A feedforward correction self-learning method and device for a coordinated control system of a thermal power unit

Abstract

The present application relates to a feedforward correction self-learning method and device for a coordinated control system of a thermal power unit, the method comprising obtaining historical operating data of the thermal power unit, and removing abnormal points from the historical operating data; using a sliding window algorithm to traverse the historical operating data after removing abnormal points to obtain the steady-state operating point of the thermal power unit; performing sufficiency and uniformity checks on the steady-state operating point of the thermal power unit; inputting the neural network model for training; importing the trained neural network model into the distributed control system of the thermal power unit; inputting the real-time load signal into the neural network model of the distributed control system to obtain a correction signal for feedforward control; inputting the correction signal into the feedforward control module of the coal feed and water feed flow in the thermal power unit to complete the feedforward control correction. The method can dynamically adjust the coal feed and water feed flow to match the current operating conditions according to the actual load of the unit, improve the coordinated control effect of the unit, and adapt to the changes in operating conditions caused by various disturbances.

Description

Technical Field

[0001] The present application relates to the technical field of coordinated control of thermal power systems, and in particular to a feedforward correction self-learning method and device for a coordinated control system of a thermal power unit. Background Art

[0002] At present, more and more renewable energy power generation is connected to the power grid, but renewable energy power generation is intermittent and unstable, and thermal power generation is usually needed to balance the power grid fluctuations caused by its connection to the grid, which requires thermal power units to be able to quickly adjust load operation over a large range under the premise of ensuring safety. However, the quality of coal entering the furnace of thermal power plants is variable, and is also affected by the operating mode and equipment aging. The feedforward control of coal and water supply in the existing coordinated control system generally adopts a fixed constant control method, which cannot adapt to changes in operating conditions autonomously, resulting in poor unit control effect and cannot meet the peak and frequency regulation requirements of the power grid.

[0003] Therefore, there is an urgent need for an online self-learning method that can correct the feedforward control of coal feed and water feed, regularly and autonomously correct the feedforward control signal, improve the control quality of the coordinated control system, and thus improve the adaptability of the control system. Summary of the invention

[0004] In order to overcome the above-mentioned deficiencies of the prior art, the present application provides a feedforward correction self-learning method and device for a coordinated control system of a thermal power unit, which specifically adopts the following technical solutions:

[0005] A feedforward correction self-learning method for a coordinated control system of a thermal power unit comprises the following steps:

[0006] Acquire historical operation data of the thermal power unit and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data;

[0007] The sliding window algorithm is used to traverse the historical operation data after removing abnormal points to obtain the steady-state operating point of the thermal power unit, wherein the steady-state operating point at least includes load data, coal supply data and water supply flow data;

[0008] Conduct sufficiency and uniformity checks on the steady-state operating points of thermal power units;

[0009] If the steady-state operating points of the thermal power unit are all verified, the verified steady-state operating points are input into the neural network model for training to obtain a trained neural network model;

[0010] Import the trained neural network model into the distributed control system of the thermal power unit;

[0011] Input the real-time load signal into the neural network model of the distributed control system to obtain the correction signal for feedforward control;

[0012] The correction signal is input into the feedforward control module of the coal feed rate and water feed flow rate in the thermal power unit to complete the feedforward control correction.

[0013] Optionally: the step of removing abnormal points from historical operation data includes:

[0014] The historical operation data of the thermal power unit is represented as three-dimensional spatial data of coal supply data, water supply flow data and load data;

[0015] The standard operating points corresponding to the unit at 30% Ne0, 40% Ne0, 50% Ne0, 60% Ne0, 70% Ne0, 80% Ne0, 90% Ne0 and 100% Ne0 are selected as sample points, where Ne0 is the rated load;

[0016] Calculate the core point C within the sample point based on the sample point and historical operation data:

[0017] C={p∈D|Num[dist(S,p)≤ε]≥MinP ts}#;

[0018] Where C represents the core point, p represents the sample point object, S represents the sample point, D represents the data set of historical operation data, dist(S,p) represents the distance between S and p, Num represents the sample count, MinP ts represents the minimum number of included points, and ε represents the scanning radius;

[0019] According to the core point C, the reachable point R and the noise point N are obtained:

[0020]

[0021] Where R represents a reachable point, N represents a noise point, p represents a sample point object, D represents a data set of historical operation data, C represents a core point, dist(p,C) represents the distance between points C and p, and ε represents the scanning radius;

[0022] By deleting the obtained noise points N, the abnormal points of the historical operation data can be eliminated.

