Load distribution method for gas-steam combined cycle unit

Abstract

The invention belongs to the technical field of load distribution of gas-steam combined cycle units, and discloses a load distribution method of a gas-steam combined cycle unit. The load distribution method comprises the steps that a steady-state working condition data set of each unit serves as data input, and a gas consumption rate prediction model of each unit is established; and taking the minimum total gas consumption rate of the whole plant as an output target, applying the gas consumption rate prediction model of each unit to obtain a global optimal solution set, and carrying out load distribution on the multiple units according to the global optimal solution set. According to the load distribution method of the gas-steam combined cycle unit, the load distribution of each unit is carried out through the generated global optimal solution, so that each unit can operate under the conditions of the most efficient and the highest energy under the condition of stable and safe operation, the performance of each unit is fully mobilized, and the load distribution efficiency of each unit is improved. And energy conservation and consumption reduction are realized while the productivity efficiency is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of load distribution of gas-steam combined cycle units, and in particular relates to a load distribution method for gas-steam combined cycle units. Background Art

[0002] The gas-steam combined cycle unit is an advanced energy generation equipment. The combined cycle unit can make full use of the waste heat of the high-temperature exhaust gas of the gas turbine, reduce production costs, greatly improve the efficiency of energy utilization, and its power generation efficiency is much higher than that of traditional coal-fired thermal power units.

[0003] However, at present, most gas-steam combined cycle power plants are single-machine automatic generation control (single-machine AGC) mode, which does not take into account the actual situation of each unit. In order to ensure the stability of the unit, low-performance units are usually used for load distribution. Units with high efficiency, low gas consumption rate and low pollution emissions cannot play their own advantages, resulting in the inability of each unit in the whole plant to autonomously allocate the optimal load according to the best economy, environmental protection and output capacity, resulting in the actual economic efficiency being limited. There are also a small number of plant-level AGCs for control. Plant-level AGCs take the entire power plant as the dispatch object. The current plant-level AGC load distribution method based on gas-steam combined cycle units is mainly based on the equal incremental rate method and linear programming method. According to the unit gas consumption rate is related to active power and ambient temperature, because it considers the impact on the unit gas consumption rate is mostly only from the perspective of ambient temperature, the factors considered are limited. In some cases, it is impossible to accurately predict the unit gas consumption rate.

[0004] Therefore, there is an urgent need to provide a load distribution method for gas-steam combined cycle units that can fully utilize the performance of each unit, save energy and reduce consumption, and have higher production efficiency to overcome the above problems.

[0005] In view of this, this application is hereby filed. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiency that the control method of the prior art cannot fully call upon the performance of each unit, resulting in limited production efficiency. The purpose is to provide a load distribution method for gas-steam combined cycle units that can fully call upon the performance of each unit and have higher production efficiency.

[0007] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is: a load distribution method for a gas-steam combined cycle unit, comprising the steps of:

[0008] S1. Taking the steady-state operating condition data set of each unit as data input and the gas consumption rate as output parameter, the XGBoost big data analysis method is used to establish a gas consumption rate prediction model for each unit;

[0009] S2. Taking the minimum total gas consumption rate of the whole plant as the output target, the dynamic programming method is adopted, and the gas consumption rate prediction model of each unit is applied to obtain the global optimal solution set under the load constraint conditions of each unit, and the load distribution of multiple units is performed according to the obtained global optimal solution set.

[0010] According to an embodiment of the present invention, before step S1, the method further includes: S0, acquiring an original data set of each unit, and performing data filtering processing on the original data set to generate a steady-state operating condition data set.

[0011] According to an embodiment of the present invention, step S0 includes:

[0012] S0.1. Obtain historical data of gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit from the SIS system, and use them as the original data set; wherein the sampling time for obtaining the historical data of the gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit is at least one year, and the interval between two adjacent data acquisitions is 1 minute;

[0013] S0.2. Perform data filtering on the original data set, and use the data after data filtering as the steady-state working condition data set.

[0014] According to one embodiment of the present invention, the parameters in the original data set include: ambient temperature, lower calorific value of fuel, gas flow, gas turbine load, steam turbine load, unit load, filter pressure difference, condenser back pressure, waste heat boiler outlet flue gas temperature, fuel heater outlet temperature, and gas consumption rate.

