This paper proposes a new Teaching–Learning Algorithm ( TLA ) that uses the chaotic map to prevent the conventional TLA from getting stuck on local optima and enhancing the convergence characteristics, due to non-repetitions nature and ergodicity of chaotic functions. Some shortcomings are encountered in the original TLA , for instance, it can be trapped in local optima. This work tries to improve it by substituting the random in the initial algorithm with chaotic sequences. At this level, the initial population is chaotically generated and chaotic values are used in both phases. The global solutions are further enhanced by adding a new third chaotic phase. To demonstrate the effectiveness of the improved Teaching–Learning algorithm ( ITLA ), a fifteen of well-known benchmark functions are used. Experimental results demonstrate that ITLA outperforms significantly the conventional TLA , in terms of the accuracy of the final solution and the speed of convergence. The enhanced optimization algorithm is employed to solve the coordinated design problem of power system stabilizers ( PSS ) and thyristor-controlled series capacitor ( TCSC ), in order to investigate the feasibility and effectiveness of the proposed method in power systems. The performance of the proposed controllers is evaluated on a multi-machine power system under large disturbance and for different operating conditions through a nonlinear time-domain simulation. At the end, the results confirm the robustness of the proposed controllers in comparison to PSS designed by ITLA ( ITLAPSS ) and TCSC designed by ITLA ( ITLATCSC ). [ABSTRACT FROM AUTHOR]