Differential evolution (DE) has proved its significance for optimizing various real-world applications and standard benchmarks. In this work, a self-adaptive version of DE is proposed namely LSHADESPA by employing three major modifications, i) proportional shrinking population mechanism for reducing computational burden, ii) simulated annealing-based scaling factor (F) for improving the exploration properties, and iii) oscillating inertia weight-based crossover rate (CR) for a balancing exploitation and exploration. The proposed algorithm has been experimentally tested on IEEE CEC 2017 and IEEE CEC 2021 benchmarks. For performance evaluation, a comparison with respect to JADE, SaDE, SHADE, LSHADE, MVMO, and others has been performed. Experimental and statistical results affirm the superior performance of the proposed LSHADESPA algorithms with respect to other algorithms.