A GIM-Based Ontological Framework and LSTM-RF Optimization for Real-Time Digital Twin Power Grid Modeling

Abstract

Given the problems associated with traditional DTPG (Digital Twin Power Grid) models in large-scale, complex power grid applications, such as the difficulty in fusing multi-source, heterogeneous data and the lack of real-time performance, this study proposes a construction and optimization method with GIM (Grid Information Model) as its core. In the model building stage, this method first realizes the standardized modeling and dynamic state encapsulation of power grid equipment through the extended GIM semantic library and holographic object encapsulation framework, and designs a five-layer dynamic mapping mechanism combined with the Levenshtein distance and BERT (Bidirectional Encoder Representations from Transformers) model to solve the problem of semantic alignment of multi-source data; then develops an event-driven architecture based on Kafka to support real-time data updates and model version backtracking; applies a spatiotemporal dynamic mapping engine, integrates GIS (Geographic Information System) and GIM topological structure, and realizes the spatial positioning and visualization of electrical events. At the optimization level, the self-learning mechanism of the LSTM (Long Short-Term Memory) and RF (Random Forest) combined algorithm is applied to realize the dynamic correction of model parameters and performance improvement. Experiments show that the proposed GIM method takes an average of 4.21 seconds to model, which is much lower than other methods. In the fusion of multi-source heterogeneous data, the field recognition accuracy averages 97.82%. In evaluating model accuracy and adaptability, the GIM method achieves a modeling accuracy of 96.5% and a prediction error rate of only 3.1%. The synchronization delay is stable at 117 to 125 milliseconds. The GIM-based digital twin power grid method in this paper improves the efficiency and accuracy of power grid digital model construction by means of standardized modeling and real-time data fusion, which can quickly respond to changes, accurately locate and warn faults, provide simulation support for dispatching, and enhance power grid reliability and intelligence.