Overhead Transmission Line Ice Thickness Prediction Data Fusion Research

Along with the rapid development of national economy, the state increased investment in the power industry and construction, a large number of high voltage or ultra-hign voltage grid were built, and are constantly into the production practice. Overhead transmission lines covered with icing and its growth is an extremely complex process associated with many factors, and the essence of the icing thickness forecasting of overhead transmission line under rough weather is to look for a mapping relationship between the many factors of input and the icing thickness of output. The on-line monitoring system of transmission lines icing in the actual operation process has accumulated a large amount of data, such as temperature, humidity, wind speed, air pressure, the tilt angle of the insulator, the angle of wind tension sensor and so on. These data are as a truly reflection of the relationship between input and output of transmission line icing process. The data fusion technology, is using the computer to analysis the sequetial number of observation information obtained, in order to complete the required decision-making and evalution tasks of information processing technology. Based on the data fusion of the prediction thickness of overhead transmission icing, it is to establish the forecast model of overhead transmission lines using some machine learning algorithm, and to assess and warn the state of the transmission line, and to formulate relevant strategy of de-icing service to related department of disaster prevention and mitigation work.This thesis mainly use machine learning algorithm of self-organizing feature map neural network (SOM) for overhead transmission lines ice level automatically, so as to determine the severity of the transmission line icing. In addition, this paper using support vector machine (SVM) for overhead transmission lines to predict the thickness of the icing, based on the the support vector machine forecasting model of overhead transmission lines icing. Specifically, this paper first introduces the research status of overhead transmission lines icing, research methods and present problems. Then starting from the existing problems, introduced the influence factors of thickness of icing on overhead power transmission lines. Highlighting of overhead transmission lines icing and its growth is an extremely complex process, which associated with multiple factors on performance for a nonlinear time series, dynamic mutability and uncertainty characteristics. Then, from the perspective of data fusion, this paper use machine learning algorithm of self-organizing feature map (SOM) neural network in order to extract the whole process of digital features of transmission line icing, and construct the clustering analysis model of ice thickness of overhead transmission lines based on meteorological information. Finally, this thesis use support vector machine (SVM) to simulate the ice thickness of one step and multiple step regression prediction of the overhead transmission lines, based on support vector machine prediction model of overhead transmission lines icing, and use the cross validation and the method of the genetic algorithm to optimize the support vector machine parameters automatically, in order to reduce the error of the prediction model, and to improve the generalization ability of the model.