CN-120085547-B - Control parameter collaborative optimization method based on hybrid heuristic optimization

CN120085547BCN 120085547 BCN120085547 BCN 120085547BCN-120085547-B

Abstract

The invention discloses a control parameter collaborative optimization method based on mixed heuristic optimization, which comprises the steps of carrying out collaborative initialization on Ns basic optimization algorithms through mixed heuristic optimization threads, carrying out parallel solution on target optimization problems based on original iteration mechanisms of each algorithm, generating feature space data containing algorithm winning communities, extracting current optimal solutions of each algorithm, constructing a cross-algorithm winning community interaction mechanism, importing the feature space data into a winning community feature library, outputting target function optimization coefficients of the algorithm, and carrying out stability verification and robustness analysis on all the target function optimization coefficients through a multi-target fusion decision unit to determine a global optimal solution. According to the invention, by combining a plurality of heuristic algorithms and realizing random exchange of winning community individuals, the influence of single algorithm defects on a solving result is avoided to the maximum extent, and the multi-algorithm collaborative rapid initialization, parallel solving and iteration can be realized.

Inventors

YUAN ZHONGYUAN
LV RUIXIN
YU WENHUA
WANG YU

Assignees

四川省艾耳能科技有限公司

Dates

Publication Date: 20260512
Application Date: 20250304

Claims (9)

