Data-Driven Algorithm for Sensor Placement Optimization

Methane is a significant contributor to climate change, and improved detection capabilities enable faster and more effective mitigation efforts. We have developed an intelligent and adaptable system designed to optimize the placement of continuous monitoring sensors on oil and gas sites for methane emissions detection. Further, our approach addresses a larger problem scale compared to previous studies and can be customized for various sensor placement objectives.

We begin with a problem: where should sensors be placed?

Our process addresses this question of optimal sensor placement, along with the tradeoffs between coverage and number of sensors. It consists of five steps:

1. Simulate potential emission scenarios based on specific wind patterns and site-specific emissions data.
2. Identify candidate sensor locations, taking into account the layout and any operational constraints at the site.
3. Simulate methane concentrations for each combination of emission scenario and potential sensor location.
4. Assess the effectiveness of emissions detection based on these simulated concentrations.
5. Select the optimal sensor locations within a defined budget using genetic algorithms and optimization techniques.

Here is what we get, when applying our approach to the METEC testing site: Figure 1 shows the site detail; Figure 2 shows the results corresponding with optimal placement both within and at the border of the site, varying by number and location of sensors; and Figure 3 shows the variance in detection coverage by budget size across multiple algorithms (ours in blue vs. greedy in yellow). 

Finally, we tested our approach at a real, anonymous oil and gas site.

This site initially had nine industry-standard sensors placed on the perimeter fenceline, following the vendors recommendation. The operator wanted to compare this setup to a new configuration using four higher-quality sensors. Given that precise placement becomes more critical with fewer sensors, our framework was used to determine the optimal locations for the new sensors under the operator constraint of restricting placement to the fenceline at a 2-meter height to facilitate installation and minimize operational interference.

The sensors were ultimately installed at the recommended locations. This case study highlights a practical implementation of our framework, while also integrating operator knowledge effectively.