Hidden Markov Model-Based State Classification in Microgrids Using Advanced Feature Extraction from PCC Measurements

The reliable classification of microgrid operating states is critical for stability, protection, and control. Conventional threshold-based and static methods are often insufficient under noisy or transient conditions, making it difficult to distinguish between Grid-connected, Islanded, and Faulted modes. This work presents a probabilistic framework based on Hidden Markov Models (HMMs) that addresses these limitations by incorporating temporal dynamics into the classification process. Point of Common Coupling (PCC) voltage signals were analyzed, and features such as RMS, peak value, energy, zero-crossing rate, and total harmonic distortion (THD) were extracted using a sliding window approach. The features were discretized via k-means clustering to generate observation sequences for HMM training. The Baum-Welch algorithm was employed for parameter estimation, and the Viterbi algorithm was used to determine the most likely sequence of hidden states. Real-time simulations on a microgrid model under controlled Grid-connected, Islanded, and Faulted scenarios demonstrated that the proposed method achieved over 90% classification accuracy. Confusion matrix analysis confirmed reliable detection of Grid-connected and Islanded states, while temporal modeling improved fault recognition.