Earth Notes: Off-Grid Stats and Equipment Change Notes

Available historic (closed) data set and snapshots:

• Historic storage stats snapshot 2009-09 (K8055 system)
• Historic Off-Grid Stats via K8055 (2007 to 2014)

Equipment Change/Upgrade Notes

• 2023-12-01: I am retiring Radbot 2T '7h' in the hallway (which has failed 'on') with a Secure Radbot1 turned down low or frost; it will not give readings or remote-control the boiler, but that radiator was barely used anyway other than on very cold days and for parties...
• 2023-11-24: after discussion, central heating on times now will be nominally 07:30 to 08:59, 13:00 to 15:59 and 19:00 to 21:29 (7h total). GB peak space/total heat demand is at ~6pm and peak ramp rate is at ~7am [watson2019peak], so both are avoided in this updated scheme.
• 2023-06-26: winterising make pages-incr to be run only in daylight, when the off-grid battery is net charging from the primary (MPPT) controller, ie in the presence of the new /run/CHARGING.flag.
• 2023-06-24: lowered usable SoC (above 50% absolute DoD discharge limit) reserve threshold from 40% to 30% since I think that the dump load is coming off prematurely and is thus not being run enough to make room for new charge, rather than there being a shortage of sun yet. New reserve should still cover ~1W of RPi3B.
• 2023-06-10: as an experiment for a week or so (well, until mid-September in the end!) now that the Thermino is unlikely to empty, the gas boiler is now turned off to avoid the unnecessary brief ignition when a tap is turned on hard.
• 2023-05-26: the (rechargeable) batteries inside the SunnyBeam appear to be dead: I had to reset the time just now. Also, the Ethernet connection to the hub from the Thermino RPi needed to be reseated after connectivity was lost; that connector does not have a clip.
• 2023-05-15: the wired USB supply does appear to be completely dead, eg as tested with a Muker USB power meter. Amazon delivery eventually turned up ~7:30pm. Borrowed an RPi4B (thank you Alan F) in the hope that I might be able to put the RPi3B card in there and have it boot, but no dice; the OS on the card may be too old. New RPi3B with no SD card shows nothing on the HDMI. Boots off existing card, but I am regretting having my monitor in portrait! Doing forced fcsk by temporarily adding to /boot/cmdline.txt the argument fsck.mode=force. Filesystems check cleanly. All looking good. Now will replace the power supply, reassemble, and cross fingers. I am taking an image of the entire SD card to my Mac first with sudo dd if=/dev/disk4 of=~/PiSDCardBackup.dmg. I seem to be getting over 22MB/sec or ~1GB/min from the SD card to my Mac's SSD for the backup (started ~8pm, over halfway done by 10pm, finished before 11:15pm, final dd stats 256087425024 bytes transferred in 11232.743693 secs (22798297 bytes/sec)). I can see why the old RPi was unhappy — the old supply (purchased 2020-06-17) seems to be letting through 13.5V+ to its USB outputs! The replacement one wired in is emitting a more dull and conventional 5.16V for now. I reassembled the board with its HAT it and reconnected the USB and Ethernet cables ready for the SD card to slot in, and connect the power 22:15Z. At 22:20Z the dump load was enabled. Manual clearing up of various log files will be needed in some cases; trimming of bad data from while the system was brought up without correct real time or sensors, possibly some of the gaps.
• 2023-05-12: the RPi3B server seemed to take a lie down just before 10am (09:00Z); and indeed the server and its power supply both seemed to have died (though the micro SD card content appears to be intact) so replacements have been ordered... I am paying something over £10 in excess of the official distributors' price for the RPi3B (~£54 vs ~£42 including VAT). Attempting to power up the RPi3B with a mains supply results in something near the connector getting very hot, and no happy flashing LEDs.
• 2023-04-28: and again adjusted powermng to more aggressively support 'dump' load when FULL has been achieved recently, reducing the L threshold from 12.45V to 12.35V; the dump load then stayed energised essentially all night. Also checked (and in a couple of cases tightened) major connections in the off-grid system; no problems were found.
• 2023-04-15: adjusted powermng to more aggressively support 'dump' load when FULL has been achieved recently, nominally lowering the excess usable charge threshold above 50% SoC from 50% to 40%, though the L cutoff limits the effect for now.
• 2023-03-15: heating off this evening, and meter readings taken.
• 2023-02-09: re-bodged the off-grid power connection for the router around noon, since good off-grid generation is otherwise going to waste. The few days back on the router's own mains adaptor showed normal levels of flakiness, which suggests that the 12V off-grid/on-grid power adaptor is not the cause.
• 2022-12-05: mail server (green) not responding after power cycling last night and this morning; may require TLC plugged into TV. Long overdue (more than a year) to move away DNS and mail and DNS... It seems that the SD card interface is glitchy (which may explain observed behaviour over a long time). A few errors had also accumulated, which fsck sorted. Now I am spending several hours with fsck -c -v checking for bad blocks (I don't know if this even works)...
• 2022-12-02: wired Ethernet link between pekoe and router fixed by builder inserting coupling and short segment of my known-good Ether cable. (His fancy tester showed all of 1 to 8 live on his cable, but just 1, 2 and 6 on my segment and on the whole thing with the coupled arrangement. We had too much of a good thing?
• 2022-12-01: heating on today (starting at noon with all doors shut and only me here to estimate response time; house currently ~15°C vs ~8°C in porch (4o) vs BBC Weather claiming ~6°C); to further trim gas use and avoid electricity grid ~9am to 10am peak, I have adjusted the on periods to be 06:30--08:59, 13:00--15:59, 19:30--21:29 (7.5h total). This may allow airing of rooms late morning.
• 2022-11-21: 11:00Z: spot-check on performance of 100Wp panel on very gloomy rainy day with a couple of sycamore leaves on and partly blocking the panel: Isc ~200mA, V (in circuit, upstream of diode ~0.3V drop and PWM) ~14V, so ~3W! Roof ~5kWp is producing ~30W so entirely respectable.
• 2022-11-20: switched powermng to using 7-day versions of grid-intensity flags as should be more representative of need.