1. A control parameter collaborative optimizing method based on mixed heuristic optimization is characterized in that, The method comprises the following steps: Carrying out collaborative initialization on a genetic algorithm, a particle swarm optimization algorithm and a firefly swarm optimization algorithm by a mixed heuristic optimization thread, and carrying out parallel solving on the energy efficiency-comfort multi-objective optimization problem of the air conditioning system based on a primary iteration mechanism of each algorithm, wherein the optimization parameters comprise an air supply temperature set value, fan frequency, a freezing water valve opening and a fresh air proportion; the mixed heuristic optimization thread is obtained by injecting a population characteristic migration mechanism and a parameter self-adaptive adjustment strategy into the reference optimization thread; when any algorithm meets a first convergence condition or the initial iteration number iter1 reaches Maxiter threshold, generating feature space data comprising temperature field distribution features, equipment energy consumption features and passenger density features, and extracting a current optimal equipment control parameter combination of each algorithm; Constructing a cross-algorithm winning community interaction mechanism, importing the feature space data into a winning community feature library containing historical working condition features, executing feature sampling and cross-population embedding operation from different winning communities based on population diversity feature coefficients, specifically calculating population diversity feature coefficients based on heat load fluctuation coefficients and passenger density change rates, and executing air supply parameter sampling and cross-population embedding operation between different winning communities according to the feature sampling and cross-population embedding operation; when a single algorithm meets a second convergence condition or the total iteration number iter2 reaches Maxiter < 2> threshold, outputting the equipment energy efficiency optimization coefficient and the temperature uniformity index of the algorithm; The stability verification of all optimization coefficients under the dynamic load working condition is carried out through a multi-objective fusion decision unit, a global optimal solution is determined based on the energy consumption fluctuation tolerance and the temperature deviation threshold value, and real-time meteorological parameters and passenger flow density monitoring data are introduced in the verification process; The feature sampling and cross-population embedding operation specifically comprises the following steps: determining the feature extraction quantity of each winning community based on a dynamic sampling proportion calculation module, wherein the dynamic sampling proportion is positively correlated with a population diversity index; Dividing a winning community into a plurality of feature subspaces according to fitness distribution by adopting a hierarchical feature sampling strategy, and executing self-adaptive sampling based on probability density in each subspace; The extracted characteristic samples are injected into the target population through a random position insertion algorithm, wherein the random position insertion algorithm comprises a gene sequence recombination verification unit and a population capacity balancing unit.
2. The hybrid heuristic optimization-based control parameter collaborative optimization method according to claim 1, wherein: the cross-algorithm winning community interaction mechanism comprises: performing multidimensional feature stitching on the feature space data by using a feature vector fusion unit to generate a winning community feature vector with cross-algorithm relevance; calculating population diversity characteristic coefficients based on the winning community characteristic vectors, and determining sampling proportion and characteristic layering intervals embedded across the population according to the coefficients; The construction method of the parameter self-adaptive regulation strategy comprises the following steps: Monitoring the change rate of objective function values of each algorithm in a continuous preset iteration window; when the change rate is lower than a dynamic adjustment threshold value, triggering a parameter disturbance module to carry out nonlinear adjustment on the core parameters of the current algorithm: performing an exponential amplification operation of the variation probability on the genetic algorithm; executing the sectional attenuation operation of the inertia weight on the particle swarm algorithm; an adaptive recalibration operation of the luciferin coefficients is performed on the firefly algorithm.
3. The hybrid heuristic optimization-based control parameter collaborative optimization method according to claim 2, wherein: The genetic algorithm, the particle swarm optimization algorithm and the native operator of the firefly swarm optimization algorithm are specifically expressed as follows: the genetic algorithm adopts a chromosome coding mechanism, and an operation operator of the genetic algorithm comprises a gene recombination unit based on a roulette selection operator, a population evolution unit of a single-point crossover operator and a gene disturbance unit of a basic mutation operator; the particle swarm optimization algorithm adopts a speed-position update model, and an operation operator of the particle swarm optimization algorithm comprises an individual optimal position tracking unit and a global optimal position fusion unit; The firefly group optimization algorithm adopts a brightness attraction mechanism, and an operation operator of the firefly group optimization algorithm comprises a relative brightness calculation unit and a dynamic distance response unit.
4. The hybrid heuristic optimization-based control parameter collaborative optimization method according to claim 3, wherein: The iterative evolutionary process of native operators comprises: The genetic algorithm executes a three-stage evolution flow of gene selection-crossing-mutation, and drives population iterative update through a fitness function evaluation unit; The particle swarm algorithm executes a speed vector update-position vector correction double-stage motion flow, and information interaction is realized through the individual history optimal solution memory unit and the global optimal solution sharing unit; And the firefly algorithm executes a three-stage bionic flow of brightness calculation, position attraction and parameter update, and intelligent evolution of the group is realized through a relative distance calculation unit and an attraction intensity adjustment unit.
5. The hybrid heuristic optimization-based control parameter collaborative optimization method according to any one of claims 1-4, wherein: The construction method of the winning community feature library comprises the following steps: Establishing a multidimensional feature tag for each winning community, wherein the multidimensional feature tag comprises an algorithm identifier, an iteration stage feature code, an objective function value change gradient and a population diversity index; and converting the multidimensional feature tag into a traceable feature vector by adopting a feature coding mechanism, and constructing a winning community feature map with space-time correlation.
6. The hybrid heuristic optimization-based control parameter collaborative optimization method according to claim 2, wherein: The parameter perturbation module comprises: the genetic algorithm variation enhancing unit adopts an exponential variation probability function; The particle swarm inertia weight adjusting unit adopts a piecewise linear decay function; and the firefly luciferin adjusting unit adopts a self-adaptive updating formula.
7. The hybrid heuristic optimization-based control parameter collaborative optimization method according to claim 1, wherein: The second convergence condition judging method comprises the following steps: Constructing a dual convergence criterion combined decision model, triggering a first convergence mark when the total iteration number iter2> the total minimum iteration number Miniter and the change quantity delta f of the objective function value is less than or equal to epsilon, wherein epsilon is a convergence threshold; triggering a second convergence mark when the continuous iteration times of the algorithm which falls into the local optimum exceeds a preset threshold value; and the dual convergence criterion combined decision model adopts a weighted voting mechanism to comprehensively judge the termination time of the algorithm.
8. A control parameter co-optimizing system based on hybrid heuristic optimization, characterized in that it is used for implementing the control parameter co-optimizing method based on hybrid heuristic optimization according to any one of claims 1-7, comprising: The system comprises a mixed heuristic optimization engine module, a winning community interaction module, a parameter self-adaptive adjustment module, a multi-objective decision module and an algorithm base library module, wherein the mixed heuristic optimization engine module is respectively connected with the algorithm base library module and the winning community interaction module and respectively transmits algorithm original iteration data and winning community feature space data; The hybrid heuristic optimization engine module comprises: The collaborative initialization unit is connected with an algorithm basic library module, performs parameter space initialization on the Ns basic optimization algorithms and establishes a parallel solving channel; the original iterative execution unit is internally provided with a genetic algorithm operator, a particle swarm optimization operator and a firefly swarm optimization operator, and drives each algorithm to execute iterative operation according to an original mechanism; the convergence monitoring unit is connected with each algorithm channel and judges the iteration termination condition through the convergence condition judging module and the Maxiter counter module; the feature extraction unit is used for generating feature space data containing algorithm winning communities and extracting current optimal solution vectors of all channels; The winning community interaction module comprises: The feature vector fusion unit is used for receiving feature space data from each algorithm channel and executing multidimensional feature splicing to generate cross-algorithm associated feature vectors; the diversity analysis unit is used for determining characteristic sampling proportion and layering interval based on the population diversity characteristic coefficient calculation module; the cross-population embedding unit is connected with the hybrid heuristic optimization engine module and used for executing the dynamic migration and population recombination operation of the winning feature vector; the parameter self-adaptive adjusting module comprises: the iterative window monitoring unit is used for counting the change rate of the objective function value of each algorithm through a sliding time window; The disturbance triggering unit is used for activating the parameter nonlinear adjusting channel when the change rate is lower than the dynamic adjusting threshold value; the algorithm regulator group comprises a genetic algorithm variation probability amplifier, a particle swarm inertia weight attenuator and a firefly luciferin calibrator; The multi-objective decision module comprises: The stability verification unit is used for carrying out robustness test on each optimization coefficient through a Monte Carlo simulation module; the decision fusion unit is used for carrying out global optimal solution screening on the pareto front solution set by adopting a fuzzy comprehensive evaluation method; The result output interface is used for generating an analysis report containing an optimized parameter configuration scheme and an objective function response curved surface; The algorithm base library module stores: The genetic algorithm core operator group comprises a selection operator library, a crossover operator library, a mutation operator library and an fitness function library; The particle swarm optimization assembly set comprises a speed update calculator, a position updater and a neighborhood topological structure template; the firefly algorithm functional package comprises a fluorescein updating module, a movement probability calculator and an attraction calculating unit; The characteristic migration operation executed by the cross-population embedding unit comprises characteristic space similarity calculation based on the mahalanobis distance, a population recombination strategy adopting a quantum rotation gate mechanism, a characteristic layered sampling method based on information entropy weight, a characteristic data processing method and a characteristic data processing method; the parameter nonlinear regulation channel comprises an exponential amplification function of genetic algorithm variation probability, a sectional attenuation strategy of particle swarm inertia weight and a self-adaptive calibration model of firefly fluorescein coefficient.
9. An electronic device, the electronic device comprising: at least one processor, and A memory communicatively coupled to the at least one processor, wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of a hybrid heuristic optimization-based control parameter co-optimization method as claimed in any one of claims 1 to 7.

