Data Filtration and Clustering for Purposes of Petroleum Quality Indicators Computation Using Situational Models

Abstract

Generally, advanced process control systems (APC-systems) base on using technological process models tolerating output quality indicators (OQI) and technical and economic indexes (TEI) forecasting “on the fly”. There are many techniques of OQI and TEI evaluation in existence, except one used to work with static information (for example, results of passive experiments) for parametric identification. In addition, control systems store operating parameter’s data in its database in a shape of time sequences without any validation or testing for homogeneity. Inhomogeneity of the data drops model quality to the state, when data makes it impossible to develop a situational model without pre-processing and cluster separation, according to which one creates the situational model. This article considers filtration and clustering techniques of APC historical data, including information about technological mode and used OQIs. Described filtering and clustering solutions based on parity models and technological measures cross-correlation techniques; their implication presented on the example of multioutput fractionating tower. Keywords—advanced control, homogeneous data situational modeling, clustering, data analysis