among fuzzy control parameters of PID controller,/for the control of the PID controller>Is an integral gain parameter among fuzzy control parameters of the PID controller,/and the like>A differential gain parameter among fuzzy control parameters of the PID controller; />Proportional gain parameter, among the neural network control parameters of the PID controller, +>Integral gain parameter of neural network control parameters for PID controller, +>Is a differential gain parameter among the neural network control parameters of the PID controller.
5. The main steam temperature adaptive control method of a supercritical thermal power generating unit according to claim 4, wherein the alpha is 1 ，α 2 And alpha 3 And the fuzzy control parameters of the PID controller, the neural network control parameters of the PID controller and the error signal group are obtained through an Actor-Critic network model based on reinforcement learning.
6. The method for adaptively controlling the main steam temperature of the supercritical thermal power generating unit according to claim 1, wherein the obtaining the control signal of the main steam temperature of the large supercritical thermal power generating unit through the adaptive PID controller according to the error signal group comprises: and according to the error signal group, obtaining a control signal of the main steam temperature of the large supercritical thermal power generating unit by the following formula:
wherein u is 0 (k) For the control signal of the main steam temperature of the large supercritical thermal power generating unit at the moment k, error (k) is an error signal group at the moment k, error (k-1) is an error signal group at the moment k-1, and k P Is the proportional gain parameter k of the adaptive PID controller I Is the integral gain parameter k of the adaptive PID controller D Is a differential gain parameter of the adaptive PID controller.
7. Main steam temperature control system of large-scale supercritical thermal power unit, characterized by comprising:
the signal processing module is used for acquiring an expected signal and an output signal of the main steam temperature of the large supercritical thermal power generating unit and calculating an error signal group of the expected signal and the output signal;
the parameter determining module is used for obtaining fuzzy control parameters of the PID controller through a fuzzy control method according to the error signal group, and obtaining the neural network control parameters of the PID controller through a pre-trained neural network control parameter model based on a dendritic neural network model;
The self-adaptive weighting module is used for carrying out self-adaptive weighted summation on fuzzy control parameters of the PID controller and control parameters of the neural network and applying the fuzzy control parameters and the control parameters of the neural network to the PID controller to obtain a self-adaptive PID controller;
and the control module is used for obtaining a control signal of the main steam temperature of the large supercritical thermal power unit through the self-adaptive PID controller according to the error signal group.
8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, realizes the steps of the main steam temperature adaptive control method of a supercritical thermal power plant according to any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the main steam temperature adaptive control method of a supercritical thermal power plant according to any one of claims 1 to 6.
CN202311630148.9A 2023-11-30 2023-11-30 Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit Pending CN117518823A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311630148.9A CN117518823A (en) 2023-11-30 2023-11-30 Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311630148.9A CN117518823A (en) 2023-11-30 2023-11-30 Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit

Publications (1)

Publication Number Publication Date
CN117518823A true CN117518823A (en) 2024-02-06

Family

ID=89741885

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311630148.9A Pending CN117518823A (en) 2023-11-30 2023-11-30 Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit

Country Status (1)

Country Link
CN (1) CN117518823A (en)

Similar Documents

Publication Publication Date Title
de Oliveira et al. Application of artificial intelligence techniques in modeling and control of a nuclear power plant pressurizer system
Shakibjoo et al. Load frequency control for multi-area power systems: A new type-2 fuzzy approach based on Levenberg–Marquardt algorithm
JPH03164804A (en) Process control system and power plant process control system
CN110285403A (en) Main Steam Temperature Control method based on controlled parameter prediction
Azeer et al. Intelligent controllers for load frequency control of two-area power system
CN111290282B (en) Predictive control method for thermal power generating unit coordination system
CN114722693A (en) Optimization method of two-type fuzzy control parameter of water turbine regulating system
CN113835342B (en) Disturbance rejection predictive control method for overheat steam temperature system
CN114428456B (en) Control method and device for thermal power generating unit control system
Prasanth et al. A hybrid technique to control superheater steam temperature in power plants using multi modeling and predictive sliding mode control
CN109062030A (en) Thermal power unit plant load prediction PID control method based on laguerre function model
JP3012716B2 (en) Process control method and its control device
CN117518823A (en) Main steam temperature self-adaptive control method and related device for supercritical thermal power generating unit
Medagam et al. Optimal control of nonlinear systems using RBF neural network and adaptive extended Kalman filter
CN116047897A (en) Gas turbine predictive control method based on parameter self-adaptive disturbance rejection controller
Li et al. Performance analysis of reheat steam temperature control system of thermal power unit based on constrained predictive control
Ma et al. ANN and PSO based intelligent model predictive optimal control for large-scale supercritical power unit
Eissa Novel fuzzy-based self-adaptive single neuron PID load frequency controller for power system
Wang et al. Improving Superheated Steam Temperature Control Using United Long Short Term Memory and MPC
Tu et al. State variable-fuzzy prediction control strategy for superheated steam temperature of thermal power units
CN112615364A (en) Novel wide-area intelligent cooperative control method for power grid stability control device
Xiao et al. Design of Liquid Level Control System of Steam Generator Based on Neural Network PID Controller
Rameshappa et al. An optimal artificial neural network controller for load frequency control of a four-area interconnected power system
Li et al. Adaptive control of steam generator feed-water system based on improved single neuron self-adaptive control
Kamarzarrin et al. A new fault-tolerant control of wind turbine pitch system based on ANN model and robust and optimal development of MRAC method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination