Forecasting Electrical Energy Consumption for Malfunction Detection in Complex Technical Systems

Anna TATARCZAK, Łukasz WIECHETEK, Adam KIERSZTYN, Marek MEDREK, Jarosław BANAS

Abstract


Issues related to monitoring and detection of an unexpected or hidden malfunction in complex technical systems become more important since the complexity of technical installations grows and the remote installations, without human supervision are widely used in many branches of industry. In the proposed solution we use detailed information on electricity consumption provided by smart energy metering technologies for monitoring and anomalies detection purposes. As the data source, we use the teletechnical installations of the telco operator network, which consists of several hundred installations of various types, each created from many standardized components like power supply, battery, air conditioner, transmitter, etc. We build individual energy consumption model of each analyzed facility, which reflects daily cycles, weekly, monthly and seasonal fluctuations. For our simulations, we use the Particle Swarm Optimization method, which allows us to parameterize the model and estimate the expected energy consumption rate. The results of simulations show very good convergence with measurement data and allow for real-time malfunction detection.

Keywords


Data mining, anomaly detection, machine learning, Particle Swarm Optimization method, short-term trend


DOI
10.12783/dtcse/optim2018/27935

Full Text:

PDF

Refbacks

  • There are currently no refbacks.