• 2022-11-13: upgraded Thermino-management RPi 'temporary' WiFi dongle to 'n' 2.4GHz Ralink MT7601U USB2.0 device from the Pi Hut, hoping to make comms a bit more robust than when using the device that came with the Maplin 'starter kit'.
• 2022-09-09: mounted the 'spare' 100Wp (~23Voc) panel near-horizontal on the shed roof today, ready to connect in to the PWM solar controller; it turns out that I slightly underestimated full cable length, but I have enough. I am connecting it via a protective rectifier: the P600A (6A) does not seem quite enough but is chunky, the MBRF1645 (16A, Schottky) looks more delicate but may do the job better even without heatsink. Saw draw of ~2.5A from new panel. Done by 13:00Z.
• 2022-07-10: finally pulled the fuse on the off-grid LVD to car adaptor to MacBook supply, which has not had any use in weeks now; will save a little juice, at most ~500mW out of ~14W typical. Not unwiring it just yet as no pressing need for LVD elsewhere.
• 2022-06-04: a quick inspection (unplug and test with multimeter) indicates that the amorphous panels (~20Wp, Voc ~20V) on the porch are still alive and generating, and also alive is the currently undeployed and rather grubby 100Wp mono (Voc measured ~23V, nameplate 26V, which appears to have M4 connectors).
• 2022-06-04: joined PVoutput.org, connected to Enphase for live readings, uploaded some initial and latest daily generation readings from SunnyBeam to prime it. May go back and upload a month at a time as 'ranges'.
• 2022-05-23: I adjusted heat battery boost from mains to be avoided below 245V; the threshold was 240V and thus unlikely to have ever triggered.
• 2022-03-27: to prolong memory card life, updating fstab for main ext4 filesystem from defaults,noatime to defaults,noatime,commit=300.
• 2022-03-09: the Sunamp Thermino was installed today, by ~4pm.
• 2022-03-08: to further trim gas use I have turned off the central heating entirely this evening just into the final 'on' period (8pm), with the intention of leaving it off, ie stopping our heating season here, unless things get very cold.
• 2022-03-08: 4o has just eaten another set of AA cells.
• 2022-03-07: to further trim gas use I have shrunk the first 'on' period; so now 06:30--10:29, 13:00--15:59, 19:30--21:29.
• 2022-03-02: raised eddi export margin (and threshold) to 100W and ET 'export threshold' to 50W from 25W (script/EnphaseBasicCapture.sh) to reduce false triggers and spill a little more to grid before storing heat.
• 2022-03-02: to further trim gas use I have shrunk the final 'on' period; so now 06:30--10:59, 13:00--15:59, 19:30--21:29.
• 2022-02-26: in order to slightly trim gas use I have brought forward the "off" time for the central heating to 21:29.
• 2022-02-23: electrical works for Thermino including Eddi done today; Eddi stats logging started. Though correctly oriented and on the correct (L) cable, the Eddi grid flow current clamp was showing readings in the wrong sense, so was swapped to the N infeed cable with same sense to get flow direction correct.
• 2022-02-15: and another flat set of 4o cells, swapped out for fully-charged little-used 2800mAh NiMHs.
• 2022-02-15: ~8am: ~6°C outside and rooms chilly with boiler on very short cycles (well under a minute flame time), so have bumped up flow temp to about 1/3rd; circuit pressure may be holding better.
• 2022-02-13: because the central heating radiator circuit pressure keeps dropping to 0 (and rising sharply, so I fear a problem with expansion vessel and/or overpressure relief valve) and the boiler is making unhappy noises at low pressure, I have turned the flow temperature down to minimum (~30°C) to see if that stresses the system less.
• 2022-02-05: and another flat set of 4o cells, swapped out!
• 2022-01-30: returned ET 'export threshold' to 25W from 125W (script/EnphaseBasicCapture.sh) before which off-grid stops dumping in OK mode, anticipating ~50W Eddi threshold/margin.
• 2022-01-20T21:00Z: air temperature probably a couple of degrees above freezing and it seems as if the Enphase AC Batteries may be refusing to fully discharge because of the cold, ie they are drawing power while not empty (298Wh) and there is demand in the house.
• 2022-01-09: 4o battery has (nearly) expired again or at least transmissions are not being received reliably: it's not much above zero this morning; will fit fresh pair of 'new' (freshly fast-charged but have sat unused in packet for several years so may be reduced capacity) 2000mAh NiMH. I think that an extra TX or three got picked up from when I decided to replace the AAs to when I actually did...
• 2022-01-05: shortened time since last FULL to stronger conservation to better protect battery against over-discharge mid-winter. See battery voltage over the month.
• 2021-12-25: 4o battery has expired again: fitting sealed pair of GP alkalines.
• 2021-12-22: observation suggests that the temperature in the porch may be consistently ~1--2°C warmer than true outside temperature because of the sheltered aspect, heat leaks from various place including batteries, etc.
• 2021-12-08: 4o cells have apparently expired yet again; it is ~8°C outside. NiMH cells from the pile on my desk swapped in.
• 2021-11-16: 4o cells have apparently expired again; it is ~5°C outside. Part-used alkalines swapped in.
• 2021-11-06: 1g swapped out for new device with fresh GP (gold) alkaline AAs noon-ish; central heating now on with times 06:30--10:59, 13:00--15:59, 19:00--21:44.
• 2021-11-04: actually switched on boiler controller (though boiler itself remains in DHW-only mode, only 1g calling for heat, giving a 1h test firing at 7pm with 2b also turned up; MHRV upstairs now on too) and "b" stats are visible from it [ "2021-11-04T09:48:43Z", "", {"@":"84F3D6D893A3F2E0","+":6,"b":1,"R":157,"SP":255} ]
• 2021-10-28: two further Enphase AC batteries installed, now ~5kWh/1kW.
• 2021-10-28: new NiMH cells in 4o (chart before replacement, radio not being received well for ~24h, though saw temp reading before new batteries put in).
• 2021-10-18: 09:50Z: appended storage.whNow to Enphase 5m stats log. First entry: 20211018T09:53Z consumption.readingTime 1634550778 consumption.net.wNow 3.893 consumption.total.wNow 414.309 production.wNow 410.417 storage.percentFull 10 storage.wNow -386 storage.readingTime 1634550273 storage.whNow 260
• 2021-10-17: corrected the SunnyBeam clock forward ~15m while taking total generation readings. Note that for today SunnyBeam reports 4.51kWh and Enphase claims 4.7kWh.
