Traditional energy-saving strategies often fail to address the nonlinear coupling relationships among components in air-conditioning water systems, frequently resulting in local optima. To solve this problem, this research proposes an improved mayfly algorithm (IMA) for global energy optimization of air-conditioning water systems. Targeting minimal total energy consumption while ensuring adequate cooling load supply at terminals, the IMA significantly improves energy efficiency by optimizing operational parameters of all system components. Firstly, the algorithm employs an enhanced Circle chaotic mapping to achieve higher-efficiency population initialization. Then, it integrates an exhaustive search method to optimize discrete decision variables, thereby reducing optimization time. Finally, a dynamic inertia weight mechanism is introduced to balance global and local search capabilities, effectively enhancing optimization efficiency. Experimental results show that the IMA achieves a 8.1% reduction in total energy consumption for a real commercial building's air-conditioning water system while converging within just 15 iterations. These findings demonstrate that the proposed IMA effectively solves energy optimization challenges in air-conditioning water systems, providing a novel solution for practical engineering applications.