关键词: 库存管理, 油品在线调和技术, 多变量广义预测控制算法, 遗传算法

Abstract: An optimal process model from a systematic perspective was constructed to solve integrated inventory management problem in refinery under demand uncertainty and on-line continuous blending technology. Firstly, demand predictions for various products were conducted; then, according to the predicted results, a Genetic Algorithm (GA) using real number coding was proposed. Based on GA, the optimized streams of intermediate oils and crude oils, the side-draw yields factors and the local optimized results of the multi-inventory were gained under the constraints of final products non-linear blending quality and crude oils linear blending quality functions in the local optimization controller . Finally, a multivariable Generalized Predictive Control (GPC) algorithm was presented to simulate the misshaped dynamic and uncertain behavior of the real system. The original models system states was updated in time, and the whole optimal operation rolling horizon strategies which could minimize the total inventory cost were executed on-line within the simulation cycle. A case study showed that the model could work with feasibility.

Key words: inventory management, on-line continuous blending technology, multivariable generalized predictive control algorithm, genetic algorithm