Explainable Reliability Modeling and Runtime Monitoring of Software Systems in Electric Vehicle Charging Infrastructure
提出一个结合贝叶斯分析和可解释人工智能的框架，用于建模电动汽车充电软件系统的可靠性，提升故障预测和诊断透明度，并通过仿真验证其效果。
The rapid expansion of electric vehicle (EV) charging infrastructure brings with it an increasing reliance on software systems for managing control logic, communication protocols, and real-time decision-making. As these systems grow more complex and interconnected, ensuring their operational reliability becomes essential—not only for individual charging stations but for maintaining broader energy grid stability and safety. This study introduces a new framework that models software reliability within EV charging systems, combining probabilistic techniques and explainable artificial intelligence (XAI) to improve failure prediction and monitoring transparency. By employing Bayesian reliability analysis and dynamic runtime observation, the proposed method identifies latent software vulnerabilities and offers interpretable diagnostic feedback, even under uncertain operating conditions. Unlike prior work focused primarily on hardware resilience or energy optimization, our research emphasizes control software robustness and the visibility of system behavior during operation. To validate the framework, we simulate an EV charging network featuring real-time data flows and multiple failure scenarios. Results show that our model enhances system stability, extends the average time between software failures, and facilitates faster issue diagnosis—all without compromising explainability. This contribution supports ongoing national efforts in clean energy transition, infrastructure modernization, and cyber-physical system safety by offering a scalable, modular, and intelligible approach to software reliability assurance in EV environments.