Secure Aggregation-Based Big Data Analysis and Power Prediction Model for Photovoltaic Systems: A Multi-Layered Approach

This study introduces a fresh approach to boost the security and accuracy of predicting photovoltaic (PV) power generation using secure aggregation techniques. The author will focus on several key stages in the PV data lifecycle, such as collecting, transmitting, storing, and analyzing data. To protect against potential attacks and prevent information leaks during these four crucial processes, we use Paillier and BVG homomorphic encryption methods. By combining TLS protocol with edge computing during data transmission, we not only improve data security but also reduce latency while tackling threats from attackers. There are two different types of attackers: honest but curious adversaries and active adversaries. Additionally, we make strategies for key management, access control, and auditing to ensure that access to information is monitored to help enhance overall system security. In the final phase of analyzing and predicting PV power output, we utilize advanced models like LSTM networks and CEEMDAN to achieve precise time-series predictions. The results indicate that these methods can effectively manage large PV datasets while keeping high prediction accuracy alongside strong security measures. The research establishes a foundation for improving homomorphic encryption, enhancing key management, and creating a big data security framework specific to photovoltaic energy production.