A method, terminal and storage medium for optimizing valve flow characteristics of thermal power unit

Abstract

The present invention provides a method, terminal, and storage medium for optimizing valve flow characteristics of a thermal power unit. The method comprises the following steps: establishing an original valve flow characteristic data set; establishing an original valve flow characteristic optimization data set based on the original valve flow characteristic data set and reasonably assuming the corresponding relationship between the total valve position command and the pressure ratio using the equivalent steam flow method; fitting and constructing an actual valve flow characteristic curve based on the original valve flow characteristic optimization data set using an improved K-means clustering analysis algorithm; and optimizing the actual valve flow characteristic curve using an improved particle swarm algorithm to obtain an optimized valve flow characteristic curve. This method avoids the problem of falling into local optimality during complex optimization and has high parameter identification accuracy and robustness.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a method, a terminal and a storage medium for optimizing valve flow characteristics of a thermal power unit. Background Art

[0002] During operation, the valve opening of a thermal power unit is constantly adjusted, causing the valve to fatigue. This can lead to discrepancies between the actual and ideal flow characteristics. If the unit control system controls the valve opening according to the set ideal flow characteristic, the desired effect will not be achieved and may even result in misadjustment or reverse adjustment, affecting the unit's primary frequency regulation performance and, in turn, reducing the economic and safety of operation. Currently, optimizing valve flow characteristics is primarily achieved through field testing, which has many drawbacks, such as time-consuming and poor accuracy.

[0003] In summary, the prior art has the following problem: how to optimize the flow characteristics of the valve. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem of how to optimize the flow characteristics of a valve.

[0005] To this end, on one hand, an embodiment of the present invention provides a method for optimizing valve flow characteristics of a thermal power plant, the method comprising the following steps:

[0006] Establish original valve flow characteristic data set;

[0007] According to the original valve flow characteristic data set, based on the equivalent steam flow method, a reasonable assumption is made about the correspondence between the total valve position instruction and the pressure ratio, and an original valve flow characteristic optimization data set is established;

[0008] According to the original valve flow characteristic optimization data set, an actual valve flow characteristic curve is constructed by fitting based on an improved K-means clustering analysis algorithm;

[0009] The actual valve flow characteristic curve is optimized based on an improved particle swarm algorithm to obtain an optimized valve flow characteristic curve.

[0010] On the other hand, an embodiment of the present invention provides a terminal for optimizing valve flow characteristics of a thermal power unit, which is used to implement the above-mentioned method for optimizing valve flow characteristics of a thermal power unit.

[0011] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the aforementioned method for optimizing valve flow characteristics of a thermal power unit.

[0012] The beneficial effects are as follows: the present invention effectively realizes data mining based on the "data addition and screening method" and the improved K-means clustering analysis algorithm to replace tedious field experiments to obtain the actual valve flow characteristic curve, effectively combines the advantages of the particle swarm algorithm (PSO), the cuckoo algorithm (CS), the seagull algorithm (SOA) and the multi-objective particle swarm algorithm (MO-PSO), increases the activity of particles, allows particles to jump out of the local area to search in a larger range, and then searches for the global optimal solution of the valve flow characteristic, avoiding the problem of easily falling into the local optimum when performing complex optimization, and has high parameter identification accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of a method for optimizing valve flow characteristics of a thermal power unit provided by an embodiment of the present invention;

[0014] Figure 2 A flow chart of a first implementation method of a method for optimizing valve flow characteristics of a thermal power unit provided by an embodiment of the present invention;

[0015] Figure 3 is the original valve flow characteristic curve of the embodiment;

[0016] Figure 4 This is the flow chart of the improved K-means algorithm;

[0017] Figure 5 : is a distribution diagram of each cluster value of the embodiment;

[0018] Figure 6 is the actual valve flow characteristic curve of the embodiment;

[0019] Figure 7 This is the flow chart of the improved particle swarm algorithm;

