CN109033489B - Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system - Google Patents

Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system Download PDF

Info

Publication number
CN109033489B
CN109033489B CN201810529748.9A CN201810529748A CN109033489B CN 109033489 B CN109033489 B CN 109033489B CN 201810529748 A CN201810529748 A CN 201810529748A CN 109033489 B CN109033489 B CN 109033489B
Authority
CN
China
Prior art keywords
clarification
fitness
energy consumption
quality
value
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.)
Active
Application number
CN201810529748.9A
Other languages
Chinese (zh)
Other versions
CN109033489A (en
Inventor
杨海东
姜梦一
徐康康
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.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
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 Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201810529748.9A priority Critical patent/CN109033489B/en
Publication of CN109033489A publication Critical patent/CN109033489A/en
Application granted granted Critical
Publication of CN109033489B publication Critical patent/CN109033489B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P40/00—Technologies relating to the processing of minerals
    • Y02P40/50—Glass production, e.g. reusing waste heat during processing or shaping

Landscapes

  • Physics & Mathematics (AREA)
  • Geometry (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Mathematics (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention discloses a horseshoe flame glass kiln energy efficiency optimization method and system based on an improved particle swarm algorithm, wherein the horseshoe flame glass kiln energy efficiency optimization method comprises the following steps of A, collecting production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping quality grades into quantized clarification factors by using the energy consumption model based on clarification quality constraint; step B, the convergence factor delta and the group fitness variance sigma are used2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm; and step C, applying the improved particle swarm algorithm in combination with a multiplier penalty function constraint processing method to an energy consumption model under clarification quality constraint to obtain an energy consumption optimal solution under the condition of ensuring the glass melting quality. And quantitatively analyzing the difficulty degree of bubbles escaping from the clarification area, analyzing key influence factors of energy consumption and quality grade, and obtaining the optimal parameter combination to guide actual production, thereby achieving the purpose of reducing energy consumption while ensuring clarification quality.

