Enhancing cybersecurity in virtual power plants by detecting network based cyber attacks using an unsupervised autoencoder approach.

The increasing adoption of the Internet of Things (IoT) in energy systems has brought significant advancements but also heightened cyber security risks. Virtual Power Plants (VPPs), which aggregate distributed renewable energy resources into a single entity for participation in energy markets, are particularly vulnerable to cyber-attacks due to their reliance on modern information and communication technologies. Cyber-attacks targeting devices, networks, or specific goals can compromise system integrity. Common attack types include Denial of Service (DoS), Man-in-the-Middle (MITM), and False Data Injection Attacks (FDIA).Among these threats, FDIA are especially concerning as they manipulate critical operational data, such as bid prices and energy quantities, to disrupt system reliability, market stability, and financial performance. This study proposes an unsupervised Autoencoder (AE) deep learning approach to detect FDIA in VPP systems. The methodology is validated on a 9-bus and IEEE-39 bus test system modeled in MATLAB Simulink, encompassing renewable energy sources, energy storage systems, and variable loads. Time-series data generated over 1,000 days is used for training, validation, and testing the AE model. The results demonstrate the model's ability to detect anomalies with high accuracy by analyzing reconstruction errors. By identifying false data, the approach ensures system reliability, protects against financial losses, and maintains energy market stability. This work highlights the importance of advanced machine learning techniques in enhancing cyber security for IoT-based energy systems and ensuring secure VPP operations.