02·PERF

Energy Performance

Seven and a half years of measured output from the array and the batteries — 97.2 MWh generated, 37.5 MWh sold back to Clark Public Utilities, and a production figure that has not moved more than three percent in six years.

956 N Noble Loop · Showing Edition · Rev A · Aug 2026

The solar array on this roof has been metered continuously since the Tesla gateway went in. Everything on this page is computed from those exports at build time — the yearly, monthly, daily and five-minute files exactly as the Tesla app produces them. No estimate, no modelled figure, no marketing arithmetic. Where the data is partial or where Tesla’s own files disagree with each other, this page says so.

97.2 MWhTotal generated

37.5 MWhSold to Clark PUD

13.5 MWhA typical year

7 yr 7 moOf metered record

Two numbers put the rest of the page in context. The array’s engineering layout, drawn before it was built, estimated 12,471 kWh a year. The measured average across the six complete steady-state years is 13,467 kWh — 8% above the design estimate, every year, for six years. And of the 97.2 MWh the roof has made, 59.7 MWh was consumed in the house as it was generated: 32% of everything this house has used since 2019 came straight off the roof, before any credit for the 37.5 MWh that was exported.

Where these numbers come from, and what to trust

Ten CSV exports from the Tesla app, taken on 5 August 2026: one lifetime-by-year file, two by-month files (2025 complete, 2026 to date), six by-day files (March–August 2026), and one five-minute file for a single day. The units are not consistent between them — the yearly file is in MWh, the daily files in kWh, and the monthly files mix the two within one header row — so every figure here is normalised to kWh before anything is added up. The 2025 totals were then re-derived from the monthly file and checked against the yearly file: solar agreed to 0.2%, household consumption to 0.3%, and the Powerwall column to the kilowatt-hour, which is what confirms that the mixed-unit header is real and not a misreading.

One column where Tesla’s own files disagree

Exported energy is the exception. Summing the daily files for March–July 2026 gives 8% more export than the monthly file for the same five months; summing the monthly files for 2025 gives 15% more than the yearly file for the same year, and 2026 behaves the same way. Generation and consumption agree across all three roll-ups to well under a percent — only the export column drifts, and always in the same direction: the coarser the file, the smaller the number. This page quotes the yearly file, the most conservative of the three, for every export figure it prints. A buyer who wants the exact billed figure should take it from Clark Public Utilities’ meter readings, not from any app.

Eight years of production

The array has produced between 13.3 and 13.7 MWh in every complete year since 2020 — a spread of 3% across six years of Pacific Northwest weather. Solar panels are usually sold on their first year; this is what the sixth looks like.

Solar generatedMWh · by year

Used in the houseSold to the grid

Six complete years between 13.3 and 13.7 MWh, of which about 40% went out to the grid rather than into the house.

Why 2019 is short, and how to read it

The 2019 row records 7.4 MWh of solar and only 9.0 MWh of household consumption, against 20.8 MWh the following year. A 3,176 sq ft house does not use 9.0 MWh in a year. The likeliest reading is that 2019 is a partial year of monitoring, not a partial year of production: the Tesla gateway that does the measuring was installed with the Powerwalls in 2019, so the record begins when the gateway did, not when the array was energised in November 2018. The two ratios — 43% of the next year’s consumption and 54% of its solar — fit roughly seven summer-weighted months of data. That is an interpretation of the numbers, not a documented fact; it is offered because the alternative reading (that the array itself was half its present size for a year) would be a material claim, and nothing in the build record supports it.

What a normal year looks like

2025 is the most recent complete year, and it is the one to look at. The shape is the whole Pacific Northwest solar story in one picture: a broad summer plateau, a deep winter trough, and a household load that barely notices either.

Roof vs houseMWh/month · 2025

Solar generatedHouse consumption

In July the roof made 59% of what the house used; in December it made 12%. The house’s own demand moves by barely a third across the year — the sun does all the moving.

Bought and soldMWh/month · 2025

Bought ↑Sold ↓

The house bought power in all twelve months and sold some back in all twelve — but April’s surplus was fifteen times December’s. What it buys hardly changes with the season; what it sells changes by a factor of fifteen.

