CN112836429A - Multi-objective optimization coal blending method based on coal quality prediction - Google Patents

Multi-objective optimization coal blending method based on coal quality prediction Download PDF

Info

Publication number
CN112836429A
CN112836429A CN202110112336.7A CN202110112336A CN112836429A CN 112836429 A CN112836429 A CN 112836429A CN 202110112336 A CN202110112336 A CN 202110112336A CN 112836429 A CN112836429 A CN 112836429A
Authority
CN
China
Prior art keywords
coal
blending
mixed
quality prediction
objective optimization
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110112336.7A
Other languages
Chinese (zh)
Inventor
李晓辰
胡涛
陈思勤
张豪庆
王文华
江辉
林义杰
史聪
茅大钧
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Electric Power University
Shanghai Shidongkou Second Power Plant of Huaneng Power International Inc
Original Assignee
Shanghai Electric Power University
Shanghai Shidongkou Second Power Plant of Huaneng Power International Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Electric Power University, Shanghai Shidongkou Second Power Plant of Huaneng Power International Inc filed Critical Shanghai Electric Power University
Priority to CN202110112336.7A priority Critical patent/CN112836429A/en
Publication of CN112836429A publication Critical patent/CN112836429A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]

Abstract

The invention relates to a multi-objective optimization coal blending method based on coal quality prediction, which comprises the following steps: 1) constructing a coal quality prediction model of the mixed coal, and determining the relationship between different component indexes and mixing proportions of each single coal and the mixed coal according to the model; 2) and constructing a multi-objective optimization coal blending model, and solving by adopting an NSGA-III algorithm to finally obtain an optimized coal blending scheme. Compared with the prior art, the method has the advantages of high prediction precision of the coal quality of the mixed coal, more accuracy and reference value of the coal blending scheme, increased flexibility of the coal blending method and the like.

Description

Multi-objective optimization coal blending method based on coal quality prediction
Technical Field
The invention relates to the technical field of coal blending of power plants, in particular to a multi-objective optimization coal blending method based on coal quality prediction.
Background
In recent years, due to the problems of coal production and distribution, and seasonal influences and other factors, the spear for supplying and demanding electric coal is prominent, and periodic 'coal scarcity' occurs. The coal quality of a plurality of coal-fired thermal power plants deviates from the designed coal quality and is unstable, so that the generating efficiency of the unit is low, the coal consumption is high, the pollutant emission exceeds the standard, and the like, and the safety, the environmental protection property and the economical efficiency of the unit operation are influenced. In order to adapt to the continuous change of coal markets, more and more coal-fired thermal power plants begin to mix some non-designed coal types.
In order to solve the situation that the coal supply of the coal fired in the thermal power plant is complicated and changeable, theoretical research on the mixed coal combustion aspect is carried out at home and abroad, certain achievements are obtained, and the coal blending combustion method mainly aiming at practical application is well popularized in engineering. The traditional power coal blending can be abstracted into a mathematical programming problem under certain constraint conditions and objective functions, generally, the coal quality and the combustion characteristic index of a boiler are taken as constraints, the lowest price of mixed coal or the minimum emission requirement of pollutants is taken as a target, and one or more program algorithms are used for solving an optimal coal blending scheme within a specified feasible region. Although the coal blending theory method is continuously developed, the problems that the additivity of the coal blending quality is not unified in evaluation, a single-target dynamic coal blending model has defects, a multi-target optimization algorithm is high in randomness and the like still exist, and therefore many existing coal blending methods lack reliability and application value.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a multi-objective optimization coal blending method based on coal quality prediction.
The purpose of the invention can be realized by the following technical scheme:
a multi-objective optimization coal blending method based on coal quality prediction comprises the following steps:
1) constructing a coal quality prediction model of the mixed coal, and determining the relationship between different component indexes and mixing proportions of each single coal and the mixed coal according to the model;
2) and constructing a multi-objective optimization coal blending model, and solving by adopting an NSGA-III algorithm to finally obtain an optimized coal blending scheme.
In the step 1), the indexes of the components of the single coal comprise calorific value, volatile matters, ash content, moisture content, sulfur content and ash fusion point.
In the step 1), the single coal and the mixed coal are in a linear relation with respect to moisture, sulfur and calorific value; for volatile components, ash content and ash fusion points, a nonlinear relation is formed between single coal and mixed coal, each component of the single coal and a corresponding index mixing proportion are used as input, component indexes of the mixed coal are used as output, and a PSO-BP neural network is adopted for prediction.
In the step 2), the targets of the multi-target optimization coal blending model comprise an economic target, a safety target and an environmental protection target.
The expression of the economic target is as follows:
Figure BDA0002919621280000021
wherein, alpha and beta are economic weight coefficients, XiIs the blending proportion of the ith single coal, PiIs the price of the ith single coal, n is the number of the single coal in the coal bunker, PCFor actual market reference coal price, QPThe predicted value Q of the calorific value of the mixed coal obtained by the coal quality prediction model of the mixed coaladThe calorific value of the designed mixed coal is obtained.
The expression of the security target is as follows:
Figure BDA0002919621280000022
wherein, gamma and delta are safety weight coefficients, QP、VP、SP、MP、STPRespectively obtaining the predicted values of the calorific value, the volatile component, the sulfur component, the moisture content and the ash fusion point of the mixed coal obtained by the coal quality prediction model of the mixed coal, Qad、VadFor the designed calorific value and volatility value of the mixed coal, Smin、Smax、Mmin、Mmax、STmin、STmaxThe minimum value and the maximum value of the sulfur content, the moisture content and the ash fusion point corresponding to each single coal are respectively.
The expression of the environmental protection target is as follows:
Figure BDA0002919621280000023
wherein epsilon and epsilon are environmental protection weight coefficients, APIs a mixed coal ash content predicted value obtained by a mixed coal quality prediction model, Amni、AmaxThe minimum and maximum moisture values for each individual coal.
