Accurate coaling blending combustion optimization method for coal-fired thermal generator set

Abstract

The invention discloses an accurate coaling blending combustion optimization method for a coal-fired thermal generator set, and the method comprises the steps: S1, obtaining the historical combustion data of each unit of a thermal power plant, and building a combustion sample database; s2, based on the combustion sample data, establishing a prediction model g (.) of the theoretical coal feed quantity wl, k, i of the coal mill; s3, based on the combustion sample data, establishing a prediction model f (.) of the heat efficiency el of unit power generation; s4, establishing a joint decision mathematical model of coal bunker-ore spot coal distribution and coal mill on / off, and setting constraint conditions; s5, establishing a variable neighborhood search algorithm for calculating a coal bunker coaling scheme and a coal mill starting and stopping scheme under the condition of minimum power generation coal consumption; and S6, carrying out coaling blending combustion operation in the time period t according to the coal bunker coaling scheme and the coal mill starting / stopping scheme calculated by the variable neighborhood search algorithm. The production and power generation operation efficiency is improved, the operation cost is reduced, and the optimal economic benefit is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of blended coal combustion, and particularly to an optimization method for precise coal feeding and blended coal combustion in coal-fired thermal power generating units. Background Art

[0002] Blended coal combustion in coal-fired power plants refers to the practice of mixing different types of coal during the power generation process to optimize the fuel characteristics and improve power generation efficiency. The background and advantages of this practice are mainly reflected in the following aspects:

[0003] (1) Resource optimization: With the gradual depletion of coal resources, the use of a single coal type may lead to inefficient resource utilization. Through blended coal combustion, the characteristics of different coal types can be fully utilized to maximize energy utilization efficiency.

[0004] (2) Cost reduction: The market prices and calorific values of different coal types vary. Through reasonable proportioning, the overall fuel cost can be reduced, and economic benefits can be improved.

[0005] (3) Environmental protection requirements: The carbon emission standards for power plants are becoming increasingly strict. Reasonable blending of different coal types helps reduce carbon dioxide and harmful gas emissions, meeting environmental protection requirements.

[0006] (4) Improving combustion efficiency: Different coal types have different combustion characteristics. Through scientific proportioning, the combustion process can be optimized, thermal efficiency can be improved, and energy losses caused by incomplete combustion can be reduced.

[0007] By burning different coal types in admixture, power plants can, while meeting power demands, reduce power generation costs and carbon emissions, achieving low-carbon and high-efficiency production operations. Generally, vertical coal-fired boilers have multiple layers of burners, multiple coal mills, and multiple coal bunkers, and there is a one-to-one correspondence between the coal bunkers, coal mills, and burners. Therefore, there are two categories of decision-making factors that affect the real-time coal types and proportions in the boiler and the boiler combustion efficiency: The first category is the types of coal stored in the coal bunkers connected to each burner and the coal mill switch selection scheme; the second category is the adjustment and control of the coal injection rate and operating parameters (such as oxygen excess, air volume, etc.) of each burner in the boiler.

[0008] Currently, power plants formulate coal feeding plans and coal mill switch control plans solely based on historical experience. The results of coal feeding and blended coal combustion control are likely to lead to adverse consequences such as high carbon emissions and high coal consumption for power generation. At the same time, existing related technologies have not considered that the coal types to be burned for improving power generation efficiency should change dynamically under different loads, and the impacts of the two types of decisions, namely the coal types stored in each coal bunker and the coal mills that are turned on and off, on the boiler combustion efficiency.

[0009] Therefore, there is an urgent need for a more scientific, accurate, reliable and comprehensive method for controlling the coal feeding of coal bunkers and the switching of coal mills in coal-fired thermal power units to guide power plants in carrying out coal blending operations, improve the operation efficiency of power generation, reduce operating costs, and achieve the optimal economic benefits. Summary of the Invention

[0010] Based on the technical problems existing in the background technology, the present invention proposes an optimized method for precise coal feeding and blending in coal-fired power generation units, which improves the operation efficiency of power generation, reduces operating costs, and achieves the optimal economic benefits.

