This research tackles the environmental concern of greenhouse gas emissions in the execution of projects, with a focus on multi-site projects where the transportation of resources is a major source of emissions. Despite growing consciousness among consumers and stakeholders about sustainability, the domain of project scheduling has often overlooked the environmental impact. This paper seeks to bridge this oversight by exploring how to reduce greenhouse gas emissions during both project activities and resource transportation. A novel approach is proposed, combining a simulation model with an improved non-dominated sorted genetic algorithm. The simulation model incorporates the stochastic nature of emission rates and costs. This method is further refined with innovative techniques such as magnet-based crossover and mode reassignment. The former is a genetic algorithm operation inspired by magnetic attraction, which allows for a more diverse and effective exploration of solutions by aligning similar ’genes’ from parent solutions. The latter is a strategy for reallocating resources during project execution to optimize efficiency and reduce emissions. The efficacy of the proposed method is validated through testing on 2810 scenarios from established benchmark libraries, 100 additional scenarios adhering to the conventional multi-site problems, and a case study. The Best-Worst Method (BWM) is applied for identifying the best solution. The findings indicate substantial enhancements compared to traditional methods with a 12.7% decrease in project duration, 11.4% in costs, and a remarkable 13.6% reduction in greenhouse gas emissions.