In the NSGA-III algorithm, coal type codes and corresponding mixing ratios of single coals are used as population individuals, and after population initialization, selection, recombination, variation, mixing and non-dominated sorting selection operations are performed, reference point setting, population standardization, association operation and individual reservation operation are performed to obtain the mixing ratios meeting target requirements, namely a coal blending scheme.
In order to adapt to a coal blending scheme, real number encoding is carried out on population individuals, for the condition of blending m kinds of single coal, in the first 2m positions of the population individual codes, the first m positions are coal codes of random m kinds of single coal in a coal warehouse, the last m positions are corresponding blending proportions, the value range of the coal codes of the single coal is defined as [1, n ] and the value range of the corresponding blending proportions is [10,80] in consideration of practical significance, wherein n is the number of the single coal in the coal warehouse.
In order to adapt to the coal blending scheme, when the number of the individual related to the edge reference point is calculated, a number larger than 2 is given, so that the number is sequentially used for increasing the convergence of the NSGA-III algorithm and reducing meaningless results after the number is selected.
Compared with the prior art, the invention has the following advantages:
according to the method, the linear weighting and PSO-BP neural network are respectively adopted to predict the characteristics of linearity and nonlinearity of different coal quality components of the mixed coal, so that the result accuracy is improved.
Secondly, multiple targets of economy, safety and environmental protection are simultaneously established by the model, the optimized coal blending scheme has higher accuracy and reference value, the weight coefficients of different targets are properly adjusted, and the flexibility of the coal blending method can be improved.
And thirdly, multiple results are usually obtained through multi-objective optimization, and in order to determine the most appropriate coal blending scheme under different conditions, a decision-making method is adopted to realize automatic coal blending.
Drawings
FIG. 1 is a diagram of operations performed by an individual in association with a reference point.
FIG. 2 is a flow chart of the NSGA-III algorithm.
Fig. 3 is a graph showing the effect of the prediction of the coal quality of the mixed coal, in which fig. 3a shows the prediction of the volatile matter of the mixed coal, fig. 3b shows the prediction of the ash content of the mixed coal, and fig. 3c shows the prediction of the melting point of the ash content of the mixed coal.
Fig. 4 is an individual solution set distribution.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
Examples
The invention provides a multi-objective optimization coal blending method based on coal quality prediction, which is used for analyzing actual coal blending data of a certain power plant as an example, establishing a coal blending coal quality prediction model according to the relation between each single coal quality and the coal blending quality, and establishing a multi-objective optimization coal blending model by analyzing the influence of coal quality characteristics on the operation economy, safety and environmental protection of a power plant unit.
The principles and steps of the present invention are described below.
1. Mixed coal quality prediction model
The premise of optimizing dynamic coal blending is that a coal quality prediction model of mixed coal is established, and the relation between different component indexes and the mixing ratio of each single coal and the mixed coal is determined.
1.1PSO-BP neural network algorithm
The PSO-BP neural network is a fused algorithm, is applied to a plurality of fields, optimizes weight and threshold values in the BP neural network by utilizing the PSO algorithm, finds out the most appropriate parameters through continuous alternation of particle positions and speeds, and assigns the parameters to the BP neural network to construct a coal quality mixed prediction model.
1.2PSO-BP neural network construction
The coal quality characteristic parameters included in the mixed coal quality prediction model comprise: moisture (M)ad) Sulfur (S)ad) Heat generation amount (Q)ad) Volatile matter (V)ad) Ash content (A)ad) Ash melting point (ST), and water content analysis of coal quality experiment result of power plantThe sulfur content and the calorific value have linear additivity, and the volatile component, the ash content and the ash fusion point have nonlinearity, so that the PSO-BP neural network is used for predicting the nonlinear relation of the coal quality components of the mixed coal.
The neural network is trained by adopting a Levenberg-Marquart algorithm, and the calculation formula of the number b of neurons in a hidden layer of the network is as follows:
Figure BDA0002919621280000041
wherein c is the number of neurons in the input layer; d is the number of neurons in the output layer; e ∈ [1,10] constant. According to a formula and experience, the numbers of hidden layer neurons of volatile components, ash content and ash melting points are respectively 7, 9 and 6.
