Power quality disturbance detection using histogram of oriented gradients with extreme learning machine

Power quality disturbance (PQD), which is caused by an increase in power electronics-based nonlinear loads, is a major concern in current power system analysis. Its detection with high classification efficiency is crucial for the smooth operation of all electrical appliances, smart grids, and industries. These days, a variety of machine learning (ML) and signal processing approaches are available for detection and classification of PQDs, providing high classification accuracy. This article presents a histogram of oriented gradients (HOG) and extreme learning machine (ELM)-based feature extraction and classification technique for PQDs. For effective classification, PQDs require many feature vectors because they are non-stationary and nonlinear, which results in high computational time. The suggested approach first uses an HOG for efficient feature extraction from power signals, then processes through ELM to classify the power quality disturbances. The first computational nature of HOG and ELM reduces the total computational burden on the controller in handling a large number of feature vectors. The test system used in this paper is used to create 18 disturbances in the MATLAB/SIMULINK environment. Short event detection, lesser computational timing, and superior classification accuracy are the key benefits of proposed technique. Additionally, classification and detection accuracy of proposed PQ classifier is validated with real-time signals generated in an experimental model.