• 2021-10-06: powermng: allowing dump to continue at a slightly lower SoC.
• 2021-10-03: turning on boiler controller (but leaving boiler itself in DHW-only mode) to start to collect stats, but stats not immediately coming through. (Pressure in rad circuit is about zero!) [2021-11-04: turns out it is useful to actually plug in the boiler controller: switching other random things on less so!]
• 2021-09-27: put 4o from being upright to on its side (top pointing towards the house) and reception reliability has improved dramatically at least to start with.
• 2021-09-16: replaced the NiMH cells in 4o (expired 2021-09-12) with part-used alkalines.
• 2021-09-16: powermng: made dumping even more aggressive: long-time-since-VHIGH now 3d+8h, long-time-since-FULL now 30d like an equalisation cycle.
• 2021-09-15: second Enphase AC battery installed today raising nominal grid-coupled storage capacity to 2480Wh.
• 2021-09-08: powermng: made dumping a bit more aggressive eg by tightening conditions for 's' mode, and pushing out 'long time since VHIGH' to over 2d.
• 2021-09-03: but about noon GMT old bank disconnected and out of the shed; new battery on-line and everything reconnected ~15:30Z.
• 2021-09-02: new Victron 12V 220Ah gel BAT412201104 arrived; will be installed in next day or three replacing 4x99Ah bank.
• 2021-08-02: replaced the NiMH cells in 4o.
• 2021-07-17: not quite an equipment note, but the foxgloves in front of one of the large garden primary array panels are finished and down.
• 2021-06-14: raised ET 'export threshold' from 25W to 125W (script/EnphaseBasicCapture.sh) before which off-grid stops dumping in OK mode to restore its own SoC, ie making off-grid system work harder to keep load off grid unless effect too tiny to matter.
• 2021-04-28 18:30Z: repurposed customer-return Radbot physically damaged as 4o replacement!
• 2021-04-23: 12:00Z: temporarily hard-wired in 12V to ~15V car adaptor (7.5A fuse in-line) and taken the MacBook off-grid.
• 2021-03-28: improved connection from 12V off-/on- grid 12V supply by jamming some wire into the recessed part of the plug. I don't expect this to end well as-is! Vigor router currently being powered off-grid (18:05Z).
• 2021-03-26: added AV1 primary array voltage (mV) to end of logged stats 2021/03/26T14:07:34Z AL 42 B1 14396 B2 -1 P 3168 BV 14162 ST VH D h A1P 6765 B1T 12 UC 100 A1V 32441 eg to track when Voc is floating above Vmpp (~28V) and thus available power is going unused.
• 2021-01-25: 22:10Z: moved the Vigor router back to mains since somehow power from the Maplin supply seemed unreliable. Maybe this managed to move ~60Wh load from mains today on a bright day... Muddied by ISP reporting total loss of service on some broadband connections at the same time.
• 2021-01-23: 14:25Z: moved the Vigor router from mains to the 12V Maplin 'dump load' supply, so will be partly off mains, ~14W observed extra draw (total ~16W). This makes our grid connection battery-backed ie with UPS!
• 2021-01-18: ~3pm: Loop system finally closed: have removed gas, electricity and network parts. Also means that there is not actually a dump load at the moment...
• 2020-12-05: suppressing most dumping if not FULL for several days (eg mid-winter) unless VHIGH or high SoC, and only for grid red/peak condition.
• 2020-12-04: 14:30Z: removed SIM900 GSM board from REV10 boiler controller to attempt to improve radio RX, in particular because 1g does not seem to be getting heard. There seem to be fewer ?RX warnings reported on serial.
• 2020-12-01: ~8am: at cusp of net generation, SunnyBeam reports ~20W when Enphase reports 0 net flow to/from generation arrays, eg 20201201T08:03Z 17 SunnyBeam recorded generation when Enphase says exactly 0 from live readings.
• 2020-12-01: ~7am: central heating turned on (under OpenTRV boiler control) with run clock set for roughly 06:45 to 15:45 and 19:15 to 21:45 to match when we're up and avoid peak (electricity) grid demand times.
• 2020-11-15: boiler controller switched on and reassociated with valves in anticipation that heating will likely be on by end of week. First light: {"@":"84F3D6D893A3F2E0","T|C16":258,"b":0,"R":133}.
• 2020-10-19: solar combiner box moved and downpipe unclogged.
• 2020-08-21: (12:35Z) USB/MODBUS connection unplugged from RPi2 system and into RPi3B; powermng on the RPi3B did not even need a rebuild to use it properly!
• 2020-08-20: (19:29Z) 5V connected up (see full I/O ribbon listing) via I/O ribbon grey wire to dump relay/driver orange wire, so dump load is operative again, now from the RPi3.
• 2020-08-18: integrated ExpanderPi GPIO support for dump load control into powermng, conditionally compiled for RPI3B target. Note that once dump load control wire is moved over power management will be split between the old and new servers until MODBUS and SolarBeam also moved over. (Purple wire from dump load SSR is +ve, blue wire is GND. Orange wire is 5V, to power a driver transistor and relay; will need connection to RPi3 board, probably grey wire, so dump is not-operative from ~19:00Z until fixed.) HAT stripboard (top) no longer in use on RPi2, but put back so as not to get lost!
• 2020-08-02: moved desklight support to RPi3B (and controlling script to /usr/local/bin), via ExpanderPi GPIO.
• 2020-07-13: cigar-lighter plug of mains adaptor used to power my MacBook Air has melted! So I am on grid for a while at least...
• 2020-06-23: moved RPi3B to off-grid system powered via wired-in JZK Car Power Converter DC 12V to 5V/ 3A Voltage Converter. (Rated 96%+ efficient, input 8--20VDC, output 5VDC 3A, 15W.)
• 2020-05-23: since the lead-acid battery bank temperature has reached about 25°C, I have taken the top layer of insulation off until autumn, to encourage cooling.
• 2020-04-12: put the solar box back up on the wall since filling with water is not an immediate risk, but moving it horizontally further from the roof edge and unblocking the downpipe still needed.