[0011] The optimized method for precise coal feeding and blending in coal-fired power generation units proposed by the present invention comprises the following method steps:

[0012] S1: Obtain the historical combustion data of each unit in the thermal power plant and establish a combustion sample database;

[0013] S2: Based on the combustion sample data, establish a prediction model g(·) for the theoretical coal feeding amount w l,k,i of the coal mill as

[0014] S3: Based on the combustion sample data, establish a prediction model f(·) for the thermal efficiency e l of unit power generation as

[0015] S4: Establish the objective function of the joint decision-making mathematical model for coal bunker - mine point coal distribution and coal mill on / off, and set the constraint conditions;

[0016] S5: Establish a variable neighborhood search algorithm for calculating the coal feeding plan of the coal bunker and the coal mill opening and closing plan under the condition of minimizing the coal consumption for power generation;

[0017] S6: Carry out coal feeding and blending operations within the time period t according to the coal feeding plan of the coal bunker and the coal mill opening / closing plan calculated by the variable neighborhood search algorithm;

[0018] Among them, H is a continuous decision variable, representing the average coal consumption for power generation within the time period t. w l,k,j is a continuous decision variable, representing the coal feeding amount required to meet the power generation demand when the coal mill i grinds the coal type k at the l-th hour within the time period t; is a 0-1 decision variable, which takes the value of 1 when the coal mill j is turned on within the l-th hour of the time period t, otherwise, it takes the value of 0; y t,k,j is a 0-1 decision variable, which takes the value of 1 when the mine point coal k is stored in the coal bunker j within the time period t, otherwise, it takes the value of 0; e l is a continuous decision variable, representing the thermal utilization rate of the coal feeding plan within the l-th hour of the time period t; the parameter p lFactors affecting the power generation efficiency in the l-th hour within the time period t; H represents the known coal consumption for power generation in the l-th hour.

[0019] Preferably, the historical combustion data in S1 includes the following data for each unit on a daily basis in hours: coal inventory data, coal property data stored in each coal bunker, the started and stopped coal mills, the coal feeding amount of each coal mill, the power generation of the unit, the load distribution data, whether heating is provided, external environment data, and coal consumption for power generation.

[0020] Preferably, the coal inventory data includes: the mining site of the coal type available for combustion in the coal yard, coal characteristics, inventory, sulfur content, moisture content, volatile matter, ash content, and low calorific value data.

[0021] Preferably, the coal property data includes: coal mining site, coal characteristics, sulfur content, moisture content, volatile matter, ash content, and low calorific value.

[0022] Preferably, the load distribution data includes: the operation time at 40%-45% load, 45%-50% load, 50%-55% load, 55%-60% load, 60%-65% load, 65%-70% load, 70%-75% load, 75%-80% load, 80%-85% load, 85%-90% load, 90%-95% load, 95%-100% load, the minimum load, and the maximum load.

[0023] Preferably, the external environment data includes: the highest temperature, the lowest temperature, weather type, and month.

[0024] Preferably, the prediction model establishment method for the theoretical coal feeding amount w of the coal mill in S2 l,k,i is as follows: Taking the coal feeding amount of the coal mill as the target and using the power generation, whether each coal mill is enabled, the sulfur content, moisture content, volatile matter, ash content, low calorific value of the coal stored in each coal bunker, the load distribution data, whether heating is provided, and the external environment data as features, an XGB regression model for predicting the theoretical coal feeding amount of the coal mill is established.

[0025] Preferably, the prediction model establishment method for the thermal efficiency e of unit power generation in S3 l is as follows: Taking the thermal efficiency as the target and using the power generation, the coal feeding amount of each coal mill, the sulfur content, moisture content, volatile matter, ash content, low calorific value of the coal stored in each coal bunker, the load distribution data, whether heating is provided, and the external environment data as features, an XGB regression prediction model for predicting the thermal efficiency of unit power generation is established.

[0026] Preferably, the constraint conditions set in S4 include:

[0027]

[0028]

[0029] Among them, z t,k,j is a continuous decision variable, representing the quantity of coal k from mining site stored in coal bunker j during time period t; parameter kwh l represents the predicted power generation in the l-th hour during time period t; parameter η t,k indicates that coal k from the mining site must be used during time period t; parameter Q t,k and h t,k and s t,k and v t,k respectively represent the inventory, lower calorific value, sulfur content, and volatile content of coal k from the mining site during time period t; parameter h lb and h ub and s ub and v lb and v ub respectively represent the lower limit of the weighted average calorific value, the upper limit of the weighted average calorific value, the upper limit of the weighted average sulfur content, the lower limit of the weighted average volatile content, and the upper limit of the weighted average volatile content of the coal in the co-firing scheme.

