Work

Safety evaluation of new energy three-electric systems using enhanced physics-informed neural networks

Year: 2026

Type: article

Abstract: Accurate safety-state measurement of the three-electric system in new energy vehicles is essential for ensuring reliable and secure operation. This paper proposes an artificial-intelligence-assisted safety measurement framework that integrates a searchable encryption mechanism with Enhanced Physics-Informed Neural Networks (EPINNs). In the proposed framework, the searchable encryption module supports privacy-preserving storage and retrieval of multi-source measurement data, while EPINNs serve as the core Artificial Intelligence (AI) model for multi-parameter state estimation and fault identification. A lightweight searchable encryption scheme based on block mapping is first designed to achieve feature-level obfuscation and protect sensitive measurement data. Then, an EPINNs-based state estimation model is developed by incorporating data-driven learning, subsystem physical constraints, and energy-consistency terms to enhance robustness under noisy sensor data. Experimental results show that the proposed method achieves improved performance compared with conventional Multi-Layer Perceptron (MLP) and Convolutional Neural Network (CNN) models in battery state-of-health estimation, motor fault detection under injected fault conditions, and microgrid load measurement, with the motor fault detection accuracy exceeding 95% under the tested conditions. These results indicate that the proposed framework provides an effective and secure technical solution for intelligent safety monitoring of new energy vehicle three-electric systems.

Source: Engineering Applications of Artificial Intelligence

DOI: 10.1016/j.engappai.2026.116500