Loading...
Loading...
Michigan (MI) · Oil & Gas
Updated June 2026
Michigan's oil and gas footprint is larger and more operationally complex than most outsiders assume. The Antrim Shale in the northern Lower Peninsula — centered on Otsego, Antrim, and Crawford counties — has roughly 10,000 active producing wells making it one of the most drilled shale plays in the continental United States by well count, despite relatively low per-well production rates. The Michigan Basin holds conventional oil production concentrated in the Trenton-Black River play near Mount Pleasant and in the Niagaran reef trend, with dozens of small independent operators running stripper wells under the oversight of the Michigan Department of Environment, Great Lakes, and Energy's Office of Oil, Gas, and Minerals (EGLE OOG). The Marathon Petroleum refinery in Detroit, one of the largest on the Great Lakes, processes roughly 140,000 barrels per day and is a major employer in the River Rouge area. And the Enbridge Line 5 dual-pipeline crossing under the Straits of Mackinac — 67-year-old infrastructure carrying 540,000 barrels per day of crude and natural gas liquids between Superior, Wisconsin and Sarnia, Ontario — has been the subject of state-federal legal dispute since Governor Whitmer issued a revocation order in 2020. Each of these segments creates its own AI demand profile, and they have almost nothing in common operationally.
Connecting AI systems to existing business infrastructure and workflows
Predictive models, data analysis, and ML pipeline development
Image recognition, object detection, video analysis, and visual inspection systems
Bespoke AI solutions, model fine-tuning, and custom model development
The Antrim Shale is an unconventional play driven by desorption of biogenic methane from organic-rich shale at shallow depths — typically 600 to 2,000 feet — which means the production decline curves, water production behavior, and artificial lift optimization requirements are quite different from Marcellus or Haynesville. ML reservoir modeling here focuses on predicting water-to-gas ratios over time, optimizing pump-off controller settings on rod-pump and plunger-lift installations, and identifying underperforming wells within multi-well lease units. Companies like Encana (now Ovintiv), Savoy Energy, and CMS Energy's gas subsidiary have historically operated large Antrim portfolios, and the fragmented nature of the play — thousands of small leases across a dozen northern Michigan counties — makes automated performance monitoring at the lease level genuinely high-value. For the Niagaran reef trend and deeper Michigan Basin conventional production, EGLE OOG's publicly available well production and injection records enable basin-wide ML models for formation pressure depletion and waterflood optimization. Central Michigan University's geology department in Mount Pleasant, located in the heart of the producing belt, has been an academic partner for several reservoir characterization studies. Operators report that ML-assisted decline curve analysis on Niagaran reef producers typically reduces the guesswork on well workovers — the difference between a successful recompletion and a dry attempt in a depleted reef carbonate often comes down to pressure depletion models most operators were running on spreadsheets until recently.
The Line 5 dispute has driven Michigan-specific pipeline AI investment that goes beyond anything you see in states without this level of political and legal scrutiny. Enbridge has deployed advanced inline inspection — magnetic flux leakage, ultrasonic, and geometry tools run annually on Line 5 segments — and the resulting gigabyte-scale inspection datasets are fed into ML anomaly classifiers that prioritize excavation and repair sequencing. The Straits of Mackinac crossing itself is subject to a separate hydrodynamic anchor-strike risk model updated after each ILI run, analyzing cumulative coating loss, cathodic protection effectiveness, and scour potential under the span. The broader implication for Michigan operators is that the Enbridge-EGLE regulatory proceeding has forced public articulation of AI-based risk methodology that other pipeline operators in the state can now benchmark against. EGLE's Pipeline Safety Regulatory Program, which enforces PHMSA regulations under a certified state program agreement, is increasingly asking operators in rate cases and permit applications to document their integrity management AI tools. For midstream operators running distribution lines in the Upper Peninsula — where Line 5 NGL supply feeds propane and heating fuel distribution — the reliability modeling question is existential: what happens to supply costs and routing if Line 5 is ultimately shut down, and how do you build an AI demand-forecasting model around a binary infrastructure risk?