[0030] Preferably, the method steps of the variable neighborhood search algorithm in S5 are as follows:

[0031] S51: Generate an initial solution

[0032] S511: Based on the load demand data during time period t, screen out the data sample with the closest load data from the combustion sample database

[0033] S512: Based on the available coal inventory data during time period t, select a coal k from the coal inventory with the closest coal quality to k', and replace the coal type k' in y' t,k',j ;

[0034] S513: Obtain the initial solution Calculate and for S0 and calculate H0;

[0035] S52: Set the termination condition of the variable neighborhood search algorithm as the maximum running time T, and execute:

[0036] S521: Perform neighborhood search on S0 to obtain a neighborhood solution that does not violate the constraint conditions

[0037] S522: Use the model g(·) to predict w' for S ’ ; t,k,j ;

[0038] S523: Predict e' using the model f(·). t , calculate H'.

[0039] S524: If H' < H0, then S0 = S', H0 = H', Return to step S521.

[0040] S53: When the running time of the variable neighborhood search algorithm exceeds T, then S * = S0, H * = H0, Terminate the algorithm.

[0041] Advantageous technical effects of the present invention:

[0042] The present invention studies the coal feeding of the coal bunker and the decision-making problem of the coal mill switch of a coal-fired thermal power generation unit, and proposes a precise coal feeding and blending control method that can reduce the power generation cost and carbon emissions. Applying the present invention can reduce the coal consumption for power generation during the production and operation of a thermal power plant to meet the power generation demand, reduce the coal consumption, reduce the coal usage cost and carbon emissions, and promote the low-carbon and efficient development of the power plant.

[0043] In the existing technical solutions of thermal power plants and the solutions proposed in existing research (such as CN111062534B, CN109404955B, and CN105117808B), either the coal types stored in each coal bunker do not change for a long time, or only a single factor (such as the carbon content in fly ash, etc.) that affects the power generation thermal efficiency or coal consumption for power generation is considered, ignoring that the coal consumption for power generation is affected by multiple factors including the coal types for blending and the coal mills that are turned on. Compared with them, the present invention fully considers that the coal types to be blended for the optimal power generation efficiency (i.e., the lowest coal consumption for power generation) should change dynamically under different loads, and the influence of the coal types stored in each coal bunker and the two types of decisions of the coal mills that are turned on and off on the coal consumption for power generation, and proposes a more advanced and comprehensive technical solution. Description of the Drawings

[0044] Figure 1 is a flowchart of the precise coal feeding and blending optimization method for a coal-fired thermal power generation unit proposed by the present invention;

[0045] Figure 2 is a logic block diagram of the variable neighborhood search algorithm proposed by the present invention. Detailed Embodiments

[0046] The present invention will be further explained below in conjunction with specific embodiments.

[0047] Referring to Figure 1 , a precise coal feeding and blending optimization method for a coal-fired thermal power generation unit proposed by the present invention includes the following steps:

[0048] S1: Obtain the historical combustion data of each unit in a thermal power plant and establish a combustion sample database. The historical combustion data includes the following data for each unit on a daily basis in hours: coal inventory data, coal property data stored in each coal bunker, the coal mills started and stopped, the coal feed rate of each coal mill, the unit power generation, the load distribution data, whether heating is provided, external environment data, and the coal consumption for power generation.

[0049] Among them, the coal inventory data includes: the mine site, coal characteristics, inventory, sulfur content, moisture content, volatile matter, ash content, and low calorific value data of the coal types available for combustion in the coal yard.

[0050] The coal property data includes: the coal mine site, coal characteristics, sulfur content, moisture content, volatile matter, ash content, and low calorific value.

[0051] The load distribution data includes: the operating hours (h) at 40%-45% load, 45%-50% load, 50%-55% load, 55%-60% load, 60%-65% load, 65%-70% load, 70%-75% load, 75%-80% load, 80%-85% load, 85%-90% load, 90%-95% load, 95%-100% load, the minimum load, and the maximum load.

[0052] The external environment data includes: the maximum temperature, minimum temperature, weather type, and month.

[0053] Establishing the combustion sample database includes: calculating the thermal efficiency of unit power generation per hour according to formula (1) based on the coal consumption for power generation, and finally establishing a combustion sample database with the thermal efficiency, coal consumption for power generation, and coal feed rate of the coal mill as the targets and other historical data as the factors.

