660mw ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method

Abstract

The present application relates to a kind of 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method, comprising: selecting the relevant characteristic variable parameter of the influence circulating fluidized bed unit bed temperature and main steam pressure;Comprehensive pre-processing is carried out to relevant characteristic variable parameter data;Select coal supply, water supply, primary air volume as input feature;Select the bed temperature value of previous time, main steam pressure value as input feature;According to the size of unit load, circulating fluidized bed unit operating condition is divided;Establish the coupling prediction model of circulating fluidized bed unit bed temperature and main steam pressure under different conditions, and parameter optimization is carried out using simulated annealing algorithm, to determine the dynamic response time step of output characteristic bed temperature and main steam pressure under different conditions;Using multi-condition model generalized predictive control algorithm, 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization are carried out.The present application can realize 660MW ultra-supercritical circulating fluidized bed boiler bed temperature high-quality prediction and control optimization.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of operation optimization of 660MW ultra-supercritical circulating fluidized bed boilers, and particularly relates to a 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method. BACKGROUND

[0002] The 660MW ultra-supercritical circulating fluidized bed boiler is one of the ways to realize clean and efficient utilization of coal under the "double carbon" target, has the advantages of multiple combustible fuel types, high operation efficiency, low nitrogen oxide emission, etc., and has broad development potential. The bed temperature is an important state parameter of the 660MW ultra-supercritical circulating fluidized bed boiler, and has a direct impact on the safe and stable operation of the boiler. If the bed temperature is too high, it will lead to a decrease in the desulfurization efficiency of the boiler, an increase in the NOx emission content in the flue gas, and an easy coking of the bed material in the furnace. If the bed temperature is too low, it will lead to a decrease in the combustion efficiency of the boiler and unstable operation of the boiler, and even cause the phenomenon of extinguishing.

[0003] There are many factors affecting the bed temperature, and the serious coupling between the factors, the large delay effect, and the gas-solid two-phase flow characteristics in the combustion process make the established mechanism model very complex, which is not suitable for the design and optimization of the power plant unit control system. At present, due to the consumption and frequency modulation tasks of new energy such as wind power and photovoltaic, the dynamic process of the 660MW ultra-supercritical circulating fluidized bed boiler is more frequent, and the performance and control difficulty of the bed temperature in the dynamic process are increased. In the process of unit operation, the bed temperature and the main steam pressure in the closed-loop system model of the 660MW ultra-supercritical circulating fluidized bed boiler under the action of the fuel quantity and the primary air quantity are seriously coupled, and the single-input single-output model and the multi-input single-output model established by many experts and scholars are difficult to meet the current high-quality prediction and control requirements. SUMMARY

[0004] The purpose of the present application is to provide a 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method to solve the above technical problems.

[0005] The present application provides a 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method, comprising the following steps:

[0006] Step 1: According to the actual process structure arrangement and operation process of the 660MW ultra-supercritical circulating fluidized bed boiler, combined with the power plant measurement point table, the related characteristic variable parameters affecting the bed temperature and the main steam pressure of the circulating fluidized bed unit are selected; the related characteristic variable parameters include the coal supply quantity, the water supply quantity, the induced air quantity, the primary air quantity, the secondary air quantity, the limestone quantity, the slag discharge quantity, and the return material quantity;

[0007] Step 2: Collecting the related characteristic variable parameter data selected in step 1 from the unit database for comprehensive preprocessing; the comprehensive preprocessing includes data cleaning processing and data filtering processing;

[0008] Step 3, based on the strong coupling relationship between the main steam pressure and the bed temperature, the relationship between the pre-processed relevant characteristic variable parameter data and the output characteristic bed temperature and the main steam pressure is analyzed from the mechanism angle, and the coal supply, the water supply and the primary air volume are selected as the input characteristics;