The particle dimension h in the particle swarm algorithm is calculated according to the formula:
h＝b+c*b+b*d+d (2)
the particle fitness function F is calculated by the following formula:
Figure BDA0002919621280000042
wherein n is the number of samples; y isijA j dimension neural network prediction value is taken as a sample i;
Figure BDA0002919621280000043
is the sample ith dimension neural network actual value.
Determining an individual extreme value and a group extreme value according to the fitness value, then updating the speed V and the position X of the particles, and the calculation formula is as follows:
Figure BDA0002919621280000044
w＝μmin+(μmax-μmin)*k+σ*q (5)
Figure BDA0002919621280000051
in the formula, the superscript t represents the tth generation; w is the velocity inertial weight; c. C1Learning factors for the individual; c. C2Is a social learning factor; r is1，r2A random number from 0 to 1; mu.sminTaking the minimum value of the random inertia weight of 0.4 mumaxTaking the maximum value of the random inertia weight to be 0.9; k is a random number from 0 to 1; σ (standard deviation) represents the degree of deviation between w and its mean, and is typically a number between 0.2 and 0.5; q is a random number in normal distribution, and random inertial weight is adopted, so that the problem that the local search capability in the early stage of iteration and the global search capability in the later stage are insufficient can be solved.
2. Multi-objective optimization coal blending model
The existing single-target coal blending model usually only considers a certain factor and ignores the overall performance of unit operation, so that the reliability of the optimal solution is low. Simple genetic algorithms deal with coal blending problems by using specific penalty functions to deal with constraints, but the determination of the penalty functions is difficult. The NSGA-II algorithm has better performance when solving the double-target optimization problem, but has great limitation when solving the three or more target optimization problems.
2.1 introduction of the NSGA-III Algorithm
Unlike NSGA-II, the NSGA-III algorithm maintains population diversity in a reference point-based manner when selecting the critical layer Pareto solution set.
2.1.1 setting of reference points
In the environment selection process, the NSGA-III determines a mapping relation with population individuals by setting a reference point to ensure diversity of the population, and usually adopts a method of constructing weight by boundary crossing to distribute the reference point on a hyperplane of (M-1) dimension, where the expression of the number P of generated reference points is:
Figure BDA0002919621280000052
where M is the dimension of the target space, i.e., the number of optimization targets, and H is the number of evenly divided fractions per target.
2.1.2 normalization of populations
Selecting a population StThe minimum value of each target range of the medium individuals is used as an ideal point of the current population
Figure BDA0002919621280000053
Figure BDA0002919621280000054
To simplify the algorithm, a translation operation is performed
Figure BDA0002919621280000055
The origin point may be set as an ideal point.
The intercept is usually determined by the extremum point, which in the multi-objective problem is the point with a larger value in only one direction of the objective function and a smaller value in the other directions, the extremum point
Figure BDA0002919621280000056
The calculation formula is as follows:
Figure BDA0002919621280000061
in the formula: l is a unit direction vector of the coordinate axis.
The relationship between the extremum point and the intercept can be calculated from the relationship between the hyperplane and the intercept, and then the normalization of the population can be expressed as:
Figure BDA0002919621280000062
in the formula:
Figure BDA0002919621280000063
aifor the intercept of each dimension, fi(x) To be planted toThe objective function value represented by the population.
2.1.3 Association operations
And performing association operation of the population individuals and the reference points after the reference points and the ideal points are constructed. As shown in fig. 1, a connection line between the ideal point and the reference point is defined as a reference line, the distance from the individual to the reference line is calculated, and the individual is connected with the reference line closest to the ideal point and is associated with the corresponding reference point.
2.1.4 Individual Retention operations
After the individuals are associated with the reference points, the individuals corresponding to the reference points which are less connected should be reserved so as to maintain diversity, and the individual reservation operation is carried out according to the principle until the population size meets the requirement.
The flow chart of the NSGA-III algorithm is shown in fig. 2.
2.2 optimization index selection
