Constructing a global energy supply network for the transition toward a post-carbon society
Read about the data analytics case study by QuantumBlack showing the construction of a global energy supply network that can help transition toward a post-carbon society.
The unprecedented impact of COVID-19 has provided a temporary respite for carbon emissions globally. Although economic recovery will likely overshadow environmental concerns, we are now in a unique position to ensure that it follows a path that is environmentally sustainable.
One way to tackle the challenges of climate change is by governing a transition of the global energy supply chain towards a net-zero carbon state. The transition will require identifying positive ‘tipping points’, also known as sensitive intervention points — SIPs. These are modest actions that can trigger an outsized response and accelerate the path to net-zero emissions.
The challenge in identifying SIPs is that the global energy system is large in scale and highly interconnected. In order to answer future research questions with evidence-based recommendations, one needs to build an interconnected, transparent, and queryable dataset of the global fossil fuel network.
In a collaborative effort with QuantumBlack data scientists and the Oxford University Smith School of Enterprise and Environment (SSEE) researchers, we built a reproducible pipeline that can create a queryable graph of the global fossil fuel supply chain by integrating a set of large geospatial data sources. This graph includes a range of fossil fuel sources (i.e. oil, gas, coal, and etc.), covers a variety of transportation and conversion modes (e.g. pipelines and railways), and connects the energy supply network to the population centres as sinks.
Method:
We used only open data to create an arrangement of the global fossil fuel supply chain. We drew on a combined 340 data sources across 11 categories of energy infrastructure. Leading sources include:
- The Department of Energy’s Global Oil & Gas Features Database
- The Natural Resources Governance Institute Resource Projects
- The Oak Ridge National Laboratory
- World Resources Institute Resource Watch
- The European Commission’s Joint Research Council Global Human Settlement Layer
- OpenStreetMaps
To prepare the data that fed the network, we performed data manipulations such as finding the coordinates for when railways (or pipelines) cross and finding the location where these connect to population centres (as sinks).
One of our goals was to avoid making simplifying assumptions before creating the graph so that researchers can impose their own assumptions when they work with it. Given the scale of the data we were working with, we relied heavily on Python packages such as GeoPandas and Shapely to efficiently implement the manipulations.
Using Kedro, we defined a pipeline with a set of nodes to handle all geospatial operations. This approach enabled us to naturally break the task down into smaller, explainable chunks, making the code much easier to read and understand.
We used Neo4j as the graph database to hold the large network, and Cypher, a specialised and efficient graph querying language. We chose Neo4j for its large ecosystem of compatible python packages and its relative ease of use.
Our Kedro pipeline was set up to produce csv data in the right format, which we then imported into the network using the bulk import tool from Neo4j. Using this, it drastically reduced the time needed to import the entire graph, allowing us to more quickly iterate on our pipeline. After they were loaded, we ran a set of Cypher queries to simplify the graph.
Outcome:
We constructed a graph that maps the paths from individual source assets, such as oil wells and coal mines, to sinks such as population centres and power stations.
The connection between the transportation networks and assets are inferred by matching assets with transportation nodes (pipeline, railway and shipping route intersections) in a few kilometre radius. To avoid having ports without sea access, we assumed all ports are connected to the nearest shipping route. To ensure all connections were realistic, we mapped the possible paths through the network as shown in the graphic below.
The raw graph has over 11 million nodes and 20 Million edges, many of which are not very interesting for research. We therefore use graph simplification methods to create “skip” edges which skip over a set of pipelines or railway nodes with a degree of 2 (see graphic below). After simplification, the graph has over 2 million nodes and 10 million edges.
What’s next?
This graph provides researchers with a holistic global view of the fossil fuel supply network, enabling a wide range of use cases that can help transition to a post-carbon society. These include connecting supply and demand by calculating the graph “flows” from energy sources to sinks, and the ability to find SIPs for effectively making the global energy system greener.
This is only the beginning, and we are planning to open source the project’s code repository. By making these assets available to all, we can empower the community to tackle impactful use cases such as:
- Evaluate different decarbonisation policies with empirical numerical simulations
- Identify test-cases and beachheads for financial institutions looking to create outsized decarbonisation impact
- Identify the ideal infrastructure clusters necessary for transitioning to a different energy carrier
- Empower activist groups with analysis and data to hold private and public sectors accountable towards decarbonisation goals
While the global pandemic continues to dominate conversations, we have an obligation to ensure that the economy follows a sustainable and environmentally conscious path of recovery. By building this network, we hope to empower researchers, private enterprise and government organisations to understand fossil fuel supply chains in a holistic way. By furthering the understanding, we can identify the proactive steps that can be taken now to assist the transition of the global energy supply chain towards a net-zero carbon state.
To find out more about the pro-bono projects that QuantumBlack undertakes, please reach out to analytics-for-social-good@quantumblack.com.
Citations:
- Kruitwagen, L. (2019) “Geospatial Sensitive Intervention Points”, a lecture for the Programme on the Transition to a Post-Carbon Economy Policy Advisory Board, Oxford Martin School, 2019–04–11.
- Kruitwagen, L., [Klaas, J. | Lakeh, A.] & [Klaas, J. | Lakeh, A.] (2020, forthcoming) “A Network Arrangement of the Global Coal, Oil, and Gas Supply Chains”.
- Farmer, J. D., et al. “Sensitive intervention points in the post-carbon transition.” Science 364.6436 (2019): 132–134.
