Crude Scheduling: How a small workaround can have a great impact on GRM...

The ability to monitor and determine the quality, composition, and availability date of the crude oil that will be processed in the crude unit is necessary for the creation of a refinery's operations plan. For the best processing margin, batches fed to the CDU need to meet both the quality constraints set by the operations department and the maximum amount of opportunity crude oils.

Ship demurrage costs can rise as a result of inefficient scheduling, which can reduce the theoretical monthly result predicted by linear programming planning models by up to 5%. As a result, optimizing crude oil logistics is crucial, particularly when crude inventory is limited and crude quality cannot be separated.

There are few companies which offers an innovative algorithm to concurrently manage crude oil logistic events (reception, transfer, and processing) and predict the hourly evolution of the status of all tanks. Crude data, simulation, and optimization techniques are integrated by an algorithm which is also able to work high-level transfer instructions where the sequences of tank loading/unloading operations are not detailed.

The calculation engine manages the service requests associated with material transfers according to specified priorities and selection criteria, to bring each event to completion as fast as possible. This simulation highlights real bottlenecks to manage to prevent operational problems. The software can solve autonomously the major components of the scheduling problem, requiring the user’s intervention only to manage significant issues requiring appropriate action.

Furthermore, integration of economically driven optimization methods generates operational plans to maximize the processing of opportunity crudes while meeting the CDU‘s feedstock quality requirements. Significant time saving enables schedulers to explore a greater number of options, resulting in a more effective and optimized scheduling plan. It is a tool that supports, in the same environment and with the same model, development of both long and short term scheduling tasks:

? Long term scheduling models can calculate and update in a few minutes the evolution of the status of all tanks and pipelines in the logistic network. It is able to define cargo arrival dates and to identify solutions in case of unforeseen operational changes.

? Short term scheduling models enable detailed instruction reports to be issued and store validated operational results (that is, actually executed operations) into a centralized historical database which can provide the information to reproduce details of past activity.

This article describes the technology and reports a case study illustrating its implementation for the scheduling of crude supply operations for an Indian petrochemical refining complex. The operator’s maritime depot receives a variety of crude oils of variable quality (in terms of sulphur content, acidity, and API). The number of available tanks does not enable proper quality segregation, therefore each tank’s evolving composition must be tracked.

Concurrently with cargo reception, some tanks are unloaded and fed in parallel (three to five pumping channels) to a pipeline. The blend resulting from parallel pumping must respect the quality constraints required for CDU processing at the other end of the pipeline. Batches exiting the pipeline can either be fed directly to the crude unit or go to refinery storage.

Modules:

There are many companies who has developed a set of technologies for crude oil characterization, plant simulation, and blending calculations and has integrated them into various modules, designed to support the most relevant planning and scheduling tasks. These modules can work standalone to perform a specific set of scheduling tasks, or share data and processes in case of extended solutions.

Crude scheduling:

Figure 1 summarizes the findings of a study carried out to identify the refining operations which mostly impact the difference between the theoretical result expected from the business plan and the actual result.

According to these results and for various reasons, the actual refinery margin can decrease, compared to the planning figure, by up to 25%. One of these reasons is improper crude logistics management (reception facilities, deposits, and pipelines) which can represent, for a mid-sized refinery, an annual $15-20 million loss: this is particularly true when variability of feedstocks forces operators to monitor the quality of their crude refinery tanks or to store different crude types separately to prevent quality contamination.

Typically, the average residence time of crude oil in refinery inventories is too low to enable proper quality segregation, and it is fundamental to improve the intelligence of tools dedicated to the scheduling of logistic assets, to manage the quality of the stocks finally fed to refinery crude units. In such cases, the crude scheduling process becomes critical, and great effort and investment is made to improve performance in this area.

There are few cases which lists the leading causes of economic losses arising from poor scheduling of crude oil logistics. In practice, LP models assume the capability to process optimal crude mixes in refinery crude units for the whole period, while in fact feedstock availability depends on the supply schedule and the logistic constraints involve unforeseen quality contamination. In this case, the refiner must exploit all degrees of freedom available during the operational planning and scheduling process to maximize the final result. Crude batches fed to the CDU must fit the quality constraints set by the operations department and maximize the content of opportunity crude oils for the best processing margin.

