The petrochemicals industry has a long history of operational and productivity improvements gained through superior process engineering and increased scale of operating assets. Over the past two years, we have seen how advanced analytics and machine learning can spur a step change in operational and financial performance.
The results of pilot projects in petrochemicals have been impressive. In operations, this could be improving yield in crackers, increasing throughput in polymerization units, or improving reliability in compressors and heat exchangers. On the commercial side, this could mean improved pricing through better integration of data on market changes or microsegmentation. These value-creation levers are possible because of the availability of an immense amount of data and improvements in processing power.
In this article, we describe the key elements required to start, accelerate, or scale up the use of advanced analytics in the petrochemicals industry: leadership commitment, high-impact initial use cases, relevant analytics capabilities, and a road map defining a systematic analytics approach. Delivering impact and scale in analytics requires companies to apply all of these elements. In our experience, employing advanced analytics could increase a petrochemicals company’s EBITDA by as much as 20 percent. And solutions are mature enough for petrochemicals companies to embrace their use immediately.
There have been significant technological advances in digital and analytics in recent years. The generation, collection, and storage of data have never been so cost-effective; at the same time, computational power is reaching unprecedented levels and at lower costs. Petrochemicals companies already possess significant amounts of data. Building expertise in data collection and analysis can create two areas of strength for petrochemicals players seeking to capture the benefits of advanced analytics:
Advanced process-control approaches that employ algorithms to stabilize operations are already widely used in the industry. These methods also generate a lot of data. The availability of high-frequency, high-quality data and a track record of productivity improvement efforts offer petrochemicals facilities—even older plants—the opportunity to use advanced analytical methods to capture significant value.
However, maximizing the benefits requires more than a focus on data assets; it also demands a broad organizational effort. Historically, control systems have been the domain of vendors rather than of petrochemicals companies developing optimization approaches by themselves. Implementing advanced analytics, from supply chain to operations and commercial processes, requires a concerted effort among operators, engineers, and other teams in the organization.
We categorize value-adding use cases into four main areas: profit per hour, asset reliability, value chains, and sales performance.
Companies can improve site-level profit per hour by optimizing yield, throughput, and energy efficiency. Depending on the use case, typical improvements range from a 5 to 7 percent rise in throughput to an increase in the yield, selectivity, and conversion of certain processes by 1 to 2 percent. These use cases can also lead to a 3 to 5 percent reduction in fuel gas, steam, and electricity consumption.
Advanced analytics can generate substantial improvement in the reliability of critical equipment such as in-line extruders or compressors. For example, the potential of predictive maintenance has long attracted attention in the petrochemicals industry, although it cannot be applied to every piece of equipment in a plant. Depending on where advanced analytics is applied, we have observed increases of 0.5 to 1.0 percent in machinery uptime, or a 1.0 to 2.0 percent reduction in maintenance costs. These improvements may seem small, but the efficiencies or savings they generate go straight to the bottom line.
Petrochemicals companies manage a network of interconnected plants with many product exchanges. Optimizing these networks has proved difficult. Now, with greater data availability and more sophisticated advanced-analytics approaches, petrochemicals companies can better carry out planning activities, allowing them to optimize overall value in their systems. This could be in the form of traditional linear-programming implementations similar to what refineries do or more advanced predictive models for intermediate products-related decisions.
Carrying out a high-impact use case has several steps: ideation, detailing, development, and implementation. Below, we elaborate on these steps using a real-world example of a petrochemicals company that developed a cracker furnace optimization model.
In this phase, the company’s advanced- analytics practitioners and focus-area experts (process engineers) gathered to discuss potential use cases. Practitioners provided perspective on what advanced analytics can offer, while focus-area professionals contributed details about the challenges of achieving better performance.
The outcome was several concrete ideas about how to improve the performance of the furnaces through dynamic control of the parameters that have an impact on chemical processes. The team focused on addressing bottlenecks in downstream components, such as cracked gas compressors and other process equipment.
The ideas were turned into designs of specific prediction and optimization problems with objectives, variables that could be controlled, and variables that couldn’t be controlled (for example, ambient variables). Teams developed a better understanding of the potential impact of the ideas in this phase. Finally, five prediction models were created and combined with a variables optimization model for the cracker furnace.
The models designed in the detailing phase were tested with real data from the past. The data were divided into training and testing data sets and used to validate and further improve the model details. This is an iterative process that can take significant time. At the end of development, a blueprint of the model was available.
Simply developing a use case that consists of prediction and optimization models on a computer and running them in real time is not sufficient to capture the full impact of advanced analytics. To get the most from optimization, operators had to redesign processes such as maintenance schedules regarding the day-to-day management of the plant. Also, operators adapted the models to reflect changes in cracker furnace operating conditions.
Following implementation, the company observed a 10 to 20 percent improvement in profitability of its cracker operations in the form of additional throughput and improved yield. In general, potential impact depends on how sophisticated the plant is at the starting point.
