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

Area of Science:

• Cybersecurity in Energy Systems
• Machine Learning Applications
• Internet of Things (IoT) Security

Background:

• The integration of the Internet of Things (IoT) in energy systems, particularly Virtual Power Plants (VPPs), increases cybersecurity vulnerabilities.
• VPPs are susceptible to cyber-attacks like False Data Injection Attacks (FDIA) that manipulate critical operational data.
• FDIA pose significant risks to system reliability, market stability, and financial performance in VPP operations.

Purpose of the Study:

• To propose and validate an unsupervised Autoencoder (AE) deep learning approach for detecting FDIA in VPP systems.
• To enhance the cybersecurity posture of IoT-based energy infrastructures.
• To ensure the reliability and stability of energy markets and VPP operations.

Main Methods:

• An unsupervised Autoencoder (AE) deep learning model was developed for anomaly detection.
• The methodology was tested on 9-bus and IEEE-39 bus systems using MATLAB Simulink.
• Time-series data spanning 1,000 days, including renewable energy sources, energy storage, and variable loads, was utilized for model training and validation.

Main Results:

• The AE model demonstrated high accuracy in detecting anomalies by analyzing reconstruction errors.
• The approach successfully identified instances of false data injection within the VPP systems.
• Validation on standard test systems confirmed the model's effectiveness in detecting FDIA.

Conclusions:

• The proposed AE deep learning approach is effective in detecting FDIA in VPP systems.
• Implementing this method ensures system reliability, mitigates financial losses, and maintains energy market stability.
• Advanced machine learning techniques are crucial for securing IoT-based energy systems and VPP operations.