• 2020-03-30: Looking at the change in electricity demand patterns, I am moving the evening "off" period for the central heating from 4--7pm to 5--9pm for now. I'll leave the Enphase weekday fake 'peak' ToU rate 4--7pm, however.
• 2020-03-25: Loop electricity transmitter seems to have gone off the air 2020-03-22T18:06Z having transmitted a reading for 18:00. Unclipping and reclipping round the wire, and power-cycling the Loop Internet receiver seems not yet to have fixed it. Neither does a reset (with the recessed button on the transmitter). This device may have died. I have better (directional) readings from the Envoy anyway.
• 2020-02-07: (09:00Z) some trimming of the large trees to SE is going on, so may see some improved morning PV generation.
• 2020-01-18: (~13:20Z) non-heatsinked new diodes are warm but not hot (a bit warmer than the axial Si rectifiers for the 12V panels) while routing ~20W each, so maybe a little under 1A each.
• 2020-01-12: using a pair of separate Taiwan Semi 100V 10A, Schottky Diode, 2-Pin TO-220AC MBR10100 without heatsink as a temporary measure. In (weak) sunlight maybe one tab getting slightly warm.
• 2020-01-04: went to investigate unexpectedly low off-grid PV generation from primary panels, even with some good direct sun around noon over last day or so, including tightening connections, and found outside junction box full of water. Diode pair from main panels destroyed/corroded, no suitable diodes (≥9A), so only one panel wired temporarily and new diodes have been ordered. RPi crashed somehow during this (~13:20Z) maybe due to supply disturbance.
• 2019-12-31: SunnyBeam stopped responding properly over USB yesterday and display mucked up today, so took batteries out and reinserted and reset date/time. Entered generation numbers from display yesterday and Enphase and sunnybeamtool readout today.
• 2019-11-23: looks as if 4o TX failed yesterday as nothing heard from it since [ "2019-11-22T12:13:46Z", "", {"@":"819C99B4B9BD84BB","+":10,"L":182,"T|C16":130,"O":1} ] and a change of battery and reassociating at the REV2 status hub have not helped.
• 2019-10-14: 12V off-grid supply seems unreliable (poor connection at the power input barrel connector?) so have reverted to mains for now.
• 2019-09-03T08:00Z: moved DrayTek Vigor 2862ac router to be powered via 12V off-/on- grid 'dump' supply; non-dump quiet draw (RPi2) ~1W, dumping to Loop ~5.5W, dumping to Loop+DrayTek ~16.5W (ie DrayTek ~11W). Power connection may be flaky.
• 2019-07-29: 12V extender switch 03040 (+2xUSB) input plug partly melted under heavy load (charging my near-empty MacBook Air), so is now useless as-is. May be worth stripping down and using as permanently wired.
• 2019-02-25: SmartThings units taken down: will try to find new location for cross-calibration w/ occupancy sensing.
• 2019-02-13: the dump relay seems to have stuck in the 'off' position, ie mains continues to be supplied to the 12V car adapter and so the networking gear is never being powered from the off-grid system. Seems to have unstuck 2019-02-17.
• 2018-12-07: even with powermng nominal status VERY LOW the SS-MPPT-15L battery status light is still green, so the battery is being looked after.
• 2018-09-28 and 29: replaced 5s, 1g with TRV2/Radbot2 T2 samples.
• 2018-09-20: replaced 2b, 3l, 6k with TRV2/Radbot2 T1 samples.
• 2018-09-18: dump relay stuck ~19:28Z, so tapped to free.
• 2018-09-09: dump relay stuck ~17:00Z, so tapped morning of 10th to free.
• 2018-08-06: Enphase AC Battery (~1.1kWh capacity) installed today: brief down time for grid-tied PV generation (and rest of electrics) during fitting.
• 2018-08-05: Loop gas meter sensor seems to have been dislodged by about noon on 2018-07-09, so zero readings between then and now: have updated reading manually.
• 2018-08-05: Loop gas meter sensor seems to have been dislodged by about noon on 2018-07-09, so zero readings between then and now: have updated reading manually.
• 2018-04-22: removed the bubble insulation from the battery bank.
• 2017-10-29: wired in direct-to-battery circuit to kitchen, fused 20A, estimated impedance ~70mΩ.
• 2017-10-20: adjusted VHIGH to also require having reached absorption stage,
• 2017-10-01: adjusted voltage thresholds down a bit since LOW was at ~100% capacity; VLOW is still above sulphation ~12.4V.
• 2017-09-23: put the bubble insulation back on the battery bank because temperatures are dropping.
• 2017-09-18: significantly raised the LOW and VLOW voltage thresholds to try to tackle sulphation and capacity loss; will mean that the networking equipment dumpload is off-grid less, especially at night.
• 2017-08-13: verified that all seems fine in outside connector box; attempted to tighten all connection screws also.
• 2016-10-31: dump relay seems to have stuck again.
• 2016-10-30: ~14:30Z: mounted the 40W amorphous panel in bay b and connected up.
• 2016-10-22: eve: dump load apparently not drawing power from off-grid system when activated. Have reseated 12V connection, but no change; only drawing power when mains switched off at wall which implies relay stuck on, though it can still be heard moving when its state is toggled. Went into service ~2014-08. May need replacing sharpish, probably with another OpenTRV boiler-control relay, though some percussive maintenance seems to have fixed it temporarily.
• 2016-10-04: pm: laid layer of plastic bubble-wrap over battery bank to attempt to reduce cooling rate at night or when otherwise could out; seemed to only drop 1°C overnight where usual has been 2 recently (graph).
• 2016-10-04: am: large tree to SE being trimmed may reduce shading of on- and off- grid systems in morning in shoulder months such as now; may be ~1kWh/d of extra generation on sunny days.
• 2016-09-12: replaced 16WW OpenTRV nodes with secure transport versions.
• 2016-09-10: parallelled +ve connection (and generally improved) from power cupboard to desk to better support heavy recharge demand from laptop, and shortened power route to RPi in hope of reducing measured supply impedance and red power light blinking (but it is still blinking when RPi is working hard).
• 2016-08-29: router seemed to hang today needing power-cycling to fix.