The coal blending model comprises coal quality characteristics and combustion characteristic parameters, for the operation of a power plant, the coal blending type and the ratio of coal can be reasonably selected according to the coal quality characteristics, the optimization process before combustion can be regarded as an optimization process, after the coal blending scheme is determined, the combustion process becomes a relatively lagged process, and when the mixed coal of a certain scheme is combusted in a boiler, the combustion characteristics of a unit are judged to be properly adjusted through a real-time state, so that the error influence existing in the coal blending scheme can be timely compensated. Because the combustion characteristics of the unit can be influenced by the difference between the coal quality of the mixed coal and the designed coal type, the invention provides the minimum absolute deviation type index and the standard deviation type index as optimization indexes by combining the characteristic parameters of the coal quality and the price of the coal type.
The NSGA-III algorithm has good effect in processing the multi-objective optimization problem without constraint conditions. The invention uses the relation between the indexes to replace the constraint condition as the optimization index, not only can control the combustion characteristic of the mixed coal through the coal quality index, but also can reduce the algorithm complexity, so the method is feasible.
2.3 objective function design
The invention classifies the indexes affecting the target function, considers the establishment of the target function from the aspects of economy, safety and environmental protection, and the indexes can repeatedly appear in the target function.
2.3.1 economic objectives
The coal blending scheme of the power plant mainly considers the heat productivity and the price of the mixed coal, and achieves the lowest coal price under the condition of ensuring the minimum deviation of the coal quality of the mixed coal and the heat productivity of the designed coal, so that the expression of the economic target is as follows:
Figure BDA0002919621280000071
in the formula, alpha and beta are economic weight coefficients; xiThe blending proportion of the ith coal is shown; piThe price of the ith coal; n is the number of single coals; pCRefer to the coal price for the actual market; qPFor the predicted value of calorific value of coal mixture, QadThe calorific value of the mixed coal is designed.
2.3.2 Security goals
In order to avoid accidents such as fire extinguishment, slagging, pipe explosion and the like caused by boiler combustion, the operation safety of the boiler is very emphasized by a power plant, and the operation safety of the power plant is ensured by controlling the combustion characteristic of the boiler through coal indexes in consideration of the fact that the operation condition of the boiler can be adjusted in the follow-up process, so that the expression of a safety target is as follows:
Figure BDA0002919621280000072
wherein, gamma and delta are safety weight coefficients; qP、VP、SP、MP、STPThe coal quality prediction value of the mixed coal is obtained; qad、VadDesigning the calorific value and the volatile value of the coal; smin、Smax、Mmin、Mmax、STmin、STmaxThe coal quality minimum and maximum values correspond to all the single coals respectively.
2.3.3 environmental protection goals
SO discharged in the coal burning process of the power plant2，NOXAnd soot, wherein NOx control is effected primarily by combustion optimization tuning, andSO2 and smoke are closely related to the coal quality of the fire coal, SO the expression of the environmental protection target is as follows:
Figure BDA0002919621280000073
in the formula: epsilon and epsilon are environmental protection weight coefficients; a. thePThe predicted value of the ash content of the mixed coal is obtained; a. themin、AmaxThe minimum and maximum moisture values correspond to all individual coals.
2.4 improvement of the NSGA-III Algorithm
2.4.1 encoding and parameter setting
In order to adapt to the particularity of the coal blending scheme, the invention adopts a real number coding method, mainly considers the situation of blending three kinds of single coal, only the first six bits are effective aiming at the operation of a genetic operator, wherein the first three bits are numbers of three kinds of random single coal in 10 coal libraries, the last three bits are corresponding blending proportion, and the variable range is defined as follows in consideration of the practical significance: min {1,1,1,10,10,10 }; max {10,10,10,80,80,80 }. And in the algorithm operation process, repeated results are deleted, and a set consisting of a non-dominant sorting level, a dominant individual number and a dominant individual is automatically generated.
2.4.2 Performance optimization of the Algorithm
The genetic crossover operator adopts a multipoint crossover mode, and randomly selects parent genes and offspring genes with discontinuous individual positions to carry out crossover operation, so that the diversity of the population can be greatly improved. It should be noted that the sum of the operators representing the blending ratio must be 100, so that normalization processing is required after the crossover operation is completed.
