Period sewage recycling vehicle routing problem based on real-time data

Abstract The period vehicle routing problem (PVRP) is an important extension of the vehicle routing problem, in which customers have a certain frequency of service. This paper studies a real-time period vehicle routing system: Daniudi gas field sewage recycling, which differs from the classic PVRP because the frequency of customer service is uncertain. A "prediction + two-stage" strategy is proposed to solve the problem. First, the integrated model predicts customers' sewage data generated that day to calculate the warning line so that customers who reach the warning line can be served on that day. Then, because the customer's sewage production changes in real time, a two-stage “pre-optimization + real-time optimization” model is proposed for each day in the period. The two-stage “pre-optimization + real-time optimization” model uses differential evolution (DE) algorithm based on niche clearing to plan the sewage recycling route. The computational results indicate that the proposed technique can reduce actual sewage recycling costs by 17.3%. By performing comparison experiments, we find that the five algorithms examined (i.e., jDE-niche, jDE, DE, GA, and ACO) reduce costs by 17.3%, 12.00%, 10.70%, 9.02% and 8.18%, respectively. Furthermore, the fifteen combined prediction models confirm the validity and effectiveness of our proposed "prediction+two-stage" strategy and jDE-niche algorithm.