The development of smart grids and distributed new energy sources has made renewable energy internet of things (REIoT) possible. However, false data injection attacks (FDIAs) during the trading process can significantly harm trading data and patterns. Existing detection methods mostly rely on labeled data, which greatly increases the difficulty and manual cost of detection. To address these issues, this paper proposes a contrastive learning-based FDIA detection method with attention timestamp masking (CLATM) for REIoT. First, data from smart grids are transformed into time series, and the parameters of these time series are masked using an attention mechanism, forcing the model to mask relatively important time series information and infer the missing information through context in subsequent processes. Then, the masked time series data are fed into a GRU layer for training, using contrastive learning to infer whether FDIAs exist in transactions. The experimental results demonstrate that the proposed model achieves higher detection accuracy.