Abstract: Aiming at the problems of low convergence precision and premature convergence in teaching and learning optimization (TLBO) algorithms, an improved TLBO algorithm based on self-learning mechanism(SLTLBO) is proposed. In the SLTLBO algorithm, students compare the differences between themselves and the teachers and the worst students, and implement various learning operations independently, so as to enhance their knowledge level and improve the convergence accuracy of the algorithm. At the same time, students make self-examination through Gaussian searching, to jump out of the local area and achieve better global search. The performance of the algorithm is tested by using 10 benchmark functions, and the experimental results verify the effectiveness of the proposed algorithm.