Aplicação digital twin para auxílio de detecção de anomalias no consumo de gás

Abstract

Gas or energy distribution companies must always pay attention to the gas or energy meters in order to ensure that they are charging for the product correctly, that is, that they are receiving for the correct amount of product made available to the customer. Eventually, there may be some problems with the devices and this measuring may be done incorrectly, either due to bad faith actions on behalf of the customer, or even due to malfunction of the meter. This work presents the development of a tool to assist anomaly detection in the measurement of gas consumption for some specific customers of a gas company. In the developed system, anomaly detection is done using linear regression to create a \textit{digital twin}. Although the final linear regression model presented the problem of autocorrelation and non-normality in the residuals, it is still possible to be improved in order to detect anomalies.