The NSGA-III algorithm estimates the density of individuals through evenly distributed reference points, in order to expect that one reference point corresponds to one individual, and to achieve even distribution of the final solution set. In order to pursue the diversity of the population, the algorithm selects the reference points containing the fewest individuals in turn to perform individual reservation operation, so that the individuals corresponding to the edge reference points are reserved preferentially, but the solution sets are extreme values on the objective function and cannot be used in practical problems at all, so that the invention provides a number larger than 2 when calculating the number of the associated individuals of the edge reference points, and the number is sequentially arranged at the back when selecting in order to increase the convergence of the algorithm and reduce meaningless results.
Although the invention has standardized operation, in order to balance the influence among the indexes, economic, safety and environmental protection weight coefficients are set, and the preference degree of the objective function is adjusted by giving a proper value to the weight coefficients, so that the flexibility of the algorithm in the actual application can be greatly increased.
3. Example analysis
The invention carries out power coal blending research on a certain power plant boiler in Shanghai, the power plant boiler designs coal types and coal quality conditions as shown in table 1, and selects actual 10 coal storage data for blending as shown in table 2.
Table 1 designs the coal quality of coal
Figure BDA0002919621280000081
TABLE 2 coal database
Figure BDA0002919621280000082
According to the data in the table, 3 kinds of single coal with different proportions are selected for mixing, and 360 groups of sample data can be obtained. And (3) establishing a PSO-BP neural network for nonlinear coal quality indexes of volatile components, ash content and ash fusion points of the mixed coal for prediction, taking 6 variables of the ratio of each single coal quality index to the mixed coal as input, taking the mixed coal quality index as output, and establishing a coal quality prediction model, wherein 340 groups are used as a training set, and the remaining 20 groups are used as a verification set. The prediction effect is shown in fig. 3, and the error between the prediction result, the weighting result, and the actual data is shown in table 3.
TABLE 3 error table for coal quality of mixed coal
Index of coal quality Predicted average relative error/%) Weighted average relative error/%)
Volatile fraction/%) 0.802 5.356
Ash content% 3.630 6.871
Ash melting Point/. degree.C 3.724 8.783
The results show that the PSO-BP neural network has a good prediction effect on the coal quality index of the mixed coal, and the accuracy is improved compared with linear weighting. And respectively storing the network weight and the threshold corresponding to the prediction processes of the volatile component, the ash content and the ash fusion point, and providing reference for the subsequent continuous optimization process.
The method adopts a multi-objective optimization coal blending model, selects economy, safety and environmental protection as optimization targets, gives equal weight under the condition of not considering target preference, respectively takes 2, 7, 8, 3, 10 and 1 for alpha, beta, gamma, delta, epsilon and epsilon, and references coal price P in actual marketCTake 920 yuan/t. The population scale is set to be 110, the iteration number is 200, each target is divided into 13 parts, the number of reference points is 105, and the distribution of the individual solution sets after simulation calculation is shown in fig. 4. It can be seen that the individual solution sets are uniformly distributed on the plane where the diagonal of the three-dimensional space is located, and the solutions corresponding to the individual in the center of the space are averaged on three targets and can be used as preference-free optimal solutions.
The coal type number and the corresponding proportion obtained by the optimization algorithm are utilized, the linear indexes are calculated by linear weighting, the nonlinear indexes are predicted under the PSO-BP neural network with well-stored weight values and threshold values, and part of individual solution sets are listed as shown in Table 4.
TABLE 4 partial individual solution set
Figure BDA0002919621280000091
As shown in Table 4, the indexes of the mixed coal are moderate, and the coal price is lower than the market reference coal price, which shows that the economic, safety and environmental protection indexes can be comprehensively considered by the multi-objective optimization coal blending method, and simultaneously, a solution set meeting the requirements is obtained. When lowest coal prices are desired, option 2 may be selected; option 5 may be selected when the highest heating value is desired; option 6 may be selected when a minimum sulfur content is desired. Coal blending personnel can adjust the weight coefficient of the index and select the optimal coal blending scheme according to the real-time condition of the power plant, the complex and changeable condition of coal types is flexibly solved, and the diversity and the practicability of the scheme are increased.

