The vehicle routing problem of refined oil secondary distribution with unloading sequence constraints is studied. Multi-compartment oil tankers with limited capacity are used to transport multi-types refined oil to gas stations from oil depot. Each compartment can load only one type of refined oil and the unloading sequence of each type of oil is fixed. Given the demand of each gas station for every type of refined oil, assuming that the demand for one type of refined oil from each gas station can only be served by one vehicle, to minimize the total costs including the fixed costs and transportation costs of the distribution vehicle, a mixed integer programming model is established, and a self-adaptive large neighborhood search heuristic(LNSH) algorithm is developed for solving the model. The effectiveness of the algorithm is verified by simulations and analysis on multiple scale of instances. The experimental results show that using the LNSH algorithm, the global optimal solutions of small size examples can be found with higher probobality; the approximate optimal solutions of middle and large scale examples can be obtained quickly, and the approximate ratio is no more than 1.2. The model and the algorithm can provide theoretical basis and decision support for oil companies to develop refined oil secondary distribution schedules.