Analyzing Electricity Theft with SAS EG

[KidCat]

Hi,

I would like to know whether any of you did an analysis of electricity theft with SAS EG. I am thankful for any hint

(users, papers, presentations).

KidCat

[TomKari]

In most analysis projects, the most time-consuming part is obtaining and preparing your data. Two questions:

1. What data are you trying to use for this analysis?

2. Do you have a clear idea of what your analysis result will look like? (For example, sketch out a graph).

That would give us something to work with.

Tom

[KidCat]

At the moment I have no specifications or restrictions. It's about clarifying who has already done something like this with SAS. a bit like collecting ideas.

[Kurt_Bremser]

It is a special case of fraud detection.

But you need to have data where you have lots of customers (the more, the better) with their contractual and other data, and a flag that states if theft was done or not. This can then be used to correlate conditions with the outcome, e.g. by multivariate regression.

The details of the process depend on the data available.

[SASKiwi]

What's the context of your question? Are you trying to analyse known electricity theft or do you want to predict electricity theft based on suspicious customer usage behaviour?

[KidCat]

Firstly, detect electricity theft. Then maybe analyze it in more detail..

[SASKiwi]

@KidCat  - That means being able to predict it. To do that you have to have examples in your data of genuine identified theft. Without that is it impossible to build a predictive model because you won't know what behaviours are different for theft versus not theft.  

[Kurt_Bremser]

I thought of some things that could be done to find suspect customers:

• sudden change in consumption up. Could mean someone tapped into that somebody's line
• sudden change down. Could mean somebody bridged the meter, or tapped into someone else's line
• If you have hierarchical data that includes metering on a district level; if the measured consumption is significantly higher than the sum of the individual measurements, someone's cheating. Could also be used to find defects