[0023] Optional: In the step of obtaining the steady-state operating point of the thermal power unit, load data and main steam pressure are used for steady-state judgment, wherein the signal sampling time of the standard operating point is set to 1s, the time span between adjacent standard operating points is set to 1 hour, and the maximum allowable deviation is set to 5%.

[0024] Optionally, in the step of obtaining the steady-state operating point of the thermal power unit, it is determined whether the load data and main steam pressure corresponding to the standard operating point satisfy the evaluation equation:

[0025]

[0026] Where P st (i) represents the main steam pressure corresponding to operating point i, N e (i) represents the unit load corresponding to operating point i, represents the mean value of the main steam pressure, represents the average load of the unit, n is the number of standard operating points, σ P and σ N are the maximum allowable deviations respectively;

[0027] If the load data and main steam pressure of the thermal power unit both satisfy the evaluation equation, the load data, coal supply data and water supply data corresponding to the standard operating point are stored.

[0028] Optional: When performing a sufficiency check on the steady-state operating points of the thermal power unit, set the threshold for the sufficiency of the steady-state operating points to 3% of the total number of operating points; when the number of steady-state operating points is greater than or equal to 3% of the total number of operating points, it is determined that the sufficiency check of the steady-state operating points of the thermal power unit has passed, and a uniformity check is performed; when the number of steady-state operating points is less than 3% of the total number of operating points, it is determined that the sufficiency check of the steady-state operating points of the thermal power unit has not passed, and the process is terminated.

[0029] Optional: When performing uniformity check on the steady-state operating points of the thermal power unit, the number of steady-state operating points in the load intervals of 20% Ne0-40% Ne0, 40% Ne0-60% Ne0, 60% Ne0-80% Ne0 and 80% Ne0-100% Ne0 are determined respectively; when the number of steady-state operating points in the four load intervals is greater than or equal to 10, it is determined that the uniformity check of the steady-state operating points of the thermal power unit has passed; when the number of steady-state operating points in any load interval is less than 10, it is determined that the uniformity check of the steady-state operating points of the thermal power unit has failed.

[0030] Optionally, the step of inputting the verified steady-state operating point into the neural network model for training to obtain the trained neural network model includes:

[0031] Input the load data, coal supply data and water supply flow data of the steady-state operating point into the neural network model;

[0032] The error back propagation algorithm is used to adjust the weights and thresholds of each hidden layer node in the neural network model so that the output value of the neural network model approaches the actual value; the adjustment formula for the output value, weight and threshold of each hidden layer node is:

[0033]

[0034] where y j Output value for the node; x jis the node input value; f() is the activation function, and the S-type function is selected; w ij is the weight between nodes i and j; b j is the threshold of node j; m is the number of hidden layer nodes; is the error function calculation formula based on Widrow-Hoff; η1 and η2 are the error coefficients respectively.

[0035] Optionally: the neural network model adopts BP neural network.

[0036] Optionally: during the training of the neural network model, the initial number of nodes in the hidden layer is set to twice the input eigenvalue; during the training process, the initial learning rate is set to 0.01 or 0.001; and the root mean square error is selected as the evaluation indicator of the neural network model.

[0037] In addition, the present application also discloses a feedforward correction self-learning device for a coordinated control system of a thermal power unit, which comprises:

[0038] A data acquisition module is used to acquire historical operation data of the thermal power unit and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data;

[0039] The data screening module is used to use a sliding window algorithm to traverse the historical operating data to remove abnormal points and obtain the steady-state operating point of the thermal power unit. The steady-state operating point at least includes load data, coal feed data and water feed flow data.

