A Comprehensive Energy System Fault Diagnosis Method Based on Multimodal Large Language Model

Abstract

This invention discloses a fault diagnosis method for integrated energy systems based on a multimodal large language model, belonging to the field of intelligent operation and maintenance and fault diagnosis technology for integrated energy systems. Addressing the issues of feature aliasing caused by multi-energy flow coupling in integrated energy systems and the scarcity of fault samples in actual operation, this method first collects multi-energy flow time-series data and operation and maintenance text to construct a dataset; then, it extracts features using a dual-tower structure and achieves cross-modal semantic alignment through a multi-positive sample contrastive learning mechanism; finally, it maps the aligned features to soft cue vectors and injects them into a pre-trained large language model, combining low-rank adaptation techniques to fine-tune and construct a generative diagnostic model. Based on the above technical solution, this invention effectively alleviates the feature aliasing problem in multi-energy flow coupling scenarios of integrated energy systems, overcomes the bottleneck of small-sample training, and directly generates semantic fault diagnosis reports, achieving a balance between high accuracy and strong interpretability.

Images