Typically in this framework, crude quality and cargo size is input from planning while scheduling can define the arrival dates of cargoes as well as handling operations throughout the logistics network up to the crude unit. In the case of definition of cargo arrival dates, the impact of this material intake on crude logistics must be foreseen.

Given the supply program, it is crucial to plan handling operations to avoid undesirable contamination, especially in the case of different operating modes based on different crude qualities, and to optimize batches by maximizing the use of opportunity crudes. The two main classes of scheduling activities differ by specific objective and time horizon; they are long term scheduling and short term scheduling. Proper execution of both activities requires the capability to simulate the evolution over time of the status of crude logistic assets (tanks, tank farms, pipelines), depending on supply program, transfer operations and processing runs.

Long term scheduling:

Long term scheduling assesses the feasibility of crude reception and transfer operations, in terms of blend quality and reliability, within a time horizon of a few months for a given schedule of crude cargo arrivals. The goal is to assure an ordered cargo arrival plan, timely unloading of cargoes at the maritime terminal with no demurrage costs, and blending to maximize consumption of heavy/opportunity crudes. A typical time horizon is 90-120 days. For example:

? M: current month, day 0-30 (Short term horizon).

■ Crude arrivals gates and dates are fixed (with three days variability)

■ Once an actual cargo reception date and time is available, the simulation must be re-run.

? M+1: next month, day 31-60 (nomination horizon).

■ Known: quality, size, and number of cargoes (nominated cargoes)

■ Unknown: arrival gates

■ Task: to define the contractual arrival dates

? M+2/M+3: following month(s), day 61 to 90-120 (buy horizon).

■ Task: suggest the quality and the size of the cargoes to be purchased With this activity, the scheduler defines the arrival dates of the crude cargoes to be nominated (typically at M+1) and finalizes cargo grades, quantity, and arrival dates for the months M+2 and M+3.

With this activity, the scheduler defines the arrival dates of the crude cargoes to be nominated (typically at M+1) and finalises cargo grades, quantity, and arrival dates for the months M+2 and M+3. For these purposes, the scheduler sets up a model of the logistic network and uses it to:

? Simulate the evolution in time of tank content considering arrivals, pipeline dispatches, and processing. The simulation aims to check the feasibility of cargoes’ reception schedule considering the status of the tanks at reception dates (volume and quality) and contemporary transfer and processing activities.

? Plan the composition of the crude batches. The scheduler must track the composition of the batches fed to pipelines and crude units to prevent issues during processing steps.

? Define contractual arrival dates for ‘nominated’ cargoes for the four months covered by the simulation.

? Estimate arrival dates for ‘to buy’ cargoes, providing feedback to the planning department and trading departments charged with the purchase of new cargoes. The long term simulation of a crude logistics operation is a time-consuming activity and the availability of a fast simulation tool is fundamental to enable the evaluation of alternative scenarios.

Short term scheduling:

Short term scheduling aims to update the status of tanks during the current month, considering deviations from the original schedule that occurred in the period. Actual operations are set in the model and recalculated to align to reality the status of the tanks and to check the feasibility of the scheduled blends. Short term scheduling supports the publication of daily operative instructions as well as the historicisation of actual movements. The typical time horizon for short term scheduling is one month (30 days).

In this case, the scheduler uses a model to:

? Re-schedule short term operations to handle unexpected events (minor changes, delays, or unforeseen shutdowns) and to update the simulation baseline accordingly.

? Calculate the status of the tanks based on actual operation accounting for schedule deviations: the logistic model can be used to keep track of the crude composition of the tanks.

? Reconcile operations: for each planned transfer, the scheduler adjusts the planned quantity to the actual one.

? Store the results of the past operation (after reconciliation) to data historian so that they can be retrieved to support future scheduling activities.

? Publish operating instructions for a horizon of about 10 days.

Even though the horizon is shorter than in the case of long term scheduling, this activity is also time consuming (the input required is much more detailed). Thus, using a tool supporting the reconciliation activities is particularly useful.

Logistic simulation:

Requirements:

Both long and short term scheduling require the availability of a model able to calculate the evolution of the status of the logistics assets depending on the events occurring in the simulation period. Tools used to develop and run such models usually feature a time consuming modelling phase (all transfer data must be manually entered), and need to update such data when any deviation from the plan occurs. Basically, instead of spending time evaluating alternative scenarios, a lot of it is wasted inputting and updating data.

