Critical paths are widely used in project schedule management, but most existing studies ignore the volatility and dynamic characteristics of project duration, which increases the risk of project implementation. This paper proposes a dynamic optimization model of random duration and the EVM (earned value method) based on critical path technology. The random duration based on a lognormal distribution is introduced into critical path technology, and the completion probability model is established to evaluate the fluctuation of the project completion schedule. Then, the earned value correction model is established and combined with performance indicators to improve the dynamic management system of proactive scheme scheduling and reactive scheme scheduling. The proposed method can fully consider the fluctuation and dynamics of the project duration, effectively reduce the errors caused by the uncertain fluctuation of the duration, and further improve scientific management and reduce management costs.