Multi-project Scheduling Optimize With Stochastic Activity Duration And Resource-constrained

With the rapid development of information technology,project management has shifted from the management of a single project to the coordination of resources across multiple projects.At the same time,uncertainties both inside and outside the project,such as a lack of project experience among management personnel or equipment failures requiring repair,can cause deviations in project activities and planned schedules,making management more difficult and increasing corresponding cost consumption.When a manager takes on a new multi-project,they must consider that the preparation work for different scheduling algorithms varies greatly.When the project execution time is tight,the manager should prioritize algorithms that can quickly formulate project scheduling plans.When there is ample preparation time for a project,the manager can choose algorithms that can bring greater benefits to develop the project scheduling plan.This article focuses on the study of the random activity duration,resource-constrained multi-project scheduling problem,and explores the performance of various scheduling algorithms under distributed multi-project resource-constrained scheduling problems through simulation experiments.Based on the existing literature,this article expounds on the research status of the multi-project scheduling problem,adapts the algorithm examples of the multi-project scheduling problem standard database(www.mpsplib.com)to meet five different activity duration distributions,proposes a multi-operation differential evolution algorithm and a deep reinforcement learning model for multi-project random scheduling,and selects 11 priority rules with better performance from various priority rules summarized in previous literature.Using the established multi-project scheduling problem dataset,this article analyzes the performance of the 11 priority rules,multi-operation differential evolution algorithm,and deep reinforcement learning algorithm in different multi-project features.Finally,an actual assembly-based building multi-project case is selected to verify and analyze the experimental results of the three algorithms.The experimental results show that: first,the performance of priority rule algorithms in multi-project problems is affected by the problem size,resource conflict degree,and uncertainty of random activity duration of the project.When the problem size of the multi-project and global resource conflict are both small,MINSLK,MINLFT,and EDD priority rule algorithms are the best choices.If the problem size of the multi-project is large and resource conflict is not significant,MINSLK and EDD priority rules can be prioritized.If the multi-project scheduling problem size is small and resource conflict is significant,managers should prioritize using the TWK-EST and TWK-LST algorithms for project scheduling.Otherwise,when the problem size and resource conflict of the project are both large,the WMDD algorithm should be given priority consideration.Second,the proposed multi-operation differential evolution algorithm has better global search ability than traditional intelligent optimization algorithms in solving resource-constrained multi-project scheduling problems,and in small-scale multi-project scheduling problems under constant activity duration,the MODE algorithm’s resistance to activity duration interference in the scheduling process shows an upward and then downward trend.Third,under constant activity duration,the MODE algorithm’s solution effect is superior to the priority rule algorithm and the DRL algorithm in all data sets,and the percentage of savings of the MODE algorithm on the priority rule algorithm is greater than that of the DRL algorithm,with the maximum percentage of savings reaching 42.8%.Under random activity duration,the DRL algorithm’s solution effect is better than that of the MODE algorithm and the priority rule algorithm,and the scheduling speed of the trained model is equal to that of the priority rule algorithm.Based on the above experimental results,project managers can choose the best scheduling algorithm for the current situation to develop a project schedule based on the urgency level,problem size,activity conflict degree,and environmental uncertainty of the multi-project they are handling.