Description

Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system
Technical Field
The invention relates to the field of glass kilns, in particular to a horseshoe flame glass kiln energy efficiency optimization method and system based on an improved particle swarm algorithm.
Background
A combustion space of the heat accumulating type horseshoe flame glass melting furnace is provided with a U-shaped flame formed by the rotation of a flame stream and a hot zone formed at the rotation position, and due to the limitation on the length of the flame and the requirement on the rotation power, the furnace is short and wide, the structure is compact, and the structural section view and the flame and airflow directions are shown in figure 1. The glass melting process comprises five stages of silicate generation, glass formation, molten glass clarification, homogenization, cooling and the like. The traditional melting technology is mainly based on manual experience to adjust relevant process parameters, and the mode is not only low in efficiency, but also difficult to adapt to diversified requirements of glass products. The bubble defect of the glass in the melting process seriously affects the product quality and brings serious loss to enterprises.
Disclosure of Invention
The invention aims to provide a horseshoe flame glass kiln energy efficiency optimization method and system based on an improved particle swarm algorithm, and the optimal parameter combination is obtained to guide the actual production, so that the aim of reducing energy consumption while ensuring clarification quality is fulfilled.
In order to achieve the purpose, the invention adopts the following technical scheme:
an energy efficiency optimization method of a horseshoe flame glass kiln based on an improved particle swarm algorithm comprises the following steps:
step A, collecting production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping quality grades into quantized clarification factors by using the energy consumption model based on clarification quality constraint; step B, the convergence factor delta and the group fitness variance sigma are used2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm; and step C, applying the improved particle swarm algorithm in combination with a multiplier penalty function constraint processing method to an energy consumption model under clarification quality constraint to obtain an energy consumption optimal solution under the condition of ensuring the glass melting quality.
Preferably, the energy consumption model building process based on the clarification quality constraint is as follows:
step A1, collecting production data of the horseshoe flame glass kiln, and obtaining a tank furnace thermal efficiency model by using a reverse balance analysis method:
Figure BDA0001676913630000021
wherein,
Figure BDA0001676913630000022
Tarchthe arch crown temperature, alpha is the air excess coefficient, x is the fuel flow, f is the aperture area of the radiation part, phi is the door aperture coefficient of the radiation part, CtThe specific heat of the radiation part is shown as the specific heat,
Figure BDA0001676913630000023
is the low calorific value of the fuel, tfireTemperature for preheating fuel, CfireTo preheat fuel to tfireAverage specific heat in time, taIs the preheating temperature of combustion air, CaPreheating combustion air to taSpecific heat capacity of time, tsmokeFor preheating the air at a temperature VsmokeVolume of compressed air, CsmokeFor preheating compressed air to tsmokeSpecific heat capacity of hour, hejectIs smoke discharge enthalpy value, lambda is heat conductivity coefficient, delta is brick thickness, F is heat dissipation area, L is combustion air amount, hkIs the enthalpy value of the cooling air;
step A2, residence time t of production streamglassAnd the time t required for the bubbles to escapebubbleAs a clarifying factor RF, then
Figure BDA0001676913630000031
Wherein,
Figure BDA0001676913630000032
ρlis the density of the glass melt, pbIs the density of the gas in the bubble, μ is the viscosity coefficient of the bubble, r is the radius of the bubble, labTo the height of the clarification zone, /)bcIs the length of the fining zone, b is the width of the tank furnace, n is the amount of gaseous material,
Figure BDA0001676913630000035
at a temperature of t0Specific heat capacity of the combustion products at that time;
step A3, a tank furnace thermal efficiency model eta (T)archα, x) and a clarification factor RF as objective functions to obtain an energy consumption model based on clarification quality constraints as
Figure BDA0001676913630000033
Preferably, the clarification factor that maps the quality level into a quantization is:
step a4, the quality ratings were five total ratings A, B, C, D and E, each with process requirements for maximum bubble diameter: grade A is that no bubbles are allowed, grade B is that the maximum bubble diameter is not more than 0.1mm, grade C is that the maximum bubble diameter is not more than 0.2mm, grade D is that the maximum bubble diameter is not more than 0.5mm, and grade E is that the maximum bubble diameter is not more than 1.0 mm;
step A5, setting temperature gradient
Figure BDA0001676913630000034
For a constant value k, the quality level is mapped into a quantized clarification factor by the relation between the bubble radius r and the clarification factor RF, resulting in:
Figure BDA0001676913630000041
step a6, the quality rating of step a5 is added to the energy consumption model based on the clarified quality constraints, namely:
Figure BDA0001676913630000042
preferably, the particle swarm optimization is improved by:
step B1, let the number of particles in the particle group be z, in the y iteration, fiIs the fitness of the ith particle, fmDenotes the fitness of the optimal particle, favgIs the average fitness and makes the fitness in the particles larger than favgHas a fitness average value of favg' then the individual optimum fitness and population are determinedThe difference in mean fitness is taken as the convergence factor Δ: Δ ═ fm-favg′|，
Figure BDA0001676913630000043
The convergence factor Δ is used to evaluate the convergence degree of the particle swarm: the smaller the convergence factor Δ, the more prone the particle to premature convergence;
step B2, by group fitness variance σ2Evaluation of the degree of dispersion of the particles in the population:
Figure BDA0001676913630000044
group fitness variance σ2The smaller the size, the more dispersed the particles, otherwise the more concentrated the particles;
step B3, the convergence factor delta and the population fitness variance sigma are calculated2As an evaluation index, the evolutionary states of the population are classified, and the inertia weight value ω is dynamically modified according to different evolutionary states:
when the particle population is fi＞favg', or favg＜fi＜f′avgThen, the inertia weight value ω ═ ω - (ω - ω) is setmin)·e-γWherein ω isminSetting the value to be 0.5, and gamma is an evolution algebra;
when the particle population is fi＜favgAnd σ is2When the inertia weight value is larger than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000051
f of particle populationi＜favgAnd σ is2When the inertia weight value is smaller than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000052
preferably, the obtaining of the energy consumption optimal solution specifically includes:
step C1, setting the energy consumption model under the clarification quality constraint as a d-dimensional target search space, wherein a population exists in the d-dimensional target search space
Figure BDA0001676913630000053
Represents m particles, of which:
Figure BDA0001676913630000054
i-1, 2, where m is a vector point of the ith particle in the d-dimensional solution space,
Figure BDA0001676913630000055
1,2, m is the search speed of the ith particle, and the optimal position searched by the ith particle is pi＝[pi1，pi2，...，pim]I 1, 2.. m, the optimal position searched by the group is pg＝[pi1，pi2，...，pim]，
vi(s+1)＝ωvi(s)+c1ε1(pi-xi(s))+c2ε2(pi-xi(s))，
xi(s+1)＝xi(s)+vi(s)，ε1，ε2Random number between 0 and 1, S is the number of iterations, learning factor
Figure BDA0001676913630000061
Learning factor
Figure BDA0001676913630000062
Step C2, randomly initializing the position and speed of the particles in the population in the solution space, and setting the current position as piSetting the optimal position in the initial group as pg；
Step C3, velocity formula v according to step C1i(s +1) and the positional formula xi(s +1) changing the velocity and position state of the particles;
and step C4, combining the penalty function and the Lagrangian function to construct a multiplier penalty function:
Figure BDA0001676913630000063
where θ is the decision variable, f (θ) is the objective function, hj(θ) ═ 0, j ═ q + 1.., m is the equality constraint;
calculating the fitness of the particles according to the multiplier penalty function;
step C5, according to the current evolution state, carrying out self-adaptive adjustment of the inertia weight value omega, and comparing the current adaptive value with the self optimal solution: if the current solution is superior to the self optimal value, setting the current adaptive value as the self optimal solution;
step C6, the fitness of the particle is compared with the historical optimal position pi＝[pi1，pi2，...，pim]The fitness of the global optimal solution is compared to decide whether the current global optimal solution needs to be updated or not;
step C7, detecting whether the evolution reaches a maximum set value or a preset convergence precision, if so, ending the evolution process to obtain an energy consumption optimal solution under the condition of ensuring the glass melting quality; otherwise, the step C4 is executed.
Preferably, the horseshoe flame glass kiln energy efficiency optimization system based on the improved particle swarm optimization algorithm comprises the following steps:
the energy consumption modeling module is used for acquiring production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping the quality grade into a quantized clarification factor by using the energy consumption model based on the clarification quality constraint; an algorithm optimization module for optimizing the convergence factor delta and the population fitness variance sigma2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm; and the energy consumption optimization module is used for combining the improved particle swarm algorithm with the multiplier penalty function constraint processing method, applying the improved particle swarm algorithm to an energy consumption model under clarification quality constraint, and obtaining an energy consumption optimal solution under the condition of ensuring the glass melting quality.
According to the horseshoe flame glass kiln energy efficiency optimization method based on the improved particle swarm algorithm, the energy consumption of a melting process is researched on the basis of ensuring the quality of a clarification process, an energy consumption optimization model under the clarification quality constraint is established by analyzing the thermal balance of the melting kiln structure and analyzing the movement rule of glass liquid flow in a clarification area and quantitatively analyzing the difficulty degree of bubbles escaping from the clarification area, key influence factors of energy consumption and quality grade are analyzed, the target model is optimized by the improved self-adaptive particle swarm algorithm, the optimal parameter combination is obtained, and the actual production is guided, so that the purpose of reducing the energy consumption while ensuring the clarification quality is achieved.
Drawings
The drawings are further illustrative of the invention and the content of the drawings does not constitute any limitation of the invention.
FIG. 1 is a structural view of a regenerative horseshoe flame glass melting furnace according to one embodiment of the present invention
FIG. 2 is a flowsheet of a glass melting furnace according to one embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further explained by the specific implementation mode in combination with the attached drawings.
Example one
The energy efficiency optimization method of the horseshoe flame glass kiln based on the improved particle swarm optimization comprises the following steps: step A, collecting production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping quality grades into quantized clarification factors by using the energy consumption model based on clarification quality constraint;
step B, the convergence factor delta and the group fitness variance sigma are used2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm; and step C, applying the improved particle swarm algorithm in combination with a multiplier penalty function constraint processing method to an energy consumption model under clarification quality constraint to obtain an energy consumption optimal solution under the condition of ensuring the glass melting quality.
According to the horseshoe flame glass kiln energy efficiency optimization method based on the improved particle swarm algorithm, the energy consumption of a melting process is researched on the basis of ensuring the quality of a clarification process, an energy consumption optimization model under the clarification quality constraint is established by analyzing the thermal balance of the melting kiln structure and analyzing the movement rule of glass liquid flow in a clarification area and quantitatively analyzing the difficulty degree of bubbles escaping from the clarification area, key influence factors of energy consumption and quality grade are analyzed, the target model is optimized by the improved self-adaptive particle swarm algorithm, the optimal parameter combination is obtained, and the actual production is guided, so that the purpose of reducing the energy consumption while ensuring the clarification quality is achieved.
Blister defects are quality defects in the presence of visible gaseous inclusions in the glass mass, with large amounts of CO occurring during batch melting and molten glass formation2And H2O and other gases still partially and incompletely escape from molten glass after the glass forming process, so that bubble quality defects are formed, the quality grade of glass products is reduced, and the mechanical strength of the glass products is influenced.
The bubble quality defect is the main defect of the working area of the tank furnace, and under the ideal state, the bubble defect caused by incomplete discharge in the clarification stage becomes a key factor influencing the quality problem of the current molten glass. As shown in fig. 2, the molten glass forms a circular flow in the melting portion, the strong backflow of the molten glass flowing to the coordination material layer in the circular flow can effectively block dross on the surface of the molten glass, the arrangement of the weir forces the production flow to turn upwards, so that the liquid flow can reach the surface of the molten glass, and then the molten glass is clarified at high temperature in a hot spot area, and after being clarified, the molten glass turns back in layers on the rear side of the weir, and enters the working portion through the liquid flow hole at the bottom, so that high-quality molten glass can be obtained. The main function of the fining area is to eliminate bubble defects, and the quality of the fining process is directly related to the quality of the glass flow. Small bubbles in the glass flow are easy to diffuse into bubbles with larger volume to increase the diameter, the rising speed is higher when the diameter is larger, and the escape time is shorter. When the diameter of the bubbles is less than 10 μm, the surface tension acts to dissolve the microbubbles in the glass stream.
Assuming that the gas inside the bubble is considered as an ideal state, and the internal components and the temperature are uniformly distributed; the shape of the bubble is approximately considered as a sphere; no chemical reaction occurs among the gases of each component in the bubbles; and the bubbles in the glass melt do not affect the heat transfer of the glass stream; the radius of the bubble is a key factor for limiting the speed of the bubble in the rising process, and the radius of the bubble changes with the change of temperature, so the relationship between the radius of the bubble and the temperature is very important.
Preferably, the energy consumption model building process based on the clarification quality constraint is as follows:
step A1, collecting production data of the horseshoe flame glass kiln, and obtaining a tank furnace thermal efficiency model by using a reverse balance analysis method:
Figure BDA0001676913630000091
wherein,
Figure BDA0001676913630000092
Tarchthe crown top temperature, alpha is the air excess coefficient, and x is the fuel flow in kg/h; f is the area of the opening of the radiation part and the unit m2(ii) a Phi is the door aperture coefficient of the radiation part, CtSpecific heat of radiation part, unit kj/(m)3·℃)；
Figure BDA0001676913630000093
The unit kj/kg is the low calorific value of the fuel; t is tfireThe temperature for preheating the fuel is 115 ℃ generally; cfireTo preheat fuel to tfireAverage specific heat in hours, unit kj/(m)3·℃)；taThe preheating temperature of combustion air is unit ℃; caPreheating combustion air to taSpecific heat capacity in kj/(kg. DEG C); t is tsmokeThe preheating temperature of the compressed air is unit ℃; vsmokeIs the volume of compressed air, in m3；CsmokePreheating compressed air to tsmokeSpecific heat capacity in terms of kj/(m)3·℃)；hejectThe unit kJ/kg is the enthalpy value of the discharged smoke; lambda is the heat conductivity coefficient, delta is the thickness of the brick material, and the unit m; f is the heat dissipation area, singlyBit m2(ii) a R is a gas constant; l is the amount of combustion air in m3；hkThe enthalpy value of cooling air is expressed in kJ/kg;
step A2, residence time t of production streamglassAnd the time t required for the bubbles to escapebubbleAs a clarifying factor RF, then
Figure BDA0001676913630000101
The larger the RF value, the longer the production flow takes to settle in the fining zone, and the shorter the bubbles escape from the glass surface, the better the fining effect;
wherein,
Figure BDA0001676913630000102
ρlthe density of the glass melt is in kg/m3；ρbThe density of the gas in the bubbles is in kg/m3(ii) a μ is the viscosity coefficient of the bubble, r is the bubble radius, in m; labHeight of the clarification zone in m; lbcLength of the clarification zone in m; b is the width of the tank furnace and the unit m; n is the amount of gaseous material in mol units;
Figure BDA0001676913630000103
at a temperature of t0The specific heat capacity of the combustion products at that time, unit kj/(kg. DEG C);
step A3, a tank furnace thermal efficiency model eta (T)archα, x) and a clarification factor RF as objective functions to obtain an energy consumption model based on clarification quality constraints as
Figure BDA0001676913630000111
The heat in the tank furnace is derived from the heat brought in by the batch and cullet as well as the atomizing medium, combustion air. The heat required by the chemical reaction in the glass melting process is the heat of effective expenditure, specifically the heat consumption when the glass liquid is heated to the melting temperature required by the glass liquid, the heat consumption required by heating the degassing product to the melting temperature generates the heat consumption of silicate, the heat consumption for forming the glass liquid, the heat consumption for evaporating water, the heat of other expenditure is the heat of exhaust smoke, the heat of the kiln body is radiated, the heat is dissipated by overflowing, and the like, and the heat of effective expenditure is shown in table 1.
Figure BDA0001676913630000112
TABLE 1