[0020] Figure 8 CV1 / 2 valve opening-total valve position command characteristic curve before and after optimization of the embodiment;

[0021] Figure 9 CV3 valve opening-total valve position command characteristic curve before and after optimization of the embodiment;

[0022] Figure 10 The total flow-total valve position command characteristic curves before and after optimization of the embodiment. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] In the embodiment of the present invention, Figure 1 , provides a method for optimizing valve flow characteristics of a thermal power unit, the method comprising the following steps:

[0025] S101: Establishing the original valve flow characteristic data set; including:

[0026] The valve flow characteristic data of M signal points of historical thermal power units are collected, and the original valve flow characteristic data set is established based on the data addition and screening method.

[0027] Based on the original valve flow characteristic data set, the data set is screened, in which the sum of the absolute values of the main steam pressure changes in the adjacent 60 seconds is less than 0.1 MPa, and the sum of the absolute values of the total valve position changes in the adjacent 60 seconds is less than 1%. As shown in formula (1), the adjacent 60 rows of data for 60 seconds can be used to obtain a valve flow characteristic data after preliminary screening, thereby obtaining the original valve flow characteristic data set.

[0028]

[0029] S102: Based on the original valve flow characteristic data set, a reasonable assumption is made based on the equivalent steam flow method regarding the corresponding relationship between the total valve position instruction and the pressure ratio, and an original valve flow characteristic optimization data set is established; specifically, based on the equivalent steam flow method, a reasonable assumption is made that there is a unique corresponding relationship between the total valve position instruction and the pressure ratio, that is, the pressure ratio ε is a function of the total valve position instruction μ, and the regulating stage pressure p1 / main steam pressure p T , characterize the valve flow characteristics of the unit with pressure ratio, and establish the original valve flow characteristic optimization data set;

[0030] S103: Optimizing the data set according to the original valve flow characteristic, and constructing an actual valve flow characteristic curve by fitting based on an improved K-means clustering analysis algorithm;

[0031] Specifically include:

[0032] Determine the k value, and determine the k value to be 35;

[0033] Select the centroid;

[0034] Calculate and compare the Euclidean distances from each sample point to all centroids to form k clusters;

[0035] Calculate the sum of squared errors and find the minimum sum of squared errors to make the total distance from the center point to each point in the cluster the shortest;

[0036] Calculate the distance between all sample points and the center point, and classify them again according to the distance;

[0037] Calculate and find the new centroid position, calculate the closest distance, reclassify, and determine whether the centroid position has changed. If the centroid position has not changed, the iteration stops; if the centroid position has changed, return to the method for obtaining the minimum sum of squared errors to make the total distance from the center point to each point in the cluster the shortest;

[0038] According to the cluster values obtained in the fitting process, the fitting degree of the actual valve flow characteristic curve is judged.

[0039] S3.1: Use the Elbow Method to determine the value of k. Based on the principle of the Elbow Method, select points on the image generated by the WCSS to determine the number of clusters k. The value of k is determined as the value of the largest inflection point.

[0040] S3.2: Randomly select k points on the plane as initial center points (centroids). The centroids do not have to be points in the original data set. The k centroids represent the division of the data into k clusters.

[0041] S3.3: Calculate and compare the Euclidean distances from each sample point to all centroids, classify them based on the proximity principle, and form k clusters;

[0042] S3.4: After the classification is completed, the first iteration is performed, and the sum of squared errors (SSE) is calculated using formula (2) to find the minimum sum of squared errors so that the total distance from the center point to each point in the cluster is the shortest;

[0043]

[0044] Where p is the sample point, Ci is the cluster, mi is the centroid of the cluster, and |p-mi| represents the distance from p to mi;

[0045] S3.5: After the first iteration, perform another iteration within the group, taking the minimum within-group sum of squares (WCSS) as the principle, calculate the distance between all sample points and the center point using formula (3), and classify them again according to the distance;

[0046]

[0047] S3.6: Perform a second iteration, calculate and find the new centroid position, calculate the closest distance, reclassify, and determine whether the centroid position has changed. If so, return to step S3.4; if not (no change), the iteration stops, the clustering result of k clusters is obtained, and the next step S3.7 is continued;

[0048] S3.7: During the algorithm iterative fitting process, the values of each cluster are obtained. If the values of each cluster are close to 1, the curve fit is good, and the actual valve flow characteristic curve is obtained by fitting. Then proceed to S4. If the values of each cluster are close to 0, the curve fit is poor, and return to S3.1.