[0009] Step 4, based on the time delay characteristic of the bed temperature response of the 660MW ultra-supercritical circulating fluidized bed boiler, the bed temperature values and the main steam pressure values at the previous time points are selected as the input characteristics;

[0010] Step 5, the operating conditions of the circulating fluidized bed unit are divided according to the size of the unit load;

[0011] Step 6, according to the operating condition division of step 5, the bed temperature and the main steam pressure coupling prediction model of the circulating fluidized bed unit under different operating conditions is established, and the simulated annealing algorithm is used for parameter optimization to determine the dynamic response time step of the output characteristic bed temperature and the main steam pressure under different operating conditions; the prediction model is:

[0012]

[0013] In the formula, T(t), P(t) represent the outputs of the bed temperature and the main steam pressure at the current time t respectively; x1(t), x2(t), x3(t) represent the inputs of the coal supply, the water supply and the primary air volume at the current time t respectively; m, n, a, b, c represent the dynamic response time steps of the bed temperature, the main steam pressure, the coal supply, the water supply and the primary air volume respectively, the dynamic response time step is related to the data sampling frequency, and the value range is 0-20, and the positive integer is taken;

[0014] Step 7, based on the multi-condition prediction model established in step 6, the multi-condition model generalized predictive control algorithm is adopted to replace the PID control, and the bed temperature prediction and control optimization of the 660MW ultra-supercritical circulating fluidized bed boiler are carried out.

[0015] Further, the data cleaning and filtering processing in step 2 comprises:

[0016] The box plot method is used to remove outlier data:

[0017] Assuming that q1 and q3 are the first quartile and the third quartile of the data, the box plot method can be expressed as:

[0018] x max =q3+1.5×(q3-q1)

[0019] x min =q3-1.5×(q3-q1)

[0020] Wherein, x maxrepresents an abnormal maximum value in the data, x min represents an abnormal minimum value in the data; if there is data less than the abnormal minimum value in the data, data greater than the abnormal maximum value is determined as outlier data and is removed;

[0021] The Kalman filtering method is used for data filtering processing, and the Kalman filtering model is shown in the following formula:

[0022] X(k)=A*X(k-1)+B*U(k)+W(k)

[0023] Z(k)=H*X(k)+V(k)

[0024] In the formula, X(k) is the system state at time k, U(k) is the control amount of the system at time k; A and B are system parameters, and for a multi-model system, A and B are matrices; Z(k) is the measurement value at time k, H is the parameter of the measurement system, and for a multi-measurement system, H is a matrix; W(k) and V(k) represent process and measurement noises respectively; the system state is optimally estimated through system input and output observation data.

[0025] Further, the working condition division is performed by using a clustering algorithm based on projection on a convex set in step 5.

[0026] Further, the clustering algorithm process is as follows:

[0027] Algorithm input: data set X, clustering number k;

[0028] Algorithm output: clustering representative set Y;

[0029] 1) Randomly select k sample objects as initial cluster centers from the entire data set X;

[0030] 2) Calculate the Euclidean distance of each sample object x m to the cluster center c i ;

[0031] 3) Find the minimum distance of each sample object x m to the cluster center c i , and classify the sample object x m into the same cluster as c i ;

[0032] 4) Calculate the weighted average value of the objects in the same cluster, and update the cluster center; wherein: the weighted coefficient of each object is calculated as follows:

[0033]

[0034] 5) Repeat steps 2) to 4) until the cluster center no longer changes.

[0035] Further, the step of parameter optimization in step 6 utilizes the simulated annealing algorithm, and the steps are as follows:

[0036] (1) select the iteration number q and the initial control temperature T q , let q = 0; select an initial state x0 from the feasible solution space, and calculate the objective function value f(x0);

[0037] (2) generate a random disturbance in the feasible solution space to generate a new state x1, and calculate the objective function value f(x1);

[0038] (3) judge whether to accept according to the state acceptance function: if f(x1) < f(x0), accept the new state x1 as the current state, otherwise, judge whether to accept x1 according to the Metropolis criterion, if accepted, let the current state equal to x1, if not accepted, let the current state equal to x0;

[0039] (4) according to the temperature cooling scheme T q+1 = CT q , C ∈ (0, 1), reduce the control temperature T q+1 ;

[0040] (5) judge whether the annealing process is terminated, when ten new solutions are not accepted in succession, or when the iteration number q is reached, terminate the algorithm, and go to step (6), otherwise go to step (2);

[0041] (6) output the current solution as the optimal solution.

