| Recently,because of the global climate change,natural disasters such as hurricanes,freezing,rain and snow have occurred frequently.These pose a great threat to the safe operation of the power grid.Especially due to the severe transmission line icing,the damage to the power grid is more serious.It will lead to serious accidents such as damage to the electrical fittings,power line break,transmission line towers collapsed,etc.,which will endanger the safe and stable operation of the power grid.So,it is very important to early warn the transmission line icing disasters for ensuring safe operation of power grids.Affected by micro-meteorology and micro-topography,icing process of transmission line presents non-linear and high dimensionality characteristics.And mathematical models are difficult to obtain.However,the accumulated data of icing monitoring system contain abundant and valuable information.We will use data-driven method to extract effective information from micrometeorological data.This thesis mainly researches the qualitative early warning method of transmission line icing as follows:1)Three algorithms based on clustering are presented to extract the disaster features of the complex system.In feature fusion level.for power transmission lines icing and its increasing,the extract of the feature is achieved.The results are helpful for using meteorology features to estimate the safety situation of the power transmission lines system in decision-making level.2)Aiming at the problem that the iteration error of quantitative forecasting model of icing load increases with the increase of forecasting step,long-term icing load can’t be forecasted,and the major icing disaster once in decades or once in hundreds of years can“t be forecasted,because the small probability of icing events are rare,and the training samples are incomplete.A model based on time neighborhood preserving embedding algorithm is presented here to predict the severity of transmission line icing process.The model trains the micrometeorological data under normal conditions and builds an early warning model.It gets a red line for early warning of serious icing,which is based on whether the prediction results exceed the limit,and then warns whether serious icing process occur.3)The above experiments were simulated by MATLAB.The historical data provided by Yunnan Power Grid was used to verify the effectiveness of the clustering model and the sequential neighbors to keep the embedded algorithm model.The Python and MATLAB hybrid programming was used to design the transmission line icing warning system. |