[0049] S104: Optimizing the actual valve flow characteristic curve based on an improved particle swarm algorithm to obtain an optimized valve flow characteristic curve.

[0050] Specifically:

[0051] S4.1: Set the population size (number of particles) to N and the maximum number of iterations to T max , population dimension D, position range limits (learning factors) are c1, c2, and speed limit (inertia weight) is ω max 、ω m i n , initial population and individual historical optimal fit, step size control factor α, discovery probability p α , search for the upper bound U b and search the lower bound L b , where N, T max , D is a positive integer greater than 0; N D-dimensional arrays are randomly generated within the upper and lower limits of the position range, and the arrays are particle populations X t , is the particle population after t iterations, t∈[0,T max ] and is an integer with an initial value of 0, which means that the particle population is randomly generated for the first time without iteration.

[0052] S4.2: At the initial time (t=0), randomly initialize the particle position x i (t＝0) =(x i1 (t＝0) ,x i2 (t＝0) ,...,x iD (t＝0) ), i=1,2,...,N, calculate the objective function value of each particle and determine the optimal position x of the initial particle pbest (t＝t) And the optimal value of the objective function x gbest (t＝t) ; Further update the particle velocity and position according to the following formula to calculate the objective function value;

[0053] v i (t＝t+1) =ωv i (t＝t) +c1r1[p i (t＝t) -x i(t＝t) ]+c2r2[g (t＝t) -x i (t＝t) ](4)

[0054] x i (t＝t+1) =x i (t＝t) +v i (t＝t+1) (5)

[0055]

[0056] Where: i∈[1,N], t∈[1,T max ];v i represents the velocity of the particle, x i represents the position of the particle; ω represents the inertia weight factor; c1 and c2 are acceleration constants, also known as learning factors; r1 and r2 are random numbers uniformly distributed between [0, 1]; p i (t ＝t) represents the best position that the i-th particle passes after t iterations, also known as the individual optimal position; g (t＝t) represents the best position found by all particles after t iterations, also known as the global optimum; v max Indicates the boundary value set by the velocity vector;

[0057] Compare the calculated optimal value of the objective function of the tth generation with that of the tth generation. If the optimal value of the objective function of the t+1th generation is better than that of the tth generation, then the optimal position of the current particle is recorded as x pbest (t＝t+1) , otherwise it is still x pbest (t＝t) ; Current particle optimal position x pbes ( t t＝t / t+1) With the discovery probability p α For comparison, if x pbest (t＝t / t+1) <p α , it means that the particle has not been eliminated. If x pbest (t＝t / t+1) >p α , it means that the particle is eliminated, the particle position is updated, the new objective function value is calculated, and the optimal position of the t+1 / t+2 generation particle is determined to be x pbest (t＝t+1 / t+2) , update the global optimal position to x gbest (t＝t+1 / t+2) Specifically, the particle position is updated using the position update method in the CS algorithm. The formula is:

[0058]

[0059] in,

[0060]

[0061] Where: x i (t＝t) is the position of the i-th bird's nest at the t-th iteration; a is the step size scaling factor; L(s,λ) follows the Lévy distribution, representing the random path of Lévy flight; s is the random flight step size; λ is the power coefficient, 1<λ<3; Γ(λ) is the standard Gamma function with unbounded mean and variance; a is 0.01 and λ is 1.5.