The software algorithm applies a paradigm change. In fact, it autonomously proposes a solution (that is, the origin and destination tank of a given service) after processing ‘high level’ instructions from the user. Setting the criteria for tank selection in different situations is much faster than preparing the detail of every transfer; this potentially reduces the time for modelling tremendously. On the other hand, the tool must provide the user with the flexibility to orientate the solution to account also for constraints which are not explicit in the model but which are taken into account while generating the schedule.

To enable the software to propose reliable and feasible solutions, software has made available in the same environment crude characterization data and MIP optimization, and has developed a simulation algorithm which is able to exploit these tools to generate the schedule automatically. Following research based on user feedback, automation was able to develop a fast and effective algorithm based on deep integration of crude assay data (bulk and fraction properties, economics), and a sophisticated simulation engine enabling, at the same time, high level set-up of transfer events, the flexibility to change and orientate the solution, and economics-driven MIP optimization models to run automatically to optimize batches.

This simulation engine takes into account operational and quality constraints, suggests the sequence of handling operations, and requires only a few minutes for a long term scheduling solution.

Crude assay data:

There are various modules for the construction and management of the crude oil characterization database. This application features a proprietary technology (multidimensional regression) specifically developed to elaborate crude assay information to produce a consistent library which can provide the data for all the properties of any fraction of any crude blend, irrespective of the source, form, and consistency of the input assay.

Model characterizes every crude oil as a mixture of pure components (C5-) and ‘pseudo-components’ (C6+), which cover the entire crude boiling range. Each pseudo-component includes pure components boiling in a narrow range of 10°C. For each property, model uses multidimensional regression to create distribution curves and allows harmonization of the shape of the resulting curves.

Any software lists the fundamental properties managed, the relative blending rule, and a meaningful boiling range (with property values estimated for each pseudo-component). Software enables the user to define the list of quality specifications to be considered to track the quality of tank contents or to set up quality constraints for tanks or crude blends (in case of quality-controlled parallel sequences). It is possible to consider almost any characterization property but, since a high number of quality specifications involves longer computing time, it is recommended to find a good trade-off between the constraints needed to characterize the tanks and the calculation resources.

Software foresees four types of quality specifications. For each type, it is necessary to define a specific set of parameters to enable their retrieval from the model library:

? Bulk: a whole crude referring property. It can be either a generic property (density, sulphur, viscosity) or the yield of a fraction.

? Cut: a crude oil’s fraction referring property (for example, the acidity of the 150-250°C fraction).

? Crude: the amount of a given crude oil in the mix.

? Crude type: the amount of a given crude type in the mix.

The simulation algorithm:

Some softwares differs from other scheduling tools because the detailed list of operations associated with each transfer event is an output of the simulation instead of an input. While other systems simulate the impact of the set of transfer operations input by the user on the status of the tanks, It can generate the set of transfer operations based on the high level instructions provided.

This difference reduces tremendously the time needed to produce/ update the scheduling plan, both for long and short term scheduling. The simulation algorithm of scheduling software can handle ‘high level’ instructions and propose a solution without needing to specify in detail the sequences of origin and destination for tanks involved in a transfer. The user can specify in detail origin, destination quantity, and flow rate (as in the case of other scheduling tools) but can also delegate to the algorithm the selection from a set of tanks of which tank to use at a given moment (‘Sequence’). In this case, it suggests the best tank for the requested service, based on user criteria.

The algorithm splits the simulation period into time slots and elaborates them in series from first to last. For each transfer event, the engine generates a list of service requests to the connected objects (tanks, tank farms, pipelines) represented in the logistic structure and manages these requests according to specified priorities and selection criteria, to bring each event to completion as fast as possible. The system manages various types of transfer events such as carrier and pipeline crude reception, transfers, and processing. The following information defines each event:

? General data (start date, flow rate, quantity, calendars)

? Origin and destination sequences (set of tanks and selection criteria)

? Linking pipes

The algorithm selects origin and destination tanks from the specified sequences and defines the transfer flow rate based on:?

