基于宽度学习的地下水水位预测研究

Research on Groundwater Level Prediction Based on Broad Learning

地下水作为水资源的重要组成部分,过度开采将导致水资源紧缺,甚至可能诱发泥石流、滑坡等地质灾害,及时掌握地下水位及其变化趋势就显得十分重要。针对当前研究中地下水位预测准确度不高且预测时间过长的缺陷,借助宽度学习算法构建了基于宽度学习的地下水水位预测模型,并利用矩阵随机近似奇异值分解对模型进行优化建立了SVDBL模型,并通过济源市地下水位历史数据对模型进行验证。结果表明,SVDBL模型的预测准确率为92.12%,且具有较强的在线训练能力;也表明将该模型用于地下水水位预测是可行的。

Groundwater is an important part of water resources.The over-exploitation of groundwater will lead to the shortage of water resources and may even induce geological disasters such as debris flows and landslides,so it is very important to understand the groundwater level and its changing trend in time.Aiming at the shortcomings of low accuracy and long prediction time of groundwater level prediction in current research,a groundwater level prediction model based on broad learning is constructed with the help of breadth learning algorithm,and the SVDBL model is established by using matrix random approximate singular value decomposition to optimize the model.The model is verified by the historical data of groundwater level in Jiyuan City.The results show that the prediction accuracy of the SVDBL model is 92.12%and it has strong online training ability,so it is feasible to use the model for groundwater level prediction.