Side-channel analysis (SCA) is a technique that extracts leaked information generated during hardware or software execution to compromise cryptographic keys. Among various approaches, profiling side-channel analysis has been proven to be a powerful method for attacking cryptographic systems. In recent years, the integration of artificial intelligence technology into profiling side-channel analysis has significantly enriched attack strategies and improved efficiency. During the profiling phase, leakage information related to the target device is typically collected by accessing a cloned device. However, practical scenarios often involve discrepancies between the cloned and target devices. Most existing studies rely on a single device for training and validation, resulting in methods that are highly environment-dependent, with limited applicability and poor portability. This study focuses on the portability challenges encountered in complex application scenarios. Challenges arising from variations in parameter settings, algorithm implementations, and hardware differences are analyzed in detail. Solutions and analysis results proposed in recent years are systematically reviewed. Based on this survey, current limitations in portability research on side-channel analysis are summarized, and potential future directions are discussed.