[0015] According to one embodiment of the present invention, the step S0.2 comprises:

[0016] S0.2.1. Filter and delete abnormal data and null value data in the original data set caused by abnormal operation of equipment and abnormal information transmission;

[0017] S0.2.2. Filter and delete the operating condition data of the unit start-up and shutdown phases in the original data set;

[0018] S0.2.3. Filter and delete the operating condition data of the variable operating condition stage in the original data set;

[0019] S0.2.4. The data after data filtering is used as the steady-state operating condition data set.

[0020] According to one embodiment of the present invention, step S2 comprises:

[0021] S2.1, dividing the steady-state operating data set of each unit into a training data set and a test data set;

[0022] S2.2. Using the XGBoost machine learning method, the gas consumption rate is used as the model to predict the output parameters, the XGBoost model is established, and the grid optimization method is used to find the optimal parameters of the model;

[0023] The established XGBoost model is a regression task, and the mean square error function is used as the loss function of the model;

[0024] S2.3. Use the training data set and the test data set to train and test the XGBoost model to obtain a gas consumption rate prediction model for each unit.

[0025] According to an embodiment of the present invention, the steady-state operating data set of each unit is divided into the training data set and the test data set in a ratio of 7:3.

[0026] According to an embodiment of the present invention, the load distribution process is divided into n stages according to the total number of units n, and the state variables and decision variables of the i-th stage are respectively X i , P i The upper and lower limits of the load of the i-th unit are P imin , P imax , the optimal value objective function F i (X i ,S i ) is represented by the total load of the first i units as X i The average gas consumption rate when the gas consumption is 2 hours; wherein n is a natural number, n≥2, i≤n, and the X i , P i Represent the total load of the first i units and the load of the i-th unit respectively;

[0027] The step S2 further comprises:

[0028] S2.4 Sequential table creation;

[0029] Step 1: F1(X1, S1) = f1(P1, S1);

[0030] Where: P 1min ≤P1≤P 1max , S1 is the data of unit 1 at the current moment, and X1 traverses its value interval [P 1min , P 1max ], while saving related data;

[0031] Step 2:

[0032] Where: P 2min ≤P2≤P2max , S2 is the data of unit 2 at the current moment; X2 traverses its value interval with a predetermined step length Combined with the calculation results of the first step, calculate F2(X2, S2) corresponding to P2 and save the relevant data;

[0033] Step 3:

[0034] Where: n≥j≥i, P jmin ≤P j ≤P jmax , S j is the data of the j# unit at the current moment. j Traverse its value range with a predetermined step size Combined with the calculation results of the previous stage, calculate P j The corresponding F j (X j , S j ), while saving related data;

[0035] Similarly, after calculating the nth unit, we can combine the calculation results of the n-1th unit to calculate X. n =P total When different P n The corresponding F n (P n , S n ) value, and finally find the optimal value;

[0036] S2.5, reverse table lookup;

[0037] The corresponding total AGC load of the whole plant is P total When the load P allocated to the nth unit after optimization is found from the table in the nth stage n ;

[0038] Similarly, in the n-1th stage, the load P allocated after optimization of the n-1th unit is n-1 , according to the state transfer relation X i =X i+1 -P i+1 , and so on, according to the dynamic programming algorithm, the optimal load distribution result of n units can be determined as follows: {P n , P n-1 ,…,P i ,…,P1}.

[0039] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0040] 1) In the present invention, by establishing and applying the gas consumption rate prediction model of each unit, taking the minimum total gas consumption rate of the whole plant as the output target, the load of multiple units is optimally allocated, and the performance of each unit is fully called upon while meeting its operating conditions, so as to improve the production efficiency of the cycle unit;

[0041] 2) In the present invention, a gas consumption rate prediction model is constructed by applying more parameters that have a direct or indirect impact on the gas consumption rate. Compared with a single factor condition, the prediction degree of the prediction model is more real and reliable, and the matching degree with the actual operation is higher, which has a positive significance for ensuring the correctness of subsequent load processing and distribution;

[0042] 3) In the present invention, by applying the data filtering and processing steps, compliant (steady-state) operating condition data are screened out, ensuring that no data anomaly errors will occur in the subsequent construction process of the gas consumption rate prediction model, thereby avoiding the occurrence of problems such as distortion in the construction of the prediction model and construction errors that lead to slow progress.

