Research On Forecast Method Of Power Grid Disaster Using Monitored GIC Data

The potential risk that geomagnetically induced currents’(GIC) effects on power grid is invaluable. Recent studies have shown that for the disaster caused by GIC, it is not the best solution that we only rely on some defensive measures and recovery management after the disasters. In this paper, through the analysis on the mechanism, calculation methods and influencing factors of power grid GIC, the forecasting methods and early warning model are investigated, supported by the project of National High Technology Research and Development Program ("863" Program),"Research on Monitoring Technology of Disastrous Space Weather Affect Communication and Power Systems",(Project No.:2012AA121005). Combined with common forecasting methods, such as time series and neural networks, the forecast models using NARX dynamic neural network and wavelet neural network are established. The models are simulated and validated by programming in MATLAB. In addition, this paper also proposes a power grid data warehouse solution based on the influencing factors of GIC in order to provide the data support for the platform of power grid GIC’s early warning. The main contents of this paper are as follows:(1) The mechanism of GIC is deeply analyzed from the source, solar activities, to the transmission network on Earth. The algorithms of calculating GIC, including calculation methods of ground induced electric field using plane wave theory, complex image theory, Earth hierarchical model and the calculation method of GIC using power grid equivalent model are introduced. The basis of early warning on power grid is also analyzed through the related data source and several common forecasting methods.(2) The establishment of power grid GIC’s early warning platform needs the support of various algorithms and models. Therefore, the prediction method based on the time series and dynamic neural network is investigated. On that basis, a power grid GIC forecasting model using NARX neural network is built and the forecast results are evaluated by comparing with the actual monitored GIC data in Guangdong Ling’ao nuclear power station.(3) Aiming at the characteristics of power grid GIC such as complex origin and saltation transiently, a prediction method based on wavelet neural network is proposed and corresponding model is established. By comparing the NARX neural network GIC forecasting model and wavelet neural network GIC model, the performance of two forecasting methods is assessed.(4) Through the analysis of mechanism and influence factors power grid GIC, building a professional GIC data warehouse is put forward to provide data support for power grid GIC warning platform, evaluation platform and decision support platform. A power grid GIC support data warehouse and application framework is studied and planned. At the same time, the data source of power grid GIC’s data warehouse data source is analyzed in detail.