基于分片线性代理模型的成品油调合优化

Product oil blending optimization based on a piecewise linear surrogate model

成品油调合对提高炼厂经济效益有着重要的作用和意义。成品油调合优化是一个非线性约束优化问题,传统的进化算法由于搜索空间大又没有结构信息,要取得期望的求解效率和解的稳定性都是具有挑战性的任务。针对上述问题,提出了一种基于分片线性代理模型的成品油调合优化方法,它包含分片线性建模和优化2部分内容。首先,利用分片线性函数模型作为成品油调合非线性调合性质指标函数的代理模型,将原非线性约束优化问题转化为一系列线性规划子问题;然后,利用差分进化算法搜索相关线性子区域来获得全局最优值,以达到提高进化算法的求解速度和避免算法陷入局部最优解的目的;最后,通过成品油调合优化案例验证了该方法的有效性。

Product oil blending has a significant impact on refinery profits. Product oil blending optimization is a nonlinear constrained optimization problem, which is challenging for traditional evolutionary algorithms, because they lack structural information and have a larger search space. This paper presents a method for product oil blending optimization based on a piecewise linear surrogate model, which includes pieeewise linear modeling and piecewise linear optimization. The nonlinear properties of the product oil blend are approximated by a surrogate model based on a piecewise linear function to transform the original problem into a piecewise linear programming problem. Then, a differential evolution algorithm is used to search linear sub regions for the globally optimal solution to improve the solution speed of the evolutionary algorithm and guide its search away from localoptimums. This method is used to solve a recipe optimization problem for gasoline blending, with the feasibility and effectiveness confirmed by the optimization results.