Anomaly detection on industrial time series for retaining energy efficiency

Improving upon or even just retaining energy efficiency at industrial plants presents a rising challenge. Energy efficiency is gradually lowered due to equipment wear and operating errors. Energy consumption increases as a result, whereas output remains nearly constant or even decreases. Maintaining energy efficiency can be achieved by continuously monitoring power consumption and taking measures accordingly. However, due to the amount of collected data in factories, employees require support in the detection of anomalies. Therefore, this paper proposes a method which is able to detect inefficiencies on univariate time series based on historical data. This enables suitable measures to be taken in order to maintain energy efficiency without the need of additional expert knowledge.