A Machine Learning Approach to Detection of Geomagnetically Induced Currents in Power Grids Open Access
Downloadable ContentDownload PDF
Solar wind, caused by charged particles which erupted from solar flares and associated coronal mass ejections into space during the peak of intensity of the sun's cycle. The shock wave and/or cloud of magnetic field generated by the solar wind interacts with the Earth's magnetic field. The pressure of the solar wind shapes Earth's magnetosphere. A major disturbance of Earth's magnetosphere and ionosphere leads to geomagnetic disturbance (GMD). Consequently geomagnetically induced currents (GICs) in the conductor surface of Earth. The flow of these currents into transmission lines among power system can cause 'half-cycle saturation' of high-voltage Bulk Electric System (BES) transformers. This phenomenon can lead to increased consumption of reactive power and create disruptive harmonics that could potentially cause very severe consequences, i.e. aging or malfunction of the electric power devices on the grid, and even a total collapse of the power system. Northern North America is especially susceptible to problems resulting from GIC as its Latitude. March 13, 1989, an exceptionally strong GMD caused major damage to electrical power equipment in Canada, Scandinavia, and the United States. Hydro-Quebec extra high voltage (EHV) transmission system experienced instability and tripping of lines carrying power to Montreal resulting in total blackout of the Hydro-Quebec system. In the United States, a voltage fluctuation of up to 4 percent was recorded on the EHV systems in Pennsylvania, New Jersey, and Maryland. On September 19, 1989, at the Salem Unit 2 nuclear power plant, a second solar storm damaged the step-up transformer. In order to limit the potential for damage which a GMD could cause, it is crucial to monitoring the GIC in a power system and mitigate the impact before GIC rise to a certain level. However the GMD doesn't indicate the GIC's impact on power system. Meanwhile directly accessing GIC which acts as a DC component in the high voltage transmission lines is quite dangerous and challenge. Conventional approaches of monitoring transmission lines relies on only measure the AC components through voltage transformer (VTs) and current transformers (CTs). Also the harmonics, which generated by nonlinear load or overloading transformers flow among the grid, will also interfere those generated by GIC when GMD level is low, those interference will make the GIC detection challenging. Therefore, an efficient and effective GIC detecting mechanism is quite demanding. The purpose of this project is to develop a GIC detection approach for the transmission system. Among various types of GIC detection method, two major time-frequency analysis techniques: wavelet transform and short-time Fourier transform are applied and their performance are also evaluated. Based on these analysis, we are developing a framework which consist of a hybrid GIC detection algorithm combined with machining learn technology, which are promising to estimate level GIC in the power system during variety of grid work on conditions.