Description

Control parameter collaborative optimization method based on hybrid heuristic optimization Technical Field The invention relates to a control parameter collaborative optimizing method based on hybrid heuristic optimization, which realizes control parameter optimization through a computer technology. Background The problem of optimizing the solution of the control parameters in the MPC controller is actually the solution of the constraint optimization problem. When the objective function and the constraint are both linear functions, the optimization problem is linear programming. When the objective function J and constraint are not entirely linear, the optimization problem is a nonlinear programming. The models involved in the optimization operation of the air conditioning system are mostly nonlinear models, and the objective function established according to the nonlinear models is also a nonlinear function. Therefore, the solution of the control parameters is a solution to the nonlinear constraint optimization problem. The commonly used solving methods of nonlinear constraint optimization problems can be roughly divided into two main types of traditional optimization methods and intelligent optimization methods. The specific optimization algorithm and the advantages and disadvantages of the various methods are shown in table 1. Table 1 nonlinear constraint optimization problem solving method As can be seen from Table 1, the conventional optimization method has the problems of low operation efficiency, easy sinking into local optimum and the like. Meanwhile, the traditional optimization method is difficult to solve the problems of non-convex feasible regions, feasible regions with a plurality of local extrema and non-communication, discrete and integer variables in whole or in part, and the like. The intelligent optimization method has the advantages of processing capacity of nonlinear problems and complex non-convex optimization problems due to strong global searching capacity, and the like, and is widely applied to the actual production process at present and achieves good effects. Heuristic algorithms in intelligent optimization methods are widely studied and applied at present. Heuristic algorithms are a class of empirically and intuitively based problem-solving methods that guide optimization and decision-making processes through heuristic rules, policies and search skills. Common heuristic algorithms include genetic algorithms, ant colony algorithms, simulated annealing algorithms, particle swarm algorithms, and the like. However, a single heuristic always suffers from some inherent drawbacks when applied. If the local searching capability of the genetic algorithm is poor, the early ripening phenomenon is easy to occur, the searching time of the ant colony algorithm is long, and the algorithm result is sensitive to parameters. The invention patent with the patent publication number of CN116300430B discloses an MPC control parameter optimizing method and application thereof in a parallel platform, and the model predictive controller can obtain better control effect by establishing an evaluation function for MPC control parameter optimizing and performing meta-heuristic search in a parameter space by using a real number coded differential evolutionary algorithm to finish parameter optimizing. But the method is still prone to trap local search traps. Disclosure of Invention The invention aims to overcome the defects of the prior art, provides a control parameter collaborative optimizing method based on mixed heuristic optimization, and combines a computer technology to furthest avoid the influence of the defects of each algorithm on a solving result and improve the probability of finding a global optimal solution. The core of the method is that individuals in winning communities of different algorithms can be randomly exchanged for optimal solution. The aim of the invention is realized by the following technical scheme: a control parameter collaborative optimization method based on hybrid heuristic optimization, the method comprising: Carrying out collaborative initialization on Ns basic optimization algorithms through a mixed heuristic optimization thread, and executing parallel solution on a target optimization problem based on a native iteration mechanism of each algorithm; When any algorithm meets a first convergence condition or the initial iteration number iter1 reaches Maxiter <1 > threshold, generating feature space data containing algorithm winning communities, and extracting the current optimal solution of each algorithm; constructing a cross-algorithm winning community interaction mechanism, importing the feature space data into a winning community feature library, and executing feature sampling and cross-population embedding operation from different winning communities based on population diversity feature coefficients; outputting an objective function optimization coefficient of the single a