Research On Improvement And Application Of Gradient-based Optimizer

Gradient-based optimizer(GBO)is a meta heuristic optimization algorithm that simulates the process of solving equations using Newton’s method.This algorithm has strong global search ability and fast convergence speed,and can efficiently solve complex practical problems.GBO has been widely applied in fields such as structural design,energy utilization,and chemical production.With the depth of research,the application fields of GBO are constantly expanding.Researchers have found that in some practical problems with constraints,GBO algorithm has bad local search ability,low convergence accuracy,and is prone to falling into local optima.This paper aims to enhance the optimization ability of GBO,analyze the characteristics of the algorithm,and propose improved algorithms to expand its application range.The main work of this article is as follows:(1)In order to improve the local search ability of gradient-based optimizer and balance the global search and local search,an improved gradient-based optimizer(IGBO)is proposed by embedding elite opposition-based learning strategy,crossover operator and a nonlinear parameter into the gradient optimizer.To verify the performance of the improved algorithm,IGBO was used to optimize the parameters of the extreme learning machine model.An IGBO-ELM framework was proposed and tested on UCI datasets.The framework was also used to predict dam seepage.Compared with other five metaheuristic algorithms,the experimental results showed that the IGBO-ELM prediction model has strong stability and prediction accuracy.(2)Simplex method is introduced to improve the worst vector and adjust the gradient direction to accelerate the convergence speed of the algorithm.An adaptive factor is introduced to enable the algorithm to adaptively adjust the search strategy and search range in the iterative process.An adaptive gradient-based optimizer based on simplex method(SMAGBO)is proposed.To evaluate the performance of the SMAGBO algorithm,14 benchmark functions and CEC-2020 benchmark test functions,as well as five engineering design optimization problems,were selected for testing experiments.Compared with eight metaheuristic optimization algorithms and existing research results,the experimental results showed that SMAGBO algorithm has strong optimization performance in function optimization problems and engineering optimization problems.(3)In order to broaden the application field of gradient-optimizer algorithm,the proposed adaptive gradient-optimizer based on simplex method(SMAGBO)is applied to power system economic dispatch problem.Under the consideration of four problem constraints,simulation experiments were conducted on two application cases to analyze and compare the optimization performance of SMAGBO with the other four algorithms in economic scheduling problems.The results showed that SMAGBO exhibited better search performance than other algorithms.