Validation of adaptive protection framework in virtualized centralized protection architectures using Hardware-in-the-Loop methodology

System (APS) integrated into a Centralized Protection and Control (CPC) platform that dynamically adjusts protection settings according to grid conditions. The CPC platform comprises an IEC 61850-3 based server designed for digital substation environments, with

Merging Units (MU), optimized virtualization enabling concurrent execution of multiple timecritical virtual Intelligent Electronic Devices (vIED), and advanced algorithms. The APS implements a two-phase design. The offline phase defines protection setting groups from analyzing over 1,200 scenarios capturing DER and load variability. Scenarios are clustered using fuzzy c-means (FCM) and optimal settings calculated using Evolutionary Particle Swarm

Optimization (EPSO) to minimize operation time preserving coordination. The online phase monitors real-time grid measurements (SV and R-GOOSE) to select setting groups using pretrained support vector machine (SVM) model. Anti-flapping logic prevents unnecessary transitions. Hardware-in-the-loop (HIL) validation at INESC TEC laboratory used real-time simulation of a 33-bus medium voltage (MV) feeder incorporating DER, reclosers, IBR, variable loads, and faults. Across more than 57,000 simulated fault cases, the APS achieved 97.4% fault-clearing success with 2.6% blinding failures under extreme conditions (high DER presence and high-resistance faults). The SVM classifier achieved 99.8% accuracy with latency below 1 ms, preserving selectivity and response. This work presents a complete methodological framework for design, implementation, and validation of a virtualized APS, addressing key gaps in literature. The vendor-agnostic architecture, suited to be executed in CPC and similar substation controllers, provides utilities a credible path to upgrade legacy protection schemes and prepare for high-renewable power systems.