[0043] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are part of the present invention and are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but do not constitute an improper limitation of the present invention. Obviously, the drawings described below are only some embodiments. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the accompanying drawings:

[0045] Figure 1 A flow chart of the main steps of a load distribution method for a gas-steam combined cycle unit in an embodiment of the present invention;

[0046] Figure 2 A partial step flow chart of a load distribution method for a gas-steam combined cycle unit according to an embodiment of the present invention;

[0047] Figure 3 A chart showing the results of data filtering processing of a specific example in an embodiment of the present invention;

[0048] Figure 4 A graph showing steady-state screening results of some data of specific examples in the embodiments of the present invention;

[0049] Figure 5 A diagram showing model results of a specific example in an embodiment of the present invention;

[0050] Figure 6 The figure is a display chart of the calculation results of a specific example in the embodiment of the present invention.

[0051] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below in conjunction with the drawings in the 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.

[0053] In the description of the present invention, it should be noted that the directions or positional relationships indicated by terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “vertical”, “inside” and “outside” are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0054] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0055] like Figure 1 and Figure 2 As shown, a load distribution method for a gas-steam combined cycle unit according to the present invention comprises the following steps:

[0056] S1. Taking the steady-state operating condition data set of each unit as data input and the gas consumption rate as output data (parameter), the XGBoost big data analysis (modeling) method is used to establish a gas consumption rate prediction model for each unit;

[0057] S2. Taking the minimum total gas consumption rate of the whole plant as the output target, the dynamic programming method is adopted, and the gas consumption rate prediction model of each unit is applied to obtain the global optimal solution set that meets the load constraints of the unit, and the load distribution of multiple units is performed according to the obtained global optimal solution set.

[0058] In the present invention, by establishing and applying a gas consumption rate prediction model for each unit, taking the minimum total gas consumption rate of the whole plant as the output target, the loads of multiple units are optimally distributed, and the performance of each unit is fully utilized while meeting its operating conditions, thereby improving the production capacity efficiency of the cycle unit.

[0059] In a specific implementation of this embodiment, a corresponding number of gas consumption rate prediction models are constructed according to the number of units.

[0060] In another specific implementation of this embodiment, only one gas consumption rate prediction model is constructed, and the gas consumption rate prediction model stores sub-models constructed corresponding to each unit.

[0061] Please refer to the attached Figure 1 In a specific implementation of this embodiment, before step S1, it also includes: S0, obtaining the original data set of each unit, and performing data filtering processing on the original data set to generate a steady-state operating condition data set.

[0062] Please refer to the attached Figure 1 In a specific implementation of this embodiment, step S0 includes:

[0063] S0.1. Obtain historical data of gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit from the SIS system (Supervisory Information System), and use them as the original data set; wherein the sampling time for obtaining the historical data of the gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit is at least one year, and the interval between two adjacent data acquisitions is 1 minute;

[0064] S0.2. Perform data filtering on the original data set, and use the data after data filtering as the steady-state working condition data set.

[0065] In a specific implementation of this embodiment, the parameters in the original data set include: ambient temperature, lower calorific value of fuel, gas flow, gas turbine load, steam turbine load, unit load, filter pressure difference, condenser back pressure, waste heat boiler outlet flue gas temperature, fuel heater outlet temperature, and gas consumption rate.

[0066] In the present invention, a gas consumption rate prediction model is constructed by applying more parameters that have a direct or indirect impact on the gas consumption rate. Compared with a single factor condition, the prediction degree of the prediction model is more realistic and has a higher matching degree with the actual operation, which has a positive significance for ensuring the correctness of subsequent load processing distribution.

[0067] Please refer to the attached Figure 1 In a specific implementation of this embodiment, step S0.2 includes:

[0068] S0.2.1. Filter and delete abnormal data and null value data in the original data set caused by abnormal operation of equipment and abnormal information transmission;

[0069] S0.2.2. Filter and delete the operating condition data of the unit start-up and shutdown phases in the original data set;

[0070] S0.2.3. Filter and delete the operating condition data of the variable load operating condition stage in the original data set;

[0071] S0.2.4. The data after data filtering is used as the steady-state operating condition data set.

