SESSION

SmartStim: Interactive, Real-Time, ML Enabled Production Optimization

OVERVIEW

EXPERIENCEIn Person
TYPELightning Talk
TRACKData Engineering and Streaming
INDUSTRYEnergy and Utilities
TECHNOLOGIESAI/Machine Learning, Delta Lake, MLFlow
SKILL LEVELBeginner
DURATION20 min

Transitioning to renewable energy is vital to improving existing oil extraction operations and minimizing unconventional methods like fracking. Shell’s SmartStim saves time and costs by suggesting process optimizations, using a Databricks streaming process to extract production data in real-time from remote global sites and run a heuristic model. An interactive Plotly Dash app is then accessible to field engineers so they can review and update model recommendations in near real-time to optimize decision-making.

 

This talk details what makes SmartStim work: extracting data from a non-streaming data source; integrating a heuristic algorithm with MLflow; streaming data into ADLS gen 2 in Delta Lake format; accessing a serverless SQL warehouse; developing the Plotly Dash UI; enabling comparisons with historicals; facilitating model parameter adjustments directly in the UI; integrating Microsoft Teams alerts; triggering MLflow model update writebacks; and N-metric plotting with dynamic y-axes scaling.

SESSION SPEAKERS

Muhammed Abdul-Hameed

/Principle IoT and Edge Computing SME
Shell