Genetic learning particle swarm optimization (GL-PSO) improves the performance of particle swarm optimization (PSO) by breeding superior exemplars to guide the motion of particles. However, GL-PSO adopts a global topology for exemplar generation and cannot preserve sufficient diversity to enhance exploration, and therefore, its performance on complex optimization problems is unsatisfactory. To further improve GL-PSO's performance and adaptability, two modifications are incorporated into the original GL-PSO. A ring topology is adopted in exemplar generation to enhance diversity and exploration, while a global learning component (GLC) with linearly adjusted control parameters is employed to improve the algorithm's adaptability. The resultant algorithm is referred to as global genetic learning particle swarm optimization with diversity enhancement by ring topology (GGL-PSOD). To validate the effectiveness of these two modifications, they are combined with GL-PSO separately and together and further tested experimentally. The comparison results on the CEC2017 test suite show that the adoption of the ring topology in exemplar generation can enhance the diversity and exploration capability of GL-PSO, while combining GLC alone with GL-PSO cannot achieve significant improvement. Incorporating both modifications into GL-PSO, the resultant GGL-PSOD exhibits high performance and strong adaptability on different types of CEC2017 functions. It outperforms seven representative PSO variants and five non-PSO meta-heuristics, including GL-PSO, SL-PSO, HCLPSO, EPSO, ABC, CMA-ES and CS.