[0040] Data verification module, used to perform sufficiency verification and uniformity verification on the steady-state operating points of thermal power units;

[0041] A data training module is used to input the verified steady-state operating point into the neural network model for training after the steady-state operating point of the thermal power unit is verified, so as to obtain a trained neural network model;

[0042] Data import module, used to import the trained neural network model into the distributed control system of the thermal power unit;

[0043] A correction signal output module is used to input the real-time load signal into the neural network model of the distributed control system to obtain a correction signal for feedforward control;

[0044] The correction control module is used to input the correction signal into the feedforward control module of the coal feed rate and water feed flow rate in the thermal power unit to complete the feedforward control correction.

[0045] Beneficial Effects

[0046] The technical solution of this application has the following beneficial effects:

[0047] The feedforward correction self-learning method of the thermal power unit coordinated control system of the present application selects steady-state operating points based on historical operating data, and combines the steady-state operating points with the BP neural network to obtain a neural network model that can calculate the coal feed data and water flow data equations, and uses the neural network model to generate a correction signal for the feedforward control in the unit DCS system, thereby realizing feedforward control correction of the coal feed and water flow. This method can dynamically adjust the coal feed and water flow to match the current operating conditions according to the actual load of the unit, improve the coordinated control effect of the unit, and can adapt to changes in operating conditions caused by various disturbances. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the feedforward correction self-learning method of the coordinated control system of the thermal power unit in Example 1 of the present application.

[0049] Figure 2 This is a schematic diagram of the structure of the neural network model in Example 2 of the present application.

[0050] Figure 3 It is a structural schematic diagram of a feedforward correction self-learning method device of a coordinated control system of a thermal power unit in an embodiment of the present application.

[0051] Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and cannot be used to limit the scope of protection of the present application. It should be noted that the following detailed descriptions are all exemplary and are intended to provide further explanation of the present application.

[0053] The Coordinated Control System (CCS) of thermal power units controls the boiler-turbine-generator set as a whole unit, so that the boiler and turbine respond to the control requirements at the same time, ensuring that the unit can quickly and stably meet the load changes and maintain stable operation. The current general control method is to directly and quickly respond to the feedforward signal representing the load or energy command, and adjust the static accuracy and dynamic compensation of the signal through closed-loop feedback control and other advanced strategies. However, the feedforward control signals of coal feed and water feed in the existing coordinated control system generally adopt a fixed constant control method, which makes the thermal power unit unable to adapt to changes in operating conditions autonomously, the unit control effect is poor, and it cannot accurately meet the peak and frequency regulation requirements of the power grid. In order to ensure that the coordinated control system of the thermal power unit has dynamic adaptability and accurately matches the peak and frequency regulation requirements of the power grid, it is necessary

[0054] like Figure 1As shown, the embodiment of the present application specifically discloses a feedforward correction self-learning method for a coordinated control system of a thermal power unit, which comprises the following steps:

[0055] Step 1, obtain the historical operation data of the thermal power unit, and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data. In the embodiment of the present application, the historical operation data of the thermal power unit in the previous three months is counted at the current time node. Since the thermal power unit may be affected by external factors during the actual operation process, some operation data in the thermal power unit may be abnormal: for example, the unit operating load is small at operating point A, and the unit operating load is large at operating point B, but the coal supply data and water supply flow data at operating point B are smaller than those at operating point A. Abnormal data in the historical operation data will affect the calculation of the subsequent feedforward control correction signal and affect the accuracy of the result, so it is necessary to remove abnormal data from the historical operation data.

