Rain Attenuation Prediction at W Band Based on Non-stationary Time-series ARIMA Model

Most existing rain attenuation prediction models are only tested at 55 GHz. There are small numbers of tested W-band rain attenuation prediction models, but these tested models have issues with high complexity and large quantities of calculations. A real-time prediction method is proposed that has a simpler structure and smaller quantity of calculations. The proposed method is based on the ARIMA model, which utilizes the relationship among the time series to establish a prediction model, conducts a stationary test on the original time series, transforms the nonstationary time series into a stationary time series by using a difference transform, and estimates the parameters of the stationary time series. This sequentially transforms the traditional nonlinear prediction into a linear prediction. First, the ARIMA (1,1,6) model is compared under the conditions of different polarizations, prediction intervals, and numbers of time series. Then the proposed model is compared with the International Telecommunication Union-R (ITU-R) and the Silva Mello rain attenuation prediction models using the conditions of vertical polarization, a prediction interval of 0.10 GHz, and a number of 50 time series. Finally, the forecasting time series with the simulant series are compared. The result shows that the prediction error between the ARIMA model, the ITU-R model and the Silva Mello model does not exceed10-3, and that the change trend of the rain attenuation is basically the same. and the goodness of fit between the forecasting and simulant time series is good, which means that the proposed model can be applied to forecasting the rain attenuation in the W band, and that it has the advantages of simpler structure and high precision in prediction.