Polycaprolactone microspheres (PCL-MS) are widely used in tissue filling, while particle size and uniformity are crucial for improving filling and therapeutic effects. Efficient optimization methods are essential for accelerating their development. Box-Behnken design was applied to investigate three factors for PCL-MS: PCL concentration (X₁), polyvinyl alcohol concentration (X₂) and water-oil ratio (WOR,X₃). Mathematical models were developed to predict particle size (Y₁) and particle size distribution width (Y₂). Multi-objective optimization of Y₁ and Y₂ was performed using the Nondominated Sorting Genetic Algorithm-II (NSGA-II) and the Multi-Objective Artificial Hummingbird Algorithm (MOAHA) to determine the optimal preparation schemes. Experimental validation was conducted to confirm the validity and reliability of the optimization schemes. After multi-objective optimization, two ideal preparation schemes were selected from the Pareto solution sets obtained by NSGA-II and MOAHA, respectively. Experimental validation showed no significant statistical difference (P>0.05) between the measured and predicted values of Y₁ and Y₂ for scheme 12 and 21 of NSGA-II and scheme 3 of MOAHA, with deviations under 5%. All protocols met target requirements, indicating their suitability for PCL-MS preparation. This study utilized Box-Behnken design and intelligent optimization to develop three alternative PCL-MS formulations, facilitating the production of microspheres with smaller particle sizes and narrower distributions, thereby advancing formulation development.