• 2016-08-19: ~10am: replaced main 20A MAXI load fuse with 20A DC breaker in fairly prominent easy-to-reach position.
• 2016-08-16: ~2am: system stopped: must fix flickering red light in case significant power issues involved; replugged microusb power cable.
• 2016-08-12: eve: using PC Meterbus adapter to SS-MPPT-15L to get more accurate battery voltage measure and deduce other parameters; eg resistance from controller to RPi appears to be ~0.1Ω.
• 2016-08-04: am: mounted and connected second CS 255W panel.
• 2016-08-03: pm: removed 40W amorphous and 20W 'floating' panel.
• 2016-08-02: dusk: fitted blocking diode for first CS 255Wp, <0.3Vf.
• 2016-08-01: mounted first CS 255W panel (actually the second in the batch).
• 2016-07-22: installed first CS 255W panel, connected to SS-MPPT-15L; 100W and 20W panels removed from 12V-nominal string.
• 2016-07-16: adjusted powermng.cpp to try to ride through brief dips in sunshine, eg from passing clouds.
• 2016-06-29: installed LS2024RP PWM 20A controller and moved ~40W of 12V-nominal solar to it; battery type set to 2 (gel); in progress ~10am/09:00Z. Noon moved 60Wp panel (~15Vmp) to PWM controller
• 2016-06-11: significantly re-arranged and tidied (and shortened) wiring from panels, now not coming into the house at all, to make separate running of ~17Vmp and ~30Vmp strings to controllers near battery bank possible; currently the two circuits are combined at the controller and the '30Vmp' circuit is still being fed by the ES-62T panel.
• 2016-05-31: wall-mounting the (10kg) ES-62T (ES62) panel; not a great location for mid-winter (will get little direct sun), nor the slickest mounting skill, but makes space for bigger things later and gets me some practice with the mounting brackets. Maybe I could add a mirror or white panel later.
• 2016-05-22: in the process of messing things around I seem to have bust the RPi B+ (it will no longer power up) so I am trying its SD card in the RPi2 B that I have to hand, and miraculously it just works!
• 2016-05-22: tidied exterior wiring; removing LiFePO4 from system (near full SoC); joined 2 external PV strings to feed LA solar controller.
• 2016-05-21: reinstated off-grid 12V MR16 4.5W desk light, RPi controlled; seem to have broken INA219 power monitoring somehow (I2C device no longer visible at 0x40), possibly by still having power across it with RPi powered off, but probably by physically breaking some weak connections.
• 2016-05-07: 10:44Z: reconnected old 40Ah battery; observing.
• 2016-05-06: trying to work out why battery bank not getting to absorption and (and having re-adjusted levels in software) then checking forward voltage on blocking diodes at wiring centre indicated that 100W panel not supplying much power; turned out to be defective MC3/4 connector, temporarily replaced by terminal block.
• 2016-05-05: battery (2A MAXI) fuse seemed to be dropping up to 150mV eg under decent charge from bright sun; replaced with 20a breaker at ~20mV.
• 2016-05-03: connection to one end of battery bank not tight; system seems to have been supported by old 40Ah gel still attached, with apparently ~14Ah remaining effective/available capacity judging by performance with dump load.
• 2016-05-02: measuring 12V power draw by Maplin N03GC adapter with inline power meter gives ~12W for modem/FTTC downstream (13.3W shown, but no draw shown as 1.2W/100mA) and a total of ~15W/1.3A (16.6W shown; mains draw indicated as 13.6W with N67HH) with Loop power monitor nominal 2.5W USB load added; not great efficiency (~80%) but not terrible; total draw from battery probably 18--20W including RPi, controller self-draw, etc...
• 2016-05-01: major rejig of 'dump' logic for: genuine dumping of excess, supporting maximum spill to grid during peak demand/intensity, and attempting to take load off-grid overnight.
• 2016-04-30: (18:00h UTC) fitted (4x) 35mm^2 30cm battery jumpers (M8 lugs) and 2x 10mm^2 (M8 lugs) battery leads to charge controller 1@~2m and 1@~1.3m+; now looks like decent implementation of Method 2; preliminary experiments suggest that the instantaneous sag at the load is now ~150mV at ~10W, but these numbers, while the B1 supply was disconnected, look like offsets from ground currents (if not leakage from the supply wire-OR):

  2016/04/30T17:30:06Z AL 170 B1 12666 B2 14336 P 1858 BV 13856 ST OK -
  2016/04/30T17:40:07Z AL 170 B1 234 B2 14251 P 1887 BV 13892 ST VL -
  2016/04/30T17:50:06Z AL 170 B1 150 B2 14336 P 1876 BV 13856 ST VL -
  2016/04/30T18:00:07Z AL 170 B1 131 B2 14345 P 1943 BV 13856 ST VL -
  2016/04/30T18:10:07Z AL 170 B1 140 B2 14036 P 1881 BV 13668 ST VL -
  2016/04/30T18:20:06Z AL 170 B1 12506 B2 13820 P 2117 BV 13324 ST OK D
  

• 2016-04-24: sample voltages in overcast afternoon with dump in progress: 12.56V near load at measurement point, 12.62V at controller battery terminals, 12.66V at one of middle batteries in pack, so 100mV losses from battery to (near) load.
• 2016-04-23: fitted (2x) 35mm^2 30cm battery jumpers (M8 lugs) in middle of 4x12V bank to avoid causing imbalance vs existing 2.5mm^2 17cm cables; found possible small drop at -ve power-take-off corner of bank where existing cable slightly damaged.
• 2016-04-18: considering addition of 60W--100W+ of semi-flexible panel such as Biard 60W Semi-Flexible to be vertically mounted on south-facing wall to help create some excess to 'dump' to move load off-grid more of the year.
• 2016-04-10: continuing to tweak powermng.c to cycle the (LA) battery less and to be more finely responsive to grid intensity as practice for larger loads and stores...