[0042] Further, the step 6 comprises:

[0043] The plurality of typical working condition point models are established to approximate the characteristics of the entire operating range of the controlled object, and a corresponding controller is designed for each sub-model;

[0044] In actual operation, the outputs of the limited number of sub-controllers are mapped to the final control bed temperature effect through switching or weighting.

[0045] By the above scheme, through the 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method, the following technical effects are achieved:

[0046] 1) The present application is based on the operation mechanism of the 660MW ultra-supercritical circulating fluidized bed boiler, combined with feature selection, operation condition division based on projection on convex set, bed temperature and main steam pressure coupling prediction, multi-condition model generalized predictive control and other intelligent methods, realizes the 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization.

[0047] 2) The present invention fully considers that during the operation of the unit, the bed temperature and the main steam pressure in the closed-loop system model of the 660MW ultra-supercritical circulating fluidized bed boiler under the action of fuel quantity and primary air volume are severely coupled (it is very difficult for the single-input single-output bed temperature models and multi-input single-output bed temperature models established by many experts and scholars at present to meet the current control requirements). A dynamic characteristic model for coupling prediction of bed temperature and main steam pressure is established, laying a foundation for the automatic control optimization of the bed temperature of the 660MW ultra-supercritical circulating fluidized bed boiler.

[0048] 3) The present invention further aims at the characteristics of strong non-linearity and large time delay of the bed temperature of the circulating fluidized bed boiler, overcomes the shortcomings of PID post-control, and combines the multi-model control strategy and generalized predictive control to improve the control quality of the bed temperature, enhancing the anti-disturbance ability and adaptive ability.

[0049] The above description is only an overview of the technical solution of the present invention. In order to be able to more clearly understand the technical means of the present invention and implement it in accordance with the content of the specification, the following takes the preferred embodiments of the present invention and combines the drawings to describe in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flowchart of the method for predicting and optimizing the control of the bed temperature of the 660MW ultra-supercritical circulating fluidized bed boiler of the present invention.

[0051] Figure 2 is an exemplary diagram of the coupling prediction model of the bed temperature and the main steam pressure of the 660MW ultra-supercritical circulating fluidized bed boiler of the present invention;

[0052] Figure 3 is an exemplary diagram of the generalized predictive control of the multi-condition model of the bed temperature of the 660MW ultra-supercritical circulating fluidized bed boiler of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The following combines the drawings and embodiments to further describe in detail the specific embodiments of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0054] Refer Figure 1 As shown, this embodiment provides a method for predicting and optimizing the control of the bed temperature of a 660MW ultra-supercritical circulating fluidized bed boiler, including the following steps:

[0055] Step 1, according to the actual process structure layout and operation process of the 660MW ultra-supercritical circulating fluidized bed boiler, combined with the power plant measurement point table, from the perspective of mechanism analysis, select the relevant characteristic variable parameters that affect the bed temperature and main steam pressure of the circulating fluidized bed unit, mainly including coal feeding amount, water feeding amount, induced draft fan air volume, primary air volume, secondary air volume, limestone amount, slag discharge amount, return material amount, etc.

[0056] Step 2, collect the data selected in step 1 from the unit database, considering that the collected measurement data may have distorted abnormal points and high-frequency noise, the collected data is comprehensively pretreated.