[0072] In the present invention, by applying the data filtering processing step, compliant (steady-state) operating condition data is screened out, ensuring that no data anomaly errors will occur in the subsequent construction process of the gas consumption rate prediction model, avoiding the occurrence of problems such as distortion in the construction of the prediction model and construction errors that lead to slow modeling progress.

[0073] In another specific implementation of this embodiment, the step S0.2.1, the step S0.2.2 and the step S0.2.3 may be performed in other orders, for example, step S0.2.2 is performed first, then step S0.2.3, and finally step S0.2.1. Of course, other orders are also possible.

[0074] In another specific implementation of this embodiment, the step S0.2.1, the step S0.2.2 and the step S0.2.3 are performed simultaneously.

[0075] Please refer to the attached Figure 1 In a specific implementation of this embodiment, step S2 includes:

[0076] S2.1, dividing the steady-state operating data set of each unit into a training data set and a test data set;

[0077] S2.2. Using the XGBoost machine learning method, taking the gas consumption rate as the prediction output parameter of the prediction model, establishing the XGBoost model, and using the grid optimization method to find the optimal parameters of the model;

[0078] The established XGBoost model is a regression task, and the mean square error function is used as the loss function of the model;

[0079] S2.3. Use the training data set and the test data set to train and test the XGBoost model to obtain a gas consumption rate prediction model for each unit.

[0080] In the present invention, the steady-state data set is divided into a training data set and a test data set, and the trained prediction model is tested to ensure the reliability of the prediction model generation.

[0081] In a specific implementation of this embodiment, the gas consumption rate prediction model trained by the training database is tested by substituting the test data set; when the data matching rate (matching rate of predicted gas consumption rate parameters and actual gas consumption rate parameters) detected is greater than a predetermined inspection standard, it indicates that the gas consumption rate prediction model is successfully constructed (tested), for example, the predetermined inspection standard is 95% of all data volumes; it can also be that the predetermined inspection standard is that the deviation value between the gas consumption rate parameters generated by the trained gas consumption rate prediction model according to other parameters (except the gas consumption rate in the steady-state data set) and the actual gas consumption rate parameters is less than the predetermined inspection standard, for example, the inspection standard is a deviation value of 5% (the allowable error value range is ±5%); it can also be a combination of the above two predetermined inspection standards, for example, when both the detected data matching rate is greater than the predetermined inspection standard and the deviation value between the gas consumption rate parameters generated by other parameters and the actual gas consumption rate parameters is less than the predetermined inspection standard, it indicates that the gas consumption rate prediction model is successfully constructed (tested).

[0082] It can be understood that the process of applying the training data set to construct the gas consumption rate model is the process of constructing a relational expression for the data (single data group) under each steady state in the training data set;

[0083] The process of using the test data set to detect the gas consumption rate model is a process of importing other parameters of the data under each steady-state condition (single data group) in the training data set except the gas consumption rate into the constructed relational expression to obtain the calculated gas consumption rate.

[0084] In a specific implementation of this embodiment, the steady-state operating data set of each unit is divided into the training data set and the test data set in a ratio of 7:3.

[0085] In the present invention, the ratio of the training data set to the test data set is 7:3, and a larger number of training data sets can ensure the accuracy and authenticity of the gas consumption rate prediction model constructed through training.

[0086] In a specific implementation of this embodiment, the load distribution process is divided into n stages according to the total number of units n, and the state variables and decision variables of the i-th stage are respectively X i , P i The upper and lower limits of the load of the i-th unit are P imin , P imax , the optimal value objective function F i (X i ,S i ) is represented by the total load of the first i units as X i The average gas consumption rate when the gas consumption is 2 hours; wherein n is a natural number, n≥2, i≤n, and the X i , P iRepresent the total load of the first i units and the load of the i-th unit respectively;

[0087] The step S2 further comprises:

[0088] S2.4 Sequential table creation;

[0089] Step 1: F1(X1, S1) = f1(P1, S1);

[0090] Where: P 1min ≤P1≤P 1max , S1 is the data of unit 1 at the current moment, and X1 traverses its value interval [P 1min , P 1max ], while saving related data;

[0091] Step 2:

[0092] Where: P 2min ≤P2≤P 2max , S2 is the data of unit 2 at the current moment; X2 traverses its value interval with a predetermined step length Combined with the calculation results of the first step, calculate F2(X2, S2) corresponding to P2 and save the relevant data;