Claims (10)

1. A multi-objective optimization coal blending method based on coal quality prediction is characterized by comprising the following steps:
1) constructing a coal quality prediction model of the mixed coal, and determining the relationship between different component indexes and mixing proportions of each single coal and the mixed coal according to the model;
2) and constructing a multi-objective optimization coal blending model, and solving by adopting an NSGA-III algorithm to finally obtain an optimized coal blending scheme.
2. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 1, wherein in the step 1), the single coal composition indexes comprise calorific value, volatile components, ash content, moisture content, sulfur content and ash fusion point.
3. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 2, wherein in the step 1), the single coal and the mixed coal are in a linear relationship with respect to moisture, sulfur and calorific value; for volatile components, ash content and ash fusion points, a nonlinear relation is formed between single coal and mixed coal, each component of the single coal and a corresponding index mixing proportion are used as input, component indexes of the mixed coal are used as output, and a PSO-BP neural network is adopted for prediction.
4. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 1, wherein in the step 2), the objectives of the multi-objective optimization coal blending model include economic objectives, safety objectives and environmental protection objectives.
5. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 4, wherein the economic objective is expressed as:
Figure FDA0002919621270000011
wherein, alpha and beta are economic weight coefficients, XiIs the blending proportion of the ith single coal, PiIs the price of the ith single coal, n is the number of the single coal in the coal bunker, PCFor actual market reference coal price, QPThe predicted value Q of the calorific value of the mixed coal obtained by the coal quality prediction model of the mixed coaladThe calorific value of the designed mixed coal is obtained.
6. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 4, wherein the expression of the safety objective is as follows:
Figure FDA0002919621270000012
wherein, gamma and delta are safety weight coefficients, QP、VP、SP、MP、STPRespectively from the coal quality of the mixed coalThe predicted values of calorific value, volatile matter, sulfur content, moisture content and ash fusion point of the mixed coal, Qad、VadHeating value and volatile value of designed coal mixture, Smin、Smax、Mmin、Mmax、STmin、STmaxThe minimum value and the maximum value of the sulfur content, the moisture content and the ash fusion point corresponding to each single coal are respectively.
7. The multi-objective optimization coal blending method based on coal quality prediction as claimed in claim 4, wherein the expression of the environmental protection objective is as follows:
Figure FDA0002919621270000021
wherein epsilon and epsilon are environmental protection weight coefficients, APThe ash content prediction value of the mixed coal obtained by the coal quality prediction model of the mixed coal, Amin、AmaxThe minimum and maximum moisture values for each individual coal.
8. The multi-objective optimization coal blending method based on coal quality prediction of claim 4, wherein in the NSGA-III algorithm, the coal type codes and the corresponding blending ratios of the individual coals are used as population individuals, and after population initialization, selection, recombination, mutation, mixing and non-dominated sorting selection operations are performed, reference point setting, population standardization, association operation and individual reservation operation are performed to obtain the blending ratio meeting the objective requirements, namely a coal blending scheme.
9. The multi-objective optimization coal blending method based on coal quality prediction according to claim 1, wherein, in order to adapt to a coal blending scheme, population individuals are subjected to real number encoding, and in the first 2m bits of the population individual encoding, the first m bits are coal type encoding of random m single coals in a coal bunker, and the last m bits are corresponding blending proportion, and in consideration of practical significance, the coal type encoding range of the single coal is defined as [1, n ], and the value range of the corresponding blending proportion is defined as [10,80], wherein n is the number of the single coal in the coal bunker.
10. The method as claimed in claim 1, wherein a number greater than 2 is assigned to the coal blending scheme in calculating the number of the individuals associated with the edge reference point, so that the number is sequentially selected to increase the convergence of the NSGA-III algorithm and reduce the meaningless results.
CN202110112336.7A 2021-01-27 2021-01-27 Multi-objective optimization coal blending method based on coal quality prediction Pending CN112836429A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110112336.7A CN112836429A (en) 2021-01-27 2021-01-27 Multi-objective optimization coal blending method based on coal quality prediction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110112336.7A CN112836429A (en) 2021-01-27 2021-01-27 Multi-objective optimization coal blending method based on coal quality prediction

