The bulk accumulation of subway passenger flow at stations during peak hours is considered to pose a significant challenge to passenger organization and train scheduling, while traveling efficiency and boarding fairness across different spatial and temporal dimensions are severely compromised by traditional uniform scheduling strategies. To address this thorny problem, an unbalanced train scheduling strategy with carriage reservation under oversaturated conditions is proposed to improve passenger traveling efficiency and boarding fairness. By introducing two groups of cumulative 0–1 variables, a matching relationship between dynamic passenger demand and scheduled trains is established, and boarding fairness for passengers located at different stations is measured by the maximum number of trains missed by each passenger. A collaborative optimization model for train scheduling and carriage reservation is developed to minimize total passenger waiting time under time-dependent passenger demand. To efficiently solve the nonlinear mixed-integer programming model, a heuristic algorithm based on adaptive large neighborhood search is developed to obtain near-optimal solutions efficiently, in which several effective destroy and repair operators are designed to accelerate algorithm convergence. Finally, numerical experiments are conducted based on the Beijing Subway Daxing Line to verify the performance and effectiveness of the proposed model and algorithm. The experimental results indicate that high-quality train timetables and carriage reservation plans can be effectively obtained within a short computing time using the proposed method. Compared with the traditional even-headway schedule, total passenger waiting time is reduced by 57.53%, and the maximum number of missed trains per passenger is decreased from 10 to 1, thereby significantly improving boarding fairness.