[0057] Step 3, considering the strong coupling relationship between the main steam pressure and the bed temperature, the relationship between the input characteristics (related characteristic variable parameters) in step 2 and the output characteristics bed temperature and main steam pressure is analyzed, and the variables with small correlation are deleted, and the coal supply, water supply and primary air volume are selected as input characteristics.

[0058] Step 4, considering that the bed temperature response of the 660MW ultra-supercritical circulating fluidized bed boiler has a large time delay, in order to establish an accurate circulating fluidized bed unit bed temperature and main steam pressure coupling prediction model, the bed temperature value and the main steam pressure value at the previous time are also used as input characteristics.

[0059] Step 5, the boiler heat storage, inertia and dynamic performance change greatly under running conditions. The circulating fluidized bed unit running conditions are divided according to the unit load size.

[0060] Step 6, according to the working condition division in step 5, the bed temperature and main steam pressure coupling prediction model of the circulating fluidized bed unit under different working conditions is established (as shown in Figure 2 ), and the simulated annealing algorithm is used for parameter optimization to determine the dynamic response time step of the output characteristics (bed temperature and main steam pressure) under different working conditions.

[0061] Step 7, considering that the control effect of the generalized predictive control (GPC) algorithm on the large lag object is obviously better than that of the PID, based on the multi-condition model established in step 6, the multi-condition model generalized predictive control algorithm is adopted to replace the PID control, and the 660MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization is realized. As shown in Figure 3 .

[0062] In this embodiment, in order to ensure the accuracy of the data in step 2, considering that the collected measurement data may have distorted abnormal points and high-frequency noise, the collected data is comprehensively pretreated, mainly including cleaning and filtering processing. The cleaned data can include null data and outlier data, and further can include shutdown condition data, that is, the data corresponding to the period when the unit load is 0 or close to 0.

[0063] The null data is the data with null value at one or more measuring points at a time, and the outlier data is the data beyond the normal range. The corresponding code is written to remove the null data, and the box plot method is used to remove the outlier data. Assuming that q1 and q3 are the 1st quartile and the 3rd quartile of the data, the box plot method can be expressed as:

[0064] xmax=q3+1.5×(q3-q1)

[0065] xmin = q3 - 1.5 x (q3 - q1)

[0066] wherein x max represents an abnormal maximum value in the data, x min represents an abnormal minimum value in the data. If there is data in the data less than the abnormal minimum value and greater than the abnormal maximum value, the data is determined as outlier data and is removed.

[0067] Next, the data is filtered.

[0068] The method of filtering the data adopts Kalman filtering method. Kalman filtering is a kind of linear system state equation, and a common model is shown in the following formula:

[0069] X(k) = A * X(k - 1) + B * U(k) + W(k)

[0070] Z(k) = H * X(k) + V(k)

[0071] In the formula, X(k) is the system state at k time, U(k) is the control amount of the system at k time. A and B are system parameters, and for a multi-model system, A and B are matrices. Z(k) is the measurement value at k time, H is the parameter of the measurement system, and for a multi-measurement system, H is a matrix. W(k) and V(k) represent the process and measurement noise respectively. The optimal estimation of the system state is obtained through the observation data of the system input and output. Since the observation data includes the influence of noise and interference in the system, the optimal estimation is regarded as a filtering process.

[0072] In the embodiment, the analysis of the relationship between the input features and the output features bed temperature and main steam pressure in step 2 is deleted in step 3. The specific coupling relationship is shown in the following table.

[0073] Variable Bed temperature Main steam pressure Coal feed rate Strong Strong Draught rate Weak Weak Water feed rate Strong Strong Primary air rate Strong Strong Secondary air rate Weak General Slag discharge rate General Weak

[0074] Therefore, from the mechanism point of view, the coal supply, water supply and primary air volume are selected as input features.