[0054]

[0055] Among them, e l represents the thermal efficiency in the l-th hour, and H represents the known coal consumption for power generation in the l-th hour.

[0056] S2: Based on the combustion sample data, establish a prediction model g(·) for the theoretical coal feed rate w l,k,i of the coal mill. Among them, w t,k,i represents the coal feed rate that should be given to meet the power generation demand when the coal mill i grinds the coal type k in the l-th hour. g(·) represents the prediction model function.

[0057]

[0058] Among them, wl,k,j is a continuous decision variable, representing the coal feeding amount that should be given to meet the power generation demand when the coal mill i grinds coal type k in the l-th hour; is a 0-1 decision variable, which takes the value of 1 when the coal mill j is started within the l-th hour of the time period t, and 0 otherwise; y t,k,j is a 0-1 decision variable, which takes the value of 1 when the coal bunker j stores the coal from the mining point k during the time period t, and 0 otherwise; the parameter p l represents the factors affecting the power generation efficiency in the l-th hour of the time period t.

[0059] The method for establishing the prediction model is as follows: Taking the coal feeding amount of the coal mill as the target, and using the power generation amount, whether each coal mill is enabled, the sulfur content, moisture content, volatile content, ash content, low calorific value, load distribution data, whether it is for heating, and external environment data of the coal stored in each coal bunker as features, an XGB regression model for predicting the theoretical coal feeding amount of the coal mill is established.

[0060] S3: Based on the combustion sample data, establish the prediction model f(·) of the thermal efficiency e of the unit power generation. l of the thermal efficiency e of the unit power generation. Among them, e l represents the thermal utilization rate of the coal combustion plan in the l-th hour of the time period t, and f(·) represents the prediction model function.

[0061]

[0062] Among them, e l represents the thermal utilization rate of the coal combustion plan in the l-th hour of the time period t.

[0063] The method for establishing the prediction model is as follows: Taking the thermal efficiency as the target, and using the power generation amount, the coal feeding amount of each coal mill, the sulfur content, moisture content, volatile content, ash content, low calorific value, load distribution data, whether it is for heating, and external environment data of the coal stored in each coal bunker as features, an XGB regression prediction model for predicting the thermal efficiency of the unit power generation is established.

[0064] S4: Based on the limitations existing in the actual operation of the power plant, establish a joint decision-making mathematical model for the coal bunker - mining point coal allocation and the opening / closing of the coal mill.

[0065] The objective function of the mathematical model is:

[0066]

[0067] The constraint conditions are as follows:

[0068]

[0069]

[0070] Among them, H represents the known coal consumption for power generation in the l-th hour; zt,k,j Denotes the quantity of the coal k from the mining site stored in the coal bunker j within the time period t; kwh l Denotes the predicted power generation in the l-th hour within the time period t; η t,k Denotes that the coal k from the mining site must be used within the time period t; Q t,k and h t,k and s t,k and v t,k Respectively denote the inventory, lower calorific value, sulfur content, and volatile content of the coal k from the mining site within the time period t; h lb and h ub and s ub and v lb and v ub Respectively denote the lower limit of the weighted average calorific value, upper limit of the weighted average calorific value, upper limit of the weighted average sulfur content, lower limit of the weighted average volatile content, and upper limit of the weighted average volatile content of the coal in the co-firing scheme.

[0071] S5: Establish a variable neighborhood search algorithm for calculating the coal bunker coal feeding scheme and the coal mill start-stop scheme under the condition of minimizing the power generation coal consumption. The algorithm structure is as Figure 2 shown and includes the following content:

[0072] S5.1: Generate an initial feasible solution, which includes the following content:

[0073] (a) Based on the load demand data within the time period t, screen out the data sample with the closest load data from the combustion sample database

[0074] (b) Based on the available coal inventory data within the time period t, select a coal k from the mining site with the closest coal quality to k' from the coal inventory and replace the coal type k' in y' t,k',j as follows:

[0075] (c) Obtain the initial solution Calculate for S0 and Calculate H0.

[0076] S5.2: Set the algorithm termination condition as the maximum running time of 3 minutes. Within 3 minutes, loop through the following steps:

[0077] (a) Perform a neighborhood search on S0 to obtain a neighborhood solution that does not violate Constraints (3) to (14)

[0078] (b) Use the model g(·) to predict and obtain w' for S ’ ; t,k,j ;

[0079] (c) Use the model f(·) to predict and obtain e' t, calculate H';

[0080] (d) If H' < H0, then S0 = S', H0 = H', return to step (a).