Companies can raise sales performance by using customer- and transaction-specific data to carry out microsegmentations, demand and price forecasting, and granular performance tracking. Customized and dynamic pricing is an important lever to improve value in commercial applications.
Investing in just a few analytics use cases through packaged solutions in isolated parts of the value chain may not produce significant benefits and sustainable value. Successful companies develop a portfolio of use cases, often employing a common approach (see sidebar “Designing a cracker furnace optimization model”).
In our experience, value maximization is possible only through a carefully designed and rigorously implemented program touching every part of the organization, with a strong emphasis on capability building and change management. Examples of potential impact through analytics in petrochemicals are presented in the exhibit.
Over the past two years, we have seen how advanced analytics and machine learning can spur a step change in productivity and financial performance.
Four elements are needed to start, accelerate, and scale up the use of advanced analytics in petrochemicals (see sidebar “How Turkey’s Petkim incorporated advanced analytics into its operations”).
In 2019, the World Economic Forum recognized Petkim, a petrochemicals company based in Turkey, as the first (and to date only) petrochemicals company in its “Lighthouse” sites for incorporating Fourth Industrial Revolution technologies into its operations. These technologies include the use of advanced analytics, artificial intelligence, and the Internet of Things.
At the center of Petkim’s operations are two major petrochemicals plants in Turkey, a mixed-feed cracker supplying ethylene and propylene to several downstream units, and a plant producing aromatic hydrocarbons for other downstream units.
In 2016, Petkim’s leadership committed to achieving a step change in performance, featuring the highest industry standards and improvements in profitability. In the first few years, Petkim focused on improving its operations, maintenance, and procurement practices. This freed up resources that could be invested in cutting-edge petrochemicals technologies.
In 2018, Petkim undertook its initial trials in advanced analytics, pursuing an ideation exercise throughout its operations and functions. Its first use case prioritized a real-time optimization of cracker operating parameters. The use case delivered substantial value, thus proving the merit of advanced analytics and mobilizing the organization to pursue the program at that plant and others.
As a result, a comprehensive road map was developed that touched all parts of the business. A central digital and analytics team was organized under the chief digital and analytics officer and grew quickly to serve the organization. Recruiting was undertaken internally and externally to develop the team further.
A digital and analytics academy offered training to all employees at Petkim; the curriculum varied according to the needs of the employees. The training encouraged these workers to think of initiatives that could benefit from advanced analytics. As a result, Petkim developed and implemented an ambitious road map of more than 40 value-creating use cases.
Implementing a successful advanced-analytics program without leadership commitment is difficult. Such commitment means the required resources will be made available to pursue an at-scale implementation rather than just introducing a few use cases. The latter could result in significantly less value capture.
Petrochemicals companies are mostly older institutions. Their workforces, while skilled, tend to fall back on standard methodologies to manage day-to-day activities. As a result, leadership commitment becomes a critical enabler of workforce desire to embrace new methodologies in daily work.
Petrochemicals companies will be dependent on external resources if they don’t develop their own analytics capabilities.
If the advanced-analytics program does not start with a few high-impact use cases, it will be hard to convince people in the organization to scale it up. The program might even grind to a halt. Therefore, showing value early with signature implementations is key for building both internal and shareholder support.
Petrochemicals companies will be dependent on external resources if they don’t develop their own analytics capabilities. Packaged solutions developed and deployed by vendors may not always align with the priorities of a company. For example, these solutions aren’t necessarily customized and therefore may not address the specific problems of an individual company. As a result, the use cases with the most impact may be missed.
There are two potential approaches to developing internal capabilities. In one, executives create a center of excellence staffed by data scientists who may have limited knowledge of petrochemicals operations or commercial activities. They would team up with focus-area experts to develop value-creating use cases.
In a second approach, the organization trains all key professionals in data science. This enables all of them to develop their own use cases. While the first approach would be faster to have impact and more practical, the second one delivers more value—but it would require more time and resources.
There is no single right approach for all companies. To get results quickly, establishing a center of excellence is often the best way forward. However, to transform an organization, upskilling at scale may be the better answer.
Petrochemicals companies need an organization-wide road map that lays out a systematic approach to analytics. The absence of a road map could result in misaligned priorities, missed value capture areas, and an incomplete transformation to an analytics mindset.
Finally, to successfully deliver impact and scale in analytics, companies need to put all four elements in place. Failure to do so could result in the development of local initiatives but no real momentum because of lack of resources; conceptual discussions but little or no traction due to lack of evidence; slow movement due to reliance on external support; and initial successes but no organization-wide scale-up.
Advanced analytics and machine learning have begun to make inroads into the petrochemicals industry. The technology is at a point that companies can confidently act to integrate it and capture value. Companies that move now to build advanced analytics into their organizations could create an enduring competitive advantage.
Never miss an insight. We'll email you when new articles are published on this topic.
Subscribed to {PRACTICE_NAME} email alerts.