An improved hidden Markov prediction model (IL-HMMs) based on incremental learning was developed,which was based on the prediction of regional groundwater table in the typical county of Dengkou,a typical arid region in Northwest China.In order to test the IL-HMMs model prediction results,the predicted results was compared with the measured data of 2013,and the results of hidden Markov model (HMMs),weighted Markov chain (WMCP) and BP neural network (BPNN) prediction model.The results showed that compared with other forecasting models,the prediction accuracy of IL-HMMs model was improved significantly,the error was smaller and the robustness was better.The groundwater depth in 2018 was predicted by using the IL-HMMs model.The prediction results showed that in 2018 the average annual groundwater depth would be increased slightly and the groundwater depth would be increased in some areas.The IL-HMMs model of groundwater depth prediction had good stability and robustness,it can provide ideas and methods for the dynamic prediction of groundwater depth,and also provide an important basis for the development,utilization and protection of groundwater resources in the region.Tracking and monitoring the change of water level,preventing the groundwater level from falling continuously and making emergency response plan can be utilized to realize the sustainable development and utilization of water resources.%以西北干旱典型县域磴口县为研究区,基于增量学习的改进隐马尔可夫预测模型(IL-HMMs),对区域地下水埋深进行了预测研究.为检验IL-HMMs模型预测效果,将模型预测结果与2013年长观井的实测数据进行了比较;同时为检验模型的优劣性,与未经增量学习的隐马尔可夫模型(HMMs)、加权马尔可夫链(WMCP)和BP神经网络(BP neural network,BPNN)预测模型的预测结果进行了比较.结果表明:与其他几种预测模型相比,IL-HMMs模型预测精度显著提高,误差更小,有较好的鲁棒性.并使用IL-HMMs模型对2018年地下水埋深进行了预测,预测结果表明,2018年地下水年平均埋深略有增加、局部区域地下水埋深增量加剧.基于IL-HMMs模型的地下水埋深预测具有很好稳定性的同时对新数据加入又有很好的鲁棒性,可为地下水埋深动态预测提供思路与方法补充,为区域地下水资源开发利用和保护提供重要依据.
展开▼
