Power Quality Disturbances Elucidation in Hybrid Systems Based on Event-Driven Variational Mode Decomposition in The Smart Grid Prospect

Abstract

Recently renewable energy (RE) sources were combined with a utility network to establish a hybrid power system to accomplish the stability of power generations. Integrating a renewable-resources-based distributed generation (DG) system into the current power grid could cause problems with power quality (PQ), system dependability, and other challenges. The power quality (PQ) disruptions assessment is essential to guarantee high-quality power generations in smart grids. The research collected effectively models different types of PQ disturbances signals. After that, these can be used for the PQ disturbances interpreting systems training and performance quantification. The recording and collection of such signals is not an easy task. A common trend is the generation of real-like signals from mathematical models to overcome this limitation. This thesis will determine the PQ disturbances model based on IEEE 1159-2019 standard. The outcome of the model will evaluate the performance of the devised system. Firstly, the signal reconstruction will be performed to realize analog-quasi signals. In this context, mature and precise cubic-spline interpolators-based signal reconstruction algorithms are sued. In the next step, the reconstructed signals will be acquired by using the MATLAB-based event-driven sensing models. The acquired signals will be segmented by using novel event-driven activity selection techniques. Afterward, the segments will be decomposed in oscillatory modes by using the adaptive-rate Variational mode decomposition (VMD). This decomposition will result in Mode updates. The pertinent features will be extracted from modes. These features used to prepare templates, testing instances and to prepare and evaluate the considered classification algorithms. The outcome of the model evaluated the performance of the devised system. In this work, we evaluated 11 classes as final results, including the following cases transient, oscillatory transient, flicker, harmonics, interruption, sag, swell, notch, harmonics with sag with flicker, harmonics with swell with flicker, and swell with oscillatory transient signals by using the MATLAB-based event-driven sensing models. The comparison of the trained data between instants PQDs and extracted data using VMD feature extraction has been studied. The average results of the VMD feature extraction method of classifiers showed an improvement percentage between 10.05% to 60.38%.Accuracy and speed have been raised. The most significant improvement in the linear SVM classifier has been shown by 60.38% of average measurements. That will help the smart grid to reach maximum power and increase its affection. Assuming that it will result in rising energy efficiency and simple hardware along with lower latency realizations compared to the classical counter sensing and processing-based approaches. This Study is well aligned with the 2030 vision of Saudi Arabia and can be well integrated into the NEOM smart metering system. The solution has potential and could be commercialized in collaboration with the authorities and industrial partners in Saudi Arabia.