Power Data Analysis and Privacy Protection Based on Federated Learning

• DOI: 10.12694/scpe.v26i2.3893
• Publication Date: 2025-02-10T19:03:31Z
• Authors: Yajie Li; Shuting Chen; Xinmiao Hu; Sen Xu; Mao Fan
• Keywords: Privacy Protection; Federated Learning; Power analysis

In order for active distribution network operators to carry out power business such as load forecasting without meter reading rights, the author proposes a research on data analysis and privacy protection based federated learning. The learning framework industry user by selecting weather time factors correlation of load. On this basis, constructed an dataset established model Long Short Term Time Series Network (LSTNet). At same time, FedML was used establish sub results indicate that: accuracy specific proposed is less than 9 p.u., theoretical maximum value SMAPE 210 indicating that method has universality can be applied in different industries. training scheme parallel, although it increases interaction 1 minute, accounts smaller proportion compared consumption (1 minute)+single (94 minutes). Conclusion: enable users conduct sharing data, support related operations while protecting electricity privacy. It better predictive performance, fewer numbers, shorter consumption.