A Neural Network-Based Security Model for Real-Time Solar Energy Monitoring Systems

Abstract

Smart grids and their increasingly linked solar power are driving the demand for more resilient, TETRA-compatible real-time cyber security solutions to protect vital data and infrastructure. This paper presents a neural network-driven security model dedicated to real-time solar energy monitoring systems for the purpose of identifying and counteracting several kinds of cyber-threats including false data injection, DoS attacks as well as unauthorized access. The model is built based on the hybrid deep learning that uses CNN to extract features, and LSTM to identify temporal patterns in data streams. Real-time data from PV monitoring devices are preprocessed by normalizing and adaptively selecting features to increase accuracy and responsiveness. Experimental results show that the proposed model outperforms standard machine learning-based classifiers in detection accuracy, false positive rates and boot response time. The embedding of the NN into the solar monitoring infrastructure guarantees that threats are made visible before an attack occurs, thus strengthening system security, energy data traceability and grid robustness. These observations demonstrate that deep neural architectures have the potential to shape autonomous and adaptive cybersecurity mechanisms addressing intelligent energy infrastructures in an era of Internet-of-Things (IoT).