An Improved Gaussian Clustering Algorithm to Accelerate the Solving of Integrated Optimization of Crude Oil Procurement and Blending

The crude oil procurement and blending process is the core link of production and operation in oil refining enterprises, and its integrated optimization is essential to guarantee blended oil property stabilization. However, the integrated model will be difficult or even impossible to solve if there are too many alternative crude oils. To handle this problem, an improved Gaussian clustering algorithm is proposed to select the appropriate crude oils for integrated optimization. Through this preselection, the search space is highly reduced, and thus the solving will be accelerated. First, an improved Gaussian Mixture Model (GMM) algorithm for crude oil clustering is detailed, and the Expectation-Maximization (EM) algorithm with price compensation function is proposed to reduce the influence of crude oil price fluctuation on the clustering results. Then, the effectiveness of the improved GMM algorithm was verified through actual cases and comparative experiments. The results showed that the improved GMM algorithm reduced the solution time by about 90%. The improved GMM algorithm can significantly improve the solving efficiency, and the EM algorithm with a price compensation function can effectively reduce the impact of crude oil price fluctuations on the clustering results. Finally, the proposed crude oil preselection-based optimization method lays the foundation and possibility for field application use.