Quarterly GADS prep cut from weeks to days

The challenge

Cascade Power & Light operates a mixed fleet of combined cycle and peaking units. Every quarter, a small performance engineering team spent the better part of two weeks pulling event logs, reconciling outage records against the historian, and hand-assembling the NERC GADS submittal. The process was spreadsheet heavy and fragile.

• Event start and end times had to be matched manually to cause codes, with frequent disagreement between operators and engineers.
• EFOR and EAF were recalculated by hand in workbooks that no one fully trusted.
• Heat rate drift went unnoticed between quarters because no one had time to trend it unit by unit.
• Auditors asked for the source of each number, and tracing it back to the historian was slow and painful.

The team knew the data existed. The problem was that assembling it into a defensible, source-cited submittal consumed the same two engineers every single quarter.

The build

Using DOTA, the performance group built a GADS Automation App that connects to the plant historian and the outage management records, classifies events against the GADS cause code taxonomy, and computes EFOR, EAF, and equivalent availability automatically. Each metric links back to the underlying time series so any figure can be opened and inspected.

The app also trends net unit heat rate continuously, flagging deviations from the expected curve so the team can see degradation between quarters rather than discovering it at audit time.

Within the first cycle the app flagged a forced derate that had been logged under the wrong cause code. It had been inflating our EFOR for two quarters, and no one had caught it by hand.

The outcome

Quarterly GADS preparation dropped by roughly 85 percent, freeing the two engineers for actual performance analysis. The misclassified derate correction pulled reported EFOR down by about 1.8 points, and every submitted value now carries a source citation back to the historian.

The team built the whole thing on DOTA in about five weeks, with no traditional software project, no procurement cycle, and no developers on the payroll.