Background: Compared with the traditional power system, the large-scale access of distributed energy resources in the new power system has a great impact on the structure and operation mode of the power grid, and it is also more susceptible to device-level and network-level FDI attacks. Objective: In order to improve the accuracy and precision of detecting false data injection attacks in distributed energy resources integration into distribution networks and to further explore time series modeling methods for measurement data, it is helpful for the FDIAs detection method to be widely adopted and applied in new power systems. Methods: To address false data injection attacks on distributed energy resource integration into distribution grids within new power systems, a data-driven time series anomaly detection method is employed. Firstly, time-aware shapelets are extracted from time series data, and then the shapelet evolution graph is constructed to capture the correlation between the shapelets. Finally, time series representation vectors are learned using segment embeddings derived from the shapelet evolution graph through the DeepWalk algorithm. These representation vectors are then input into a BO-XGBoost anomaly detector, facilitating the detection of FDIAs. Results: After multiple rounds of parameter tuning, the parameters of Shapelet quantity (K=40) and segment length (L=4) achieved an accuracy of 92.8% in FDIA detection. Comparative experimental results with different algorithms indicate that, compared to other unsupervised learning methods, this approach exhibits an accuracy improvement of 20-40%. In the case of BOXGBOOST, it achieves a 5% increase in accuracy compared to the unmodified XGBOOST. Conclusion: The experimental results indicate that this method can effectively detect false data injection attacks on the integration of distributed energy resources into distribution grids within new power systems. This model significantly enhances detection accuracy and precision while also imparting physical significance to the dynamic evolution of time series models.