[0056] In the embodiment of the present application, the abnormal data processing adopts the density-based clustering (DBSCAN) anomaly detection algorithm, which mainly finds low-density abnormal samples in the sample space through clustering. Specifically, the steps of removing abnormal points from the historical operation data in the embodiment of the present application include:

[0057] Firstly, the historical operation data of the thermal power unit is represented as three-dimensional spatial data of coal supply data, water supply flow data and load data;

[0058] Secondly, the standard operating points corresponding to the unit at 30% Ne0, 40% Ne0, 50% Ne0, 60% Ne0, 70% Ne0, 80% Ne0, 90% Ne0 and 100% Ne0 are selected as sample points, where Ne0 is the rated load;

[0059] Then the core point C within the sample point is calculated based on the sample point and historical operation data:

[0060] C={p∈D|Num[dist(S,p)≤ε]≥MinP ts}#;

[0061] Where C represents the core point, p represents the sample point object, S represents the sample point, D represents the data set of historical operation data, dist(S,p) represents the distance between S and p, Num represents the sample count, MinP ts represents the minimum number of included points, and ε represents the scanning radius; it should be noted that, in the embodiment of the present application, the parameters ε and MinP ts The specific value of is related to the disturbance of the unit and the amount of data. The parameters ε and MinP tsIt can be determined according to the actual disturbance degree of the unit and combined with the trial and error method. The initial value of the parameter is generally set to a smaller value. If the clustering result is scattered, the parameter value is increased, otherwise the parameter value is reduced until a satisfactory clustering effect is achieved; if the unit is disturbed to a large extent, the parameter value is appropriately reduced.

[0062] The reachable points R and noise points N are obtained again based on the core point C. For all core points, the points within the neighborhood of the core point ε are called reachable points, and the points outside the neighborhood are called noise points, that is, abnormal points that need to be removed; the reachable points R and noise points N can be expressed as:

[0063]

[0064] Where R represents a reachable point, N represents a noise point, p represents a sample point object, D represents a data set of historical operation data, C represents a core point, dist(p,C) represents the distance between points C and p, and ε represents the scanning radius;

[0065] Finally, the obtained noise points N are deleted to complete the elimination of abnormal points in the historical operation data.

[0066] Step 2: Use a sliding window algorithm to traverse the historical operation data after removing abnormal points to obtain the steady-state operating point of the thermal power unit, where the steady-state operating point at least includes load data, coal feed data, and water feed flow data;

[0067] It should be noted that in the embodiment of the present application, in the process of obtaining the steady-state operating point of the thermal power unit, load data and main steam pressure are used for steady-state judgment, wherein the signal sampling time of the standard operating point is set to 1s, the time span of adjacent standard operating points is set to 1 hour, and the maximum allowable deviation during steady-state judgment is set to 5%.

[0068] Specifically, when performing steady-state judgment in the embodiment of the present application, it is necessary to judge whether the load data and main steam pressure corresponding to the standard operating point satisfy the evaluation equation:

[0069]

[0070] Where P st (i) represents the main steam pressure corresponding to operating point i, N e (i) represents the unit load corresponding to operating point i, represents the mean value of the main steam pressure, represents the average load of the unit, n is the number of standard operating points, σ P and σ N are the maximum allowable deviations respectively;

[0071] If the load data and main steam pressure of the thermal power unit both satisfy the evaluation equation, the current standard operating point is judged to be a steady-state operating point, and the load data, coal supply data and water supply data corresponding to the standard operating point are stored.

[0072] If any of the load data and main steam pressure of the thermal power unit does not satisfy the evaluation equation, it is judged that the current standard operating point is not a steady-state operating point, and the standard operating point is discarded.

[0073] Step 3: Perform sufficiency and uniformity checks on the steady-state operating points of the thermal power units.

[0074] In the embodiment of the present application, when the sufficiency check is performed on the steady-state operating point of the thermal power unit, the threshold value of the sufficiency of the steady-state operating point is generally set to 3% of the total number of operating points; for example, in the embodiment of the present application, the operation data at an interval of 1h is used as a operating point, and there are nearly 2,200 operating points in three months. Considering the time of unit start-up and shutdown, maintenance, and load lifting, in order to ensure that the number of selected steady-state operating points is sufficient, the selection threshold of the steady-state operating point is generally set to about 3% (about 60) of the total number of operating points. When the number of steady-state operating points is greater than or equal to 3% of the total number of operating points, it is determined that the sufficiency check of the steady-state operating point of the thermal power unit has passed, and a uniformity check is performed; when the number of steady-state operating points is less than 3% of the total number of operating points, it is determined that the sufficiency check of the steady-state operating point of the thermal power unit has not passed, and the process of the present application method is terminated, waiting for the next self-learning.

