<p>Reliability concerns in islanded AC microgrids have led to the interconnection of microgrids, demanding advanced protection schemes for secure operation of parallel interconnected AC microgrids (PIAC-MGs). This paper proposes a novel, communication-independent, single-ended, current-based protection scheme for ultra-fast fault detection and classification. The method addresses the key challenge of discriminating internal from external faults in PIAC-MGs with AC tie-lines, a topic rarely covered in prior work. Using a superimposed modal transformation index, faults are detected within one sample period locally, eliminating the need for communication infrastructure and reducing cost. Fault classification leverages Clarke transformation, cascaded mathematical morphology filtering, and a tree-based rule classifier to identify fault types (LG, LLG, LL, LLL, and LLLG) and distinguish load changes. The scheme is robust against mutual coupling effects, insensitive to fault current levels, and scalable for practical deployment. Validated on a PIAC-MG test system via EMTP-RV and MATLAB simulations, results confirm high accuracy, speed, and resilience. This work advances PIAC-MG protection by offering a fully autonomous, cost-effective, and comprehensive solution addressing limitations of prior methods and supporting reliable, communication-free operation of modern interconnected microgrids.</p>