Dynamic Multi-objective Optimization Algorithm based on Transfer Learning for Environmental Protection

In the life, there are many dynamic multi-objective optimization problems related to the optimization objectives and dynamic change factors. The solution of this problem requires that the algorithm used can quickly find the optimal solution in the new environment. Most methods ignored the independence and distribution of individual data, especially for dynamic optimization problems with variable amplitudes, which lack rationality. Based on the classical algorithm NSGA-II, this paper proposes a dynamic multi-objective optimization algorithm based on transfer learning called TR-FC-NSGA-II. Especially, for the dynamic multi-objective optimization problem with changeable amplitude, paper first proposes kind of angle change, which embodies the change range concretely. And then we carry out different strategies for different individuals affected by the change amplitude. The transfer learning algorithm is introduced to screen the optimized population twice, and more reasonable individuals are selected to solve the problem. Typical test function problems are used to test the algorithm, and the algorithm is compared with DNSGA-II and TRNSGA- II. The experimental results show that the improved algorithm can quickly find the optimal solution suitable for the current environment in dynamic environment. Compared with the traditional algorithm, the efficiency of the solution is greatly improved. The improved algorithm can not only be applied to optimize the use of raw materials in production, reduce resource waste and pollution, but also optimize the traffic routes, reduce congestion accidents and protect the social environment.