Subsidy plan modeling and optimization for the promotion of green technologies in maritime transportation

Abstract

Environmental protection and sustainable development are recognized as one of the major issues in the maritime industry. Various green technologies have been developed to alleviate the ship emission problem. In this thesis, we focus on the promotion of two of the comparatively common technologies, namely shore power and using liquefied natural gas (LNG) as marine fuel. Government subsidies are powerful tools for the promotion, but existing studies on them stop at qualitative analysis or policy evaluating. In this thesis, we aim to fill the gap and study the subsidy plan modeling and optimization for the promotion of shore power and LNG as marine fuel. The thesis consists of three studies. Chapter 2 focuses on the application of subsidies in the promotion of shore power and obtains the optimal subsidy design for the port to encourage ships to use shore power while berthing. The trade-off between the environmental benefits and subsidy expenses is the main issue to be addressed. Considering the characteristics of the port, including the unit environmental benefits of emission reduction, the electricity price, and the historical data of ship visits, a stochastic model was built to describe the problem. We make full use of existing data of ship visits to make an approximation of the visits in the coming year, which is a closer estimate than that of existing relevant studies. Taking advantage of the problem structure, we convert the model into a deterministic one by applying sample average approximation (SAA) and binomial distribution. Next, without the loss of generality, we reformulated the model and made the model tractable so that it could be solved by CPLEX. Abundant numerical experiments were conducted to validate the model and show the influence of values of crucial parameters on the optimal solution. We summarized useful managerial insights from the numerical experiment results and sensitive analysis. Chapter 3 investigates the government subsidy plan optimization for LNG as fuel for maritime transportation. In this problem, the government provides subsidies for ports and ship operators to cover part of the LNG bunkering station construction cost and ship conversion cost. With the aim to maximize the net benefit, namely the environmental benefit minus the subsidy expenditure, the government needs to decide the amount of subsidies to be offered. Given the relationships between different parties, we abstract the problem into a trilevel programming model that consists of the government, port, and ship levels. Taking advantage of the behavior rules of ship operators, we convert the bilevel (port level and ship level) problem into an equivalent single-level problem. Then, with a enumeration algorithm, we identify the optimal subsidy plan for the government. The proposed model and solution method were validated by a series of numerical experiments with realistic parameters. Chapter 4 explores the LNG bunkering station deployment problem. Due to the limited annual budget, the government cannot build LNG bunkering stations at all ports in the area at a time. In practice, it will take several years to complete the building of LNG bunkering system. Despite that the ports at which LNG bunkering stations will be built are predetermined, the specific construction sequence are flexible, as well as the construction situation in each period. Considering that ship routes in the area have different port of calls, ship emission of each route also varies with the construction sequence. Therefore, this chapter aims to identify the optimal construction sequence that minimizes the total ship emission in the construction period. A two-stage method is proposed to solve the problem. In the first stage, we reduce the number of potential optimal solutions of each ship route, and in the second stage, decision matrices are adopted to indicate the choice of shipping lines and convert the bilevel problem into a single-level problem that can be solved by CPLEX after linearization. Comparison between the results of the two-stage method and a greedy algorithm shows the superiority of our method.