? Volume and pumping (load/ unload) constraints

? Status (volume and content) and availability

? Handling operations (drainage, measurements)

? Quality specifications (quality constraints set for the receiving tank)

? Batch quality targets (in case of parallel pumping)

? Pipeline quality tracking

? Tank selection logics and exceptions. The calculation run generates the details of the operations associated with each transfer event as well as the time evolution of the status of any asset in the logistic network.

Batch optimization engine:

Software uses an optimized blending function, enabling it to determine origin tanks and blending ratios in the case of automatic parallel origin sequences. When required, the algorithm automatically formalizes a mixed integer programming optimization problem applying the constraints set, and solves it to find the origin tank of each pumping channel and the related pumping flow rate. The simulation algorithm then uses this result to elaborate on the transfer event until the solution is applicable. When for some reason the solution becomes unfeasible (for example, one of the origin tanks empties) a new problem is formalised based on the updated scenario.

Model set-up:

The simulation model foresees the preliminary set-up of a logistic network; it features the following modelling objects:

? Tank: modelled with their actual geometrical properties, pumping constraints, and quality specifications.

? Pipeline: connecting tanks can be one way, two-way or a simple connection

? Tank operations: to model operations triggered by loading or unloading

? Docks: cargo mooring points

? Ships’ templates: different types of available crude carriers

? Crude types: crude oil associations

? Quality specifications: modelling the quality of crude oil batches

? Calendars: defining timetables for tank operations or events

The logistic network groups tanks in tank farms which can be actual or logical, depending on the simulation. It is easy to modify and update: Software graphical user interface enables the user to customize the behavior of each object to reproduce actual operating procedures. Once the logistics are defined, it is necessary to define the transfer events occurring during the simulation period. It features four types of events that model all types of operation; each event is defined by general information (start date, flow rate, quantity, calendars), origin sequence (set of origin tanks and selection criteria), destination sequence (set of destination tanks and selection criteria), and linking pipes.

The algorithm uses this information to select, for each time slot, the origin and destination tanks as well as transfer flow rate. The events editor enables the definition of each transfer as well as the related parameters (type, calculation mode, origin, destination, volume, flow rate, and quality constraints). In this environment, it is also possible to set high level transfer instructions.

Results:

The simulation calculates, with hourly resolution, the evolution of the status of all the tanks as well as the detail of material transfers. An exhaustive report section is also featured. Examples of the different types of reports generated by the simulation – all exportable in Excel – include:

? Gantt chart reporting the events with transfer origin and destination?details and highlighting transfer issues of any kind.

? Quality, composition, and origin tanks of all the crude oil batches fed to the pipelines and CDUs during the simulated period, both on a daily average and a batch base.

? Status (volume, composition, and quality) of all inventories at a selected date/time with hourly resolution.

? Graphical representation of ship discharge events with detail of the activity of the receiving tanks (loading, unloading, locked, on hold).

? Tank status table with graphical and tabular evolution of the status of all tanks.

? Tank graph with overall evolution of the composition of the selected tank.

? Pipeline line fill at any selected time during the simulation period.

Benefits:

The economic objective function applied by the automatic optimization engine available in software tends to maximize the use of low value opportunity crude oils, corresponding to an increase in refinery margin by 1-2 cents/bbl. The automatic calculation algorithm enabled updating the scheduling plan in a few minutes enabling: a precise cargo arrival sequencing plan for uninterrupted crude availability in optimum cargo parcels; and a quick analysis of different scenarios for an effective response to unexpected events such as a cargo’s arrival delay, or a supplier confirms for a loading date different to the nominated one. This helps to minimize demurrage costs by 1-2 days every quarter. The algorithm enables a quick blend feasibility check for timely decisions on opportunity cargo purchases.

Conclusion:

Integrating crude characterization, simulation, and optimization technologies, It is a decision support system to optimize management of ship discharge and coastal storage tanks automatically with minimum user intervention. It assesses optimal crude blends for the pipeline or CDU, maximizing the use of heavy/opportunity crude oils in compliance with feed quality constraints.

The system enables:

? CDU/pipeline feed quality tracking.

? Automatic calculation of a scheduling solution and CDU product yields and qualities.

? Obtaining a solution to crude scheduling problems in less time.

? Analysis of alternative scenarios.

? Better economics compared to other scheduling tools.

? Using the same tool and model can be used for long and short term scheduling.