Traceability Notes (V1) ================== DHD20170215 DOCUMENT HISTORY CREATED: "DRAFT" Damon Hart-Davis 20170215 REVIEWED: Mark Hill 20170221 ISSUED "V1": update 20170221 after review MH20170221 16WW Loop Meter --------------- See 20160819LoopErrorSample.txt for one sample of Loop meter tracking the (calibrated) gas meter and indicating and error of 0.07% (4th SF affected). New data point (reports vs actual) for the same Loop meter: As of 2017 02 15 T 10 13 Z: * Actual reading 2769.39 (m^3) * Loop server data reports 2769.27 (m^3) thus further slip (increased tracking error) of 10Wh (0.01kWh) in 5 months. A total error of 0.12kWh over more than 12 months and ~405kWh (0.03%). Tracking seems well within acceptable error bounds over typical time scales and consumption volumes for this household. [DHD20170407 16:44 BST: further data point: actual 2282.12, Loop 2822.01.] Second Loop Meter ----------------- As a different sanity check over a shorter period in a trial house (code 10), using paper installation records and the Loop data dump: * At 2016-11-21T16:30Z the Loop interface/site showed 41957.96 m^3 (internally the Loop data shows 9374.65kWh since start). (Reading taken from the Loop Web site while selecting the home for a visit some days prior to the visit.) * At 2016-11-25T11:00Z (approx) the physical gas meter read 41994.32 m^3 (internally the Loop data shows 9743.29kWh since start). (Reading taken on the premises during the visit.) Nominally the m^3 readings indicate (at 11.1kWh/m^3) use of 404kWh in that interval, whereas the Loop internal data suggests 369kWh consumption, ie 35kWh or ~10% difference, though it should be noted that neither time is required to be exact for the installation records. However, note that the physical meter reading is entered at the Loop Web site as the Loop device is installed, with '0' accumulated kWh. Thus the tracking error is better computed as 35/9743 or ~0.35%, noting again that the timestamps are not exact and the gas consumption rate is quite high on these cold days (1C to -3C outside). Additional Verifications Possible --------------------------------- There are sufficient current records available to verify a large fraction (~25%) of the Loop meters' tracking accuracy as above, historically and of those in current use. All the Loop meter failures seen to date have been gross, ie no further tracking, eg because of communication failures in the Loop chain, or in sime cases evidently the Loop meter's optical sensor has become detached or similar. See 'Further Note' below. In general, the simplest way to verify total tracking error is to stand in front of the physical meter periodically (or at the end of the trial) during a site visit and note the physical meter reading and the value that Loop reports at that moment on its Web site. Neither of the worked examples here is cause for concern. Additional Issues to be Explored -------------------------------- 1) Allowed tolerance in the calibrated energy supply meter and sensitivity of test results to that. This would affect the slope but not the efficacy measurement (ratio of the slopes with and without energy savings enabled) providing that the supply meter error/tolerance does not drift significantly over the course of the study. 2) Sensitivity of test results to temperature differences between the weather station measurement point and the households in the trial. If the difference (or the major component of it) were steady in one direction then this would have the same effect as an offset in the base temperature, whose sensitivity is discussed elsewhere. A more complex issue is where the microclimate around the test homes differs significantly from that at the weather station, including such effects as different wind speeds where unplanned ventilation is a significant extra heat loss factor for example. This does not seem to be an impediment to broad use of heating degree days currently, and so is probably not a critical issue, but there is the scope (for example) to fit simple IoT-based instrumentation for temperature, wind and insolation at some test homes to quantify the effects. Further Note for Improvement ---------------------------- It would probably be useful to add an explicit filter to the computations to discard data where heating fuel demand is 0/flat regardless of apparent calls for heat as this may indicate gross failure of the meter reading device.