To enhance the performance of the competitive swarm optimization (CSO) algorithm in solving complex high-dimensional optimization problems, an improved CSO algorithm is proposed based on the partition strategy and convergence speed controller (PCSCCSO). First, a dynamic partition strategy driven by fitness change rate is introduced to improve the convergence and search efficiency. Then, a fast CSO strategy is designed with a triple competition mechanism to strengthen the optimization capability. Winning particles are updated via an opposition-based learning strategy, defeated particles learn from the average position of the winning subgroup and inferior particles undergo mutation-enhanced local search. These mechanisms can effectively balance global exploration and local exploitation, and can enhance the optimization efficiency. Finally, an adaptive convergence speed controller is developed by integrating the cosine similarity between particles and the global best solution with stagnation counting, which can dynamically regulate particle search behaviors to avoid local optima and accelerate global convergence. Theoretical analysis demonstrates the stability and convergence of the proposed algorithm. Experiment results indicate that, compared to other improved algorithms, the PCSCCSO algorithm can obtain superior convergence accuracy and efficiency in solving complex high-dimensional optimization problems.