Publications (1)

Publication Number Publication Date
CN112836429A true CN112836429A (en) 2021-05-25

Family

ID=75931975

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110112336.7A Pending CN112836429A (en) 2021-01-27 2021-01-27 Multi-objective optimization coal blending method based on coal quality prediction

Country Status (1)

Country Link
CN (1) CN112836429A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113591389A (en) * 2021-08-09 2021-11-02 工数科技(广州)有限公司 Coal proportioning determination method for power generation process of thermal power plant
CN113723668A (en) * 2021-08-02 2021-11-30 华院计算技术(上海)股份有限公司 Coal blending method, system, equipment and storage medium based on robust optimization
CN113741359A (en) * 2021-08-09 2021-12-03 华能国际电力股份有限公司上海石洞口第一电厂 Thermal power plant safety monitoring system and method in mixed coal blending combustion mode
CN116307513A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Thermal power plant coal blending scheme optimization method based on improved MOEA/D algorithm
CN116777537A (en) * 2023-05-26 2023-09-19 华能国际电力股份有限公司上海石洞口第二电厂 Method and system for calculating and optimizing electricity-measuring coal cost and blending coal based on coal quality characteristics

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109376945A (en) * 2018-11-13 2019-02-22 华能国际电力股份有限公司上海石洞口第电厂 A kind of coal mixing combustion optimization system based on more coals
CN110232497A (en) * 2019-04-25 2019-09-13 华电国际电力股份有限公司技术服务分公司 A kind of coal mixing combustion intelligent management and system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109376945A (en) * 2018-11-13 2019-02-22 华能国际电力股份有限公司上海石洞口第电厂 A kind of coal mixing combustion optimization system based on more coals
CN110232497A (en) * 2019-04-25 2019-09-13 华电国际电力股份有限公司技术服务分公司 A kind of coal mixing combustion intelligent management and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
夏季: "火电机组配煤掺烧全过程优化技术研究与应用", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 *
夏季等: "基于非支配排序遗传算法的无约束多目标优化配煤模型", 《中国电机工程学报》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113723668A (en) * 2021-08-02 2021-11-30 华院计算技术(上海)股份有限公司 Coal blending method, system, equipment and storage medium based on robust optimization
CN113723668B (en) * 2021-08-02 2022-05-10 华院计算技术(上海)股份有限公司 Coal blending method, system, equipment and storage medium based on robust optimization
CN113591389A (en) * 2021-08-09 2021-11-02 工数科技(广州)有限公司 Coal proportioning determination method for power generation process of thermal power plant
CN113741359A (en) * 2021-08-09 2021-12-03 华能国际电力股份有限公司上海石洞口第一电厂 Thermal power plant safety monitoring system and method in mixed coal blending combustion mode
CN116307513A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Thermal power plant coal blending scheme optimization method based on improved MOEA/D algorithm
CN116307513B (en) * 2023-02-01 2023-12-22 华能国际电力股份有限公司上海石洞口第二电厂 Thermal power plant coal blending scheme optimization method based on improved MOEA/D algorithm
CN116777537A (en) * 2023-05-26 2023-09-19 华能国际电力股份有限公司上海石洞口第二电厂 Method and system for calculating and optimizing electricity-measuring coal cost and blending coal based on coal quality characteristics
CN116777537B (en) * 2023-05-26 2024-04-12 华能国际电力股份有限公司上海石洞口第二电厂 Method and system for calculating and optimizing electricity-measuring coal cost and blending coal based on coal quality characteristics

Similar Documents

Publication Publication Date Title
CN112836429A (en) Multi-objective optimization coal blending method based on coal quality prediction
CN107203687B (en) Multi-target cooperative intelligent optimization control method for desulfurization process of absorption tower
CN106991539B (en) Energy system optimal scheduling method and device
CN108954375B (en) Coal-saving control method for boiler
Li et al. Day-ahead city natural gas load forecasting based on decomposition-fusion technique and diversified ensemble learning model
Lu et al. Scenarios modelling for forecasting day-ahead electricity prices: Case studies in Australia
CN112330179B (en) Fuzzy comprehensive evaluation method for coal blending combustion based on improved entropy weight method
CN110598929A (en) Wind power nonparametric probability interval ultrashort term prediction method
CN114036855A (en) Dynamic coal blending method, system, equipment and storage medium for thermal power plant
CN109190817B (en) Two-layer decision optimization method for coal-fired coupled biomass emission reduction power generation
CN114330869A (en) Multi-energy day-ahead scheduling optimization method for iron and steel enterprise considering multi-level indexes
Wei et al. A novel approach to hybrid dynamic environmental-economic dispatch of multi-energy complementary virtual power plant considering renewable energy generation uncertainty and demand response
CN114117925A (en) Comprehensive energy system planning method considering high-proportion distributed photovoltaic consumption
Hu et al. Hybrid prediction model for the interindustry carbon emissions transfer network based on the grey model and general vector machine
CN116307507A (en) Multi-target particle swarm coal blending method based on superheater wall temperature prediction
Fu et al. A multi-fuel management model for a community-level district heating system under multiple uncertainties
CN114091720A (en) Multi-target coal blending method based on decision maker preference
Mehta et al. Optimization of combined economic emission dispatch problem using artificial bee colony method
CN110728031B (en) Multi-objective optimization method for balancing complex petrochemical process production energy based on ANN modeling
CN109506248B (en) Case reasoning boiler combustion optimization method based on online optimization
Wang et al. Multi-layer encoding genetic algorithm-based granular fuzzy inference for blast furnace gas scheduling
CN117391734B (en) Power generation cost prediction method based on support vector machine regression model
CN111520740B (en) Method for coordinately optimizing operation of multiple porous medium combustors
CN108073784A (en) A kind of method for solving of the optimization of ethylene feedstock scheduling scheme of Kernel-based methods figure
CN111783988B (en) Multi-target gray prediction evolution method and system for solving environmental economic scheduling problem

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20210525

RJ01 Rejection of invention patent application after publication