[0062] S4.3: Particle population after resetting the position of the tth generation Substitute it into the valve flow characteristic function, and combine it with the optimization data set established in S2) to calculate the error between the function output value and the historical data. Use the error to obtain the reset population particle fitness and the population average fitness. Then use the optimization algorithm (PSO-CS-MO) that combines the nonlinear inertia weight in PSO-CS and MO-PSO to update the classification position. Substitute the reset particle population into the valve flow characteristic function, and use the total valve position command signal data in the identification data set as input. The corresponding output is obtained through transfer function calculation. fzzr , calculate the particle error value after the tth generation reset position by the following formula:

[0063]

[0064] Where, O zr To identify the valve opening signal value in the data set, is the calculated value of valve opening after the particle at the ith reset position of the tth generation is substituted into the transfer function, for The corresponding error value, i is any integer value between 1 and N;

[0065] Through the following formula Calculate the average and get the average error value of the valve opening after the particles after the tth generation N reset positions are substituted into the transfer function Will and the average error In comparison, the error value is lower than The first group is higher than The second group is:

[0066]

[0067] The particle population of the second group is returned to S4.2, and the particle population of the first group is updated in position by combining the nonlinear inertia weight in the CS algorithm. The formula is as follows:

[0068]

[0069] V i (t＝(t+1)*) =ωV i (t＝t*) +c1r1(p i (t＝t*) -X i (t＝t*) )+c2r2(g (t＝t*) -X i (t＝t*) )(12)

[0070] Where, X i (t＝t+1) is the i-th particle of the t+1th generation; is the i-th particle after the position is reset in the t-th generation; is the global optimal particle; x, y, z are the parameters related to the spiral attack behavior of the CS algorithm; are the moving speeds of the i-th particle after the position is reset in the t+1th and tth generations respectively; is the optimal individual particle after the tth generation reset position; r1 and r2 are random numbers between 0 and 1; c1 and c2 are constants between 0 and 1, and ω is the nonlinear inertia weight factor. The calculation formula is:

[0071]

[0072] Where, ω (t＝t) represents the nonlinear inertia weight at iteration t, ω max 、ω min Represent the maximum and minimum inertia weights respectively.

[0073] S4.4: Combining population crowding and crossover mutation in MO-PSO, the objective function value of each particle in the population is obtained through power flow calculation. The dominance relationship between the generated new particle population and the corresponding individual in the particle's individual optimal position is determined, and the individual optimal position is updated using the crossover rule. The individual crowding density of each grid is calculated using the adaptive grid method, and a non-dominated sorting is performed to determine the global optimal position. The population crowding is calculated, and the crossover mutation probability is calculated based on the population crowding. The particle dimension information is selected based on the obtained crossover mutation probability. The particles are crossover-mutated and denormalized, and the speed and position are further updated iteratively.

[0074] S4.5: Determine whether the maximum number of iterations has been reached: If not, return to S4.2; if yes, output the optimal value of the population, that is, the optimized total flow-total valve position and each valve opening-total valve position fitting curves.

[0075] The sum of squares of the calculation errors is calculated using the following formula:

[0076]

[0077] Among them, p is a sample point, Ci is a cluster, mi is the centroid of the cluster, and |p-mi| represents the distance from p to mi.

[0078] The following calculation formula is used to calculate the distance between all sample points and the center point:

[0079]

[0080] The actual valve flow characteristic curve is optimized based on the improved particle swarm algorithm to obtain an optimized valve flow characteristic curve, which specifically includes:

[0081] Based on the population congestion and crossover mutation in MO-PSO, the objective function value of each particle in the population is obtained through power flow calculation;

[0082] Determine the dominance relationship between the generated new particle population and the corresponding individual in the individual optimal position of the particle, and update the individual optimal position using the crossover rule;

[0083] The individual crowding density of each grid is calculated by the adaptive grid method, the global optimal position is determined by non-dominated sorting, and the population crowding degree is calculated;

[0084] Calculating the crossover mutation probability according to the population crowding degree;

[0085] The particle dimension information is selected according to the obtained cross-mutation probability, the particles are subjected to cross-mutation operations and denormalization, and the speed and position are iteratively updated.

