Refinery tank farm simulation: flowing vs accumulative tank farm modeling

       In refinery and gas processing systems, tank farms are typically implemented in two fundamental configurations: flowing and accumulative. These two types define how hydrocarbon streams are buffered, distributed, and controlled within refinery logistics networks and directly determine the behavior of storage and transfer operations in refinery simulation models.
       In this article, the primary focus is on the key properties, structural behavior, and functional differences between two main types of tank farms used in refinery and gas processing systems. The analysis is aimed at comparing their operational logic and highlighting how each type behaves differently when implemented in industrial simulation models.
       Tank farms in refinery systems serve as a key buffering and coordination layer between continuous processing units and discrete logistics operations. In a refinery digital twin, they are not only physical storage assets but also decision-driven simulation entities that synchronize material flows with production planning and transportation constraints. This dual role requires separating real-time operational behavior from planning-level inventory and scheduling logic, which is the basis for distinguishing between flowing and accumulative tank farm models.

       Refinery logistics systems cannot be accurately represented using a single generic storage model, as industrial tank farms operate under fundamentally different operational logics depending on their role within the flow chain.
       Two types of tank farms are required because they operate at different levels of refinery system control and serve different operational objectives. Flowing tank farms are typically used for real-time, throughput-oriented operations where continuous material movement and immediate process coupling are critical. In contrast, accumulative tank farms are driven by production planning logic and focus on inventory control, batch accumulation, and shipment scheduling.
       This separation reflects two fundamentally different optimization goals in refinery systems: real-time flow stabilization versus strategic inventory and logistics management within production planning frameworks.

       Tank farms in refinery systems act as a critical synchronization layer between continuous production processes and discrete logistics operations. Within a refinery digital twin, they represent both physical storage infrastructure and a computational control entity that ensures consistency between real-time material flows and planning-level decisions.
       From a logistics perspective, tank farms decouple upstream processing units from downstream transportation constraints such as pipeline capacity, terminal availability, and shipment schedules. This buffering function allows refineries to maintain stable throughput while adapting to variability in demand, supply disruptions, and operational constraints.
       In simulation environments such as PRL-based models, tank farms are explicitly separated into flowing and accumulative logic to reflect this dual role. Flowing tank farms operate at the operational level, supporting continuous balancing of input and output streams. Accumulative tank farms operate at the planning level, where inventory levels are managed according to production plans, storage policies, and shipment requirements.
       Together, these two representations enable a consistent modeling of refinery logistics, linking real-time process dynamics with production planning, inventory optimization, and decision-making within a unified digital twin framework.

       A flowing tank farm model represents a key element in refinery simulation systems, primarily designed to smooth out fluctuations in outgoing material flows directed to downstream processing units, while also providing temporary storage when reception or dispatch is temporarily unavailable. Such tank farms are typically closely integrated with process units and other storage systems that collectively form the total inflow to the facility.
       As a result of varying operating modes of process units, the presence of associated and off-spec streams, maintenance activities, and other operational factors, the inlet flow to the tank farm is inherently time-dependent and variable. In refinery simulation and digital twin models, this dynamic behavior is represented as a time-varying flow rate, which directly influences system stability and buffering requirements.
       A flowing tank farm typically consists of multiple storage tanks (groups) that are hydraulically or logically connected in parallel. During operation, these tanks are synchronized so that an approximately identical residual level is maintained across all units. In this configuration, the tanks are filled and discharged simultaneously, acting as a single coordinated storage system rather than independent units. This synchronized behavior ensures uniform utilization of capacity and stable flow redistribution across the refinery logistics network.
       Such an approach provides effective smoothing of fluctuations in the incoming flow relative to the incoming stream dynamics. In addition, it enables a controlled and gradual system response when the incoming flow drops to zero, ensuring either a smooth shutdown of downstream units or their stable continuation of operation based on remaining available flow and system inertia.

       An accumulative tank farm model represents a key subsystem in refinery simulation environments, focused on stepwise accumulation, storage, and subsequent preparation of product for dispatch in accordance with production and logistics planning requirements. Unlike flowing tank farm configurations, this type of system is not primarily intended for flow smoothing, but rather for the controlled formation of product residuals required for downstream operations, including conditioning, blending, certification, and scheduled shipment execution. Accumulative tank farms are typically closely integrated with loading systems and distribution terminals, which impose strict constraints on amount and discharge rates.
       Such a tank farm consists of multiple storage tanks; however, its defining feature is a sequential organization of operational states, including accumulation, conditioning, certification, and dispatch. At any given time, one tank is actively engaged in accumulation, while other tanks operate in different states such as certification, or waiting. Additionally, one tank is typically assigned to active discharge operations. The roles of tanks are dynamically exchanged over time depending on operational requirements.
       During operation, special attention is given to maintaining a sufficient volume of product that is ready for dispatch in order to ensure compliance with production schedules and logistical commitments.

Refinery tank farms operate at two distinct decision layers. The real-time operational layer is represented by flowing tank farms, focusing on continuous material balancing and throughput stability. The planning and scheduling layer is represented by accumulative tank farms, where inventory levels, shipment execution, and production plans are managed over time.

