Dynamic Forecasting Method of Multi-Dimensional Potential of Green Electricity Industrial Parks Considering Privacy Preserving

The green electricity industrial park (GEIP) serves as a critical approach for energy conservation and carbon reduction, effectively enhancing the proportion of renewable energy in terminal energy consumption and improving the stability of the new-type power system. Currently, GEIPs are in a nascent developmental stage characterized by ambiguous interaction mechanisms and lacking a comprehensive evaluation system and potential forecasting methodology for green development. In this context, this study proposes a dynamic forecasting method of multi-dimensional potential to provide theoretical support for the development of GEIPs. Firstly, a multi-dimensional evaluation index system and forecasting model are established to quantitatively assess the renewable energy absorption ratio and regulation potential of GEIPs. Considering the substantial privacy-sensitive data involved, the Cheon-Kim-Kim-Song (CKKS) scheme is implemented to address user privacy concerns while streamlining encryption procedures. Subsequently, a Multi-head Attention-Multilayer Perceptron (MA-MLP) training framework based on fully homomorphic encryption is developed to enhance the accuracy of the potential forecasting model. Case studies demonstrate that the proposed method effectively measures the developmental potential of GEIPs. The CKKS-MAMLP method significantly improves the reliability and accuracy of potential forecasting across various parks, providing empirical evidence for their sustainable development. This approach offers actionable guidance for efficient operation and healthy evolution of the new-type power system under the “dual carbon” goals.