[0086] According to the cluster values obtained in the fitting process, the fitting degree of the actual valve flow characteristic curve is judged, including:

[0087] During the iterative fitting process of the algorithm, the values of each cluster are obtained. If the values of each cluster are close to 1, the curve fitting is good, and the actual valve flow characteristic curve is obtained by fitting, and the actual valve flow characteristic curve is optimized. If the values of each cluster are close to 0, it means that the curve fitting is poor, and the k value is determined again.

[0088] In an embodiment of the present invention, an information processing terminal is further provided, and the information processing terminal is used to implement the aforementioned method for optimizing flow characteristics of a valve of a thermal power unit.

[0089] In an embodiment of the present invention, a computer-readable storage medium is further provided, comprising instructions, which, when executed on a computer, enable the computer to execute the aforementioned method for optimizing valve flow characteristics of a thermal power unit.

[0090] The present invention effectively realizes data mining based on the "data addition and screening method" and the improved K-means clustering analysis algorithm to replace tedious field experiments to obtain the actual valve flow characteristic curve, effectively combines the advantages of the particle swarm algorithm (PSO), the cuckoo algorithm (CS), the seagull algorithm (SOA) and the multi-objective particle swarm algorithm (MO-PSO), increases the activity of particles, allows particles to jump out of the local area to search in a larger range, and then search to obtain the global optimal solution of the valve flow characteristic, avoiding the problem of easily falling into the local optimum when performing complex optimization, and has high parameter identification accuracy and robustness.

[0091] Example 1:

[0092] The technical solution of the present application is as follows: historical signals of valve flow characteristics of thermal power units are collected, and an original valve flow characteristic data set is established based on the "Data summation filtering method"; an optimized original valve flow characteristic data set is established based on the equivalent steam flow method; the actual valve flow characteristic curve is fitted based on an improved K-means clustering analysis algorithm; and an optimized valve flow characteristic curve is obtained by optimization based on an improved particle swarm fusion search algorithm (PSO-CS-SOA-MO).

[0093] The present invention provides a method for optimizing the flow characteristics of a valve of a thermal power unit, the method comprising the following steps:

[0094] S1. Collect valve flow characteristic data of M signal points of historical thermal power units and establish an original valve flow characteristic data set based on the "data summation filtering method";

[0095] S2, based on the data set established in S1, it is reasonable to assume that there is a unique corresponding relationship between the total valve position instruction and the pressure ratio based on the equivalent steam flow method, that is, the pressure ratio ε (regulating stage pressure p1 / main steam pressure p T ) is a function of the total valve position command μ, the valve flow characteristics of the unit are characterized by the pressure ratio, and the original valve flow characteristic optimization data set is established;

[0096] S3. Based on the optimized data set established in S2, the actual valve flow characteristic curve is further fitted and constructed based on the improved K-means cluster analysis algorithm;

[0097] S4. Optimize the actual valve flow characteristic curve based on the improved particle swarm optimization algorithm (PSO-CS-MO) to obtain the optimized valve flow characteristic curve;

[0098] Example 2:

[0099] The present invention provides a method for optimizing the flow characteristics of a thermal power unit valve, such as Figure 2 As shown, the specific steps of the embodiment of the present invention are as follows:

[0100] Step 1: Collect 50,000 valve flow characteristic data signals with 1s intervals from a 300MW unit, including the unit's total valve position instruction, valve opening before optimization, main steam flow, main steam pressure, and regulating stage pressure signals. Based on the "data summation filtering method", the main steam pressure of each moment data is subtracted from the main steam pressure of the previous moment data according to formula (1), and the difference is taken as the "main steam pressure change". Then, the main steam pressure changes of the data 30s before and after each moment data are summed up, and the sum is taken as the "60s main steam pressure change sum". Data with a 60s main steam pressure change sum less than 0.1 are filtered. Then, the same main steam pressure processing method is used to filter data with a "60s DEH total valve position instruction change sum" less than 1%. After filtering, more than 3,000 valve flow characteristic data signals with 1s intervals are obtained, and the original valve flow characteristic data set is established.

