Calibrating microscopic traffic simulation models is a prerequisite for simulation applications. This study proposes three novel methods to improve the accuracy and interpretability of the calibration model. The proposed approach involves selecting the calibration parameter, refining the model parameter system, and optimizing the calibration results. The first method expands the single-point mean into a multi-point distribution. The cumulative distribution curve of delay was selected as the calibration parameter. The second method divides the parameter system into global and local parameters. Global parameters were calibrated using NGSIM measured data, and local parameters were calibrated through intelligent algorithms. The third method proposes a methodology of parameter clustering recursion based on the genetic algorithm results, with information entropy selected as the analysis index. To evaluate the effectiveness of the proposed optimization methods, this study used NGSIM trajectory data as a case study. Eight simulation schemes based on the three optimization methods were designed, and simulation experiments were conducted using the VISSIM platform. The results show that the accuracy of the multi-point distribution calibration and parameter value optimization method is significantly higher than the default method. Additionally, the optimization method with calibration of both global and local parameters was more consistent with actual driving characteristics. This study provides a theoretical foundation for improving the practical application of traffic simulation technology, which has significant implications for transportation planning and management.