The Marathon Petroleum refinery in Detroit's southwest side processes Canadian crude delivered via Great Lakes pipeline and barge routes, and the refinery's footprint in the Delray neighborhood has made it one of the most closely monitored industrial facilities in Michigan under EGLE's Air Quality Division. AI applications at Marathon Detroit mirror those at peer refineries nationally: process optimization models on crude unit and FCC throughput, predictive maintenance on heat exchangers and compressors, and advanced process control on reformer and hydrotreater units. What is specific to the Michigan market is the EGLE continuous emissions monitoring data-sharing requirement and the overlay of USEPA Region 5's environmental justice enforcement posture, which has intensified scrutiny of fence-line air quality at this facility since 2022. For AI procurement decisions at facilities like Marathon Detroit, the compliance AI layer — continuous emissions monitoring data validation, predictive NOx and SO2 exceedance flagging, CEMS data reconciliation for EPA Title V permit reporting — is often the initial entry point because the regulatory ROI is direct and auditable. The shortlist criterion here is demonstrated work under EPA Region 5 compliance frameworks, not just general refinery AI experience. Process optimization AI built on Aspen Technology or Honeywell Forge platforms has been deployed across Marathon's refinery network, and the Detroit facility participates in that enterprise AI rollout. Local integrators in Detroit with both refinery process experience and Michigan EGLE compliance familiarity are rare — most engagements draw from Houston or Chicago-based consultants.
Production optimization AI for Antrim Shale should address three specific challenges: water production forecasting (Antrim wells produce large volumes of formation water that must be managed under EGLE discharge permits), pump-off controller optimization on rod-pump installations, and anomaly detection on the dense telemetry from thousands of low-rate wells. Platforms like Ambyint, SparkCognition, and Corva have been deployed on comparable unconventional shallow plays. Implementation for a 200-well Antrim portfolio typically runs $150K–$400K including SCADA integration and 12 months of model training on historical production data. EGLE OOG well records are publicly available and can supplement operator-provided data for basin-level benchmarking.
Enbridge runs annual inline inspection on Line 5 and feeds ILI data into ML-based anomaly classifiers that prioritize excavation and repair based on metal loss severity, coating condition, and proximity to high-consequence areas including the Straits crossing. The company's integrity management program submitted to EGLE documents the AI-assisted prioritization methodology, which has become a de facto standard for PHMSA compliance under Michigan's certified state pipeline safety program. Outside vendors providing ILI data analytics — TDW Inspection Services, ROSEN Group, Baker Hughes — all support ML-enriched reporting that feeds these models. Michigan operators running smaller diameter distribution lines can adopt scaled-down versions of the same anomaly-classification approach for their own EGLE-required integrity management plans.
Michigan's EGLE Office of Oil, Gas, and Minerals does not yet mandate AI-specific data systems, but its 2021 updated Part 615 rules (Oil and Gas Exploration and Production Act) require enhanced reporting on well integrity, injection operations, and production data that are natural inputs for AI analytics platforms. EGLE OOG's MiWell online database contains production history back decades, and operators using AI reservoir models routinely pull this public data to supplement their own records. The EGLE pipeline safety program under Act 16 maps directly to PHMSA federal requirements, so federal IMP documentation standards effectively set the AI methodology bar for Michigan operators.
Michigan Basin stripper well operators — often running 5 to 50 wells with lean field crews — have the most to gain from low-cost AI monitoring that prevents deferred maintenance from becoming an abandonment event. Cloud-based production surveillance tools like Wellview, Cortex, or P2 Energy's analytics modules can process basic SCADA and gauge-board data to flag unusual production decline, increasing GOR, or pressure anomalies that indicate casing integrity issues. At the scale of a typical Michigan Basin independent, the total AI investment is $20K–$80K annually for SaaS surveillance plus periodic consulting support. The EGLE OOG orphan well remediation program context adds urgency: operators who catch declining-well problems early avoid the regulatory jeopardy that comes with unplugged abandoned wells.
The Michigan Oil and Gas Association (MOGA), headquartered in Lansing, is the primary industry group and holds an annual conference attended by both operators and service vendors. The Great Lakes Energy Institute at Michigan Tech University in Houghton has been a research partner for Antrim Shale reservoir characterization studies. The Michigan section of SPE (Society of Petroleum Engineers) runs technical workshops in Mount Pleasant that are well-attended by Antrim and Michigan Basin operators. These are the primary venues where AI vendors targeting Michigan independent operators gain traction — Houston-style megaconference appearances rarely reach the small independent operator community that makes up most of Michigan's producing-well count.
Join other experts already listed in Michigan.