Data Anomaly Detection in Cyber-Physical Energy Systems

The data is created in an agent-system, used for controlling distributed energy systems. The agents follow the Lightweight Power Exchange Protocol to negotiate whenever an agent detects a planning problem. The protocol is  explained in "A lightweight distributed software agent for automatic demand—supply calculation in smart grids" by Veith, Steinbach and Windeln. Different kinds of anomalies were created by manipulating an agent accordingly: anomalies in the values of the exchanged messages and anomalies in the communication behavior. The implementation of the power exchange used for the data generation can be found here: https://gitlab.com/mango-agents/mango-library/-/tags/Integration_of_the_LPEP.

Anomalies in the values are manipulated by 500 % of the original value (dataset 500_p.csv).

Anomalies in the communication behavior are anomalously started every minute for 8996 seconds in the future (dataset 1m_8996_50p.csv) and every 15 minutes for 2012 seconds in the future (15m_2012_50p.csv).

For anomalies in the communication topology, an agent was chosen to which selected agent does not send any messages, although the agent is part of the neighborhood (Topology anomalies/agent_removed.csv). Furthermore, an agent was manipulated to send messages to an agent which is normally not part of its neighborhood (Topologie anomalies/agent_added.csv).

Normal data is also given.