Building an energy-impact action queue for a solar portfolio

A portfolio dashboard becomes operationally valuable when it helps a team decide what to investigate first. Raw alert count is a weak priority signal: one high-energy persistent event can matter more than dozens of short warnings. A defensible action queue combines estimated impact with confidence, duration, recurrence, and the cost of delay.

1. Separate known impact from unknown impact

Calculate potential lost energy only for intervals where expected and actual production are both valid and comparable. A site with missing telemetry may have substantial risk, but its loss is unknown rather than zero. Keep communications and coverage problems in the queue with a distinct reason so data restoration competes fairly for attention.

2. Normalize the site context

Absolute energy loss helps rank business impact, while relative performance helps identify severity independent of system size. Use both. Record each site’s capacity, energy value assumption, timezone, model version, and data coverage for the same reporting period. Without consistent boundaries, a portfolio ranking can reward better instrumentation rather than better operation.

• Potential lost energy over the selected period
• Performance percentage relative to expected energy
• Data coverage and operational availability
• Duration and recurrence of the probable issue
• Classification confidence and recommended inspection action
• Estimated energy value using an explicit site-level assumption

3. Build an explainable priority score

A priority score should be decomposable into understandable inputs. Start with estimated impact, then apply documented weights for confidence, persistence, severity, and operational status. Avoid opaque precision: the score is a sorting aid, not a physical measurement. The interface should expose the reasons a site moved above or below another site.

4. Design the queue around decisions

1. Show the highest potential impact with its data-confidence qualifier.
2. Keep unknown-impact communications issues visible in a separate state.
3. Group repeated events so alert volume does not dominate the ranking.
4. Attach the probable cause, supporting evidence, and inspection recommendation.
5. Allow operators to change the reporting period without changing metric definitions.
6. Record resolution outcomes so future prioritization can be reviewed and improved.

5. Establish a review cadence

Use short periods for active operations and longer periods for persistent-loss review. A daily queue can direct immediate investigation; monthly and quarterly views reveal chronic underperformance and recurring data problems. Preserve the underlying interval evidence so an executive summary can always be traced back to the measurements and assumptions that produced it.