• 2016-04-06: checked wiring from battery to controller to load; slump of ~0.2-0.3V at 1A/12W seems excessive:

  2016/04/19T04:30:06Z AL 168 B1 12525 B2 13247 P 1847 BV 12880 ST L -
  2016/04/19T04:40:06Z AL 168 B1 12525 B2 13247 P 1858 BV 12880 ST L -
  2016/04/19T04:50:07Z AL 168 B1 12525 B2 13247 P 1858 BV 12880 ST L -
  2016/04/19T05:00:06Z AL 168 B1 12525 B2 13247 P 1876 BV 12880 ST OK D
  2016/04/19T05:10:06Z AL 168 B1 12290 B2 13229 P 2159 BV 12848 ST OK D
  2016/04/19T05:20:07Z AL 168 B1 12272 B2 13229 P 2102 BV 12860 ST OK D
  2016/04/19T05:30:06Z AL 168 B1 12225 B2 13229 P 2079 BV 12860 ST OK D
  2016/04/19T05:40:06Z AL 168 B1 12253 B2 13229 P 2184 BV 12848 ST OK D
  

  Seems to be ~60mV in cables from controller to load (~4m one way, 4mm^2 cabling: 4mm2 is 9.22mOhm/m (supply and return) so 4m should be 36.88mOhms or 37mV drop at ~1A), ~20mV through controller, ~40mV across (each?) 20A MAXI fuse (either side of controller), with some possible extra losses on -ve battery wiring side to fix and generally beef up. (20A MAXI fuse resistance should be ~3.1mΩ 3mV drop at ~1A.)
• 2016-03-31: now have off-grid system take FTTC/router load also when grid intensity is high, and here it is responding to morning peak.
• 2015-09-19: running the MacBook Air gently (1 core at 70%, so no fan needed) for SETI@Home uses ~15W, and ~40Wh per unit (~200TFLOPS), when running as a manual dump load on the Li battery and its PV. All cores flat out hits ~26W.
• 2015-08-20: moved FTTC adapter (~4W) to same 12V supply as Internet router (~8W) thus bringing total load that can be switched off-grid to ~12W ie ~0.3kWh/day (2015-09-18: FTTC adapter and Internet router load measured at mains via 12V supply to be ~10W with N67HH; rising to ~16W with Fairphone (FP1) recharging via USB.)
• 2015-07-26: upgraded (REV2) OpenTRV stats hub to use interrupt-driven OTRFM23BLink code which should improve reception reliability.
• 2015-06-07: Internet router (8W) kept off-grid completely overnight (saving 0.2kWh grid draw); should now extend potential battery support to additional 4W FTTC device which would also give resilience in a power cut.
• 2015-05-04: MacBookAir running directly from Li battery via Mikegyver.com custom adapter lead works well and can run indefinitely off-grid given current levels of sunshine.
• 2014-12-06: adjusted solar panel layout in garden to better capture winter sun angles. Tweaked Li powermng voltage thresholds too very slightly.
• 2014-11-14: adjusted LA powermng voltage thresholds to allow discharge to ~70% for dumping given that getting over 80%SoC is relatively inefficient and requires lots of full sun which may not happen often in winter.
• 2014-10-13: moved ADSL entry point and this WiFi router and entire server downstairs to kitchen cupboard; note gaps and glitches in various data streams ~5pm.
• 2014-10-11: adjusted dump algorithm to try to keep load off grid in winter weekday evening peak ~15:30 to 19:30.
• 2014-08-09: ADSL power dump hardware reconstructed using modified (NC rather than NO) OpenTRV boiler relay control; draws ~250mW to energise and LA (B1) voltage drops ~0.1V at the sense point for assumed total ~8W load including relay overhead.
• 2014-08-07: see how spiky some days' PV generation can be compared to others.
• 2014-07-21: (9:35am) Sunny Beam continuously plugged in as an experiment; expected ~150mW extra load (partly because display is permanently on.
• 2014-07-20: RPi +B ~1.5W typical power draw measured from upstream of 12V-to-5V converter.
• 2014-07-16: SheevaPlug tentatively off 21:33BST, with RPi providing services.
• 2014-07-13: official battery voltage monitoring now on RPi and live power monitoring has been added with high-side AdaFruit INA219 I2C board.
• 2014-07-11: moving services from SheevaPlug to RPi; removed ADSL dump load for now.
• 2014-06-08: increased the basic battery/power sampling interval from 2m to 10m, with intermediate (5m) samples taken and logged when extra energy likely available (eg not overnight or mid-winter).
• 2014-05-04: moved outside (OpenTRV) sensor from inside shed where it was getting rather warm, to porch out of direct sunlight (~3pm); may be ~2°C higher than EGLL temperatures. Bruno's screw slots worked just as designed! Updated s/w on all nodes 8pm--9pm which caused interesting temperature wobbles.
• 2014-04-27: ADSL auto-off-grid hardware finally wired up today after keeping an eye on s/w control behaviour for a while.
• 2014-04-26: all OpenTRV sensors now collecting data, including new one outside under cover.
• 2014-04-10: large (weeping willow) tree more or less due east of us ~30m taken down today; may get less shadowing and thus more generation in the early morning through the summer.
• 2014-03-16: ADSL off-grid given SLA bank showing sign of decent charge (including float); considering auto-changeover between mains and off-grid using latching relay (eg RS 686-7212 / Omron G6EK134PSTUS5DC) on mains side and possibly DC side also controlled by digital outputs under main system s/w control or on basis of (say) 13V/14V thresholds and PICAXE/AVR control. Manually switching on/off grid as necessary for now.
• 2014-03-16: note that 60Wp extra 'winter' panel set not deployed this winter.
• 2013-07-07: removed some wire-OR diodes where panels already have built-in blocking diode and leakage/failure of wiring/diode/panel seems low-risk; may be good for a few % more output when feeding MPPT controller, eg as much as 0.7V/14V.
• 2013-07-07: ADSL back off-grid (for a couple of weeks) after a couple of good sunny days and more predicted (will need taking off for poor days and/or SSR-driven mechanism eg to drop mains input so reg uses 12V when not at 14V for a day).
• 2013-07-01: SLA seems to be struggling with ASDL + SheevaPlug load, so ADSL back to mains for time being.
• 2013-06-25: reconnected dump load for LiFePO4 battery and un-parked MotorWind turbine upstream of LiFePO4 battery. Also transferred SheevaPlug and ADSL modem off-grid (~0.3kWh/d).