The radiation part comprises a charging opening, a nozzle brick hole, a temperature measuring hole, a small furnace mouth, a working part and the like, but because the space temperature of the radiated parts such as the charging opening, the nozzle brick hole, the temperature measuring hole and the like is lower, and the door hole coefficient phi of the radiation part is lower, the heat loss of radiation to the molten glass working part is only considered. The reverse balance method can easily determine the magnitude of various heat losses, and further know the operation condition of the tank furnace, so the thermal efficiency is solved by using a reverse balance analysis method:
Figure BDA0001676913630000121
wherein:
Figure BDA0001676913630000122
q2＝xCfiretfire，q3＝xLCata，q4＝xVsmokeCsmoketsmoke，
Figure BDA0001676913630000123
according to the model of the thermal efficiency of the tank furnace, the factors influencing the thermal efficiency of the tank furnace are many and complex, and the main factors are as follows:
arch top temperature TarchThe crown temperature is usually controlled as a characteristic parameter of the actual temperature of the molten glass during the glass melting process. Increasing the crown temperature increases the melting rate to some extent, but with the increase in crown temperature, the heat dissipation loss of the kiln body is increased, so thatThere is a need for refractory wall materials and insulation materials with better insulation properties, thereby increasing costs.
The air excess coefficient alpha is the ratio of the actually consumed air amount to the theoretically required air amount, and CO and O in the flue are analyzed by a flue gas analyzer2、CO2The content of (A) is calculated as:
Figure BDA0001676913630000124
ω(O2) Is the percentage of oxygen in the flue gas. If the air excess factor α is too large, the heat loss of the flue gas exhaust increases, whereas if it is too small, the heat loss of the incomplete combustion increases. An inappropriate air excess factor a also affects the pressure and temperature inside the kiln. Therefore, the air excess coefficient alpha must be adjusted within an optimal range during the operation of the melting furnace.
The fuel flow x, which is a large energy source consumed during the operation of the kiln plant, directly affects the thermal efficiency of the kiln as an energy source. In addition, in order to effectively improve the operating efficiency of the kiln in the actual operation of the kiln, the ratio between the fuel and the combustion air is often controlled according to the actual production condition.
Preferably, the clarification factor that maps the quality level into a quantization is:
step a4, the quality grades were divided into A, B, C, D and E for a total of five grades, each having process requirements for the maximum diameter of the bubbles: grade A is that no bubbles are allowed, grade B is that the maximum bubble diameter is not more than 0.1mm, grade C is that the maximum bubble diameter is not more than 0.2mm, grade D is that the maximum bubble diameter is not more than 0.5mm, and grade E is that the maximum bubble diameter is not more than 1.0 mm;
step A5, setting temperature gradient
Figure BDA0001676913630000131
For a constant value k, the quality level is mapped into a quantized clarification factor by the relation between the bubble radius r and the clarification factor RF, resulting in:
Figure BDA0001676913630000132
step a6, the quality rating of step a5 is added to the energy consumption model based on the clarified quality constraints, namely:
Figure BDA0001676913630000133
the diameter of the maximum bubble is an important index for glass quality grade division, a mountain-shaped temperature curve is adopted in the embodiment, and the temperature system has the advantages that the hot spot is prominent, and the bubble boundary line is clear and stable. The part of the area before the hot spot range has a relatively slow temperature rise due to the heat transfer resistance caused by the batch material covering the surface of the molten glass, and the temperature rise is relatively fast as the heat absorption increases with the distance from the batch material covering area. However, the potential of the melting furnace is difficult to be fully exerted by the temperature system, in order to enhance the melting capacity of the batch, measures need to be taken to increase the premelting temperature in the area covered by the batch, and meanwhile, the temperature gradient of the clarification homogenization area behind the batch area needs to be ensured to meet the production process requirement of clarification homogenization, so the temperature gradient of the clarification area is required to meet the process requirement, and the investigation can know that the temperature gradient is ideal to be a fixed value when the process standard is established.
Preferably, the particle swarm algorithm is improved specifically as follows:
step B1, let the number of particles in the particle group be z, in the y iteration, fiIs the fitness of the ith particle, fmDenotes the fitness of the optimal particle, favgThe average fitness is obtained, and the fitness in the particles is more than favgHas a fitness average value of favgTaking the difference between the individual optimal fitness and the population average fitness as a convergence factor delta: Δ ═ fm-favg′|，
Figure BDA0001676913630000141
The convergence factor Δ is used to evaluate the convergence degree of the particle swarm: the smaller the convergence factor Δ, the more prone the particle to premature convergence;
step B2, by group fitness variance σ2Evaluation of the degree of dispersion of the particles in the population:
Figure BDA0001676913630000142
group fitness variance σ2The smaller the size, the more dispersed the particles, otherwise the more concentrated the particles;
step B3, the convergence factor delta and the population fitness variance sigma are calculated2As an evaluation index, the evolutionary states of the population are classified, and the inertia weight value ω is dynamically modified according to different evolutionary states:
when the particle population is fi＞favg', or favg＜fi＜f′avgThen, the inertia weight value ω ═ ω - (ω - ω) is setmin)·e-γWherein ω isminSetting the value to be 0.5, and gamma is an evolution algebra;
when the particle population is fi＜favgAnd sigma2When the inertia weight value is larger than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000151
f of particle populationi＜favgAnd sigma2When the inertia weight value is smaller than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000152
the particle swarm optimization adopts a 'speed-displacement' evolutionary model, is simple to operate, has few parameters, has a memory function capable of dynamically tracking the search condition, and is a parallel search optimization algorithm with high efficiency. However, the algorithm has certain limitations: after the particle swarm is evolved for a plurality of generations, all particles are probably gathered at a local optimal position, and the phenomenon of the gathering of the particle swarm is premature convergence, so that how to enable the particle swarm algorithm to escape from the local optimal position in time in the process of searching a solution space is the current problem.
The particle swarm algorithm is adaptively improved, so that the inertial weight value omega of the particle swarm needs to be changed independently according to different evolution states in different stages of the evolution, the population evolution efficiency is improved through the adaptive inertial weight mode, and the global optimal solution of the population is obtained. The solution space of the particle swarm algorithm is the value range of the solution of the actual optimization problem, and the particle may exceed the set boundary range when updating the speed and the position of the particle to become an invalid solution. The calculation times of the adaptive values of the invalid solutions increase the workload of the algorithm, so that the boundary processing needs to be performed on the particles beyond the boundary, and the particles are guaranteed to return to the effective search space and continue to be searched.
Figure BDA0001676913630000153
Wherein [ x ]min，j，xmax，j]Is the definition range of the j-th dimension particle. The above-described boundary processing means that the updated particle reaches a new position, and if it is detected that the position exceeds the boundary, the particle is translated to the boundary and its velocity is updated again.
Reasonably judging the early convergence degree of the algorithm is important for reasonably improving the algorithm. The convergence factor delta reflects the similarity degree between individuals with the maximum fitness in the population, avoids some adverse effects possibly brought by poor particles, and accurately describes the premature convergence degree of the population individuals. However, the convergence factor delta does not distinguish whether the population is in a dispersion stagnation state or a local optimal state at present, so that certain disadvantages exist when the evolution state of the population is judged, and therefore, the population fitness variance sigma is introduced2Evaluating the dispersion degree of the particles in the population and integrating the variance sigma of the population fitness2And reasonably evaluating the population evolution state under the condition of a convergence factor delta. The inertial weight value ω is used to adjust global exploration and local optimization capabilities. The larger inertia weight value omega accelerates the global development capability of each particle in the population, accelerates the convergence rate of the algorithm, and the smaller inertia weight value omega has moreThe method is beneficial to the local exploration capability of each particle and can increase the uniformity of the solution. Therefore, the inertia weight value omega needs to be dynamically modified according to different evolution states, and the particles can still have certain searching capacity at any time of evolution, so that the particles can effectively jump out of local optima.