[0075] In the embodiment, the operation condition division of the circulating fluidized bed unit according to the unit load size in step 5 is achieved by using a clustering method based on Projections onto Convex Sets (POCS). In mathematics, a convex set is a set in which the line segment between any two points in the set is also in the set. Projection is an operation of mapping a point to a subspace in another space. Given a convex set and a point, the operation can be performed by finding the projection of the point on the convex set. The projection is the point in the convex set closest to the point, which can be calculated by minimizing the distance between the point and any other point in the convex set. The clustering operation is achieved by mapping the features to the convex set in another space through projection.

[0076] The working principle of the algorithm is similar to that of the classic K-Means algorithm, but there is a difference in the way each data point is processed: the K-Means algorithm weights each data point equally, but the POCS-based clustering algorithm weights each data point differently, which is proportional to the distance of the data point to the cluster center. The overall process of the algorithm is as follows:

[0077] Algorithm input: data set X, number of clusters k

[0078] Algorithm output: cluster representative set Y

[0079] Step 1: randomly select k sample objects from the entire data set X as initial cluster centers;

[0080] Step 2: calculate the Euclidean distance of each sample object x m to the cluster center c i ;

[0081] Step 3: find the minimum distance of each sample object x m to the cluster center c i , and assign the sample object x m to the same cluster as c i ;

[0082] Step 4: calculate the weighted average of the objects in the same cluster and update the cluster center; where: the weighted coefficient of each object is calculated as follows:

[0083]

[0084] Step 5: repeat steps 2-4 until the cluster center no longer changes.

[0085] In the embodiment, the bed temperature and main steam pressure coupling prediction model of the circulating fluidized bed unit under different operating conditions is established according to the operating condition division in step 5. The model is as follows:

[0086]

[0087] In the formula, T(t), P(t) represent the output of the bed temperature and the main steam pressure at the current time t respectively; x1(t), x2(t), x3(t) represent the input of the coal supply, the water supply, and the primary air volume at the current time t respectively; m, n, a, b, c represent the dynamic response time step of the bed temperature, the main steam pressure, the coal supply, the water supply, and the primary air volume respectively, which are related to the data sampling frequency, and generally range from 0 to 20, and are positive integers.

[0088] In step 6, the simulated annealing algorithm is used for parameter optimization to determine the dynamic response time step of the output characteristics (bed temperature and main steam pressure) under different working conditions. The simulated annealing algorithm is a heuristic random search method that not only introduces appropriate random factors, but also introduces the natural mechanism of the physical system annealing process. In the iteration process, it not only accepts points that make the objective function value "better", but also can accept points that make the objective function value "worse" with a certain probability, and the acceptance probability gradually decreases with the decrease of the temperature, which can make the algorithm jump out of the local optimal solution and obtain the global optimal solution, which is beneficial to improve the reliability of obtaining the global optimal solution. The solving steps are as follows:

[0089] Step 1: Select the iteration number q and the initial control temperature T q (Let q = 0) and select an initial state x0 from the feasible solution space, and calculate the objective function value f(x0);

[0090] Step 2: Generate a random disturbance in the feasible solution space to generate a new state x1, and calculate the objective function value f(x1);

[0091] Step 3: Determine whether to accept according to the state acceptance function: if f(x1) < f(x0), accept the new state x1 as the current state, otherwise, determine whether to accept x1 according to the Metropolis criterion, if accepted, set the current state equal to x1, if not accepted, set the current state equal to x0;

[0092] Step 4: According to the temperature cooling scheme T q+1 = CT q , C ∈ (0, 1), reduce the control temperature T q+1 ;

[0093] Step 5: Determine whether the annealing process is terminated, when ten new solutions are not accepted in a row, or when the iteration number q is reached, terminate the algorithm and go to step 6, otherwise go to step 2;

[0094] Step 6: The current solution is output as the optimal solution.

