| Research Abstract |
Kagoshima Prefecture has been suffering from natural disasters by typhoons repeatedly. They hit power systems very badly and sometimes cut off electricity. To ensure the rapid restoration of electricity supply, one needs to predict the amount of damage by typhoon accurately. The purpose of this project is to design a predictor, which is able to quantify the electric power damage caused by incoming typhoon, from its weather forecasts. From the Meteorological agency, we collect the following data which are considered as inputs of the predictor : track of typhoon, central barometric pressure, wind speed, progressive velocity, radius of storm, etc. The outputs of the predictor are the number of damaged distribution lines and poles, etc. The track, size and strength of typhoon vary from hour to hour and have wide ranges from a quantitative point of view. So we reform the whole date to design a skilled algorithm of the prediction by weighted averaging and normalizing techniques. The track of typhoon is turned out to be strongly relative to the amount of damage. It is evaluated by using a sum of such as Gaussian functions whose parameters are suboptimally selected with the aid of GA. The predictor consists of two stages prediction. The first stage is main, and its prediction error is decreased at the nest stage. For example, the 2^<nd> order polynomials prediction, which is equivalent to a simplified GMDH, is used at the first stage. At the second stage, one of regression models, Neural networks, and RBF networks is employed. This predictor is applied to 29 typhoons that hit Kagoshima in the past 13 years. Simulation results show that the proposed predictor enables us to estimate the amount of damage by typhoon well.
|