[0081] S5.3: When the algorithm running time exceeds 3 minutes, then S * = S0, H * = H0, terminate the algorithm.

[0082] S6: The coal-fired power plant conducts coal feeding and blending operations within the time period t according to the coal bunker coal feeding plan and coal mill start / stop plan calculated by the algorithm. Finally, it can greatly reduce the coal consumption and coal usage cost and reduce carbon emissions while meeting the power generation demand.

[0083] Application Example

[0084] Taking the optimization solution of a 630MW unit of a certain power plant on April 10, 2023 as an example, the specific implementation manner of the present invention will be described:

[0085] S1, S2, S3, S4, and S5 have been completed according to the above content.

[0086] Collect the available coal inventory data, load data, etc. of the power plant on April 10. The available coal inventory of the power plant on this day is as follows:

[0087] Mine site Coal type Coal quantity (tons) Net calorific value Sulfur content Volatile matter Moisture Mine site 1 High heat 30250 4585 0.0034 0.4115 0.0657 Mine site 2 High heat 2200 4919 0.0085 0.4179 0.05 Mine site 3 High heat 6600 5085 0.0053 0.3107 0.1146 Mine site 4 High heat 770 4556 0.0093 0.3974 0.042 Mine site 5 Low heat 10120 4210 0.0039 0.4205 0.0505 Mine site 6 Low volatile matter 6050 4474 0.0031 0.1937 0.0894 Mine site 7 Low volatile matter 1210 5301 0.0027 0.1041 0.0796 Mine site 8 Low ash fusion temperature 3520 5972 0.0052 0.3772 0.1264 Mine site 9 Low ash fusion temperature 1980 5382 0.0045 0.3814 0.1354 Mine site 10 Low ash fusion temperature 4400 5441 0.003 0.3315 0.1244 Mine site 11 Low ash fusion temperature 6820 5664 0.0044 0.323 0.122 Mine site 12 High sulfur 3630 4442 0.0085 0.3802 0.1071 Mine site 13 High sulfur 1430 5130 0.0126 0.1665 0.0708

[0088] Divide the 24 hours of this day into 6 time periods, and each time period contains 4 consecutive hours. The specific usage process is as follows:

[0089]

[0090]

[0091] The parameter values in the above constraints are h lb = 4300, h ub = 6000, s ub = 0.011, v lb = 0.1, v ub = 0.5. Calculate using the above algorithm based on the data, and compare the algorithm results with the results of the power plant using its own plan as follows:

[0092]

[0093] It is known that the actual coal consumption for power generation of this power plant on April 10, 2023 was 290.99 g / kWh, the actual coal consumption cost was 4.3949 million yuan, and the carbon emission cost was 0.5575 million yuan. Compared with the results of the free plan of the thermal power plant on that day, the average coal consumption for power generation of the solution of the present invention can be reduced by 1.47 g / kWh, the coal consumption can be saved by 146 tons, which is equivalent to about 24,000 yuan of coal cost, and the carbon emission cost is reduced by 20,000 yuan.

Claims

1. A method for optimizing precise coal blending and combustion of coal-fired power generation units, characterized in that: The steps are as follows: S1: Obtain the historical combustion data of each unit in the thermal power plant and establish a combustion sample database; S2: Based on the combustion sample data, establish the theoretical coal feed rate w of the coal mill l,k,i The prediction model g(·) is S3: Based on the combustion sample data, establish the thermal efficiency e of the unit l The prediction model f(·) is S4: Establish the objective function of the mathematical model for the joint decision-making of coal bunker-mine point coal allocation and coal mill on / off And set constraints; S5: Establish a variable neighborhood search algorithm for calculating the coal bunker loading scheme and coal mill start-up and shutdown scheme under the condition of minimizing the coal consumption of power generation; S6: Carry out coal loading and blending operation within time period t according to the coal bunker loading plan and coal mill start / shutdown plan calculated by the variable neighborhood search algorithm; Among them, H is a continuous decision variable, which represents the average coal consumption for power generation in time period t; w l,k,j is a continuous decision variable, which indicates the amount of coal that should be fed to meet the power generation demand when coal mill i is grinding coal type k in the first hour; is a 0-1 decision variable. When coal mill j is turned on within the lth hour of time period t, it takes the value of 1, otherwise, it takes the value of 0; t,k,j is a 0-1 decision variable. When the coal k from the mine is stored in the coal bunker j during the period t, the value is 1, otherwise, the value is 0; e l is a continuous decision variable, indicating the thermal utilization rate of the coal-fired scheme in the first hour of time period t; parameter p l It represents the factors that affect the power generation efficiency in the first hour within the time period t; H represents the known coal consumption for power generation in the first hour.

2. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 1 is characterized in that: The historical combustion data in S1 includes the following data for each unit in units of one hour every day: coal inventory data, coal property data stored in each coal bin, coal mills that are turned on and off, coal feed rate for each coal mill, unit power generation, load distribution data, whether heating is provided, external environment data, and coal consumption for power generation.

3. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 2 is characterized in that: Coal inventory data include: mining locations, coal characteristics, inventory quantity, sulfur content, moisture content, volatile matter, ash content and low calorific value of coal types available for combustion in the coal yard.

4. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 2 is characterized in that: Coal attribute data include: coal mine location, coal characteristics, sulfur content, moisture, volatile matter, ash content and lower calorific value.

5. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 2 is characterized in that: The load distribution data includes: 40%-45% load operation time, 45%-50% load operation time, 50%-55% load operation time, 55%-60% load operation time, 60%-65% load operation time, 65%-70% load operation time, 70%-75% load operation time, 75%-80% load operation time, 80%-85% load operation time, 85%-90% load operation time, 90%-95% load operation time, 95%-100% load operation time, minimum load and maximum load.

6. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 2 is characterized in that: External environment data includes: maximum temperature, minimum temperature, weather type and month.

7. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 1 is characterized in that: Theoretical coal feed rate of S2 coal mill w l,k,i The method for establishing the prediction model is as follows: taking the coal feed rate of the coal mill as the target, and taking the power generation, whether each coal mill is enabled, the sulfur content, moisture, volatile matter, ash content, low calorific value of the coal stored in each coal bin, load distribution data, whether heating is provided, and external environmental data as features, an XGB regression prediction model for predicting the theoretical coal feed rate of the coal mill is established.

8. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 1, characterized in that: Thermal efficiency of power generation in S3 l The method for establishing the prediction model is as follows: taking thermal efficiency as the goal, and taking power generation, coal feed amount of each pulverizer, sulfur content, moisture, volatile matter, ash content, low calorific value of coal stored in each coal bin, load distribution data, whether heating is provided, and external environmental data as features, an XGB regression prediction model for predicting the thermal efficiency of the unit is established.

9. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 1, characterized in that: The constraints set in S4 include: Among them, z t,k,j is a continuous decision variable, indicating the amount of coal k stored in coal bunker j in time period t; parameter kwh l Indicates the expected power generation in the first hour of time period t; parameter η t,k Indicates that coal k at the mine must be used within time period t; parameter Q t,k 、h t,k 、s t,k 、v t,k They represent the inventory, lower heating value, sulfur content, and volatile matter of coal k at the mine site in time period t; parameter h lb 、h ub 、s ub 、v lb 、v ub They respectively represent the lower limit of the weighted average calorific value of coal in the blending plan, the upper limit of the weighted average calorific value, the upper limit of the weighted average sulfur content, the lower limit of the weighted average volatile matter, and the upper limit of the weighted average volatile matter.

10. The method for optimizing accurate coal-loading and blending of coal-fired power generation units according to claim 1, characterized in that: The method steps of the variable neighborhood search algorithm in S5 are as follows: S51: Generate initial solution S511: Based on the load demand data in period t, select the data sample closest to the load data from the combustion sample database. S512: Based on the available coal inventory data in period t, select a coal with the same quality and k from the coal inventory. ′ The closest mine point coal k, to y t ′ ,k′,j Type of coal in k ′ Make a replacement; S513: Get initial solution Calculate S0 and Calculate H0; S52: Set the termination condition of the variable neighborhood search algorithm to a maximum running time of T, and execute: S521: Perform a neighborhood search on S0 to obtain a neighborhood solution that does not violate the constraints S522: To S ’ Use model g(·) to predict w t ′ ,k,j ; S523: Use model f(·) to predict e t ′ , calculate H ′ ; S524: If H ′ <H0, then S0 = S ′ , H0 = H ′ , Return to step S521; S53: When the running time of the variable neighborhood search algorithm exceeds T, S * =S0,H * =H0, Terminate the algorithm.

Images