North Plains Connector Advances Toward Approval. Here Is the Seam It Would Cross.

On August 28, the U.S. Department of Energy and the Montana Department of Environmental Quality released the Final Environmental Impact Statement for the North Plains Connector: a 420-mile, up-to-525-kV high-voltage direct-current line from a substation near Colstrip, Montana to two connection points in North Dakota, near Center and St. Anthony. The headline number is 3,000 MW of bidirectional transfer between the Eastern and Western Interconnections.

The more revealing number is 0.102. That is the correlation between hourly real-time prices at Colstrip and at Glendive, Montana, 124 miles to the northeast. Ask the Grid matched 14,511 hours across the two nodes from January 2025 through August 2026. Colstrip is priced in CAISO's Western Energy Imbalance Market; Glendive is priced in SPP. Against a Northern California trading hub more than 900 miles away, Colstrip correlates at 0.612; against Glendive, at 0.102. The two Montana nodes sit on opposite sides of the interconnection boundary the proposed line would cross.

Grid United and ALLETE are the joint developers. The project is described as an approximately $6 billion investment, up from the $2.5 billion estimate in the 2023 development agreement. Major construction is expected to begin in 2028 and service in 2032. The FEIS closes the federal and state environmental review; it is not a Record of Decision, a financing award or a construction start.

What the connector would join is unusual by the standards of transmission projects. The United States runs on three largely independent alternating-current systems, and they cannot be tied together directly because they do not run in synchrony. Power crosses between the Eastern and Western Interconnections only through back-to-back DC converter stations. In Montana there is exactly one, the Western Area Power Administration's Miles City station, rated at up to 200 MW. Nationally, seven such links carry a combined 1,320 MW.

At 4 p.m. Mountain Time on December 17, 2025, the Colstrip node cleared at $836.22/MWh. In the same hour Glendive cleared at -$15.29/MWh: generators there were paying to keep producing. Across the Mountain-Time day, Colstrip averaged $134.55/MWh and Glendive averaged 31 cents. Glendive settled below zero in 15 of 24 hours.

Colstrip's price rose three times between 4 and 8 p.m. while Glendive's stayed at or below zero for most of the day. The gap between them reached $851.51/MWh in the 4 p.m. hour, the widest in the twenty-month record. These are the settled prices; the operational reasons for each side's position — unit outages, wind output, the loading of the Miles City tie — are not established by the price series alone.

A price gap 120 miles wide can have ordinary causes: a binding constraint, an outage, a unit tripping. To separate those from the seam, Ask the Grid ran the same hourly correlation on six other pairs — two inside the Eastern Interconnection, two inside the Western, two more that cross the boundary. The Eastern comparators are Thunder Spirit Wind in North Dakota and the SPP North hub; the Western ones are CAISO's NP15 trading hub and the NorthWestern/Bonneville interface.

The ranking does not follow distance. Two Eastern nodes correlate at 0.958. Colstrip and CAISO's NP15 hub, at opposite ends of the Western Interconnection, correlate at 0.612. Colstrip and Glendive, the closest pair in the set, correlate at 0.102. The average absolute gap across the seam is $25.71/MWh, against $4.72 for the tightest same-side pair. Every pair that crosses the boundary is less coupled than every pair that does not. The comparisons are consistent with persistent separation across the interconnection boundary, although they do not isolate the contribution of individual transmission constraints.

Seven pairs could still be a coincidence of which seven. So Ask the Grid widened it to every priced node in the corridor that carries a coordinate in its grid registry — 96 locations, one per half-degree cell, drawn from 604 Western Energy Imbalance Market nodes across Montana (atg.caiso_prices_rt) and 9 SPP nodes in eastern Montana and North Dakota (atg.spp_prices_rt) — and correlated all 4,559 pairs.

The two populations do not overlap anywhere. The weakest same-side pair in the sample correlates at 0.307; the strongest across-seam pair correlates at 0.125. Nothing lands between them. The farthest same-side pair in the sample, 501 miles apart inside the Western Interconnection, correlates at 0.57; the closest across-seam pair, 139 miles apart, correlates at 0.10. Distance moves these points hardly at all; which grid a node sits on separates them into two bands.

