基于粒子群算法实现原油加工配方优化
Optimization for processing formula of crude oil based on particle swarm algorithm
蒋立沫
作者单位：东南大学
中文关键字：原油调合；非线性；粒子群算法；罚函数法；线性加权法
英文关键字：crude oil blending; non-linear; particle swarm algorithm; penalty function; linear weighted sum method
中文摘要：原油在线调合优化技术的发展与应用，改变了依据经验制定原油加工配方的传统方法。原油调合过程中存在非线性控制指标，传统的线性优化技术难以计算出原油加工的最优配比。本文基于原油基本性质和常减压侧线控制指标，采用粒子群算法实现原油加工配方的优化。首先将各原油数据归一化处理，然后以原油各性质偏差和总成本最低为优化目标，将约束条件以罚函数形式归入目标函数，以线性加权法构造增广目标函数，建立非线性原油调合模型，通过粒子群算法求解模型。实例验证了该求解方案优化出的原油加工配方，在符合各项控制指标的前提下，既符合企业生产加工路线，又使得调合后的原油总价格最低，实现了经济指标的优化。
英文摘要：The development and application of crude oil blending optimization method has changed the traditional method based on experience. There are nonlinear control indexes in the crude oil blending process, using the traditional linear optimization technology is difficult to calculate the optimal proportion of crude oil. Based on the basic properties of crude oil and the indicators of crude distillation unit, this paper employs particle swarm algorithm to optimize processing formula of crude oil. First of all, the crude oil data were normalized, and then the crude oil blending optimization method took the minimum deviation of blending properties and the lowest total cost as the optimization goal. Penalty function was added to object function to convert constrained optimization into unconstrained optimization, and the augmented objective function was constructed by linear weighted sum method. Finally, the non-linear blending model was solved by the particle swarm algorithm. The case of crude oil blending confirmed the rationality and validity of the approach. At the premise of compliance with various control constraints, this solving approach not only met the requirements of all production and processing routes, but also achieved the optimization of economic indexes.