[0101] Step 2: Based on the data set established in step 1, it is reasonable to assume that there is a unique corresponding relationship between the total valve position instruction and the pressure ratio based on the equivalent steam flow method, that is, the pressure ratio ε (regulating stage pressure p1 / main steam pressure p T ) is a function of the total valve position instruction μ, and the pressure ratio ε corresponding to the full range of the total valve position instruction (μ = 100%) in the unit operation data is approximately 0.88, that is, the pressure ratio reference value is 0.88. The valve flow characteristics of the unit are characterized by the pressure ratio, and the original valve flow characteristic optimization data set is established. The relationship curve between the equivalent steam flow and the total valve position instruction is plotted, and the original valve flow characteristic curve is obtained as shown in the attached figure. Figure 3 shown.

[0102] Step 3: Based on the optimized data set established in step 2, i.e. the original valve flow characteristic curve, the improved K-means algorithm process is as shown in the attached figure. Figure 4 As shown:

[0103] (1) The Elbow Method is used to determine the value of k. Based on the principle of the Elbow Method, the number of clusters k is determined by selecting points from the image generated by the WCSS. The maximum value of the inflection point is set as the k value, and k = 35 is obtained.

[0104] (2) Randomly select 35 points on the plane as the initial center points (centroids). The centroids do not have to be the points in the original data set. The 35 centroids represent the division of the data into 35 clusters.

[0105] (3) Calculate and compare the Euclidean distances between each sample point and all centroids, cluster them based on the proximity principle, and form 35 clusters;

[0106] (4) After clustering is completed, the first iteration is performed, and the sum of squared errors (SSE) is calculated using formula (8) to make the total distance from the center point to each point in the cluster the shortest;

[0107] (5) After the first iteration, perform another iteration within the group. Based on the principle of minimizing the sum of squares (WCSS) within the group, calculate the distance between all sample points and the center point using formula (9), and classify them again according to the distance.

[0108] (6) Perform the second iteration, calculate and find the new centroid position, calculate the closest distance, and reclassify;

[0109] (7) If the center of mass position does not change, the iteration stops;

[0110] (8) The clustering results of 35 clusters were obtained. The values of each cluster were obtained during the iterative fitting process of the algorithm as shown in the attached figure. Figure 5 As shown, the values of each cluster are greater than 0.8;

[0111] (9) The curve fitting is good, and the actual valve flow characteristic curve obtained by fitting is as shown in the attached figure. Figure 6 shown.

[0112] Step 4: The improved particle swarm algorithm process is as follows Figure 7 As shown:

[0113] (1) Based on the actual valve flow characteristic curve obtained by fitting in step 3, the total flow-total valve position characteristic curve and the valve opening-total valve position characteristic curve before optimization are constructed;

[0114] (2) Set the population size N to 500 and the maximum number of iterations T max is 500, the population dimension D is 11, ω max is 0.8, ω min is 0.2, the position range is limited to [0,100], the speed limit is set to [-1,1], c1 and c2 are both 0.5, and the probability of finding p is α＝ 0.3, set the initial population fitness and individual historical optimal fitness to infinity, randomly generate particle population X according to the range limit, and start iteration;

[0115] (3) When the 500 particle swarms are continuously iterating, the velocity and position of each particle are updated by the PSO algorithm, i.e., equations (10) and (11), to obtain the optimal position of a group of particles; then, the optimal position x of the current particle is updated. pbest (t＝t / t+1) With the discovery probability p α Make comparisons;

[0116] (4) If x pbest (t＝t / t+1) <p α , indicating that the particle has not been eliminated, if x pbest (t＝t / t+1) >p α , it means that the particle is eliminated, and the optimal position of the current particle is substituted into equations (13) and (14) for updating; the eliminated particle group will increase the update and calculation of the CS algorithm once, and the difference in iteration time can be ignored.