[0075] After the adequacy check of the steady-state operating point is passed, it is necessary to further perform a uniformity check on the steady-state operating point of the thermal power unit. The embodiment of the present application respectively determines the number of steady-state operating points in the load intervals of 20% Ne0-40% Ne0, 40% Ne0-60% Ne0, 60% Ne0-80% Ne0 and 80% Ne0-100% Ne0; when the number of steady-state operating points in the four load intervals is greater than or equal to 10, it is determined that the uniformity check of the steady-state operating point of the thermal power unit has passed; when the number of steady-state operating points in any load interval is less than 10, it is determined that the uniformity check of the steady-state operating point of the thermal power unit has failed.

[0076] Step 4: If all steady-state operating points of the thermal power unit have passed the verification, the verified steady-state operating points are input into the neural network model for training to obtain a trained neural network model.

[0077] Input the load data, coal supply data and water supply flow data of the steady-state operating point into the neural network model;

[0078] The error back propagation algorithm is used to adjust the weights and thresholds of each hidden layer node in the neural network model, so that the output value of the neural network model is close to the actual value; combined with Figure 2As shown, the adjustment formula for the output value, weight and threshold of each hidden layer node is:

[0079]

[0080] where y j Output value for the node; x j is the node input value; f() is the activation function, and the S-type function is selected; w ij is the weight between nodes i and j; b j is the threshold of node j; m is the number of hidden layer nodes; is the error function calculation formula based on Widrow-Hoff; η1 and η2 are the error coefficients respectively.

[0081] It should be noted that the neural network model in the embodiment of the present application adopts BP neural network. BP neural network is a multi-layer feedforward neural network trained according to the error back propagation algorithm, which has powerful self-learning and adaptive capabilities and can efficiently handle various nonlinear problems.

[0082] During the training process of the neural network model, the initial number of nodes in the hidden layer is set to twice the input eigenvalue, and then the number of nodes is appropriately changed according to the training results; during the training process, the initial learning rate is set to 0.01 or 0.001, and the learning rate can be gradually reduced during the training process according to the complexity of the steady-state operating point data, so as to converge faster in the initial stage, and then slow down the update speed when approaching the optimal solution; the root mean square error is selected as the evaluation indicator of the neural network model, and multiple trainings are performed, and the value of the hyperparameter is adjusted each time to find the optimal training model.

[0083] Step 5: Import the trained neural network model into the distributed control system of the thermal power unit.

[0084] Step 6: Input the real-time load signal into the neural network model of the distributed control system to obtain the correction signal of the feedforward control. In the present application, the neural network model of the distributed control system inputs the load data of each steady-state operating point, and then outputs the corresponding coal feed data and water flow data; the output coal feed data and water flow data are respectively compared with the actual coal feed data and actual water flow data corresponding to the original signal to obtain the incremental value, and the incremental value is used as the correction signal of the feedforward control.

[0085] Step 7: Input the correction signal into the feedforward control module of the coal feed rate and water feed rate in the thermal power unit to complete the feedforward control correction.

[0086] like Figure 3 As shown, in addition, the present application also discloses a feedforward correction self-learning device for a coordinated control system of a thermal power unit, which comprises:

[0087] A data acquisition module is used to acquire historical operation data of the thermal power unit and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data;

[0088] The data screening module is used to use a sliding window algorithm to traverse the historical operating data to remove abnormal points and obtain the steady-state operating point of the thermal power unit. The steady-state operating point at least includes load data, coal feed data and water feed flow data.

