Earth Notes: On Website Technicals (2026-08)
Updated 2026-09-08.: Low Holiday Imports
Various of us away at different times this month has reduced residual imports even as generation (and a little rain, finally) fell.
uptime
17:27:07 up 363 days, 4:57, 1 user, load average: 0.30, 0.39, 0.42
: FUELINST Intensity Updates
I have updated the GB Grid Carbon Intensity intensity weights for interconnectors for calculations using the direct emissions for the country at the other end as computed by Electricity Maps: thank you!
Since intensities are falling in general, for better accuracy I have also made all of the 'fuels' non-zero gCO2w/kWh for calculations based on [UNECE2022electricity]:
NSPHYD(11gCO2w/kWh)NUCLEAR(5gCO2w/kWh)WIND(13gCO2w/kWh)
: powermng USoC wobble
Strangeness in the usable state of charge (USoC/UC) computation... When there is a lot of (primary array) solar generation, more than ~100W:
2026/08/28T10:20:06Z AL 1858 B1 12430 B2 -1 P 40398 BV 11679 ST OK D e A1P 23207 B1T 22 UC 100 A1V 28962 2026/08/28T10:30:06Z AL 1762 B1 12461 B2 -1 P 40461 BV 11652 ST OK D e A1P 21909 B1T 22 UC 100 A1V 29206 2026/08/28T10:40:06Z AL 6291 B1 13193 B2 -1 P 25094 BV 12775 ST OK - S A1P 82710 B1T 22 UC 0 A1V 17389 2026/08/28T10:50:06Z AL 8480 B1 13624 B2 -1 P 11622 BV 13319 ST H - t A1P 114765 B1T 23 UC 0 A1V 28562 2026/08/28T11:00:07Z AL 10664 B1 13654 B2 -1 P 11757 BV 13697 ST H - S A1P 152813 B1T 23 UC 0 A1V 27100 2026/08/28T11:10:06Z AL 3127 B1 13141 B2 -1 P 13470 BV 11924 ST OK D e A1P 40918 B1T 23 UC 88 A1V 26029 2026/08/28T11:20:06Z AL 11186 B1 14060 B2 -1 P 40296 BV 12459 ST VH D T A1P 134740 B1T 23 UC 0 A1V 26587 2026/08/28T11:30:06Z AL 1689 B1 12708 B2 -1 P 43042 BV 10959 ST OK D T A1P 21758 B1T 23 UC 100 A1V 26511 2026/08/28T11:40:06Z AL 9881 B1 14054 B2 -1 P 40785 BV 12495 ST VH D T A1P 138530 B1T 23 UC 31 A1V 31287 2026/08/28T11:50:06Z AL 9888 B1 14039 B2 -1 P 24288 BV 12767 ST VH - S A1P 135737 B1T 23 UC 0 A1V 30936
I thought it was overflow in the calculation, so I have made that essentially impossible by doing part in 32-bit.
It seems to be overestimating the actual sag/rise in the battery from my deliberately somewhat high battery impedance estimate:
Battery extra USoC reserve 5% for poor forecast and/or VHIGH / FULL not recent. Battery 1 effective/usable SoC 16%, dump thresholds min 25% higher 39% (uncompensated USoC 100%); usable to dump = 0; batt1FmV=14057 sagB1mV=1820 _T_batt_temp_comp_mV=-45. ... LASTDATA: 2026/08/28T11:38:19Z AL 10309 B1 13913 B2 -1 P 42115 BV 12495 ST VH D T A1P 144509 B1T 23 UC 16 A1V 31662
So I now ignore the sag estimate when the battery voltage is high enough for the battery to be about full:
Battery extra USoC reserve 5% for poor forecast and/or VHIGH / FULL not recent. Battery 1 effective/usable SoC 100%, dump thresholds min 25% higher 39% (uncompensated USoC 100%); usable to dump = 1; batt1FmV=14081 sagB1mV=1819 _T_batt_temp_comp_mV=-45. ... LASTDATA: 2026/08/28T12:04:47Z AL 7584 B1 13270 B2 -1 P 4074 BV 12960 ST OK - t A1P 107910 B1T 23 UC 100 A1V 32380
I also cleared a bit of shading for one of the panels so that they can both generate at once more of the time!
: BibTeX patent inventor
During (re)submission of my journal paper it became clear that I ought to allow a inventor field in @patent records, so for now, if there, it must be identical to author.