[0117] (5) The actual valve flow characteristic curve obtained by fitting in step 3 is taken as input with 1% total valve position instruction as the interval, and 100 total valve position instructions are taken as input. After the transfer function, 100 corresponding valve opening calculation values ​​O are output. fzzr , the measured value of valve opening O zr With O fzzr Calculate the error e' using formula (15) and record it; average the valve opening error values after substituting the 100 reset particles into the transfer function using the following formula to obtain the average valve opening error value e' avg :

[0118]

[0119] (6) Compare the opening error value e' of each valve with the average error value e' avg For comparison, e' is lower than e' avg The first group is higher than e' avg The second group of particle populations returns to step 3 for the next iteration;

[0120] (7) The first group of particle populations continues to combine the nonlinear inertia weight in the MO-PSO algorithm, that is, formula (19), to update the position after classification. The position update formula is as shown in formulas (17) and (18). In formula (19), T max is the maximum number of iterations, which is 500; ω max and ω min The values are 0.8 and 0.2 respectively.

[0121] (8) It is determined that the maximum number of iterations has been reached, the iteration is terminated, and the optimization result is output, that is, the optimized total flow-total valve position characteristic curve and each valve opening-total valve position characteristic curve are obtained;

[0122] The optimized valve opening-total valve position characteristic curve is shown in the attached figure. Figure 8 and attached Figure 9 The optimized total flow-total valve position characteristic curve is shown in the attached Figure 10 As shown. Figure 8 and attached Figure 9 It can be seen that the relationship curves between the valve opening and the total valve position command before optimization have a large deviation from the ideal curve, while the relationship curves between the valve opening and the total valve position command after optimization are very close to the ideal curve. Figure 10 It can be seen that the total flow rate and the total valve position command before optimization show a nonlinear relationship, while the total flow rate and the total valve position command after optimization are essentially proportional, very close to the theoretical total flow rate-total valve position curve. This shows that the optimization accuracy of the algorithm of the present invention is very high, proving the rationality and feasibility of the algorithm for optimizing valve flow characteristics.

[0123] Compared with the prior art, the present invention has the following beneficial effects:

[0124] 1. The present invention performs data preprocessing through a "data summation filtering method" to eliminate bad points in massive operating data and preliminarily screen out valve flow characteristic data that can better represent different operating conditions.

[0125] 2. The present invention clusters massive operating data through an improved K-means clustering analysis algorithm, and further fits data points that can better represent different operating conditions to draw the actual valve flow characteristic curve. Ultimately, the data mining method based on the improved K-means clustering analysis algorithm replaces the cumbersome field test method to obtain the actual valve flow characteristic curve.

[0126] 3. The present invention increases the vitality of particles by utilizing the excellent optimization ability, global search ability and local jump-out ability of the PSOCS-MO fusion search method, allowing particles to jump out of the local area to search in a larger range, and then search for the global optimal solution, further optimizing the optimization model.

[0127] The present invention effectively realizes data mining based on the "data addition and screening method" and the improved K-means clustering analysis algorithm to replace tedious field experiments to obtain the actual valve flow characteristic curve, effectively combines the advantages of the particle swarm algorithm (PSO), the cuckoo algorithm (CS), the seagull algorithm (SOA) and the multi-objective particle swarm algorithm (MO-PSO), increases the activity of particles, allows particles to jump out of the local area to search in a larger range, and then search to obtain the global optimal solution of the valve flow characteristic, avoiding the problem of easily falling into the local optimum when performing complex optimization, and has high parameter identification accuracy and robustness.