[0093] Step 3:

[0094] Where: n≥j≥i, P jmin ≤P j ≤P jmax , S j is the data of the j# unit at the current moment. j Traverse its value range with a predetermined step size Combined with the calculation results of the previous stage, calculate P j The corresponding F j (X j , S j ), while saving related data;

[0095] Similarly, we can calculate the nth unit and combine the calculation results of the n-1th unit to calculate X. n =P total When different P n The corresponding F n (P n , S n ) value, and finally find the optimal value;

[0096] S2.5, reverse table lookup;

[0097] The corresponding total AGC load of the whole plant is P total When the load P allocated to the nth unit after optimization is found from the table in the nth stage n ;

[0098] Similarly, in the n-1th stage, the load P allocated after optimization of the n-1th unit is n-1 , according to the state transfer relation X i =X i+1 -P i+1 , and so on, according to the dynamic programming algorithm, the optimal load distribution result of n units can be determined as follows: {P n , P n-1 , ..., P i , ..., P1}.

[0099] By applying a load distribution method for a gas-steam combined cycle unit provided in the present application, a gas consumption rate prediction model for each unit is constructed and applied by utilizing sufficient sampling data and data filtering and processing processes, with the goal of minimizing the gas consumption rate of the entire plant. Through the generated global optimal solution, the load distribution of each unit is carried out, so that each unit can operate under the most efficient and highest energy conditions under stable and safe operation, fully mobilizing the performance of each unit, making more effective use of energy, improving production efficiency and achieving energy conservation and consumption reduction.

[0100] In a specific implementation of the load distribution method provided in the application, the process data can be found in the attached Figure 3 To Attachment Figure 6 :

[0101] The Guangzhou Nansha gas-steam combined cycle power plant is taken as the research object. The power plant contains two gas-steam combined cycle units. The data sampling time is from 2023-01-01 to 2023-12-31, a total of one year of data, with a sampling interval of 1 minute. A total of 1,051,200 data samples are obtained from the two units, and the original data set of a single unit has 525,600.

[0102] The invalid data in the original data of the two units are filtered out by using the data filtering method, and samples with abnormal values ​​and null values ​​are deleted, samples of the unit start-up and shutdown conditions are deleted, and samples of the unit variable load conditions are deleted. The results of data filtering are as follows: Figure 3 As shown, the results of steady-state screening of some data (partial steady-state operating condition data set) are as follows Figure 4 As shown;

[0103] The steady-state operating data set of each unit is divided into a training data set and a test data set in a ratio of 7:3, and input into the gas consumption rate prediction model based on the XGBoost algorithm for training. RMSE (Root Mean Square Error) and R2 (coefficient of determination) are used as evaluation indicators. The prediction results of the prediction model are as follows: Figure 5 As shown;

[0104] The given 9 AGC loads of the whole plant are optimized and calculated. The gas consumption rate prediction model of the two units established above is used to predict the gas consumption rate of each unit. The dynamic programming method is used to solve the global optimal solution under the constraints. The calculation results are as follows: Figure 6 As shown; the test results show that the unit load optimization allocation method based on big data analysis reduces the total gas consumption rate of the whole plant, proving that this method is accurate and practical.

[0105] The above description is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this patent can make some changes or modifications to equivalent embodiments of equivalent changes by using the technical content suggested above without departing from the scope of the technical solution of the present invention. The implementation scheme in the above embodiment can also be further combined or replaced. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the solution of the present invention.

Claims

1. A load distribution method for a gas-steam combined cycle unit, characterized in that: The load distribution method comprises the steps of: S1. Taking the steady-state operating condition data set of each unit as data input and the gas consumption rate as output parameter, the XGBoost big data analysis method is used to establish a gas consumption rate prediction model for each unit; S2. Taking the minimum total gas consumption rate of the whole plant as the output target, the dynamic programming method is adopted, and the gas consumption rate prediction model of each unit is applied to obtain the global optimal solution set under the load constraint conditions of each unit, and the load distribution of multiple units is performed according to the obtained global optimal solution set.

2. A load distribution method for a gas-steam combined cycle unit according to claim 1, characterized in that: Before step S1, the method further includes: S0, acquiring an original data set of each unit, and performing data filtering processing on the original data set to generate a steady-state operating condition data set.

