独家优惠奖金 100% 高达 1 BTC + 180 免费旋转

Setting Context For Models

A rather obvious concept that underpins kWIQly is that performance can change for a variety of reasons.

The long-term history of a meters’ consumption may no longer represent what is achievable today.

To be most helpful kWIQly must make the right decisions on how to process data. Sometimes this requires user assistance — because there are things kWIQly cannot know or even guess.

However, when advised correctly, kWIQly makes light of otherwise very tedious work.

In the following sections we will look at four common cases, and how to handle them:

Let’s consider an issue arising out of a phenomenal savings initiative and how best to handle it (using a real world case — anonymous — from a major food wholesaler).

Our start position is that kWIQly sees change but cannot explain it.

Under each of ‘off-track’, ‘key-intervals’ and ‘diagnostic’ views the change is made very obvious. And, anticipating that a significant change may have happened, kWIQly starts a savings analysis. But all is not as it should be…

Sometimes changes are great news, below we see savings tabulated for a contrast between the two periods (in this case the impact of a successful store refit) and it shows a massive 40.1% electrical savings after seasonal adjustment.

kWIQly is assuming that the average of these two periods is what is ‘normally achievable’. In effect it is giving a permanent ‘atta-boy’ for something that changed in the past — potentially long ago.

We would want the calculation and credit for the change to be available. But if load now rises by say 5% it represents a new problem and not a reduction in the benefit from the change.

kWIQly can handle this easily, but makes some observations. These can be seen in the ‘off-track’ view once the meter is selected where kWIQly figures out a few facts, which at first seem a little complicated, but are easily explained:

Statistics for the current meter:

Looking at the ‘key intervals’ view and tabbing to ‘Σ Deviation’ which representing cusum error relative to a zero base-line we see a history of major change with one very dominant change:

As kWIQly notes this occurred on 2016–07–1

In this case kWIQly suggests

This tells us that the trend after the change above is entrenched. Normally if you have a good reliable target — and random changes you would expect the outcome to be above or below like a fair coin toss.

The average run of continual heads or tails is 2 = 1+0.5 + 0.25 + 0.125 …

In effect kWIQly is implying the current model has bias because it sees average ‘runs’ of 89 days.

Common sense tells us that if we are trying to spot waste events we want to consider a ‘new-normal’. which takes place after the implementation of change.

Just under these statistics — we see how we can achieve this:

If we select a range (or single day) that encompasses the implementation period , often a re-furbishment project, we achieve several objectives.

Sometimes consumption increases or decreases relative to a current context temporarily. Simple examples might include a seasonal sales period or holiday, or due to a plant failure or control mishap. These changes although representing deviations from normal, should be rectified, or may fix themselves.

If the same adjustment technique as above were used in error (representing permanent change). The correct history would be ignored and a savings or impact calculation made.

In this case it is appropriate to flag these days (or periods) as abnormal or deviant as shown below

Doing this will have three effects:

If kWIQly is not being fed data regularly — it gets hungry and complains ! (Actually it creates a record in a non-comms. list)

This list is maintained automatically — so when problems for a particular meter are fixed the list is cleared down. Some older AMR meters are reliant on periodic (eg monthly) communications. These should not be counted as failed.

Others have occasional communication problems eg when a lorry is parked between the transmitted and the radio tower it is trying to talk to.

To be implemented: It will be possible to specify at a portfolio level (or by individual meter) a threshold for this diagnosis. Obviously long term data loss does not help energy management.

In another case (particularly for Gas Meters — which are harder to monitor) some suppliers have problems getting scaling right, or the meter can ‘flat-line’ ie incorrectly report zero consumption.

Obviously any such data will be highlighted by kWIQly as problematic. However, it can be ignored by selecting the correct action, it can either be excluded as above — or if it is a past problem the Initialise change clears all data until the end of the Date Range. This will not result in savings calculations as it would be inappropriate. It will however set a ‘new-normal’ context (that is not corrupted by the old inaccurate data). This will be reflected in KPIs and at a portfolio level after nightly update

If a meter has been disposed of, disconnected ,capped off, or sold but your AMR provider has not yet advised kWIQly , Disposal is the answer. This archives all data (unless deletion is requested) and makes corrective changes to all statistics (including summary histories), to ensure period / period comparisons are ‘like-with-like’.

Seasonal changes are normal —Waste events may arise from many meters when there is a change of weather and an over or under reaction by users or control systems! These are genuine and should be corrected by action in the field. The data should not be changed (example with explanation below)

Over-reactions compromise comfort and/or waste energy

kWIQly draws a model that best captures what is occurring, but as seen below during shoulder periods approximately 5–15 °C:

Add a comment

What is your view?

Send

Related posts:

Do Black Men Even Like Black Women?

I have heard several black men complain bitterly against Black women for all sorts of things. It seemed like they blamed all of their problems on them. Then I remembered. I snapped my fingers as the…

Yes! Anger is my Strength

I have heard many of us saying that they are worried of their anger. A number of people visit anger management sessions, read blogs, etc. They are right on their front because they have ever felt…

Buat Slot Deposit 5000 Yang Akan Dibanggakan Kerabat Anda

Manfaat bermain di Slot Deposit 5000 dibandingkan kasino darat benar-benar mulai menumpuk ketika Anda melihatnya dengan cermat. Selain keuntungan bermain di rumah atau dalam perjalanan, ada banyak…