Critical infrastructure such as electricity, water, gas and heat networks are central to a country’s social and economic wellbeing and their disruption or destabilization can have a drastic impact on the population. One of the biggest challenges for utilities is dealing with fraudulent behavior. It is estimated that 30-50% of the energy produced worldwide is lost through non-technical losses such as electricity theft and unmetered or unbilled consumption. This corresponds to a financial loss of 80 to 100 billion US$ per year. This money is not available to carry out regular maintenance, finance investments in new systems and services and drive forward necessary innovations. In addition, electricity theft can cause considerable damage to infrastructure, for example through short circuits and fires.
The aim of the joint project is to develop the FraudDetect data analysis software, which can be used to protect critical infrastructures from electricity theft and other forms of fraud. FraudDetect uses consumption data from customers for the analysis. This is provided by the ZONOS smart meter software already developed by Cuculus. The data is then analyzed using the latest methods of explainable AI and causal linkage analysis. As a result, FraudDetect provides operators of critical infrastructures with key information for combating fraud. On the one hand, this is a global estimate of the extent of fraud in the company and, on the other, a detailed list of potentially fraudulent customers. For energy suppliers, the use of FraudDetect is therefore a major financial incentive. In addition, its use leads to a significant increase in grid stability. This fosters economic growth, promotes social well-being and leads to more resource-efficient operation of critical infrastructures and a more sustainable energy supply.
In the joint project, Cuculus is working together with TU Ilmenau and AiVader Gmbh. Among other things, Cuculus GmbH is developing a data generator for energy and water networks based on simulated consumption data for various fraud scenarios and is responsible for the overall coordination of the project. In the project, the partners are investigating three different analysis methods, with the focus of the specialist area being on researching the combination of AI-based methods with causal analysis methods.
The project will run for three years. The project is funded by the Ministry of Economics, Science and Digital Society of the Free State of Thuringia, co-financed by the European Union (EFRE) and coordinated by the Thüringer Aufbaubank.