Much of our current world would not exist without industrial refining. The availability of well-defined, high-purity chemicals enables the production of just about every product we use. Industrial refining is also resource-intensive; in the United States alone, petroleum refineries consume over 800 trillion watts per year. Companies that operate refineries benefit from making sure they are getting the most out of their inputs, and given the scale of industrial refining, even small changes can have large impacts. This is where refinery optimization comes into play. Refinery optimization is the data-focused process of making a refinery more efficient. It is considered a subdiscipline of production optimization.
Refineries already take advantage of a variety of resource-saving strategies. For example, the heated gases boiled off during distillation pass through reboilers to heat the incoming mixture of liquids. Flammable waste from the refining process fuels on-site energy generation. Steam systems transfer heat from various processes throughout the facility. Critically, many companies have also already wired their refinery operations with a suite of sensors. These sensors provide live data about the facility’s operation and can be connected to refinery modeling software to give a bird’s-eye overview of the refinery at work. The process simulation may have been created when the refinery was originally built, or it may have been outlined later.
Refinery operators already understand that to get the most return on investment, they must make daily and even hourly adjustments to the mix of products being refined. The cut point, or the temperature used to define one fraction of a distillate from another, will be adjusted according to market demands.
Ultimately, all decisions about refinery operations have to answer to the market. All refineries must remain profitable to stay in business: changes in the cost of inputs and the sale price of outputs define important limits on how a refinery can operate. Fortunately, refiners are accustomed to adjusting in response to shifting economic imperatives. The product slate, or the desired output, depends on the cost of inputs and the refinery’s capabilities. Light, sweet crude is in increasingly short supply, requiring refiners to make more allowances for heavy crude as an input.
If a refinery keeps running into the same difficulties when adjusting their product slate, the refinery optimization process may indicate it’s time for upgrades. Adjusting the cut point can only go so far for any given input before a physical upgrade is required.
Identifying what to produce at what quantities, as well as where and how to sell the final products, are key factors in refinery optimization. Spot contracts, or agreements for only one, near-term product delivery, provide a refinery the chance to exploit short-lived opportunities in the market. For longer-term planning, term contracts can lock in a set income stream, with the risk that conditions may change and leave the refiner losing money on the contract due to increased input prices.
Refiners face a daunting task when setting their product slate. A hierarchy of tools and models can help make the best use of a refinery, starting with market and operating conditions as the inputs for a linear program model. The refinery linear program model guides short-term decisions, taking into account near-term conditions and treating the capacity and yield of the refinery as fixed. The scheduling, blending and delivery optimization models all take their cues from the linear program, and as result inherit the short-term focus.
Industrial AI, the application of artificial intelligence to industrial problems, is well suited to help navigate this thorny issue. Artificial intelligence works by training an algorithm known as a neural network on large volumes of data. The algorithm links the pieces of information together in software, and makes adjustments to the linkages until the input data resembles the known outputs of said data. For instance, the links might automatically tie the amount of fuel being fed into the system with the rising temperature in the stack. This may be a trivial for a human operator, but the artificial intelligence algorithm will also find links that might escape human attention.
AI for oil and gas can add a longer-term dimension to refinery optimization. An artificial intelligence algorithm can deduce patterns in operations data and integrate the maintenance needs into the forecasting refiners perform. That is, the equipment degradation that results from running different cut points at different capacities can be taken into account when setting the refinery optimization parameters.
How is refinery optimization used to increase margins?
Refiners can get better prices for different mixes of products depending on market conditions. Forecasts based on fuel choice, feed stock and cut points allow refiners to adjust to create the optimal product slate and increase profits as demand shifts.
How is refinery optimization a part of the smart refinery?
A smart refinery continuously performs refinery optimization as part of a data-driven process of increasing efficiency. A refinery that doesn’t perform refinery optimization cannot be said to be a smart refinery, but refinery optimization by itself doesn’t make a refinery smart. A smart refinery lets the data lead the process.
The Digital Twin and the Smart Enterprise