Research On Fuzzy Decision Strategy And Bilevel Optimization Mechanism For Large-Scale Optimization Problems

In the real world,there are a large number of multiobjective optimization problems,that is,optimization problems with two or more related conflicting objective functions.Many practical problems are NPC problems.When using deterministic algorithms and original mathematical methods to solve such problems,they are usually tricky and helpless.Therefore,evolutionary algorithms are widely used to solve NPC problems and have been a hot research topic in recent years.With the rapid development of information industrialization,engineering problems in the industry are becoming more and more complex,especially when there are many optimization variables involved.Large-scale multiobjective optimization problems reach more than 100 dimensions by being considered as decision variables.In large-scale multiobjective optimization problems,a large number of decision variables lead to a huge search space in the optimization problem.Existing large-scale multiobjective optimization evolutionary algorithms can be divided into four categories: 1)evolutionary algorithms based on co-evolution;2)evolutionary algorithms based on decision variable grouping;3)evolutionary algorithms based on problem conversion;4)evolutionary algorithm based on particle learning mechanism.The main mechanism of the above four types of methods is dimensionality reduction,but there is a natural shortcoming: the algorithm will optimize in the decision space after dimensionality reduction so that parts of the original decision space can never be explored.Therefore,this paper proposes a method that can find the best in the original decision space and reduce the search range of the algorithm in the decision space to improve the convergence of the algorithm and effectively solve the large-scale multiobjective optimization problem.This paper proposes an evolutionary algorithm framework for large-scale multiobjective optimization of fuzzy decision variables.The framework divides the entire evolution process into two main stages: fuzzy evolution and precise evolution.In fuzzy evolution,we perform fuzzy processing on the decision variables of the original solution to reduce the search range of the evolutionary algorithm in the decision space and make the evolutionary population quickly converge.The degree of fuzzification gradually decreases with the evolution process.Once the population has approximately converged,the framework will shift to precise evolution.In precise evolution,the actual decision variables of the solution are directly optimized to increase the diversity of the population,thus getting closer to the true Pareto optimal front.We have conducted experiments on various large-scale multiobjective problems with up to 500 to 5000 decision variables.Four different types of representative multiobjective evolutionary algorithms have been embedded in the proposed framework and compared with their original evolutionary algorithms.The proposed framework has been compared with three popular largescale multiobjective optimization frameworks.The Competitive Group Optimizer(CSO)has been embedded into the proposed framework and compared with four other state-of-the-art large-scale multiobjective optimization algorithms.Experimental results show that in largescale multiobjective optimization,the framework proposed in this paper can significantly improve the performance and computational efficiency of multiobjective optimization algorithms.Subsequently,this paper combines a multiobjective optimization algorithm with a bilevel optimization framework to solve a practical engineering problem—the energy hub system planning problem.Energy hub system planning is a large-scale discrete multiobjective problem and also belongs to a master-slave game.It is difficult to obtain a solution to this problem in a limited time through deterministic algorithms.In order to solve the above problems,we designed a multiobjective bilevel optimization framework based on preference selection.The framework is divided into a lower-level optimizer and an upper-level optimizer.The lower-level optimizer is composed of a multiobjective evolutionary algorithm and Lagrangian interpolation method,and the upper-level optimizer is composed of a trisection search method.The preference-based selection mechanism can ensure that the lower-level optimizer makes deterministic decisions.In actual problems,the upper level optimizes the optimal capacity of energy equipment,and the lower level optimizes the real-time operating power of each energy equipment.Compared with commercial optimizers,this method makes up for the shortcomings of commercial optimizers that cannot solve nonlinear discrete problems.The method we propose helps to solve the planning,design,and operation scheduling problems of complex energy hub systems and multienergy complementary systems.This method provides a theoretical basis for further research on the optimal scheduling of the entire life cycle of the energy hub system.Finally,this paper applies the theoretical research results of multiobjective optimization to practical engineering problems to solve the problem of intelligent garage parking space allocation.NSGA-II is very suitable for solving multiobjective nonlinear discrete problems.It avoids complex mathematical model analysis and directly optimizes the original model,which can perfectly balance the two objectives and optimize the optimal solution at the same time.When the number of warehousing vehicles is very large,the parking space allocation problem is extended to a large-scale multiobjective optimization problem.The dimensionality reduction method based on decision variable grouping can be considered to improve the optimization performance of NSGA-II.In this paper,simulation experiments are carried out on the proposed optimization method,and the experimental results are in line with the principle of vehicle allocation and are reasonable and efficient.