Comprehensive and practical optimal delivery planning system for replacing liquefied petroleum gas cylinders

In the daily operation of liquefied petroleum gas service, gas providers visit customers and replace cylinders if the gas is about to run out. The plans should both prevent gas shortages and realize the minimum working time. Existing research has two limitations: the absence of a comprehensive system and the difficulty of solving large-scale problems. In the former limitation, existing research tackled the partial problems of making plans for cylinder replacement, such as planning delivery routes given gas consumption forecast or determining the customers for visiting without obtaining the route. It does not consistently achieve gas shortage prevention and short working hours even when combining individual optimal methods. In the latter limitation, most existing studies have difficulty solving the problem within a reasonable time if there are many customers. This is because they simultaneously determined the customers for visiting and route planning by preparing binary variables representing customer-to-customer travel. In this study, we construct a comprehensive and practical system from gas consumption forecast to determine delivery routes for cylinder replacement with large-scale customers. To address these challenges, our method takes two steps: determining which customers to visit within several days and a single-day route. Moreover, we mitigate gas shortages among customers with poor forecast performance by considering the uncertainty of the gas consumption forecast. A field test involving over 1000 customers in Japan confirmed that the system is operationally viable and capable of preventing gas shortages and realizing short working time.