[0095] In the embodiment, the multi-condition model establishment based on step 6 described in step 7 adopts a multi-condition model generalized predictive control algorithm to replace the PID control, so as to realize the bed temperature prediction and control optimization of the 660 MW ultra-supercritical circulating fluidized bed boiler. The specific scheme is to approximate the characteristics of the whole running interval of the controlled object by using the multiple typical condition point models established in step 6, and to design a corresponding controller for each sub-model. In actual operation, the outputs of these limited sub-controllers are mapped to the final control bed temperature effect through switching or weighting.

[0096] The bed temperature prediction and control optimization method of the 660 MW ultra-supercritical circulating fluidized bed boiler has the following technical effects:

[0097] 1) Based on the operation mechanism of the 660 MW ultra-supercritical circulating fluidized bed boiler, combined with the characteristics selection, operation condition division based on projection on convex set, bed temperature and main steam pressure coupling prediction, multi-condition model generalized predictive control and other intelligent methods, the bed temperature prediction and control optimization of the 660 MW ultra-supercritical circulating fluidized bed boiler is realized.

[0098] 2) The bed temperature and main steam pressure coupling prediction dynamic characteristic model is established by fully considering the serious coupling of the bed temperature and the main steam pressure in the closed-loop system model of the 660 MW ultra-supercritical circulating fluidized bed boiler under the action of the fuel quantity and the primary air quantity during the operation of the unit (the single-input single-output bed temperature model and the multi-input single-output bed temperature model established by many experts and scholars at present are difficult to meet the current control requirements), which lays a foundation for the bed temperature automatic control optimization of the 660 MW ultra-supercritical circulating fluidized bed boiler.

[0099] 3) In view of the strong nonlinearity and large delay of the circulating fluidized bed boiler bed temperature, the multi-model control strategy and the generalized predictive control are combined to overcome the shortcomings of the PID post-control, so as to improve the bed temperature control quality and enhance the anti-disturbance ability and self-adaptive ability.

[0100] The above only describes the preferred embodiments of the present application and is not used to limit the present application. It should be noted that for ordinary skilled persons in the technical field, several improvements and modifications can be made without departing from the technical principles of the present application, and these improvements and modifications should be regarded as the protection scope of the present application.

Claims

1. A method for bed temperature prediction and control optimization of a 660 MW ultra-supercritical circulating fluidized bed boiler, characterized in that, The method comprises the following steps: Step 1, according to the actual process structure arrangement and operation process of the 660 MW ultra-supercritical circulating fluidized bed boiler, combined with the power plant measuring point table, the related characteristic variable parameters affecting the circulating fluidized bed unit bed temperature and main steam pressure are selected; The related characteristic variable parameters include coal supply amount, water supply amount, induced draft amount, primary air amount, secondary air amount, limestone amount, slag discharge amount and return material amount; Step 2, the related characteristic variable parameter data selected in step 1 are collected from the unit database for comprehensive pretreatment; the comprehensive pretreatment includes data cleaning processing and data filtering processing; Step 3, based on the strong coupling relationship between the main steam pressure and the bed temperature, the relationship between the pretreated related characteristic variable parameter data and the output characteristic bed temperature and main steam pressure is analyzed from the mechanism angle, and the coal supply amount, water supply amount and primary air amount are selected as input characteristics; Step 4, based on the time delay characteristic of the bed temperature response of the 660 MW ultra-supercritical circulating fluidized bed boiler, the bed temperature values and main steam pressure values at previous time points are selected as input characteristics; Step 5, the circulating fluidized bed unit operation condition is divided according to the unit load size; Step 6, according to the condition division of step 5, the circulating fluidized bed unit bed temperature and main steam pressure coupling prediction model under different conditions is established, and the parameters are optimized by using the simulated annealing algorithm to determine the dynamic response time step of the output characteristic bed temperature and main steam pressure under different conditions; the prediction model is as follows: In the formula, T(t), P(t) represent the outputs of the bed temperature and the main steam pressure at the current t time point respectively; x1(t), x2(t), x3(t) represent the inputs of the coal supply amount, water supply amount and primary air amount at the current t time point respectively; m, n, a, b, c represent the dynamic response time steps of the bed temperature, main steam pressure, coal supply amount, water supply amount and primary air amount respectively, the dynamic response time step is related to the data sampling frequency, and the value range is 0-20, and the positive integer is taken; Step 7, based on the multi-condition prediction model established in step 6, the multi-condition model generalized predictive control algorithm is adopted to replace the PID control, and the bed temperature prediction and control optimization of the 660 MW ultra-supercritical circulating fluidized bed boiler are carried out.

