基于时间序列分析的水位短期预测模型仿真

摘要

To more effectively predict the water level,a method was proposed combining time-series analysis and Kalman filte-ring to optimize the prediction.By analyzing the time series of related water level station using the autoregressive distributed lag model,the lag length of each site was determined.After obtaining the coefficients of every variable,initial predicted value was calculated using the related history data,and then the Kalman filtering method was utilized to correct the initial prediction,the final predicted value was got,and the water level forecasting model was established.This hybrid model was used to simulate the prediction process of real water level station.Results of simulation show that the model can effectively predict the short-term trend of the water level station with stable prediction capability and high prediction accuracy.%为能有效预测水位，提出一种结合时间序列分析和卡尔曼滤波的优化方法。通过自回归分布滞后模型对站点的水位数据进行分析，得到各站点的滞后期长度，求得各变量的系数后，结合相关水位数据计算得到初始预测值，利用卡尔曼滤波对预测结果进行修正，获得最终预测值，建立优化的水位预测模型。利用该组合模型进行水位预测实例仿真，仿真结果表明，该模型能有效地预测水位短期内的趋势，预测能力稳定、预测精度高。