Hengyi Refinery: Combining Multiple Data Sources to Extract Richer Insights

Hengyi Refinery: Combining Multiple Data Sources to Extract Richer Insights
Picture and satellite imagery of Brunei’s Hengyi Refinery of Pulau Muara Besar Island, Brunei Bay
Picture and satellite imagery of Brunei’s Hengyi Refinery of Pulau Muara Besar Island, Brunei Bay

Even by today’s efficient refinery standards, this project was sensational. Within six months of its launch, China’s Hengyi Petrochemical Co Ltd was operating at near full-capacity at its 160,000 barrels-per-day refinery in the oil-rich kingdom of Brunei, Reuters reported on Jan. 2.

Located in Pulau Muara Besar, an island off Brunei Bay, the Hengyi plant is one of only four greenfield refineries in Asia. Greenfield refineries use clean process technologies and produce cleaner emissions. Greenfield projects are also typically built from scratch on undeveloped land previously used for agriculture or landscape design. According to an Oxford Business article, the Brunei government leased a 260-hectare greenfield site on the island to secure Hengyi’s investment in the 30:70 joint-venture valued at $3.4 billion.

Prior to the launch, Hengyi said the refinery will produce nearly 6 million tonnes of gasoline, diesel and aviation fuel a year combined. The Brunei plant includes a one-million-tonne-per-year aromatics plant and another unit producing 500,000 tonnes per year of benzene.

According to the Reuters report, exports of refined fuels from the refinery, including gasoline, diesel and liquefied petroleum gas, have been running smoothly since launch, and aviation fuel from the plant has even reached the U.S. west coast.

The Brunei government leased a 260-hectare greenfield site on the Pulau Muara Besar island to secure Hengyi’s investment in t
The Brunei government leased a 260-hectare greenfield site on the Pulau Muara Besar island to secure Hengyi’s investment in the 30:70 joint-venture valued at $3.4 billion

GO Gives OI Clients On-the-Ground, Real-Time Truths on Brunei Hengyi

While Reuters can claim exclusivity to its story, the truth is Orbital Insight’s GO customers will know about global refinery construction and maintenance as they occur. Our objective is simple: provide real-time, on-the-ground truths on any subject to sector analysts, investors and other professionals, and allow them combine those with other intelligence to obtain a comprehensive, multi-perspective view.

Our AI-driven geospatial tracking tool GO, complete with vessel tracking, has been monitoring progress at the Pulau Muara Besar project from the moment construction began and transmitting the data in real-time to our clients, including the delivery of the first crude oil shipment to the refinery — which was only reported by the media four months later. We essentially had what’s known in the media business as the “scoop” on Brunei Hengy.

In today’s machine-dominated capital markets, where computers make nano-second decisions on multi-million dollar trades of stocks to commodities, investors need to know almost everything as it happens. While cable and wire news services are ubiquitous sources of information, they do not have the immediacy of GO’s instantaneous data.

Geospatial Data Is Real-Time, And Bias- And Tamper-Free

Also, unless they are witness to the incident, news services rely on intermediary sources — from government officials to company executives or ordinary people — to tell the story. GO’s data, meanwhile, are narratives in themselves, made up of indisputable imagery and on-ground truths that are free of bias and tampering.

The Hengyi refinery will produce nearly 6 million tons of gasoline, diesel and aviation fuel in a year combined. The plant in
The Hengyi refinery will produce nearly 6 million tons of gasoline, diesel and aviation fuel in a year combined. The plant includes a one-million-tonne-per-year aromatics plant and another unit producing 500,000 tonnes per year of benzene.

In the analysis on Hengyi’s Brunei refinery, energy analysts leveraged the GO platform and its wide range of geospatial information sources — including cell phones, ship-based AIS sensors, optical satellites and radar satellites — to study daily progress at the complex.

The analysis was performed across a variety of locations in Pulau Muara Besar, including the Hengyi refinery’s various structures, loading berths, crude storage containers and the island as a whole. The end result was a unique data set that provided a daily pulse into the construction to processes at the plant. While this particular example covers the Hengyi plant, energy analysts and commodity traders are also using GO to track global refineries, LNG, and petrochemical plants across the world with the same scalable methodology.

Daily Normalized Foot Traffic

By plotting the normalized foot traffic across the refinery and the island as a whole, it is possible to observe broad trends within the ramp up of the facility. For this refinery, traffic began to build during Q4 2017, with significant ramp-up over the course of 2018.

The foot traffic data detects ramp down of crews at the site during the first week of July 2019. Bloomberg confirmed the data in a July 11, 2019 report that said construction has been completed and equipment tested at the site.

Hourly Normalized Foot Traffic (Indexed)

On factory floors, hourly foot-traffic data provide user the ability to extract information such as shift patterns. GO analysis users can zoom into individual days to get an understanding of when workers arrive at a plant each day, and whether crews were working overtime shifts.

Identifying Anomalies

Anomalies in foot traffic patterns provide key insights into whether or not a facility was going to come online when expected, and if it was experiencing downtime once production was up and running.

Daily Normalized Weekend (Friday/Sunday) Traffic

Generally traffic across the refinery was very consistent, but major spikes were detected on weekends during September and October 2018. The traffic spikes were particularly pronounced on September 23, October 7, and October 28.

