Define a first set of alerts

[saroele]

Let's use this issue to list all possible alerts we can think of. What do you want to be alerted of? Which alerts are useful? Which are possible? At the meeting of 5 dec 2016 we will discuss those, prioritize and create some specific issues for some alerts.

A few to get us started:

• water leak
• the consumption of a sensor over the past week/month is higher than x% of the population (benchmarking, x is sensor-dependent setting)
• the consumption of a sensor over the past week/month increased by x% since the previous period
• last night, your standby consumption was unexpectedly high
• we think you can save x (kWh, €, %, CO2, ...) by defrosting your freezer
• your boiler is takting too much, + advice on what to do
• your thermostat seems to be badly configured
• we thinkyou don't have a water saving shower? You can save x (kWh, €, %, CO2, ...) by installing a shower with flowrate of 8l/min
• turn of x! (outdoor light, basement light, radio, ...)
• the insulation of your house is in the worst x% of the houses with similar age.
• your house is leaky. Some air strips can save you x (kWh, €, %, CO2, ...)
• time to replace your fridge/freezer/boiler/... you can save x (kWh, €, %, CO2, ...)
• congratulations, your consumption is lower than x% of the population! Would you like to get an alert when it raises ==> configure here

[WolfBerwouts]

Hi Roel,

In attachment you can find a list that we created in the beginning of ENCO. You will see that not everything is feasible/relevant but maybe it is usefull.

Greetings Scenario's - Blad1-2.pdf

[Ryton]

Another trivial-to-implement, but urgently needed alert for filtering out invalid data for other analyses (& notifying the user to rectify):

• a sensor does not generate valid/variational data during X hours/days (zero or constant flatline signal).
• a sensor does generate abnormal (invalid) data at time Z (e.g. water use > 100 l/min, electr > 100 kW residentially)

Reason: I regularly (5-6 times per year) loose water or electricity meter readings, and often only discover that a few weeks after...

Other suggestions:

• PV output does not follow solar profile.
• Heating energy use (gas/electricity): Estimation of the slope of the energy signature (kWh / Heating Degree Day) and weather-independent baseload (Estimated consumption at 0 degree days).

I tried the last one with my own data already, but there are quite a lot of outliers, so it might be hard in practice, to generate an unsupervised yet robust estimate.

[saroele]

I love these!! Monitoring the monitoring is quite important :-) And I think many PV owners will like the PV alert. Nice!

[saroele]

@WolfBerwouts zitten ook een paar goed in jouw lijstje. tnx!

[fixersf]

• verbruik gas is groter dan nul tussen 2u en 4u 's nachts bij buitentemperaturen hoger dan 5°C (=nachtregime kan beter)
• kwartierverbruik water is gedurende 24uur niet 0 geweest (=lek) *klimaatgecorrigeerd gasverbruik voorbije maand meer dan 10% hoger dan klimaatgecorrigeerd verbruik zelfde maand vorig jaar (=stijging van het gasverbruik, vermoedelijk door minder goede instelling van kamerthermostaat, of meer thuis)
• elektriciteitsverbruik voorbije maand is 10% minder dan elektriciteitsverbruik zelfde maand vorig jaar (= besparing)

grts, Sam

[fixersf]

• gas consumption is higher than zero between 2 and 4 o'clock at night at outside temperatures higher than 5°C (=night consumption can be lowered)
• 1/4th hour water consumption hasn't been 0 during the last 24 hours (=water leakage)
• climate normalised gas consumption from the past month is 10% higher than the climate normalised consumption from the same month last year
• electricity consumption last month is 10% higher than electricity consumption same month last year

[vi-di]

Quick personal proposal as discussed on Dec 5th Opengrid meeting:

For me general reporting themes are:

• Climat goals
• Benchmarking/comparison
• M&V, project results, efficiëncy
• Anomaly detection.

In general, an analysis gives a result we want to report. Based on the severity of the result, it could give an alarm or just a remark in a periodic report (see bottom)

For me the most important alarms are:

• Waterleak
• PV panels not working properly
• Solar thermals system not working
• Rainwater system not working
• High standby Electrical
• High Peak Power Electrical
• Heating not in summer mode (only DHW of cooking) of holiday if not at home
• Big change in way of using HVAC
• More general alerting based on deviation real vs predicted data. Severity of deviation defines way of notification

Then a second category falls into automated consultancy:

Typical shower flow l/min Typical toilet flush volume Analysis of efficiency of heating installation: modulation, start-stop, correct sizing of power, ‘stooklijn’,... If we could detect Appliances you could analyse consumption of e.g. fridge and monitor evolution of their performance over time % Autoconsumption PV Max capacity of grid connection elec

—> General remark on alerting: way of alert (periodic reporting, sms, alarm,…) could be based on KPI’s coming from analysis result. Severity score could be based on something like Risk analysis:

Methode van Kinney (Fine & Kenneth)

[Pasted Graphic.tiff]

Best regards,

Vincent Dierickx +32 486 34 91 62 @vincentdierickx, @energieID EnergieID cvba-so www.energieID.behttp://www.energieID.be

Op 3 dec. 2016, om 19:44 heeft fixersf notifications@github.com<mailto:notifications@github.com> het volgende geschreven:

• gas consumption is higher than zero between 2 and 4 o'clock at night at outside temperatures higher than 5°C (=night consumption can be lowered) • 1/4th hour water consumption hasn't been 0 during the last 24 hours (=water leakage) • climate normalised gas consumption from the past month is 10% higher than the climate normalised consumption from the same month last year • electricity consumption last month is 10% higher than electricity consumption same month last year — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[saroele]

Your note can double roughly as a meeting report.

From the meeting, I learned that we have to focus on making crystal clear analyses, and many of them will return a KPI (like a benchmark or a deviation of an expected value).
Then, for each deviation or KPI, we fix a severity score, cfr risk analysis. Based on this score and the deviation, an alert level can be obtained (eg. debug - info - warning - error - critical).

Finally, the user can specify his desired alert level for each analysis. When the analysis returns a severity score higher than the setting of the user, the alert is sent. Else, the result is reported in the scheduled report.

A nice way of reporting is by creating some kind of score card, like this example: https://connectedsavings.com/products/energy-efficiency/ (see the scorecard for smart meter interval).