Reference

The conversational smart-meter

Automatic meter readings tie consumption to a time-stamp in a data base.

Since logged readings answer - " How much energy was used at time X ? ", 

this really cannot be considered smart (even if it does involve a pretty chart).

So what is considered smart ?

Useful services can "wrap around" energy data to begin to appear "smart". So here is a scale of increasing "smartness"

Services may;
provide a stream of interpretation or a display
-  but this can easily be listened to or ignored so call this "smart",

slightly better they may;
react conditionally to thresholds
- less likely to be ignored, but not very good without user context
 so call this "smarter",
better yet they might allow;
a user to define their context and then answer appropriately to that context
so call this "smartest", This starts to addresses the audience question of my previous post
Note : way beyond "smartest" there is also kWIQly  - but that isn't the point of this post!

An example to illustrate might be helpful (the first level is rather obvious but the others may not be):

Illustrations; Fairly dumb smart, smarter, smartest:

1) smart - a stream of data displayed...




Have you ever noticed when a clock stops ticking ?

Humans have highly developed skills to "tune-out" noise, even from their own children. 


Is this a stream of fascinating insights ? 

or perhaps it is a gripping read ?
We know anything "always on" but not threatening or vital - is soon ignored
(isn't exactly that the cause of much energy waste)


For example - do you have anything held to your fridge with a magnet ? 

If so think back to the last time you noticed what it says ?


Easy to ignore ?

"To be effective a smart-meter needs to be more compelling than a screaming kid"

Yes, perhaps that is "a big ask" but it is also a reasonable objective - so let's look again to see if this passes...

This just "screams" energy waste doesn't it !

2) smarter - interactively addressing specific questions

Now imagine a smart-meter where you can set thresholds and above the threshold an alert is sent (to a designated listener !). 

There are several SaaS companies that let you do this. However, if your use case is even remotely complex (you use more gas in winter for heating), it needs to be maintained (whether by you or the service) or it becomes worse than useless - You have configured something designed to interrupt - but it interrupts at the wrong time! 

There are also some simple ways around this (eg tracking recent mean deviations and volatility) but unless you have a good underlying model (understanding what constitutes a valid alert you will end up with a version of the boy that cried wolf - as in Aesop's fable

English: The illustration by Francis Barlow of...

English: The illustration by Francis Barlow of the fable "The Boy who Cried Wolf", 
called by him DE PASTORIS PUERO ET AGRICOLIS, 1687 (Photo credit: Wikipedia)


This is obviously a way of inspiring great music - but not so good for a smart meter.


By the way - why not do yourself a favour - 

watch this video with your children this evening 

(bookmark it if you want)

 Warning: frightening for the very young
They will love you all the more (your time inspires their growth) - it delivers a simple cool morality message - they will learn of marvellously inspiring classical music that really talks to the story and 
you can enjoy the superb animation !
Something for everyone, - as they say in US of A land : "What's not to like ? !"



    By the way if you love it - please share it !

As I was saying before I so rudely interrupted myself...

"- interrupting yourself is not always a good idea from a smart-meter"

So we move on to what we can consider the smartest - 

3) Smartest - "a user can define their context and the "smart-meter" can answer questions appropriately to that context"

When looking at smart functionality it is important to isolate different aspects of "smart".  For example the interface presented to the user may be elegant, snappy and easy to use, or ugly, slow and a nightmare (I won't show any examples here).

It might be equally be text based, numeric, iconic or graphic, but presentation issues do not determine the "smartness" of the capability. (Though the user experience or "UX" as it is known is unquestionably an extremely aspect of any product or service that aspires to frequent use).

So we do not "care" (for the purpose of this conversation) how information is represented or conveyed, we care about the underlying "intelligence". Ie what work the smart-meter can do that is "smart".  This is like saying when we choose a doctor we are more concerned by his or her training and experience than their social standing, looks, golf handicap or bedside manner (even though these may help doctors market themselves).

The most famous determination of the quality of artificial intelligence is the Turing Test from wikipedia

"The Turing test is a test of a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. In the original illustrative example, a human judge engages in natural language conversations with a human and a machine designed to generate performance indistinguishable from that of a human being."

So we might consider the "conversations" that a smart-meter could have with the user, and judge whether these convincing or not. If they are - it is smart.  

Note: kWIQly while can deliver one side of this conversation - as a pre-requisite - solutions are built on top. We also do not claim to have invented any intelligence here - merely an artificial simulation in a limited area that is constrained to asking and answering very stylized questions

I want you to imagine yourself as "user" in the following dialogue.  This starts to represent what is now possible for the smartest of smart meters.  The user represented (that is you) is assumed to be pretty well versed in energy management - it would be possible to repeat this for a much more lay-user ("who only just gets a thermostat")

User: "Please note that the following days are school holidays:"
List of holidays goes here (and is applied to all the meters in the school district)

Smart-Meter(s): "Got it"

User: "During weekends the school is closed, during holidays unless flagged the school is closed, during school-days occupancy is from 08:00 to 16:45"

Smart-Meter(s): "Got it"

User: "Gas use relates to cooking, heating and domestic hot water, comfort heating is required during occupancy and to protect fabric in school holidays under extreme weather conditions, but not at weekends"

Smart-Meter(s): "Got it" - 

Note this was all setting context and was like data entry - now it gets interesting

User: "Which school has the most individual days of holiday where consumption is more consistent with school days than holidays?"

Smart-Meter from one school : "Check this out;


User: "OK - So on days when  temperatures are in the range seen last week - 

at what time of day is most energy wasted (please band the answer by level of waste) across all schools ?"

Smart-Meters ; "You mean something like this;"

"On the most wasteful days most energy is wasted before opening"

User: "OK - I see that if the boiler capacity was lower less waste would even be possible - so how big does the heating capacity have to be for the fist school you picked out school to ensure warmth in temperatures as low as X:

Smart-meter: "Easy - that would be around 45 kW, but would require a working optimum start algorithm to bring heating on earlier on really cold days"

Closing note from kWIQly ; The above transactions are possible, but to make it readable we did style the conversation to make it "human-like"- So our client implementations typically support standardized enquiries which are triggered by button clicks etc. rather than a typed interface

Enhanced by Zemanta