Hourly Normalized Daily Weekend (Friday/Sunday) Traffic

Zooming in on the hourly data, end-users extract more information about what happened on these anomalous days. The data showed that for both September 23 and October 7, crews arrived early, at around 7AM.

However, on the 23rd, much of the workforce had left by mid-morning. On the 7th, the traffic remained at the facility until the mid-afternoon.

Overnight Traffic Only

By analyzing the hourly data for this facility, deeper trend analysis revealed that the majority of work traffic occurred between 6AM and 11PM. By filtering out these times from the dataset, it was possible to observe key dates where shifts are working late into the night. One such date, July 3, 2019, occurred shortly before the construction of the facility is complete.

Heatmap, Geospatial

Geolocation Data Heatmaps

Heatmaps give users the ability to zoom in on a smaller time range of data and observe where devices have clustered around the facility. This information is important as it can provide insight into what part of the facility is being worked on.

Heatmap, Geospatial

Monthly Heatmap

Monthly aggregations show where crews have been working around the site over construction activity. The majority of the activity at the Brunei Hengyi plant was focused around the center of the refinery. But as construction neared completion, there was a reduction in device activity around the storage tanks and other parts of the facility. One interesting observation was the spike in activity near the docks that occurred in October of 2019.

Heatmap, Geospatial

Zoom in to Assess Activity Spikes

When analyzing the anomalous days at the facility during the September-October 2018 period, the activity was focused on the central part of the facility. Notwithstanding that, there was also a spike in activity at the eastern end of the plant.

Workforce Trends by Country of Origin

Analyzing the origin for devices at a facility provide analysts with the ability to monitor for teams coming in from abroad. In certain cases, these teams will be specialized, corresponding to a specific job function or stage in the construction process.

Mapping New Buildings Detected by Computer Vision

Land use maps assess the status of individual buildings across a site. These maps are produced with medium resolution satellite imagery, enabling clients to get up-to-date views of the buildings or roads that have been constructed. Mapping using medium resolution imagery provides clients with the highest possible cadence (weekly) to assess changes to a facility.

Mapping Infrastructure Change: 2017

Instead of analyzing a single point in time, Orbital Insight GO’s land use change detection algorithm alerts customers to relevant changes over time, such as the construction of new roads, and oil tanks, represented as a unique geometry by data.

In 2017, the majority of the Hengyi site construction was focused around building roads to access the different parts of the site, and buildings around the center of the facility.

Changes Detected in 2018

2018 saw the construction of storage tanks around the facility, along with the remainder of the buildings.


AIS Data

Brunei Hengyi Refinery Berths

Arrival of ships at the smaller berth 3 in early May coincided with the observation of crude oil storage levels increasing at the facility.

All Ship Traffic

Tanker Traffic Leaving the Refinery from Berths 1, 4, and 5

Berths 1, 4, and 5 are deep draft ports specialized for handling tanker traffic at the facility. Shipments to these berths began in July with the delivery of additional crude oil. The first shipments of refined product leaving the facility were detected on November 3rd. This was not reported in the media until November 6th.

Oil Data

Brunei Floating Roof Oil Storage

In addition to monitoring tanker traffic at the facility, Orbital Insight uses satellite imagery to observe the fill rates for crude oil stored in the region. The data mined by Orbital Insight showed a steep increase in the volume of crude oil stored in Brunei beginning May 2019. This aligned with both AIS ship activity observed in early May and a May 7 news report that the refinery had received its first crude shipment.

Hengyi refinery — flaring activity

Flare intensity

The Orbital Insight team leverages short wave infrared (SWIR) satellites to detect high intensity releases of heat, also referred to as flares. These flares are a by-product of refinery operation as the facilities burn off unwanted gases. These flares are observable on a high cadence and can provide insight into whether or not a facility is operational. In this example, the Orbital Insight team detected flaring at Brunei Hengyi on October 26 and December 30th.


The Hengyi facility has vast implications on the Southeast Asian region’s refining capacity. Analysts are particularly keen to understand commodity movements and unreported maintenance activity at facilities like these.

We look at this with many different sources of geospatial data, which in combination, are vastly more insightful than any single source on its own.

Contact sales@orbitalinsight.com to learn more.




Perspective is a powerful thing.

Recommended from Medium

The Future of Agri-Food In The Post COVID-19 Era — Part IV-Online Delivery

Could Stitch Fix Stock Really Increase by a Whopping 57%?

Peloton recalls treadmills after injuries and child’s death

Peloton recalls treadmills after injuries and child’s death

First to Market Advantage?

This Company Wants to Turn Everyone Into a Software Developer

How is AI Transforming the Restaurant Industry in 2022?

Gamestop: beware the madness of crowds

Free-floating in Chicago (carshare that is)

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Orbital Insight

Orbital Insight

Geospatial analytics for an interconnected world.

More from Medium

Spatial sampling and resampling for Predictive Mapping with Machine Learning: A tutorial in R

Estimating Travel Cost from OpenStreetMap

Mapping the geospatial patterns of broadcasted news

AI in Agriculture: Opportunities and Challenges