F of particle populationi＞favgWhen, it means that the particle is a relatively good particle in the population, approaching global optima, oravg＜fi＜favgWhen the particle is detected to be in the optimal state, the particle is detected to be in the optimal state; therefore, a smaller inertia weight value omega is set, and the local optimizing capability is enhanced.
When the particle population is fi＜favgAnd sigma2When the value is larger than the preset range value, the algorithm is in a stagnation state, the particle distribution is loose, and at the moment, the inertia weight value omega needs to be properly reduced, so that the local optimization capability is enhanced; when the particle swarm approaches the local optimal position, the speed is updated mainly by ω vi(s), but generally the inertia weight value ω of the particle group is smaller than 1, so the velocity of the particle becomes smaller and smaller, which greatly increases the probability of premature convergence. So here the inertial weight value ω is chosen to decrease from 1.2.
When the particle population is fi＜favgAnd sigma2When the value is smaller than the preset range value, the algorithm is in a local convergence state, the particle distribution is concentrated, the inertia weight value omega is properly increased, the global detection capability of the particles is increased, and therefore the local optimal state is jumped out as soon as possible.
Preferably, the obtaining of the energy consumption optimal solution specifically includes:
step C1, setting the energy consumption model under the clarification quality constraint as a d-dimensional target search space, wherein a population exists in the d-dimensional target search space
Figure BDA0001676913630000171
Represents m particles, of which:
Figure BDA0001676913630000172
i is 1,2, and m is a vector point of the ith particle in a d-dimensional solution space,
Figure BDA0001676913630000173
1,2, m is the search speed of the ith particle, and the optimal position searched by the ith particle is pi＝[pi1，pi2，...，pim]I 1, 2.. m, the optimal position searched by the group is pg＝[pi1，pi2，...，pim]，
vi(s+1)＝ωvi(s)+c1ε1(pi-xi(s))+c2ε2(pi-xi(s))，
xi(s+1)＝xi(s)+vi(s)，ε1，ε2Random number between 0 and 1, S is the number of iterations, learning factor
Figure BDA0001676913630000174
Learning factor
Figure BDA0001676913630000175
Step C2, randomly initializing the position and speed of the particles in the population in a solution space, wherein the solution space is the value range of the solution of the practical optimization problem, and the current position is set as piSetting the optimal position in the initial group as pg；
Step C3, velocity formula v according to step C1i(s +1) and the positional formula xi(s +1) changing the velocity and position state of the particles;
and step C4, combining the penalty function and the Lagrangian function to construct a multiplier penalty function:
Figure BDA0001676913630000176
where θ is the decision variable, f (θ) is the objective function, hj(θ)＝0，j＝q + 1.. m is an equality constraint, vjIs a Lagrange multiplier adopted in the jth iteration;
calculating the fitness of the particles according to the multiplier penalty function;
step C5, according to the current evolution state, carrying out self-adaptive adjustment of the inertia weight value omega, and comparing the current adaptive value with the self optimal solution: if the current solution is superior to the self optimal value, setting the current adaptive value as the self optimal solution;
step C6, the fitness of the particle is compared with the historical optimal position pi＝[pi1，pi2，...，pim]The fitness of the global optimal solution is compared to decide whether the current global optimal solution needs to be updated or not;
step C7, detecting whether the evolution reaches the maximum set value or reaches the preset convergence precision, if so, ending the evolution process to obtain the optimal solution of energy consumption under the condition of ensuring the glass melting quality; otherwise, the step C4 is executed. The multiplier penalty function method and the improved adaptive particle swarm optimization algorithm are combined to solve the energy consumption model under the clarification quality constraint, and the optimal value of the heat efficiency meeting the clarification factor value range is obtained, so that theoretical guidance is provided for actual production.
In the optimization process of the constraint optimization problem, the balance between an objective function and a constraint is ensured, and if the optimization process is excessively limited by a constraint condition, an area with the best fitness of the objective function can be missed; if the quality of the solution is overtaken, the feasible domain may be exceeded. Therefore, the key to dealing with the constraint optimization problem is dealing with the constraint, and the most widespread method of dealing with the constraint is the penalty function. The penalty function method is based on the sequence unconstrained minimization principle, constructs a proper penalty function according to the specific characteristics of constraint conditions, and adds the penalty function item in the objective function, thereby converting the constrained problem into the unconstrained optimization problem.
The penalty function is described as:
Figure BDA0001676913630000181
h (y) is the penalty function strength, y is the algorithm currentThe number of iterations, H (θ), is a penalty factor. The probability that the optimal solution is obtained at the boundary in the constraint optimization is considered to be larger, so that a multiplier penalty function is constructed on the basis that the boundary value is used as a reference item and the constraint condition is fully guaranteed.
The particles in the particle swarm algorithm represent possible solutions of an objective function, the particles are searched iteratively in a solution space, and the searching speed can be dynamically adjusted according to self experience and social group experience. Learning factor c1Learning the factor c for self-experience1The larger the size, the more the search in the self-domain can be guided, so the learning factor c needs to be set1The method is large in the initial stage of evolution and small in the later stage of evolution; learning factor c2Group experience expressed, learning factor c2When the size is large, the particles can be guided to carry out global search, so that the learning factor c needs to be set2The early stage of evolution is small, and the later stage of evolution is large.
Example two
In order to verify the effectiveness of the improved particle swarm optimization, a single constraint function and a multi-constraint function are respectively used for testing and comparing.
Goldstein-Price single constraint test function of
f(x)＝[1+(x1+x2+1)2(19-14x1+3x1 2-14x2+6x1x2+3x2 2)]
×[30+(2x1-3x2)2(18-32x1+12x1 2+48x2-36x1x2+27x2 2)]
-2≤x1，x2≤2
The optimal solution for this test function is known to be 3.0 and not at the boundary position. The experiment was performed 15 times with 50 particles, and the maximum number of iterations was set to 300. The experimental results are shown in table 2, and it can be known from table 2 that the optimal solution is obtained after 77 generations by using the conventional particle swarm algorithm, while the optimal solution is obtained after 63 generations by using the improved particle swarm algorithm, and the convergence performance of the improved particle swarm algorithm is improved.
Figure BDA0001676913630000191
Figure BDA0001676913630000201
TABLE 2
A high-dimensional multi-constraint function of
f(x)＝(x1-10)2+5(x2-12)2+x3 4+3(x4-11)2+10x5 2
+7x6 2+x7 2-4x6x7-10x6-8x7
s.t.-127+2x1 2+3x2 4+x3+4x4 2+5x5≤0
-282+7x1+3x2+10x3 2+x4-x5≤0
-196+23x1+x2 2+6x6 2-8x7≤0，-10≤xi≤10，i＝1，...，7
4x1 2+x2 2-3x1x2+2x3 2+5x6-11x7≤0
The optimal solution for the known high-dimensional multi-constraint function is 680.630057 and is not at the boundary position. The experiment was performed 15 times with 50 particles, and the maximum number of iterations was set to 300. The experimental results are shown in table 3, and it can be known from table 3 that the optimal solution is obtained after 82 generations using the conventional particle swarm algorithm, while the optimal solution is obtained after 66 generations using the improved particle swarm algorithm, and the convergence performance of the improved particle swarm algorithm is improved.
Figure BDA0001676913630000202
TABLE 3
EXAMPLE III
In this embodiment, the energy consumption model based on the clarification quality constraint in the first embodiment is solved by using a Matlab platform to obtain an optimal solution. Before solving, preparation work needs to be carried out on simulation environment and parameters, and the steps are as follows:
(1) setting an experimental environment: simulation experiments were performed in an environment of Intel (R) core (TM) i7-5500U CPU, 2.40GHZ, 12GB, MATLAB2017 a.
(2) Crown top temperature T of handlearchThe air excess factor alpha and the fuel flow x are used as variables to be optimized, the quality level of the clarification factor RF is selected as level B, and
Figure BDA0001676913630000211
writing an m-file as an objective function.
(3) And setting an energy consumption model based on clarification quality constraint through an SMOPSO1 function, setting related parameters of the improved particle swarm algorithm, and returning and storing the values to a parameter structure body. The related parameters are configured as follows: the population number is 150, the maximum iteration number is 500, omegamin＝0.3，ωmax＝0.9。