[0128] The above description is merely an illustrative embodiment of the present invention and is not intended to limit the scope of the present invention. The various components of the present invention may be combined with each other without conflict, and any equivalent changes and modifications made by a person skilled in the art without departing from the concept and principles of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A method for optimizing valve flow characteristics of a thermal power unit, characterized in that: The method comprises the following steps: Establish original valve flow characteristic data set; According to the original valve flow characteristic data set, based on the equivalent steam flow method, a reasonable assumption is made about the correspondence between the total valve position instruction and the pressure ratio, and an original valve flow characteristic optimization data set is established; According to the original valve flow characteristic optimization data set, an actual valve flow characteristic curve is constructed by fitting based on an improved K-means clustering analysis algorithm; Optimizing the actual valve flow characteristic curve based on an improved particle swarm algorithm to obtain an optimized valve flow characteristic curve; The optimizing data set according to the original valve flow characteristic and constructing the actual valve flow characteristic curve based on the improved K-means cluster analysis algorithm specifically includes: Determine the k value; Select the centroid; Calculate and compare the Euclidean distances from each sample point to all centroids to form k clusters; Calculate the sum of squared errors and find the minimum sum of squared errors to make the total distance from the center point to each point in the cluster the shortest; Calculate the distance between all sample points and the center point, and classify them again according to the distance; Calculate and find the new centroid position, calculate the closest distance, reclassify, and determine whether the centroid position has changed. If the centroid position has not changed, the iteration stops; if the centroid position has changed, return to the method for obtaining the minimum sum of squared errors to make the total distance from the center point to each point in the cluster the shortest; According to the cluster values obtained in the fitting process, the fitting degree of the actual valve flow characteristic curve is judged; The method of optimizing the actual valve flow characteristic curve based on the improved particle swarm algorithm to obtain the optimized valve flow characteristic curve specifically includes: Based on the population congestion and crossover mutation in MO-PSO, the objective function value of each particle in the population is obtained through power flow calculation; Determine the dominance relationship between the generated new particle population and the corresponding individual in the individual optimal position of the particle, and update the individual optimal position using the crossover rule; The individual crowding density of each grid is calculated by the adaptive grid method, the global optimal position is determined by non-dominated sorting, and the population crowding degree is calculated; Calculating the crossover mutation probability according to the population crowding degree; The particle dimension information is selected according to the obtained cross-mutation probability, the particles are subjected to cross-mutation operations and denormalization, and the speed and position are iteratively updated.

2. A method for optimizing valve flow characteristics of a thermal power unit according to claim 1, characterized in that: The sum of squares of the calculation errors is calculated using the following formula: Among them, p is a sample point, Ci is a cluster, mi is the centroid of the cluster, and |p-mi| represents the distance from p to mi.

3. A method for optimizing valve flow characteristics of a thermal power unit according to claim 2, characterized in that: The following calculation formula is used to calculate the distance between all sample points and the center point:

4. A method for optimizing valve flow characteristics of a thermal power unit according to claim 1, characterized in that: The establishing of the original valve flow characteristic data set includes: The valve flow characteristic data of M signal points of historical thermal power units are collected, and the original valve flow characteristic data set is established based on the data addition and screening method.

5. A method for optimizing valve flow characteristics of a thermal power unit according to claim 2, characterized in that: The method of determining the fitting degree of the actual valve flow characteristic curve according to each cluster value obtained in the fitting process includes: During the iterative fitting process of the algorithm, the values of each cluster are obtained. If the values of each cluster are close to 1, the curve fitting is good, and the actual valve flow characteristic curve is obtained by fitting, and the actual valve flow characteristic curve is optimized. If the values of each cluster are close to 0, it means that the curve fitting is poor, and the k value is determined again.

6. A method for optimizing valve flow characteristics of a thermal power unit according to claim 2, characterized in that: Determine the k value to be 35.

7. A terminal for optimizing valve flow characteristics of a thermal power unit, characterized in that: The terminal is used to implement the method for optimizing valve flow characteristics of a thermal power unit as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that The storage medium stores instructions, and when the instructions are executed on a computer, the computer is caused to execute the method for optimizing flow characteristics of a valve of a thermal power unit according to any one of claims 1 to 6.

Images