Multi-objective optimization method and system for dynamically using neural network

Abstract

The invention discloses a multi-objective optimization method and system for dynamically using a neural network, and relates to the technical field of multi-objective optimization algorithms, and the method comprises the steps: obtaining a decision variable of a first-generation individual of a genetic algorithm, and putting the obtained decision variable into a simulator to obtain a corresponding performance index; training the neural network model; converting a performance index corresponding to a first generation of individuals of the genetic algorithm into a fitness function, obtaining a dominating relationship between a feasible solution set and a non-feasible solution set based on a rapid non-dominating sorting algorithm, and then selecting a next generation of parent individuals from non-dominating leading edges of the feasible solution set and the non-feasible solution set by using an accessibility algorithm; obtaining a decision variable of the current generation individual, and putting the decision variable into a simulator to obtain a performance index; the feasible parent individuals and the infeasible parent individuals of the next generation are selected through a reachability algorithm, and the number of the individuals generated by each operator of the next generation is dynamically allocated. According to the method, the problem of multi-objective optimization can be solved under the condition that physical simulation is relatively complex and computing resources are limited.

Description

technical field [0001] The invention relates to the technical field of multi-objective optimization algorithms, in particular to a multi-objective optimization method and system using a neural network dynamically. Background technique [0002] Genetic algorithms in multi-objective optimization methods (such as NSGA-II, MOEA / D, SPEA, etc.) have shown good performance in many engineering optimization problems. These algorithms draw on the evolutionary methods of natural organisms and combine the process algorithms of biological evolution. By using operators such as selection, mutation, and crossover in each generation, competitive individuals can be obtained and non-dominated frontiers can be formed. In the field of physical optimization, many problems require complex simulations to obtain performance indicators, which results in a very limited number of individuals that genetic algorithms can obtain in limited computing resources. At this time, if the decision space in the o...

AI Technical Summary

Problems solved by technology

The multi-objective genetic algorithm combined with the neural network has been proven to have a faster convergence speed, but these algorithms are completely dependent on the predicted results after the neural network starts training. The small-scale training set of the neural network can easily cause the prediction accuracy of the neural network to fail to meet the needs of such constraints

Images