(4) According to the number and the value range of the set variables, aiming at the target function and the parameter structure body, and calling the SMOPSO2 function to optimize the target function and the parameter structure body, so as to obtain an optimal solution, as shown in Table 4.
Figure BDA0001676913630000212
TABLE 4
When the selected quality grade is B, the value range of RF satisfies 3.12 < RF < 3.55, the result accords with the quality grade standard as can be seen from the optimization result given in Table 4, and the thermal efficiency is obviously improved by the improved particle swarm algorithm under the condition of ensuring the product quality through comparing and analyzing the result with the data collected on site. And the improved algorithm is proved to be very effective. The same applies when the quality class selects the other classes.
Example four
The horseshoe flame glass kiln energy efficiency optimization system based on the improved particle swarm optimization algorithm comprises the following steps:
the energy consumption modeling module is used for acquiring production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping the quality grade into a quantized clarification factor by using the energy consumption model based on the clarification quality constraint;
an algorithm optimization module for optimizing the convergence factor delta and the population fitness variance sigma2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm;
and the energy consumption optimization module is used for combining the improved particle swarm algorithm with the multiplier penalty function constraint processing method, applying the improved particle swarm algorithm to an energy consumption model under clarification quality constraint, and obtaining an energy consumption optimal solution under the condition of ensuring the glass melting quality.
Preferably, the energy consumption modeling module comprises:
submodule A1 for collecting the production data of horseshoe flame glass kiln, and obtaining the thermal efficiency model of the tank furnace by using an inverse balance analysis method:
Figure BDA0001676913630000221
wherein,
Figure BDA0001676913630000222
Tarchthe crown top temperature, alpha is the air excess coefficient, and x is the fuel flow in kg/h; f is the area of the opening of the radiation part and the unit m2(ii) a Phi is the door aperture coefficient of the radiation part, CtSpecific heat of radiation part, unit kj/(m)3·℃)；
Figure BDA0001676913630000223
The unit kj/kg is the low heating value of the fuel; t is tfireThe temperature for preheating the fuel is 115 ℃ generally; cfireTo preheat fuel to tfireAverage of timeSpecific heat, unit kj/(m)3·℃)；taThe preheating temperature of combustion air is unit ℃; caPreheating combustion air to taSpecific heat capacity in kj/(kg. DEG C); t is tsmokeThe preheating temperature of the compressed air is unit ℃; vsmokeIs the volume of compressed air, in m3；CsmokePreheating compressed air to tsmokeSpecific heat capacity in terms of kj/(m)3·℃)；hejectThe unit kJ/kg is the enthalpy value of the discharged smoke; lambda is the heat conductivity coefficient, delta is the thickness of the brick material, and the unit m; f is the heat dissipation area in m2(ii) a L is the amount of combustion air in m3；hkThe enthalpy value of cooling air is expressed in kJ/kg;
submodule A2 for adjusting the residence time t of the production streamglassAnd the time t required for the bubbles to escapebubbleAs a clarifying factor RF, then
Figure BDA0001676913630000231
Wherein,
Figure BDA0001676913630000232
ρlis the density of the glass melt in kg/m3；ρbThe density of the gas in the bubbles is in kg/m3(ii) a μ is the viscosity coefficient of the bubble, r is the bubble radius, in m; labHeight of the clarification zone in m; lbcLength of the clarification zone in m; b is the width of the tank furnace and the unit m; n is the amount of gaseous material in mol units;
Figure BDA0001676913630000241
at a temperature of t0The specific heat capacity of the combustion products at that time, unit kj/(kg. DEG C);
and submodule A3 for modeling the thermal efficiency of the tank furnace eta (T)archα, x) and a clarification factor RF as objective functions to obtain an energy consumption model based on clarification quality constraints as
Figure BDA0001676913630000242
Preferably, the energy consumption modeling module further comprises:
submodule a4 for classifying the quality classes into A, B, C, D and E for a total of five classes, each class having process requirements for maximum bubble diameter: grade A is that no bubbles are allowed, grade B is that the maximum bubble diameter is not more than 0.1mm, grade C is that the maximum bubble diameter is not more than 0.2mm, grade D is that the maximum bubble diameter is not more than 0.5mm, and grade E is that the maximum bubble diameter is not more than 1.0 mm;
and submodule A5 for setting a temperature gradient
Figure BDA0001676913630000243
For a constant value k, the quality level is mapped into a quantized clarification factor by the relation between the bubble radius r and the clarification factor RF, resulting in:
Figure BDA0001676913630000244
submodule a6 for adding the quality level of submodule a5 to the energy consumption model based on the clarified quality constraints, namely:
Figure BDA0001676913630000251
preferably, the algorithm optimization module comprises:
submodule B1 for setting the number of particles of the group of particles to z, in the y-th iteration fiIs the fitness of the ith particle, fmDenotes the fitness of the optimal particle, favgIs the average fitness and makes the fitness in the particles larger than favgHas a fitness average value of favgTaking the difference between the individual optimal fitness and the population average fitness as a convergence factor delta: Δ ═ fm-favg′|，
Figure BDA0001676913630000252
The convergence factor Δ is used to evaluate the convergence degree of the particle swarm: the smaller the convergence factor Δ, the more prone the particle to premature convergence;
submodule B2 for passing through the variance σ of population fitness2Evaluation of the degree of dispersion of the particles in the population:
Figure BDA0001676913630000253
group fitness variance σ2The smaller the size, the more dispersed the particles, otherwise the more concentrated the particles;
and a sub-module B3 for dividing the convergence factor Δ and the population fitness variance σ2As an evaluation index, the evolutionary states of the population are classified, and the inertia weight values are dynamically modified according to different evolutionary states:
when the particle population is fi＞favg', or favg＜fi＜f′avgThen, the inertia weight value ω ═ ω - (ω - ω) is setmin)·e-γWherein ω isminSetting the value to be 0.5, and gamma is an evolution algebra;
when the particle population is fi＜favgAnd sigma2When the inertia weight value is larger than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000261
when the particle population is fi＜favgAnd sigma2When the inertia weight value is smaller than the preset range value, setting an inertia weight value:
Figure BDA0001676913630000262
preferably, the energy consumption optimization module comprises:
submodule C1 for setting the energy consumption model under the clarification quality constraint as a d-dimensional target search space in which a population is present
Figure BDA0001676913630000263
Represents m particles, of which:
Figure BDA0001676913630000264
i is 1,2, and m is a vector point of the ith particle in a d-dimensional solution space,
Figure BDA0001676913630000265
1,2, m is the search speed of the ith particle, and the optimal position searched by the ith particle is pi＝[pi1，pi2，...，pim]I 1, 2.. m, the optimal position searched by the group is pg＝[pi1，pi2，...，pim]，
vi(s+1)＝ωvi(s)+c1ε1(pi-xi(s))+c2ε2(pi-xi(s))，
xi(s+1)＝xi(s)+vi(s)，ε1，ε2Random number between 0 and 1, S is the number of iterations, learning factor
Figure BDA0001676913630000266
Learning factor
Figure BDA0001676913630000267
Submodule C2 for randomly initializing the position and velocity of the particles in the population in the solution space, setting the current position to piSetting the optimal position in the initial group as pg；
Submodule C3 for calculating the velocity formula v from step C1i(s +1) and the positional formula xi(s +1) changing the velocity and position state of the particles;
submodule C4 for constructing a multiplier penalty function combining the penalty function and the lagrange function:
Figure BDA0001676913630000271
where θ is the decision variable, f (θ) is the objective function, hj(θ) ═ 0, j ═ q + 1.., m is the equality constraint;
calculating the fitness of the particles according to the multiplier penalty function;
the sub-module C5 is configured to perform adaptive adjustment on the inertia weight value ω according to the current evolution state, and compare the current adaptive value with the optimal solution of the sub-module: if the current solution is superior to the self optimal value, setting the current adaptive value as the self optimal solution;
submodule C6 for matching the fitness of the particle with the historical optimum position pi＝[pi1，pi2，...，pim]The fitness of the global optimal solution is compared to decide whether the current global optimal solution needs to be updated or not;
the submodule C7 is used for detecting whether the evolution reaches a maximum set value or reaches a preset convergence precision, if the requirement is met, the evolution process is ended, and an energy consumption optimal solution under the condition of ensuring the glass melting quality is obtained; otherwise sub-module C4 is executed.
The technical principle of the present invention is described above in connection with specific embodiments. The description is made for the purpose of illustrating the principles of the invention and should not be construed in any way as limiting the scope of the invention. Based on the explanations herein, those skilled in the art will be able to conceive of other embodiments of the present invention without inventive effort, which would fall within the scope of the present invention.

