The Constrained Differential Immune Clonal Selection Algorithm And Its Application In Optimization Of Gasoline Blending Operation

The petroleum refining industry is one of the pillar industries of our country, which is also an important guarantee to achieve the rapid development of China’s national economy. However, after decades of development, the crude oil always still can not meet the national standard after only one refining, due to the limitations of the technology and process. For this reason the oil blending is emerge. The oil blending is the last step of petroleum products, its effects have a direct impact to the benefit of the refinery. Thence, the research for the optimization of oil blenfing is with high economic value.However, as oil blending process is a complex constrainted multi-objective optimization problem, the traditional approach has been unable to adapt to the requirements of low-cost, low-quality. Therefore, starting from the multi-objective nature of the problem, combined with the existing constraint handling mechanism with multi-objective optimization algorithm, this paper propse a new type of constrained multi-objective optimization algorithm named constrained differential immune clonal selection algorithm. This algorithm uses a new constraint handling mechanism, which overcomes the weak point of Constrained Tournament Method. And it alos uses a two-branch structure, which uses the differential evolution’s global search capability and immune clonal selection’s local search ability to improve the search capabilities of the algorithm. Experiments show that the proposed algorithm has a good convergence and distribution on handling with those several constrained multi-objective optimization problems.At the last of this paper, we choose the gasoline blending optimization to be the representative of the oil blending optimization problems. In this section, a constrained multi-objective optimization model with the low-cost, low-excessed-octane-values objectives for gasonline blending optimization problem is proposed. And with the use of the propsed algorithm, we get a satisfactory optimization result.