With a… GPT chat about electricity theft, i.e. by using artificial intelligence, DEDDIE is trying to limit the problem of electricity theft from the grid, the cost of which is borne by other consumers, who pay the value of the stolen energy but also the Administrator himself.

DEDDIE recently organized an innovation competition aimed at developing machine learning algorithmic models for the detection of rheumatism, which gathered an impressive participation from 267 teams, (companies and individuals).

Last week the winners were announced: they are the AETHON Engineering, Silentech and Dataminds teams who shared prizes with a total value of 30,000 euros.

The contestants attempted to discover the services where electricity theft is (or has a high probability of) being done, by delving into a large amount of data provided by the Operator to the contestants for domestic and business uses, regarding:

1. Supply Details, such as their electrical power, their geographical location and their use.

2. Details of the supplier

3. Requests submitted to DEDDIE for change of supplier, interruption of electricity supply due to debt, etc. and the dates of submission of the requests.

4. Consumption level and dates of receipt of indications.

5. Data for identifying electricity theft and date when the finding was made

As the Director of the Directorate of Research & Innovation of DEDDIE, Giorgos Loukos, explains to APE, the request was the development of an algorithmic model which will identify the services in which there is a greater possibility of electricity theft.

According to the results, the models developed by the competition winners had over 70% success ratesthat is, the consumers who were indicated as potential power thieves had actually, in this percentage, been “arrested” by the DEDDIE services for power theft.

An evaluation element for the selection of the best proposals was also the identification of services in which power theft does not occur.

At a later stage, the models that had the best performance will be tested on larger population samples to confirm their reliability. The final goal of DEDDIE is to identify, with the help of artificial intelligence, the facilities in which there is a greater possibility of electricity theft so that, accordingly, more targeted and effective controls can be carried out.

According to the data of the Regulatory Authority, electricity theft in recent years corresponds to 4.5% of annual energy consumption, while it was 1-2% before the financial crisis. In addition, with the energy crisis, the cost of energy that is “lost” has skyrocketed. The Authority has set specific targets for reducing rheumatic thefts with the aim of eliminating them by 2031.

As Mr. Loukos points out, the HEDNO Datathon competition was an action of the first Greek sandbox introduced by DEDDIE in the energy sector (that is, a limited, controlled environment in which tests of an innovative application can be carried out) which is, as he emphasizes, an important tool to attract and accelerate innovation.

The next field in which innovative solutions will be attempted on the part of the Administrator, is the creation of a digital model of the distribution network in which they can e.g. to “run” supply and demand scenarios to find solutions to potential problems. For example, how can cuts in renewable energy production be minimized during periods of low demand, a problem that grids are increasingly facing as “green” generation increases.