Case study · Solar operations

Hail in fourteen minutes. Nine to move the plant.

Albedo, operations software for utility-scale solar. It shows what a storm will cost before it arrives, stows each block just in time, and follows the damage all the way to a fixed plant.

Scope
Flow, product, rules
Surface
Control room web
Hard part
A plant of 4,592 trackers, drawn live
Deliverable
Coded screens, recorded flows, a live plant
00Overview

A warning with minutes in it, and a plant that needs nine.

At 15:24 a hail cell was heading for Los Arcos, 248 MWp in Badajoz. Every tracker stowed too early is power the plant stops selling. Every one stowed too late is a row of cracked glass.

0

Minutes a block of trackers needs to reach 60°

0

Trackers moved, block by block, each before its hail

0

MWh given up on purpose, shown before the click

3

Trackers caught flat, replayed with the new rule, down from 3

01The afternoon

Each block stows just before its hail.

Not all at once. D at 15:25, C at 15:45, B at 15:53, so the rest of the plant keeps producing while the cell crosses. Drag the afternoon.

Los Arcos · 248 MWp · Thursday 17 September

15:20

Output
198.6 MW
Trackers at 60°
0 / 2,758

Wide strips are tracking the sun. A thin blue line is a tracker at 60°. Orange is moving, or not answering.

02The product

One storm, every screen.

Every number matches across screens: 248 MWp, 2,758 trackers, 58 MWh, 14 silent, 41 cracked, 0 on the ones that stowed.

Fourteen plants, one cell on the radar
The warning, with its cost and a plan per block
The plan running: D at 60°, C moving, B next
The day after: where 144 MWh went
Thermal imagery of the 14 trackers that went silent
A work order with the day it should happen
Rules, replayed on the last 90 days
Close up of the stow card: the countdown, a stow and hail line for each block, and what the plan gives up

A plan per block, with the hail on each line

Close up of the running plan: Block D confirmed at 60°, Block C moving with 14 silent trackers, Block B still producing

Confirmed, moving, silent, still producing

03The day after

58 MWh was a choice. The other 86 were not.

Losses read as a chart you can open, so Friday starts with the one that pays to fix: two blocks the storm never rinsed.

Cleaning A and E pays back in 7 days. The rain forecast says nobody else will do it.

04The proof

41 cracked modules. 0 on the trackers that stowed.

A drone flies the 14 trackers that went silent. Three never got the command. The eleven beside them are clean.

The same flight becomes the insurer's evidence pack

05By Monday

A fixed plant, and a rule so it never happens twice.

The work order picks the day the parts and the weather allow. The rule pages the site manager the minute a comms box goes quiet in a warning.

Monday, because the breaker arrives at 08:00

Replayed on Thursday: 3 trackers caught flat become 0

06The board

The storm on one board, before a pixel.

One FigJam: the brief, Thursday as it stood, the tensions, Thursday rebuilt, the screen map and who hears what. Every number on it made it into the product.

Thursday rebuilt: every moment points at the screen that handles it

As it stood

Three tensions

Seven screens, three ways in

Only two moments interrupt anyone

07Decisions

Three calls made before a screen was drawn.

Each one has a price, and the price is written down.

  1. 01

    Stow each block just in time, not the whole plant

    Albedo times every block to reach 60° four minutes before its hail, so the rest keeps selling power. The cost: if the cell speeds up, those four minutes are the whole margin.

  2. 02

    Show what stowing costs next to the button

    Energy and revenue given up sit beside the command. The cost: on a real warning, the operator reads a number they would rather not see.

  3. 03

    A tracker that does not answer is never counted as safe

    Silent trackers stay orange until a person or a drone confirms them. The cost: more alerts on days when a comms box flickers.

08What we test next

Three assumptions that need evidence.

With control room operators and site managers, on recorded storms, with tasks rather than opinions.

  1. Assumption 1

    The margin

    Is four minutes enough when a cell speeds up, or do operators stow early anyway?

  2. Assumption 2

    The cost line

    Does seeing €2,780 next to the button make anyone hesitate on a warning that is real?

  3. Assumption 3

    The page

    Will a site manager act on rule 03 at 15:03, or wait for the control room to call?

Which decision in your product runs against a clock?

We find the decision your screen should be making, and hand it back designed in five days.