• 2013-06-24: reconnected LiFePO4 battery; having to assume that C- is (-ve) charge and D- is (-ve) discharge. Secondary voltage sense line and power to desk also reconnected.
• 2013-06-21: primary 12V panel set reconnected to charge controller and thus batteries (and primary voltage sense line); battery state LED moved from amber to green shortly thereafter. Rest of system will be reconnected over next few days.
• 2013-06-16: 12V system still not restored after building works, and MPPT charged showing red for connected old 40Ah gel, so made direct connection from ~24Wp on shed to charger and reconnected main gel set to charger also.
• 2013-04-29: temporarily taking SheevaPlug and ADSL on-grid and disconnecting panels/controllers/etc, for rewiring.
• 2013-04-21: put away nominal-60Wp winter off-grid PV panel set.
• 2013-04-11: ADSL modem/router taken off-grid this evening; ~0.25V drop (non-drooping) in overnight LA battery voltage, presumably from cable losses.
• 2013-01-26: noticed that solar controller amber LED was lit mid-morning with SLA voltage ~12.2V; sunshine forecast for a good chunk of today and tomorrow and voltage already showing ~12.7V on charge. Server has stayed off-grid so far this winter.
• 2013-01-01: lowered LiFePO4 upper threshold to 13.1V to allow for more measurement noise and the fact that it is 'sacrificial'.
• 2012-11-30: winter ~60Wp amorphous panel set put out after run of dark weather and SLA battery consistently under 13V overnight.
• 2012-10-05: ADSL modem moved on-grid until spring on the basis of SLA @ 12.3V at 9am and only intermittent sunshine forecast over next week or so. (Immediate jump to 12.6V after!)
• 2012-09-14: finally installed Maplin [defunct link "http://www.maplin.co.uk/12v-40w-solar-panel-kit-510795"] N27JL 40W 12V amorphous panel in afternoon via own 5A Si diode; may want to shorten/fatten connection cables at some point to minimise losses.
• 2012-09-05: given another day with the SLA voltage only reaching ~13.2V (@<25°C) I have experimentally removed the ADSL modem load (back on grid instead) to see what difference that makes beyond instant 0.2V rise. Tightened all connections for external and internal screw-blocks (nothing was worryingly loose this time). Note that as of today only 60Wp UniSolar panel and MotorWind are charging the LiPO4 battery with everything else directed to the SLA. (Just after noon 2 days later SLA voltage at ~14.4V ie getting good charge.) Taking back off grid morning of 2012-09-09 at least briefly as SLA battery has been at 13.2V all night. I am currently of the opinion that the ADSL's power management had stopped working properly and power cycling it is the main fix, but giving the SLA a good charge was fine anyway.
• 2012-09-04: some evidence of poor connections from 100Wp panel (poor charging even in good sun, silly/low voltages at connection box; seemed better after breaking and remaking in-line connections, but will need monitoring.
• 2012-08-05: removed dead 12Wp panel (nearest house, likely to capture least energy anyway) from front porch / shed roof 36Wp set currently running to LiFePO4 as spotted 2012-02-21. Attempted to clean two remaining panels and also very slightly increase tilt of one now nearest the house.
• 2012-07-25: adjusted SLA 'high and charging' down further to 13.2V (from 13.3V) to allow for temperature/voltage compensation at ~30°C ambient.
• 2012-07-08: adjusted SLA 'high and charging' down further to 13.3V (from 13.4V) to allow for temperature/voltage compensation.
• 2012-07-07: bought Maplin [defunct link "http://www.maplin.co.uk/12v-40w-solar-panel-kit-510795"] N27JL 40W 12V amorphous panel £80 (ie £2/Wp) on offer to install vertically on south-facing wall in September.
• 2012-06-17: retightened all connections, topped-up stand-alone 3.2Ah SLA direct with 1.5W panel in sunshine to ~13.1V, adjusted SLA 'high and charging' to 13.4V (from 13.5V) as previously slightly too conservative I think.
• 2012-05-13: put away 'summer' 60Wp amorphous panel set connected to SLA for the summer.
• 2012-03-15: ADSL modem moved off-grid until autumn; SLA voltage now seen dipping below LiFePO4 ~0.2V at night.
• 2012-02-21: systematically tightened all terminal-block connections because observed deep voltage dips/sags suggested poor connections and high-resistance somewhere. Found one very loose (SheevaPlug ground) and plug rebooted at some point suggesting at least one very bad connection that did indeed need fixing. Noted also that one 12W panel at front seems to be dead.
• 2011-12-11: measurements at noon show voltages at wiring centre 13.05V, but 12.65V at sensor wire terminals upstairs and reading via k8055 as 12.6V/12.7V, which suggests higher sensor input impedances may be needed for accuracy (and battery drain not as dire as advertised), so adjusting pots ~12:50 to get true reading as at voltage probes downstairs, not locally, so about 2% high locally given 1kΩ probe resistance and 50kΩ input pots in series. At 15:30 reading 12.6V/12.7V on k8055 vs 13.0xV downstairs possibly due to drop in 0V side, so have tweaked a little further to compensate. As of 2011-12-14 just after sunny noon reading of ~14.9V (LA) and ~13.6V (Li) corresponded to ~14.5V and ~13.3V at the wiring centre/probe.
• 2011-12-05: used sensor wiring already in place to measure battery voltages closer to source (yellow for Li, red for LA) via 1kΩ resistors tapping wiring centre downstairs, to suffer less from the voltage drop to SheevaPlug at higher supply currents. May require voltage threshold adjustments.
• 2011-11-06: added 'extra' nominal 60Wp south-facing ground-mounted panels to off-grid SLA-charging circuit as the system has missed coming out of low-power mode a couple of times in the last week or so. One of the panels seems to be effectively dead (~2Voc or less) but the others are still showing a respectable 16--23Voc, so swapped in separate 15W panel of same dimensions from nearby (though its rear glass plate is cracked and so performance is probably below nominal) and replaced that with a new [defunct link: "http://www.maplin.co.uk/solar-powered-12v-15w-battery-charger-508575"] Maplin N70JK 15W with metal stand.