       Within the Petroleum Refining Library (PRL), two fundamental types of tank farm components are implemented to model distinct operational regimes of hydrocarbon storage and transportation in refinery and gas processing simulation systems. These components form a unified modeling layer for tank farm infrastructure within the AnyLogic environment and serve as standard reusable building blocks in digital twin architectures.
       The first type, the flowing tank farm (RpFlowing), is designed for buffer-oriented storage systems where the primary objective is to smooth fluctuations in incoming and outgoing material streams. In PRL, this component implements continuous redistribution of flows through a synchronized set of active tanks. RpFlowing is used in scenarios where the dominant requirement is stabilization of material transfer between process units, without explicit long-term accumulation or planning logic.
       The second type, the accumulative tank farm (RpAccumulative), implements a more complex state-driven logic focused on product accumulation, conditioning, and planned dispatch. In PRL, this component supports a multi-stage representation of product states, including accumulation, certification (passeportization), and shipment preparation. Unlike the flowing model, RpAccumulative is tightly integrated with logistics and production planning subsystems, enabling execution of scheduled shipments, control of product flows, and management of operational constraints.
       Both tank farm types are implemented as parameterized and reusable components, enabling their deployment across a wide range of simulation scales—from localized process units to full refinery digital twins and distributed logistics networks. In both flowing and accumulative tank farm types, additional flow streams may be present, including loss flows, additives and intentionally destroyed flows, which are treated as separate components of the overall material balance. Their combined use enables hybrid modeling architectures, where part of the system operates under real-time flow stabilization logic, while other parts follow structured accumulation and planning-driven behavior.

       Hybrid modeling in refinery digital twin architectures combines flow-driven and state-driven paradigms to represent both continuous transport processes and discrete storage dynamics within a unified simulation framework. This approach is essential because refinery systems include fundamentally different subsystems that cannot be accurately modeled using a single abstraction.
       Flow-oriented elements, such as sources, pipelines and flowing tank farms, are represented through continuous transfer and buffering logic, where system behavior is dominated by real-time flow stabilization. In contrast, accumulative tank farms rely on state-based state of product and flow balance representation that describe the evolution of product flows and remaining material in the system, including accumulation, conditioning, and scheduled dispatch according to operational plans.
       Within PRL and AnyLogic-based implementations, hybrid modeling is achieved by coupling RpFlowing and RpAccumulative components. Both of these components inherit from a common base class ReservoirPark, which defines shared structural properties, interface contracts, and core material balance logic. This inheritance ensures consistency in how different tank farm types integrate into larger refinery simulation models.
       The main advantage of this approach is multi-scale consistency: RpFlowing ensures short-term flow stability, while RpAccumulative supports long-term planning and control of product flows and residual material states. This significantly increases the realism of refinery digital twins and improves the accuracy of logistics, production planning, and scenario analysis under operational constraints.

       The analysis of flowing and accumulative tank farms demonstrates that refinery storage systems cannot be reduced to a single modeling abstraction without losing essential behavioral distinctions. These two structures represent fundamentally different operational regimes in refinery logistics: flow stabilization versus product-driven planning.
       Flowing systems function as dynamic buffers that mitigate short-term fluctuations between process units and ensure continuous, stable material transport. In contrast, accumulative systems represent structured product-handling nodes where hydrocarbons are accumulated, conditioned, and released according to operational schedules and logistical constraints.
       From a digital twin perspective, separating these paradigms is essential for preserving the accuracy of flow dynamics, storage behavior, and system-wide coordination. The combined use of both approaches enables the construction of realistic hybrid models that capture both real-time process behavior and higher-level planning logic within refinery networks.

1. Why is a single tank farm abstraction insufficient in refinery simulation models?
Refinery systems exhibit two fundamentally different operational regimes: continuous flow stabilization and structured product handling. Flowing tank farms (RpFlowing) operate as dynamic buffers for stream stabilization, while accumulative tank farms (RpAccumulative) represent state-driven product transformation and controlled dispatch. A unified abstraction cannot capture both transient flow behavior and plan-driven material evolution without significant loss of simulation fidelity.

2. What is the fundamental modeling distinction between RpFlowing and RpAccumulative?
RpFlowing represents a continuous material transfer system where dynamics are governed by flow balancing between connected units. RpAccumulative represents a state-transition system where product evolves through operational stages such as accumulation, conditioning, and shipment execution. The key distinction lies in flow continuity versus state-based transformation of product.

3. How do these tank farm types affect refinery digital twin behavior?
RpFlowing primarily influences short-term system dynamics by stabilizing fluctuations in material transport. RpAccumulative determines long-term system behavior by controlling how product flows are accumulated, processed, and released according to operational constraints. Their interaction defines both real-time responsiveness and strategic production behavior of the digital twin.

4. What modeling errors occur if only one tank farm type is used?
Using only RpFlowing leads to loss of structured product handling, making planning and shipment behavior unrealistic. Using only RpAccumulative distorts real-time flow dynamics, eliminating realistic buffering effects between process units. In both cases, the result is incorrect material balance evolution and reduced predictive accuracy of the model.

5. Why is inheritance from a common ReservoirPark class important in PRL?
Both RpFlowing and RpAccumulative inherit from a shared base class ReservoirPark, which defines common structural interfaces, material balance logic, and integration rules. This ensures architectural consistency across PRL components and allows both tank farm types to be seamlessly integrated into unified refinery simulation models.

6. What is the advantage of combining RpFlowing and RpAccumulative in hybrid models?
Combining both components enables multi-scale simulation: RpFlowing ensures stable real-time flow propagation, while RpAccumulative manages structured product evolution under operational constraints. This hybrid structure improves the realism of refinery digital twins by aligning short-term process dynamics with long-term operational and logistical behavior.