3. A load distribution method for a gas-steam combined cycle unit according to claim 2, characterized in that: The step S0 comprises: S0.

1. Obtain historical data of gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit from the SIS system, and use them as the original data set; wherein the sampling time for obtaining the historical data of the gas consumption rate parameters and parameters that directly or indirectly affect the gas consumption rate of the unit is at least one year, and the interval between two adjacent data acquisitions is 1 minute; S0.

2. Perform data filtering on the original data set, and use the data after data filtering as the steady-state working condition data set.

4. A load distribution method for a gas-steam combined cycle unit according to claim 2 or 3, characterized in that: The parameters in the original data set include: ambient temperature, fuel lower calorific value, gas flow, gas turbine load, steam turbine load, unit load, filter pressure difference, condenser back pressure, waste heat boiler outlet flue gas temperature, fuel heater outlet temperature, and gas consumption rate.

5. A load distribution method for a gas-steam combined cycle unit according to claim 4, characterized in that: The step S0.2 comprises: S0.2.

1. Filter and delete abnormal data and null value data in the original data set caused by abnormal operation of equipment and abnormal information transmission; S0.2.

2. Filter and delete the operating condition data of the unit start-up and shutdown phases in the original data set; S0.2.

3. Filter and delete the operating condition data of the variable operating condition stage in the original data set; S0.2.

4. The data after data filtering is used as the steady-state operating condition data set.

6. A load distribution method for a gas-steam combined cycle unit according to claim 4, characterized in that: The step S2 comprises: S2.1, dividing the steady-state operating data set of each unit into a training data set and a test data set; S2.

2. Using the XGBoost machine learning method, taking the gas consumption rate as the prediction output parameter of the prediction model, establishing the XGBoost model, and using the grid optimization method to find the optimal parameters of the model; The established XGBoost model is a regression task, and the mean square error function is used as the loss function of the model; S2.

3. Use the training data set and the test data set to train and test the XGBoost model to obtain a gas consumption rate prediction model for each unit.

7. A load distribution method for a gas-steam combined cycle unit according to claim 6, characterized in that: The steady-state operating data set of each unit is divided into the training data set and the test data set in a ratio of 7:

3.

8. A load distribution method for a gas-steam combined cycle unit according to claim 6, characterized in that: According to the total number of units n, the load distribution process is divided into n stages. The state variables and decision variables of the i-th stage are X i , P i The upper and lower limits of the load of the i-th unit are P imin , P imax , the optimal value objective function F i (X i , S i ) is represented by the total load of the first i units as X i The average gas consumption rate when the gas consumption is 2 hours; wherein n is a natural number, n≥2, i≤n, and the X i , P i Represent the total load of the first i units and the load of the i-th unit respectively; The step S2 further comprises: S2.4 Sequential table creation; Step 1: F1(X1, S1) = f1(P1, S1); Where: P 1min ≤P1≤P 1max , S1 is the data of unit 1 at the current moment, and X1 traverses its value interval [P 1min , P 1max ], while saving related data; Step 2: Where: P 2min ≤P2≤P 2max , S2 is the data of unit 2 at the current moment; X2 traverses its value interval with a predetermined step length Combined with the calculation results of the first step, calculate F2(X2, S2) corresponding to P2 and save the relevant data; Step 3: Where: n≥j≥i, P jmin ≤P j ≤P jmax , S j is the data of the j# unit at the current moment. j Traverse its value range with a predetermined step size Combined with the calculation results of the previous stage, calculate P j The corresponding F j (X j , S j ), while saving related data; By analogy, we can calculate the nth unit, and combine the calculation results of the n-1th unit to calculate X n =P total When different P n The corresponding F n (P n , S n ) value, and finally find the optimal value; S2.5, reverse table lookup; The corresponding total AGC load of the whole plant is P total When the load P allocated to the nth unit after optimization is found from the table in the nth stage n ; Similarly, in the n-1th stage, the load P allocated after optimization of the n-1th unit is n-1 , according to the state transfer relation X i =X i+1 -P i+1 , and so on, according to the dynamic programming algorithm, the optimal load distribution result of n units can be determined as follows: (P n , P n-1 ,…,P i ,…,P1}.

Images