2. The 660 MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method according to claim 1, characterized in that, The data cleaning processing and data filtering processing in step 2 include: Outlier data is removed by using the box plot method: Suppose that q1 and q3 are the first quartile and the third quartile of data, and the box plot method can be expressed as: x max = q3 + 1.5 x (q3 - q1) x min = q3 - 1.5 x (q3 - q1) Wherein, x max represents an abnormal maximum value in the data, x min is an abnormal minimum value in the data; if there is data less than the abnormal minimum value in the data, data greater than the abnormal maximum value is determined as outlier data and is removed; Data filtering processing is carried out by using the Kalman filtering method, and the Kalman filtering model is shown in the following formula: X(k) = A * X(k-1) + B * U(k) + W(k) Z(k) = H * X(k) + V(k) In the formula, X(k) is the system state at k time point, U(k) is the control amount of the system at k time point; A and B are system parameters, and for a multi-model system, A and B are matrices; Z(k) is the measurement value at k time point, H is the parameter of the measurement system, and for a multi-measurement system, H is a matrix; W(k) and V(k) represent the process noise and measurement noise respectively; the system state is optimally estimated through system input and output observation data.

3. The 660 MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method according to claim 1, characterized in that, The clustering algorithm based on projection on convex sets is used for condition division in step 5.

4. The 660 MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method according to claim 3, characterized in that, The clustering algorithm process is as follows: Algorithm input: data set X, cluster number k; Algorithm output: cluster representative set Y; 1) randomly select k sample objects from the whole data set X as initial cluster centers; 2) compute the Euclidean distance of each sample object x in the dataset to the cluster center c m i to the cluster center c​ 3) find the minimum distance of each sample object x m to cluster center c i and put this sample object x m into the same cluster as c i ; 4) calculate the weighted average of objects in the same cluster and update the cluster center; wherein: the weighted coefficient of each object is calculated as follows: 5) repeat steps 2) to 4) until the cluster center no longer changes.

5. The 660 MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method according to claim 1, characterized in that, The step of parameter optimization in step 6 by using simulated annealing algorithm is as follows: (1) Select the iteration number q and the initial control temperature T q Let q = 0; select an initial state x0 from the feasible solution space, and calculate the objective function value f(x0); (2) generate a random disturbance in the feasible solution space to generate a new state x1, calculate its objective function value f(x1); (3) judge whether to accept according to the state acceptance function: if f(x1) < f(x0), accept the new state x1 as the current state, otherwise, judge whether to accept x1 according to the Metropolis criterion, if accepted, let the current state equal to x1, if not accepted, let the current state equal to x0; (4) according to the temperature cooling scheme T q+1 = CT q , C e (0, 1), to reduce the control temperature T q+1 ; (5) judge whether to terminate the annealing process, when ten new solutions are not accepted in succession, or when the iteration number q is reached, terminate the algorithm, turn to step (6), otherwise, turn to step (2); (6) output the current solution as the optimal solution.

6. The 660 MW ultra-supercritical circulating fluidized bed boiler bed temperature prediction and control optimization method according to claim 1, characterized in that, The step 6 comprises: A plurality of typical working condition point models are established to approximate the characteristics of the whole operating range of the controlled object, and a corresponding controller is designed for each sub-model; In actual operation, the outputs of the limited sub-controllers are mapped to the final control bed temperature effect through switching or weighting.

Images