The Vancouver BibTeX style seems not to support record types @www, @dataset and @report, so I am rewriting them to @misc for this paper.
I also slightly improved error reporting for the conversion to HTML sub-pages.
Myenergy dashboard broken again
It will not authenticate with by normal SSO and so I cannot see real-ish-time data via it.
As of the evening of it seems to be working again...
: Eddi Export Margin API?
It seems as though the Eddi's Export Margin is adjustable in real time. That is a feature I would very much like to use to smooth grid response (eg) to storage levels and grid carbon intensity, rather than the 'bang-bang' STOP that I use now.
But I will have to discover that API somehow if twonk does not get there first!
Instead of STOP I might set the Export Margin to ~3700W, ie the G83/G98 limit, for example, for diversion.
: Paper Tiger
This morning I submitted paper revisions in response to round 1 of reviews, a day ahead of deadline! ... And got a response back the same day, asking for more major revisions!
uptime
10:28:57 up 348 days, 21:59, 1 user, load average: 0.58, 0.48, 0.52
: Greener Net Negative
The rising net export (negative import) figure over ~1 year (374 days) is good!
Winter USoC dump thresholds
Thinking ahead again to mid-winter, I have set powermng now to nominally allow 2 days' dumping (keeping the router off-grid), eg in state 'e' even without grid imports, by adjusting the upper usable SoC threshold for dumping to accommodate those 2 days' router load, assumed to be 30% USoC per day. I have set more than one day' router load, as rarely is every day going to get the battery fully charged then.
This has lowered the upper dumping threshold from ~60% to ~40%, with the lower threshold remaining at 20% of USoC, ie about 60% full SoC.
Heatwaves
I added to the spelling dictionary used by the EOU site build heatwave, heatwaves (why yes, I am writing about them more!) and microinverter.
uptime
17:33:30 up 342 days, 5:04, 1 user, load average: 0.71, 0.60, 0.55
: preprint
Also added preprint to the dictionary, given the excitement of seeing my paper cited in one [agyeman2026HaaS].
: Hot Diodes
~Z I checked the connections in the back garden main PV connection box; all seem fine. But the Schottky diodes used to parallel the 255Wp panels were painfully hot, though probably not hot enough to harm them. Both diodes were hot even though one panel is partially shaded.
The microinverter will be adding a bit over 1A load, or maybe 0.5W extra heat, to that.
Maybe I should find small clip-on heatsinks for them. I think I have some.
: Paper Review Updates
I am working on first-round updates/corrections to my paper currently in flight with a journal.
One improvement asked for by the editors and reviewers is higher figure quality
. I take this to be both technical presentation/resolution and content/labelling, etc.
Many of the figures are gnuplot line graphs. For HTML pages I am now using SVG where possible because scalable and often much more compact than PNG, especially brotli-compressed. But pdflatex with which I am preparing this paper does not handle SVG, so initially I had reverted to PNG, but there are visible artefacts from the 640x480 pixel raster format used.
I have started to generate PDF output alongside SVG and PNG, as an alternative scalable vector format (ie without raster artefacts) that pdflatex does accept. 
I hope that this sufficiently addresses the technical issue for those graphs that I can easily regenerate from data. Some historic PNG snapshots (eg screenshots) may be harder to improve.
I have also changed the axes to be a more conventional journal style, with more descriptive axis labels and the axis tick values purely numeric.
: End of Month Data Fun
I was away at a music festival for a few days and while obviously I logged in from my phone via ssh to commit the bulk of the EOM data, some parts of the ritual work better locally, while sober, maybe... I think everything is now done.
More carbon negative
Net electricity utility energy kgCO2e over the last 12 full calendar months: -64
Net kgCO2e for each of the last 12 full calendar months: -40 -8 11 27 42 52 28 -2 -22 -40 -52 -60
At 16WW over 367 days from 2025-08-01 to 2026-08-03: electricity import 3.1kWh/d (net -2.2kWh/d).
uptime
% uptime 15:50:25 up 337 days, 3:21, 1 user, load average: 0.39, 0.41, 0.44
References
- [agyeman2026HaaS] Heat as a service as a viable decarbonisation strategy and energy cost affordability in the UK
- [UNECE2022electricity] Carbon Neutrality in the UNECE Region: Integrated Life-cycle Assessment of Electricity Sources
(Count: 2)