• 2011-09-10: wired old 40Ah gel into main battery bank, direct to MPPT controller via long-ish cable and self-resetting thermal fuse with aim of it being slightly-higher impedance route than to main bank, thus avoiding 40Ah taking too much strain in parallel with much newer ~400Ah set.
• 2011-08-20: ADSL modem back on-grid as SLA voltage getting low.
• 2011-08-13: currently getting about 0.2V drop on 12V line upstairs powering ADSL modem (~9W measured) and 0.1V drop powering server (@~4W).
• 2011-07-25: moved connection of 36Wp amorphous panels (at front) to main charger/battery in part to help keep ADSL off-grid for now and to simplify setup. Older 40Ah gel may optionally be paralleled with new bank, possibly via own fuse. (Old 40Ah gel at 13.23Voc today, 2011-09-10 12:00 GMT ~25°C ~13.2Voc.)
• 2011-07-11: fitted Morningstar Remote Temperature Sensor (RTS) to MPPT controller to help protect the LA battery bank. Fixed the sensor to the -ve lug of a battery in the middle of the bank: have thermally insulated it (with an old sponge as it happens) from the air to ensure that it reflects battery core temperature as accurately as possible.
• 2011-07-04: ADSL off-grid again; 0.20V drop from controller battery terminals to bus upstairs near power take off for modem at 12.78V (bat).
• 2011-04-22: found and fixed wiring problem which had smallest (2Wp) panel reversed and fighting against 60Wp winter panel set, so not contributing to SLA input.
• 2011-04-22: moved 12Wp panel from LiFePO4 to SLA input so former is dumping less and latter is full more, ie better balancing.
• 2011-04-16: wired in dump load (currently only ~16R/15W) for LiFePO4 battery.
• 2011-04-15: adjusted turbine crowbar trip to 18V, ie above max output voltage from LiFePO4 battery and below Voc of the PV inputs, so presumably 'safe' for the BMS input, to try to reduce spurious and potentially damaging repeated operation a little when the battery is full.
• 2011-04-11: took down (60Wp amorphous) 'winter' solar panel set.
• 2011-04-10: experimentally taking ADSL router off-grid using lap-top adapter (previously using mains via same adapter). Initially until 2011-05-12 when battery getting a bit low.
• 2011-01-14: rearranging system to accommodate new SLA bank.
• 2010-11-20--2010-11-21: inserting 30A automotive 'MAXI' fuse into SLA +ve connection to Morningstar controller, then next day 20A in load side from controller, thus SLA side voltage reading glitches to 0V, while LiFePO4 side keeps the SheevaPlug running.
• 2010-10-11: moved MotorWind turbine over-voltage crowbar circuit to outside junction box and plumbed output via blocking diode to secondary/LiFePO4 array circuit, ie any wind power now charges the LiFePO4 battery unless already full.
• 2010-10-11: disconnected wind-power-available sensor (Input3).
• 2010-10-11: split main PV (motley) array into two isolated parts, in particular the lower-voltage Uni-Solar triple-junction (Voc ~ 17V) is now in the section feeding the LiFePO4 battery.
• 2010-10-11: added 100Wp mono panel to main off-grid PV array.
• 2010-10-09: upgraded battery voltage sensing to include second battery and reduce noise, and improve accuracy using external 50kΩ trimpots.
• 2010-10-05: added LiFePO4 12V 20A battery to system fed from a sub-set of the panels and powering the SheevaPlug from a diode-OR with the normal SLA output.
• 2010-10-01: now reducing sampling rate (to no more than one sample each 10 minutes from previous sample each even minute) if battery voltage/SoC very low so as to save energy associated with the sampling, logging to disc, etc...
• 2010-09-30: reconnection of approx 60Wp of south-facing ~70°-tilt panels as battery crept down to 'amber' at controller (<12V) after several successive gloomy days, plus other minor reconfiguration for robustness/winter.
• 2010-04-05: temporary removal of approx 60Wp of south-facing ~70°-tilt panels now battery is reaching float every day and turbine 'emergency brake' is repeatedly operating!
• ~2009-10-15: addition of approx ~72Wp of south-facing ~70°-tilt amorphous panels from Maplin.
• ~2009-09-21--2009-10-03: major reconfig with laptop off battery, SheevaPlug permanently on battery, battery voltage measured via analogue input rather than via two opto digital inputs (LVD/high), and solar controller(s) being upgraded to a single Morningstar MPPT device (SS-MPPT-15L).
• 2008-07-01 (~0900UTC) new laptop/server internal battery fitted.
• 2008-06-26 Input1 NOW UNALLOCATED; no longer 'DC-DC converter output OK'. Current allocation/meaning of inputs:
  • "Battery Voltage High" (Input0) reports essentially when the battery is full and still charging, so the system should dump some energy.
  • "LVD Output On" (Input2) reports when the battery is fairly full and is running the server/laptop.
  • "Wind Power" (Input3) reports when a turbine is turning, though not necessarily generating useful energy.
• 2008-06-24 laptop internal battery died so 2008-06-26 replaced mains/off-grid switchover mechanism with an 'instant' diode-based scheme.
• 2008-06-05 (~1600UTC) rewiring to reduce impedance and improve behaviour under heavy load.
• 2008-06-01 (~1700UTC) addition of 2x12Wp amorphous panels (Maplin N32CX, TopRay TPS-102-12) on east-facing near-flat porch roof, bringing off-grid total to ~120Wp.
• 2008-05-24 (~1200UTC) addition of 12Wp amorphous panel (Maplin N32CX, TopRay TPS-102-12) on east-facing near-flat porch roof (in place of 13Wp test panel).
• 2008-05-24 (~0700UTC) addition of tiny 2.4Wp (Maplin N59FU, Nikkai) amorphous panel on west-facing near-flat shed roof.
• 2008-05-10 (~1430UTC) to 2008-05-24 (~1200UTC): temporary/intermittent addition of 13Wp horizontal PV on east side of house to catch morning sun and extend off-grid period for server.
• 2008-04-17 (~0800UTC) to 2008-04-23 (~1100UTC): major rewiring.
• 2008-04-10 (~2000UTC): disconnection of 2x1W turbines for wiring rearrangement.
• 2007-12-01 (~1600UTC): installation and connection of MotorWind 8-turbine device.