Optimization of Compressed Natural Gas Refueling Station Distribution: Swiss Case Studies

For a successful transition from today's status as niche market product to a mass-market product compressed natural gas (cng) cars need a dense compressed natural gas refueling station network. In the next few years, the number of cng refueling stations in Switzerland will most likely be increased from today’s 50 up to 350 in 2020 in several steps by the natural gas industry. Therefore 300 out of the existing 3470 petrol filling stations have to be chosen for this investment, preferably at lowest possible social cost. This leads to an optimization problem with about 10300 possible different distributions of these 300 cng filling stations. The paper reports and discusses solutions for this problem employing the Simulated Annealing heuristic and appropriate objective functions. In the first case study the objective function uses aggregated data based on the locations of the existing gas stations and municipalities. The second case employs an objective function which approximates the social costs for a specific distribution of cng refueling stations. The approach models the response of the car owners to the distribution by predicting the resulting number of cng cars in each municipality with a multinomial logit model, calibrated to current Swiss ownership patterns.