Integration of feedstock quality monitoring, logistics fluctuations, and production recipe management for creating disturbance-resilient adaptive next-generation oil refineries

Abstract

The article examines the formation of disturbance-resilient adaptive control loops for an oil refining enterprise based on end-to-end integration of feedstock quality monitoring, logistics dynamics, and management of blending production recipes. It is shown that under high market volatility, degradation of the feedstock base, and tightening environmental constraints, traditional schemes relying on static recipes and discrete laboratory measurements systematically reproduce information gaps between technological, logistics, and accounting subsystems, which leads to material balance deviations at the level of several percent and an increase in the share of off-spec products. Based on multiparametric processing of heterogeneous time series and the construction of verified digital twins of key units, a hybrid architecture is proposed that combines physicochemical modeling, probabilistic representation of logistics fluctuations, and predictive algorithms for proactive adjustment of operating modes and recipes. Quantitative results demonstrate that the transition to continuous quality control and dynamic correction of setpoints in gasoline blending radically reduces the dispersion of the octane number while simultaneously decreasing the “quality buffer” and reducing the share of expensive components, thereby shifting control from a reactive to a proactive mode without loss of compliance with specifications. For primary processing, a saturating dependence of technological stability on the forecast horizon of feedstock properties is identified: the most significant reduction in fluctuations of key parameters and an increase in light product yield are achieved when forecasting several hours ahead, aligned with furnace inertia and hydraulic delays. The conducted stress testing under simultaneous supply disruptions and measurement failures confirms that data reconciliation and cross-validation of indirect indicators sharply reduce anomaly identification time, the volume of off-spec production, and the duration of regime recovery. A synergistic effect of subsystem integration is shown, manifested in increased adherence to the production schedule and in the identification of hidden relationships between feedstock variability and catalyst life, which expands the practical applicability of the approach to tasks of margin optimization, industrial safety, and equipment reliability.