That flat top row is worth a sentence. What the house buys from Clark PUD stays between 1.7 and 2.3 MWh in every month of the year, because the household load rises through the summer at almost exactly the rate the array does. The solar surplus mostly leaves the property rather than displacing a winter bill — which is precisely why the design estimate targeted a 71.9% usage offset rather than 100%, and why Clark PUD’s March 31 credit true-up matters. Both are explained on the Electrical & Energy page.

Every day for five months

Annual and monthly figures flatter solar. Days do not. Here is every single day from 1 March to 4 August 2026 — 157 of them, unsmoothed, with a seven-day mean drawn over the top so the season is visible through the weather.

Daily productionkWh/day · Mar–Aug 2026

One day7-day mean

The best day made 78.6 kWh; the worst made 4.2 — a 19-fold spread inside one season. The trend is smooth; any individual day is weather.

Across those 157 days the roof produced something on every one of them, exported to the grid on 145, and out-produced the entire household on 9. The worst day of the run, 11 March, made 4.2 kWh — about what a single electric-car charge draws. This is the honest shape of solar at 45° north, and it is why the Powerwalls and the grid connection both still matter.

One day, five minutes at a time

The finest-grained export available is a single day at five-minute resolution: 5 August 2026, pulled at lunchtime. It shows the mechanism the yearly charts only imply — the house waking up on grid power, the array coming up at 06:45, crossing the household load at 07:40, and running the house on sunlight with a surplus going out to the meter for the rest of the record.

One day, 5-minutekW · 5 Aug 2026

Solar outputHouse demand

From 07:40 the roof covered everything the house was doing and sent the rest to the meter — 14.9 kWh exported before one o’clock.

The biggest load of the day ran in the dark. 13.4 kW at 04:20 — the signature of a car charging overnight, which is also what the rising annual consumption on this page looks like. See EV charging.
Peak solar output was 6 kW on an 11.7 kW DC array. That is normal and expected: these are five-minute averages, in August, on a system whose AC point of interconnection is placarded at roughly 7.6 kVA. Instantaneous DC nameplate is not what a roof delivers.
The Powerwall column reads zero at every one of the 156 intervals, and the state-of-charge column reads 100% at every reading, all day. The batteries neither charged nor discharged — they simply stood there, full. That is not an anomaly; it is the posture this system has been run in for years, and it is the subject of the next section.

The Powerwalls: from daily cycling to standing reserve

The batteries have a documented history, and it is not the one a brochure would choose. It is more interesting than that.

Drawn from the batterieskWh · by year

About 1.5 MWh of everyday cycling in 2019 and 2020, a 920 kWh burst in 2022, and essentially nothing since — 10 kWh in the whole of 2025.

Read plainly: the Powerwalls did roughly 1.5 MWh of work a year in 2019 and 2020, then stopped. From 2021 onward the only material discharge in the record is 920 kWh in 2022. In 2025 the batteries gave up 10 kWh across the entire year — and the monthly file puts all of it in December.

A battery that is not discharging is either broken or held in reserve. This one is held in reserve, and the five-minute file is the direct evidence: state of charge pinned at 100% for every reading of a summer day with a 4.5 kW surplus available to it. That is what a backup reserve set at or near 100% looks like in the data — the owners chose insurance over arbitrage. The energy that would otherwise have cycled through the batteries each evening went to the grid instead, which is why the export figures on this page are as large as they are.

What that posture bought — and the caveat on it

The owners’ account is that across years of grid outages the switchover has been undetectable in essentially every event: no flicker, no clocks to reset, and the news usually arriving from a neighbour or from the app. One perceptible blink, early on, is the single exception they recall. The backup covers the whole house — there is no “essential circuits” sub-panel — which is unusual for a Powerwall install. This is owner-reported. The data on this page corroborates the posture (a battery held full and idle) but not the events themselves: the Tesla app’s Backup History log is the objective record, and it is being archived for the buyer’s edition. Until it is here, treat the outage record as testimony and the reserve behaviour as measurement.