Claims (6)

1. A horseshoe flame glass kiln energy efficiency optimization method based on an improved particle swarm algorithm is characterized by comprising the following steps:
step A, collecting production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping quality grades into quantized clarification factors by using the energy consumption model based on clarification quality constraint;
step B, the convergence factor delta and the group fitness variance sigma are used2As an evaluation indexEvaluating the evolution state of the particle swarm and improving the particle swarm algorithm;
step C, the improved particle swarm algorithm is combined with a multiplier penalty function constraint processing method and applied to an energy consumption model under clarification quality constraint to obtain an energy consumption optimal solution under the condition of ensuring the glass melting quality;
the energy consumption model establishing process based on clarification quality constraint is as follows:
step A1, collecting production data of the horseshoe flame glass kiln, and obtaining a tank furnace thermal efficiency model by using a reverse balance analysis method:
Figure FDA0003534747970000011
wherein,
Figure FDA0003534747970000012
Tarchthe arch crown temperature, alpha is the air excess coefficient, x is the fuel flow, f is the aperture area of the radiation part, phi is the door aperture coefficient of the radiation part, CtThe specific heat of the radiation part is shown as the specific heat,
Figure FDA0003534747970000013
is the low calorific value of the fuel, tfireTemperature for preheating fuel, CfireTo preheat fuel to tfireAverage specific heat of time, taIs the preheating temperature of combustion air, CaPreheating combustion air to taSpecific heat capacity of time, tsmokeFor preheating the air at a temperature VsmokeVolume of compressed air, CsmokePreheating compressed air to tsmokeSpecific heat capacity of hour, hejectIs smoke exhaust enthalpy value, lambda is heat conductivity coefficient, delta is brick thickness, F is heat dissipation area, L is combustion air amount, hkIs the enthalpy value of the cooling air;
step A2, residence time t of production streamglassAnd the time t required for the bubbles to escapebubbleAs a clarifying factor RF, then
Figure FDA0003534747970000021
Wherein,
Figure FDA0003534747970000022
ρlis the density of the glass melt, pbIs the density of the gas in the bubble, mu is the viscosity coefficient of the bubble, r is the radius of the bubble, labTo the height of the clarification zone, /)bcIs the length of the fining zone, b is the width of the tank furnace, n is the amount of gaseous material,
Figure FDA0003534747970000024
at a temperature of t0Specific heat capacity of the combustion products at that time;
step A3, a tank furnace thermal efficiency model eta (T)archα, x) and a clarification factor RF as objective functions to obtain an energy consumption model based on clarification quality constraints as
Figure FDA0003534747970000023
The clarification factor that maps the quality level to quantization is:
step a4, the quality grades were divided into A, B, C, D and E for a total of five grades, each having process requirements for the maximum diameter of the bubbles: grade A is that no bubbles are allowed, grade B is that the maximum bubble diameter is not more than 0.1mm, grade C is that the maximum bubble diameter is not more than 0.2mm, grade D is that the maximum bubble diameter is not more than 0.5mm, and grade E is that the maximum bubble diameter is not more than 1.0 mm;
step A5, setting temperature gradient
Figure FDA0003534747970000031
For a constant value k, the quality level is mapped into a quantized clarification factor by the relation between the bubble radius r and the clarification factor RF, resulting in:
Figure FDA0003534747970000032
step a6, the quality rating of step a5 is added to the energy consumption model based on the clarified quality constraints, namely:
max η(Tarch,α,x)
s.t.RF(Tarch,α,x)
Figure FDA0003534747970000033
2. the horseshoe flame glass kiln energy efficiency optimization method based on the improved particle swarm optimization algorithm according to claim 1, wherein the particle swarm optimization algorithm is improved by:
step B1, let the number of particles in the particle group be z, in the y iteration, fiIs the fitness of the ith particle, fmDenotes the fitness of the optimal particle, favgThe average fitness is obtained, and the fitness in the particles is more than favgHas a fitness average value of favgIf the individual optimum fitness and the population fitness are larger than favgFitness average value f ofavgThe difference of' is taken as the convergence factor Δ: Δ ═ fm-favg'|，
Figure FDA0003534747970000041
The convergence factor Δ is used to evaluate the convergence degree of the particle swarm: the smaller the convergence factor Δ, the more prone the particle to premature convergence;
step B2, by group fitness variance σ2The degree of dispersion of the particles in the population was evaluated:
Figure FDA0003534747970000042
group fitness variance σ2The smaller the size, the more dispersed the particles, otherwise the more concentrated the particles;
step B3, converging the factorΔ and population fitness variance σ2As an evaluation index, the evolutionary states of the population are classified, and the inertia weight value ω is dynamically modified according to different evolutionary states:
when the particle population is fi＞favg', or favg＜fi＜favgIn this case, the inertia weight value ω ═ ω - (ω - ω) is setmin).e-γWherein ω isminSetting the value to be 0.5, and gamma is an evolution algebra;
when the particle population is fi＜favgAnd sigma2When the inertia weight value is larger than the preset range value, setting an inertia weight value:
Figure FDA0003534747970000043
when the particle population is fi＜favgAnd sigma2When the inertia weight value is smaller than the preset range value, setting an inertia weight value:
Figure FDA0003534747970000044
3. the horseshoe flame glass kiln energy efficiency optimization method based on the improved particle swarm optimization algorithm according to claim 2, wherein the energy consumption optimal solution is obtained specifically by:
step C1, setting the energy consumption model under the clarification quality constraint as a d-dimensional target search space, wherein a population exists in the d-dimensional target search space
Figure FDA0003534747970000051
Represents m particles, of which:
Figure FDA0003534747970000052
for a vector point of the ith particle in the d-dimensional solution space,
Figure FDA0003534747970000053
the optimum position searched by the ith particle is p for the search speed of the ith particlei＝[pi1,pi2,...,pid]I 1, 2.. m, the optimal position searched by the group is pg＝[p′i1,p′i2,...,p′id]，
Figure FDA0003534747970000054
Figure FDA0003534747970000055
ε1,ε2Random number between 0 and 1, s is the number of iterations, learning factor
Figure FDA0003534747970000056
Learning factor
Figure FDA0003534747970000057
Gamma is evolution algebra;
step C2, randomly initializing the position and speed of the particles in the population in the solution space, and setting the current position as piSetting the optimal position in the initial group as pg；
Step C3, according to the velocity formula in step C1
Figure FDA0003534747970000058
And position formula
Figure FDA0003534747970000059
Changing the speed and position state of the particles;
and step C4, combining the penalty function and the Lagrangian function to construct a multiplier penalty function:
Figure FDA00035347479700000510
where θ is the decision variable, f (θ) is the objective function, hj(θ) ═ 0, j ═ q + 1.., m is the equality constraint;
calculating the fitness of the particles according to the multiplier penalty function;
step C5, according to the current evolution state, carrying out self-adaptive adjustment of the inertia weight value omega, and comparing the current adaptive value with the self optimal solution: if the current solution is superior to the optimal value of the current solution, setting the current adaptive value as the optimal solution of the current solution;
step C6, the fitness of the particle is compared with the historical optimal position pi＝[pi1,pi2,...,pid]The fitness of the global optimal solution is compared to decide whether the current global optimal solution needs to be updated or not;
step C7, detecting whether the evolution reaches the maximum set value or reaches the preset convergence precision, if so, ending the evolution process to obtain the optimal solution of energy consumption under the condition of ensuring the glass melting quality; otherwise, the step C4 is executed.
4. A horseshoe flame glass kiln energy efficiency optimization system based on an improved particle swarm algorithm is characterized by comprising:
the energy consumption modeling module is used for acquiring production data of the horseshoe flame glass kiln, establishing an energy consumption model based on clarification quality constraint, and mapping the quality grade into a quantized clarification factor by using the energy consumption model based on the clarification quality constraint;
an algorithm optimization module for optimizing the convergence factor delta and the population fitness variance sigma2As an evaluation index, evaluating the evolution state of the particle swarm, and improving the particle swarm algorithm;
the energy consumption optimization module is used for combining the improved particle swarm algorithm with a multiplier penalty function constraint processing method, applying the improved particle swarm algorithm to an energy consumption model under clarification quality constraint, and obtaining an energy consumption optimal solution under the condition of ensuring the glass melting quality;
the energy consumption modeling module comprises:
submodule A1 for collecting production data of horseshoe flame glass kiln, obtaining thermal efficiency model of tank furnace by using reverse balance analysis method:
Figure FDA0003534747970000061
wherein,
Figure FDA0003534747970000071
Tarchthe arch crown temperature, alpha is the air excess coefficient, x is the fuel flow, f is the aperture area of the radiation part, phi is the door aperture coefficient of the radiation part, CtThe specific heat of the radiation part is shown as the specific heat,
Figure FDA0003534747970000072
is the low calorific value of the fuel, tfireTemperature for preheating fuel, CfireTo preheat fuel to tfireAverage specific heat of time, taIs the preheating temperature of combustion air, CaPreheating combustion air to taSpecific heat capacity of time, tsmokeFor preheating the air at a temperature VsmokeVolume of compressed air, CsmokeFor preheating compressed air to tsmokeSpecific heat capacity of hour, hejectIs smoke exhaust enthalpy value, lambda is heat conductivity coefficient, delta is brick thickness, F is heat dissipation area, L is combustion air amount, hkIs the enthalpy value of the cooling air;
submodule A2 for adjusting the residence time t of the production streamglassAnd the time t required for the bubbles to escapebubbleAs a clarifying factor RF, then
Figure FDA0003534747970000073
Wherein,
Figure FDA0003534747970000074
ρldensity of glass melt, pbIs the density of the gas in the bubble, mu is the viscosity coefficient of the bubble, r is the bubble halfDiameter, labTo the height of the clarification zone, /)bcIs the length of the fining zone, b is the width of the tank furnace, n is the amount of gaseous material,
Figure FDA0003534747970000081
at a temperature of t0Specific heat capacity of the combustion products at that time;
and submodule A3 for modeling the thermal efficiency of the tank furnace eta (T)archα, x) and a clarification factor RF as objective functions to obtain an energy consumption model based on clarification quality constraints as
Figure FDA0003534747970000082
The energy consumption modeling module further comprises:
submodule a4 for classifying the quality classes into A, B, C, D and E for a total of five classes, each class having process requirements for maximum bubble diameter: grade A is that no bubbles are allowed, grade B is that the maximum bubble diameter is not more than 0.1mm, grade C is that the maximum bubble diameter is not more than 0.2mm, grade D is that the maximum bubble diameter is not more than 0.5mm, and grade E is that the maximum bubble diameter is not more than 1.0 mm;
and submodule A5 for setting a temperature gradient
Figure FDA0003534747970000083
For a constant value k, the quality level is mapped into a quantized clarification factor by the relation between the bubble radius r and the clarification factor RF, resulting in:
Figure FDA0003534747970000084
submodule a6 for adding the quality level of submodule a5 to the energy consumption model based on the clarified quality constraints, namely:
max η(Tarch,α,x)
s.t.RF(Tarch,α,x)
Figure FDA0003534747970000091
5. the system for optimizing the energy efficiency of the horseshoe flame glass kiln based on the improved particle swarm optimization algorithm according to claim 4, wherein the algorithm optimization module comprises:
submodule B1 for setting the number of particles of the group of particles to z, in the y-th iteration fiIs the fitness of the ith particle, fmDenotes the fitness of the optimal particle, favgThe average fitness is obtained, and the fitness in the particles is more than favgHas a fitness average value of favgIf the individual optimum fitness and the population fitness are larger than favgAverage value f of fitness ofavgThe difference of' is taken as the convergence factor Δ: Δ ═ fm-favg'|，
Figure FDA0003534747970000092
The convergence factor Δ is used to evaluate the convergence degree of the particle swarm: the smaller the convergence factor Δ, the more prone the particle to premature convergence;
submodule B2 for passing through the variance σ of population fitness2The degree of dispersion of the particles in the population was evaluated:
Figure FDA0003534747970000093
group fitness variance σ2The smaller the size, the more dispersed the particles, otherwise the more concentrated the particles;
and a sub-module B3 for applying a convergence factor Δ and a population fitness variance σ2As an evaluation index, the evolutionary states of the population are classified, and the inertia weight values are dynamically modified according to different evolutionary states:
when the particle population is fi＞favg', or favg＜fi＜favgIn this case, the inertia weight value ω ═ ω - (ω - ω) is setmin).e-γWherein ω isminSetting the value to be 0.5, and gamma is an evolution algebra;
when the particle population is fi＜favgAnd sigma2When the inertia weight value is larger than the preset range value, setting an inertia weight value:
Figure FDA0003534747970000101
when the particle population is fi＜favgAnd sigma2When the inertia weight value is smaller than the preset range value, setting an inertia weight value:
Figure FDA0003534747970000102
6. the system for optimizing the energy efficiency of the horseshoe flame glass kiln based on the improved particle swarm algorithm according to claim 5, wherein the energy consumption optimizing module comprises:
submodule C1 for setting the energy consumption model under the clarification quality constraint as a d-dimensional target search space in which a population is present
Figure FDA0003534747970000103
Represents m particles, of which:
Figure FDA0003534747970000104
for a vector point of the ith particle in the d-dimensional solution space,
Figure FDA0003534747970000105
the optimum position searched by the ith particle is p for the search speed of the ith particlei＝[pi1,pi2,...,pid]I 1, 2.. m, the optimal position searched by the group is pg＝[p′i1,p′i2,...,p′id]，
Figure FDA0003534747970000106
Figure FDA0003534747970000107
ε1,ε2Random number between 0 and 1, s is the number of iterations, learning factor
Figure FDA0003534747970000108
Learning factor
Figure FDA0003534747970000109
Gamma is evolution algebra;
submodule C2 for randomly initializing the position and velocity of the particles in the population in the solution space, setting the current position to piSetting the optimal position in the initial group as pg；
Submodule C3 for calculating the speed formula according to step C1
Figure FDA0003534747970000111
And position formula
Figure FDA0003534747970000112
Changing the velocity and position state of the particles;
submodule C4 for constructing a multiplier penalty function combining the penalty function and the lagrange function:
Figure FDA0003534747970000113
where θ is the decision variable, f (θ) is the objective function, hj(θ) ═ 0, j ═ q + 1.., m is the equality constraint;
calculating the fitness of the particles according to the multiplier penalty function;
the submodule C5 is configured to perform adaptive adjustment on the inertia weight value ω according to the current evolution state, and compare the current adaptive value with the self optimal solution: if the current solution is superior to the self optimal value, setting the current adaptive value as the self optimal solution;
submodule C6 for matching the fitness of the particle with the historical optimum position pi＝[pi1,pi2,...,pid]The fitness of the global optimal solution is compared to decide whether the current global optimal solution needs to be updated or not;
the submodule C7 is used for detecting whether the evolution reaches a maximum set value or reaches a preset convergence precision, if the requirement is met, the evolution process is ended, and an energy consumption optimal solution under the condition of ensuring the glass melting quality is obtained; otherwise sub-module C4 is executed.
CN201810529748.9A 2018-05-29 2018-05-29 Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system Active CN109033489B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810529748.9A CN109033489B (en) 2018-05-29 2018-05-29 Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810529748.9A CN109033489B (en) 2018-05-29 2018-05-29 Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system

