A genetic algorithm-dung beetle optimization (GADBO) is proposed to address the problem of low convergence accuracy of the dung beetle optimization algorithm due to being trapped in local optima. This improved algorithm combines the strategies of initial population and mutation disturbance to escape local optima and enhance global search capabilities. Through the testing and comparison of 10 benchmark functions
the effectiveness of the GADBO algorithm improvement has been verified
and GADBO has higher optimization accuracy than other swarm intelligence optimization algorithms. The GADBO algorithm is applied to the hyperparameter optimization of long short-term memory (LSTM) neural networks
and a photovoltaic power generation prediction model is established. The simulation results show that the LSTM prediction model optimized by the GADBO algorithm has a fitting coefficient of 98.45%
and the average absolute error and average absolute percentage error are also minimized. The accuracy and precision of the model have been improved
verifying the applicability and effectiveness of GADBO in neural network optimization.
ZHANG X Q, ZHANG Y Y, MING Z F.Improved dynamic grey wolf optimizer[J]. Frontiers of information technology & electronic engineering, 2021, 22(6): 877-890.
XUE J K, SHEN B.Dung beetle optimizer: a new meta-heuristic algorithm for global optimization[J]. The journal of supercomputing, 2023, 79(7): 7305-7336.