[0089] Data verification module, used to perform sufficiency verification and uniformity verification on the steady-state operating points of thermal power units;

[0090] A data training module is used to input the verified steady-state operating point into the neural network model for training after the steady-state operating point of the thermal power unit is verified, so as to obtain a trained neural network model;

[0091] Data import module, used to import the trained neural network model into the distributed control system of the thermal power unit;

[0092] A correction signal output module is used to input the real-time load signal into the neural network model of the distributed control system to obtain a correction signal for feedforward control;

[0093] The correction control module is used to input the correction signal into the feedforward control module of the coal feed rate and water feed flow rate in the thermal power unit to complete the feedforward control correction.

[0094] The device provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented by the method embodiment will not be described again here.

[0095] like Figure 4 As shown, the embodiment of the present application also provides an electronic device, including a processor and a memory, a program or instruction stored in the memory and executable on the processor, and when the program or instruction is executed by the processor, the following is achieved: Figure 1 The various processes of the method embodiment shown in the figure can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0096] The embodiment of the present application also provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, the above Figure 1 The various processes of the method embodiments described above can achieve the same technical effect, and will not be described again here to avoid repetition.

[0097] The present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above Figure 1The various processes of the method embodiments described above can achieve the same technical effect, and will not be described again here to avoid repetition.

[0098] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned sequence numbers of the embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0099] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0100] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another device, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0101] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0102] In addition, all functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0103] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, etc., various media that can store program codes.

[0104] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application can essentially or in other words, the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a device (which can be a terminal or a platform, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0105] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A feedforward correction self-learning method for a coordinated control system of a thermal power unit, characterized in that: The steps include: Acquire historical operation data of the thermal power unit and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data; The sliding window algorithm is used to traverse the historical operation data after removing abnormal points to obtain the steady-state operating point of the thermal power unit, wherein the steady-state operating point at least includes load data, coal supply data and water supply flow data; When the steady-state operating point of the thermal power unit is obtained, it is determined whether the load data and main steam pressure corresponding to the standard operating point satisfy the evaluation equation: ； in represents the main steam pressure corresponding to operating point i, represents the unit load corresponding to operating point i, represents the mean value of the main steam pressure, Indicates the average load of the unit. n is the number of standard working condition points, and are the maximum allowable deviations respectively; If the load data and main steam pressure of the thermal power unit satisfy the evaluation equation, the load data, coal supply data and water supply data corresponding to the standard operating point are stored; Conduct sufficiency and uniformity checks on the steady-state operating points of thermal power units; If the steady-state operating points of the thermal power unit are all verified, the verified steady-state operating points are input into the neural network model for training to obtain the trained neural network model: Input the load data, coal supply data and water supply flow data of the steady-state operating point into the neural network model; The error back propagation algorithm is used to adjust the weights and thresholds of each hidden layer node in the neural network model so that the output value of the neural network model approaches the actual value; the adjustment formula for the output value, weight and threshold of each hidden layer node is: ； in Output value for the node; Enter a value for the node; For the activation function, choose the S-type function; For Node i and j The weight between For Node i The threshold value of m is the number of hidden layer nodes; is the error function calculation formula based on Widrow-Hoff; and are the error coefficients respectively; Import the trained neural network model into the distributed control system of the thermal power unit; Input the real-time load signal into the neural network model of the distributed control system to obtain the correction signal for feedforward control; The correction signal is input into the feedforward control module of the coal feed rate and water feed flow rate in the thermal power unit to complete the feedforward control correction.

2. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 1 is characterized in that: The step of removing abnormal points from historical operation data includes: The historical operation data of the thermal power unit is represented as three-dimensional spatial data of coal supply data, water supply flow data and load data; Select the unit in , , , , , , and The corresponding standard operating point is taken as the sample point, where is the rated load; Calculate the core point C within the sample point based on the sample point and historical operation data: ； in C Indicates the core point, p Represents a sample point object, S represents the sample point, D A dataset representing historical operation data, express S , p The distance between two points, represents the sample count, Minimum number of included points, Indicates the scanning radius; According to the core point C Get reachable points R and noise points N : ； in R Represents a reachable point, N represents the noise point, p Represents a sample point object, D A dataset representing historical operation data, C Indicates the core point, express C , p The distance between two points Indicates the scanning radius; The noise points obtained N Delete to eliminate abnormal points in historical operation data.

3. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 1 is characterized in that: In the step of obtaining the steady-state operating point of the thermal power unit, load data and main steam pressure are used for steady-state judgment, wherein the signal sampling time of the standard operating point is set to 1s, the time span of adjacent standard operating points is set to 1 hour, and the maximum allowable deviation is set to 5%.

4. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 1 is characterized in that: When performing sufficiency check on the steady-state operating points of the thermal power unit, the threshold for the sufficiency of the steady-state operating points is set to 3% of the total number of operating points; when the number of steady-state operating points is greater than or equal to 3% of the total number of operating points, the sufficiency check of the steady-state operating points of the thermal power unit is determined to have passed, and a uniformity check is performed; when the number of steady-state operating points is less than 3% of the total number of operating points, the sufficiency check of the steady-state operating points of the thermal power unit is determined to have failed, and the process is terminated.

5. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 4 is characterized in that: When checking the uniformity of the steady-state operating point of the thermal power unit, the , , and The number of steady-state operating points in the load range; when the number of steady-state operating points in the four load ranges is greater than or equal to 10, it is determined that the uniformity check of the steady-state operating points of the thermal power unit has passed; when the number of steady-state operating points in any load range is less than 10, it is determined that the uniformity check of the steady-state operating points of the thermal power unit has not passed.

6. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 1 is characterized in that: The neural network model adopts BP neural network.

7. The feedforward correction self-learning method of the coordinated control system of a thermal power unit according to claim 1 is characterized in that: During the training process of the neural network model, the initial number of nodes in the hidden layer is set to twice the input characteristic value; during the training process, the initial learning rate is set to 0.01 or 0.001; and the root mean square error is selected as the evaluation index of the neural network model.

8. A feedforward correction self-learning device for a coordinated control system of a thermal power unit, characterized in that: include: A data acquisition module is used to acquire historical operation data of the thermal power unit and remove abnormal points from the historical operation data, wherein the operation data at least includes load data, coal supply data and water supply flow data; A data screening module is used to traverse the historical operation data of the abnormal points by using a sliding window algorithm to obtain the steady-state operating point of the thermal power unit, wherein the steady-state operating point includes at least load data, coal supply data and water supply flow data; When the steady-state operating point of the thermal power unit is obtained, it is determined whether the load data and main steam pressure corresponding to the standard operating point satisfy the evaluation equation: ； in represents the main steam pressure corresponding to operating point i, represents the unit load corresponding to operating point i, represents the mean value of the main steam pressure, Indicates the average load of the unit. n is the number of standard working condition points, and The maximum allowable deviation If the load data and main steam pressure of the thermal power unit satisfy the evaluation equation, the load data, coal supply data and water supply data corresponding to the standard operating point are stored; Data verification module, used to perform sufficiency verification and uniformity verification on the steady-state operating points of thermal power units; The data training module is used to input the verified steady-state operating point into the neural network model for training after the steady-state operating point of the thermal power unit is verified, so as to obtain the trained neural network model: Input the load data, coal supply data and water supply flow data of the steady-state operating point into the neural network model; The error back propagation algorithm is used to adjust the weights and thresholds of each hidden layer node in the neural network model so that the output value of the neural network model approaches the actual value; the adjustment formula for the output value, weight and threshold of each hidden layer node is: ； in Output value for the node; Enter a value for the node; For the activation function, choose the S-type function; For Node i and j The weight between For Node i The threshold value of m is the number of hidden layer nodes; is the error function calculation formula based on Widrow-Hoff; and are the error coefficients respectively; Data import module, used to import the trained neural network model into the distributed control system of the thermal power unit; A correction signal output module is used to input the real-time load signal into the neural network model of the distributed control system to obtain a correction signal for feedforward control; The correction control module is used to input the correction signal into the feedforward control module of the coal feed rate and water feed flow rate in the thermal power unit to complete the feedforward control correction.

Images