Averaged across all 14,511 matched hours, Colstrip runs $14 to $17/MWh above Glendive overnight, when Dakota wind is producing into low regional demand with limited ways out and the Western system still has load to serve. Through the late afternoon the sign reverses, and Glendive runs about $21/MWh above Colstrip at the 4 and 5 p.m. peak.

Broken out by month, the pattern is a fixture of the clock rather than of the season. Two vertical bands run the whole height of the record: the Western end is dearer overnight in nineteen of the twenty months, the Eastern end through the working afternoon in nineteen of twenty. Nineteen cross zero twice in their average day. January 2026 and May 2026 are the two exceptions, one in each direction.

On average, these historical node prices favour east-to-west transfers overnight and west-to-east transfers in the afternoon. That pattern illustrates why bidirectional transfer capability is relevant here; it is not a simulation of the proposed connector's dispatch.

Counting only the starkest case — one node below zero while the other is above it — there were 1,719 hours with Glendive negative and Colstrip positive, and 571 hours the other way. Together that is 2,290 hours, 16 percent of the record. Over the same window Glendive settled below zero in 14.2 percent of five-minute intervals against Colstrip's 6.3 percent. Their averages across the matched hours are nearly identical — $26.99 at Colstrip, $26.76 at Glendive. Neither end is persistently cheap or persistently dear. They are persistently out of step.

The developers' case rests partly on resource diversity, and that is separately measurable. Using ERA5 reanalysis at 100 meters for January through August 2026, hourly wind speed at the Colstrip end and at the Center, North Dakota end correlates at 0.373. In 466 hours North Dakota was windy while Montana was calm; in 115 hours the reverse. Both ends were calm in 7.6 percent of hours. The comparison suggests potential resource diversity between the two ends; quantifying a reliability benefit would require analysis of generation availability, demand and transmission constraints during stressed hours, which this does not attempt.

The scale is also unusual relative to what it lands in. NorthWestern Energy's Montana balancing area peaked at 2,043 MW across the window (caiso.load_actual), so the connector is rated at roughly 1.5 times the peak demand of the system at its western end. It is continental infrastructure that happens to terminate at Colstrip.

For generators and developers on either side of the seam, the correlation table is a map of basis risk. A Dakota wind project's revenue and a Montana coal or gas unit's revenue are set by two markets that barely speak to each other today; a 3,000 MW link would make each side's afternoon and overnight prices a function of the other's, and hedges written against one hub would need to account for that.

For utilities and offtakers in the Northwest and Upper Midwest, the wind-speed and price patterns are where a reliability case would start, not where it ends. That the two ends are rarely calm in the same hour, and that their average prices peak at different hours, are conditions worth testing against generation availability and demand in a stressed-hours study.

For regulators and planners, the finding is that the seam is structural rather than episodic. It has a daily shape, it appears in every month of the record, and it survives the same-side control. Whatever the line's final cost, the gap it targets is not a feature of one bad week.

Ask the Grid matched hourly averages of five-minute real-time prices at COLSTRIP34_5_NWMTSHAREGNODE in atg.caiso_prices_rt and PSGO_MDUM_GLENDC1 in atg.spp_prices_rt from January 1, 2025 through August 31, 2026, the longest overlap available because SPP's real-time history in the warehouse begins January 1, 2025. Correlations are Pearson on shared hours. The Colstrip node represents NorthWestern Energy's share of the Colstrip plant and carries that generator's local congestion; over the same 14,511 matched hours the Avista and Bonneville interface prices for the same balancing area average $28.44 and $28.98 against the node's $26.99, which is consistent with a regional pattern rather than a single-generator artifact. Every average, share and correlation quoted here is computed on that one window and grain. A price gap is evidence that two locations separated, not of which facility bound. No dollar benefit is estimated: a 3,000 MW flow would compress the spread it is measured against.

Explore the underlying datasets: CAISO real-time prices, SPP real-time prices, balancing-authority load and the ERA5 weather grid. The Western node has its own page at Colstrip 3 and 4, NorthWestern share. The figures are frozen at publication. Open SPP on the live map to inspect current plants, prices and transmission across eastern Montana and the Dakotas.