Publications (2)

Publication Number Publication Date
CN109033489A CN109033489A (en) 2018-12-18
CN109033489B true CN109033489B (en) 2022-06-07

Family

ID=64611462

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810529748.9A Active CN109033489B (en) 2018-05-29 2018-05-29 Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system

Country Status (1)

Country Link
CN (1) CN109033489B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111259540B (en) * 2020-01-14 2023-04-11 广东工业大学 Energy efficiency analysis and optimization method for photovoltaic glass calendering and molding process
CN114638467B (en) * 2022-01-31 2023-04-07 南通鑫鑫医药药材有限公司 Production quality detection system for medical glass container

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SU1201236A1 (en) * 1982-10-25 1985-12-30 Предприятие П/Я М-5314 Method of heating bath glassmaking furnace
US5764544A (en) * 1995-11-16 1998-06-09 Gas Research Institute Recuperator model for glass furnace reburn analysis
WO2016001310A1 (en) * 2014-07-01 2016-01-07 Magma Giessereitechnologie Gmbh A method and algorithm for simulating the influence of thermally coupled surface radiation in casting processes
CN106863136A (en) * 2017-01-15 2017-06-20 复旦大学 CCOS glossing full frequency band converged paths planing methods
CN107301303A (en) * 2017-07-14 2017-10-27 吴康 3D glass heat bender mold heating system colony intelligence Optimization Designs

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8573896B2 (en) * 2007-10-17 2013-11-05 Remco International, Inc. Method of dynamic energy-saving superconductive transporting of medium flow

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SU1201236A1 (en) * 1982-10-25 1985-12-30 Предприятие П/Я М-5314 Method of heating bath glassmaking furnace
US5764544A (en) * 1995-11-16 1998-06-09 Gas Research Institute Recuperator model for glass furnace reburn analysis
WO2016001310A1 (en) * 2014-07-01 2016-01-07 Magma Giessereitechnologie Gmbh A method and algorithm for simulating the influence of thermally coupled surface radiation in casting processes
CN106863136A (en) * 2017-01-15 2017-06-20 复旦大学 CCOS glossing full frequency band converged paths planing methods
CN107301303A (en) * 2017-07-14 2017-10-27 吴康 3D glass heat bender mold heating system colony intelligence Optimization Designs

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Dual least squares support vector machines based spatiotemporalmodeling for nonlinear distributed thermal processes;processesKang-Kang Xu;《Journal of Process Control》;20171231;第54卷;第81-89页 *
玻璃窑炉节能优化设计初探;邵景楚;《玻璃》;20110625;第38卷(第6期);第20-22页 *

Also Published As

Publication number Publication date
CN109033489A (en) 2018-12-18

Similar Documents

Publication Publication Date Title
CN109033489B (en) Improved particle swarm algorithm-based horseshoe flame glass kiln energy efficiency optimization method and system
CN102867220B (en) A kind of method of real-time estimate ladle refining furnace refined molten steel temperature
CN103882176B (en) The online dynamic control method of a kind of convertor steelmaking process based on data-driven
Rout et al. Dynamic model of basic oxygen steelmaking process based on multi-zone reaction kinetics: Model derivation and validation
CN102176221B (en) Coke furnace temperature predicting method based on dynamic working conditions in coke furnace heating and burning process
CN104899463A (en) Blast furnace molten iron silicon content four-classification trend prediction model establishing method and application
Wang et al. Prediction of endpoint phosphorus content of molten steel in BOF using weighted K-means and GMDH neural network
US20240211662A1 (en) Calculating and real-time monitoring method for boundary of blast furnace tuyere raceway
CN107038307A (en) Mechanism predicts integrated modelling approach with the Roller Conveying Kiln for Temperature that data are combined
Li et al. Burden surface decision using MODE with TOPSIS in blast furnace ironmkaing
CN118016203B (en) LF refining temperature forecasting method based on mechanism model and XGBoost algorithm
CN205376668U (en) Lithium ion battery ization becomes system
CN102492964A (en) Device for heat exchange of electrolytic bath
Chen et al. A synergetic immune clonal selection algorithm based multi-objective optimization method for carbon fiber drawing process
CN110567534B (en) Method for predicting flow of combustion air outlet in glass melting furnace and related device
CN116702613A (en) Multi-objective optimization method and system for product performance and sintering energy consumption of ternary positive electrode material
CN112329269A (en) Sintering ignition temperature modeling prediction method based on working condition identification
CN109034523A (en) A kind of Rotary Kiln operating condition operation mode optimization method of fusion least square method supporting vector machine recurrence and particle group optimizing
CN116070369A (en) Particle swarm optimization-based injector structure optimization method
CN111427388B (en) Control system and method for iron condensing layer of blast furnace hearth
Wu et al. Multi-objective optimization of key parameters of stirred tank based on ANN-CFD
CN110059410B (en) Belly air intake hypersonic precursor optimization design method based on B spline
Yang et al. A molecular simulation study on the sintering mechanism of TiO2
Zhang et al. A combustion control strategy of hot blast stove based on kernel fuzzy c-means (FCM)
Yang et al. Billet temperature soft sensor model of reheating furnace based on RVM method

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
GR01 Patent grant
GR01 Patent grant