Two practical consequences for whoever lives here next. First, that reserve is a slider, not a fixture — moving it down would put those evening kilowatt-hours back through the batteries instead of onto the grid, at the cost of cushion. Second, the batteries have therefore done remarkably few cycles for their age, which is the variable that governs battery life. The Electrical & Energy page covers the setting itself; the power-outage runbook covers what to do when it earns its keep.

The record, year by year

Every figure charted above, in full. Energy in MWh except the Powerwall column, which is small enough to want kWh. “Used on site” is generation minus export — solar the house consumed as it was made.

Year	Generated	Exported	Used on site	House used	Bought in	From battery (kWh)
2019 partial	7.4	1.8	5.6	9.0	3.7	1,500
2020	13.6	5.7	7.9	20.8	13.2	1,500
2021	13.7	6.4	7.3	23.3	16.2	0
2022	13.4	5.3	8.1	25.0	17.3	920
2023	13.4	5.8	7.6	26.1	18.7	0
2024	13.3	5.2	8.1	31.4	23.4	0
2025	13.4	4.2	9.2	32.7	23.7	10
2026 to 5 Aug	9.0	3.1	5.9	19.0	13.3	0
Total	97.2	37.5	59.7	187.3	129.5	3,930

And 2025, the most recent complete year, month by month.

Month 2025	Generated	House used	Bought in	Sold out	Roof covered
January	0.63	2.40	1.90	0.18	26%
February	0.52	2.30	1.90	0.14	23%
March	0.87	2.40	1.90	0.31	36%
April	1.50	2.50	1.70	0.74	60%
May	1.70	2.80	1.80	0.71	61%
June	1.80	3.00	2.00	0.67	60%
July	2.00	3.40	2.10	0.71	59%
August	1.60	3.30	2.20	0.48	48%
September	1.30	3.10	2.30	0.44	42%
October	0.82	2.90	2.30	0.27	28%
November	0.41	2.30	2.00	0.11	18%
December	0.28	2.40	2.10	0.05	12%

Reading your own data

Every number on this page came out of the free Tesla app, and it will keep producing them for the next owner. Nothing here needs a subscription, an installer or a service call.

Live: the app’s power-flow screen — sun, house, battery, grid, and which way the energy is moving right now. That is the screen that makes the whole system legible in about four seconds.
History: the energy graphs, switchable between day, week, month, year and lifetime. The month and year views are the ones worth a monthly glance; the day view at five-minute resolution is what the last chart on this page is drawn from.
Export: each history view has a download control that produces exactly the CSV files this page is built from. Pull one a year, keep them, and you will have a maintained performance record of the house — the single most useful thing you can hand the next owner.
Outages: the app keeps a Backup History log of every grid event the gateway rode through. It is being archived into this manual as the objective counterpart to the owners’ account above.

The one comparison worth making

Once a year, compare the year’s generation figure against the eight rows in the table above. A year that lands inside 13.3–13.7 MWh is a healthy system. A year materially below that, with no shading or weather explanation, is worth a service call — and the most common cause is one of the two inverters having quietly failed, which halves production and nothing else. That diagnosis, and what to do about it, is on the solar production looks low runbook.

Electrical & Energy — the system itself: what the equipment is, how backup works, net metering, and the account transfer.
Runbook — Power outage — what the Powerwalls do for you, and the few things you should still do.
Runbook — Solar production looks low — how to tell weather from a fault.
EV charging — the largest single load in the record above.

Tesla — the Tesla app for energy
Power flow, energy graphs, and the history download that produces the CSV files this page is built from.
Tesla — Backup History
The app’s log of grid events the gateway carried the house through — the objective record behind the outage account above. To be archived.
Clark Public Utilities — solar net metering
What happens to the exported energy on this page, and the March 31 annual true-up.

Provenance

Source data: ten CSV exports from the Tesla app for this address, taken 5 August 2026 and archived in the manual’s source tree. Charts and figures are generated from those files at build time — this page cannot disagree with its own data. Nothing here identifies an account, a meter or a site.