基于进化支持向量机的滑坡地下水位动态预测
来源期刊：中南大学学报(自然科学版)2012年第12期
论文作者：彭令 牛瑞卿 叶润青 赵艳南
文章页码：4788 - 4795
关键词：地下水位；预测；滑坡；支持向量机；遗传算法
Key words：ground water level; prediction; landslides; support vector machine; genetic algorithm
摘 要：地下水位动态预测对滑坡稳定性评价具有关键作用。滑坡地下水位演化过程是一个受水文地质条件控制，并受降雨、库水和气温等多种影响因素综合作用而发展演化的非线性动力系统，地下水位与其影响因素之间存在非线性响应关系。以三峡库区白家包滑坡地下水位监测数据为例，在深入分析滑坡地下水位变化特征及其与影响因素响应关系的基础上，利用非线性智能遗传算法和支持向量机建立进化支持向量机耦合模型，并对地下水位进行预测，其预测结果的均方差和相关系数的平方分别为0.013和0.929，说明预测结果与实测值较吻合。选择神经网络模型进行对比，耦合模型的均方差小154%，而相关系数的平方大10%。综合表明进化支持向量机耦合模型具有较好的拟合和泛化能力，是一种行之有效的滑坡地下水位预测方法。
Abstract: Prediction of ground water level is ignificant in evaluation of landslide stability. The evolution process of the ground water level in landslides is a nonlinear dynamic system which is controlled by the hydrogeology condition and to suffers from comprehensive effects by multiple influential factors such as rainfall, reservoir water level, temperature and so on. There is the nonlinear response between ground water level and its influencing factors. According to ground water level data of Baijiabao landslide in the Three Gorges reservoir area, the response relationship between influential factors and ground water level variation was analysed, and the characteristics of ground water level in the landslide were discussed. Using a nonlinear genetic algorithm and support vector regression (GA-SVR) model, the values of ground water level in landslides is predicted. Predicted values of the GA-SVR model are consistent with the measured values. The mean squared error of the GA-SVR model is only 0.013, which is less than those of radial basis function artificial neural network (RBF-ANN) model by 154%. And the squared correlation coefficient of the GA-SVR model reaches 0.929, which is more than those of RBF-ANN model by 10%. It is indicated that the GA-SVR model has a great fitting and generalization ability. It is an effective method for prediction of ground water level in landslides.
彭令1，牛瑞卿1，叶润青2，赵艳南1
(1. 中国地质大学 地球物理与空间信息学院，湖北 武汉，430074；
2. 三峡库区地质灾害防治工作指挥部，湖北 宜昌，443000)
摘 要：地下水位动态预测对滑坡稳定性评价具有关键作用。滑坡地下水位演化过程是一个受水文地质条件控制，并受降雨、库水和气温等多种影响因素综合作用而发展演化的非线性动力系统，地下水位与其影响因素之间存在非线性响应关系。以三峡库区白家包滑坡地下水位监测数据为例，在深入分析滑坡地下水位变化特征及其与影响因素响应关系的基础上，利用非线性智能遗传算法和支持向量机建立进化支持向量机耦合模型，并对地下水位进行预测，其预测结果的均方差和相关系数的平方分别为0.013和0.929，说明预测结果与实测值较吻合。选择神经网络模型进行对比，耦合模型的均方差小154%，而相关系数的平方大10%。综合表明进化支持向量机耦合模型具有较好的拟合和泛化能力，是一种行之有效的滑坡地下水位预测方法。
关键词：地下水位；预测；滑坡；支持向量机；遗传算法
PENG Ling1, NIU Rui-qing1, YE Run-qing2, ZHAO Yan-nan1
(1. Institute of Geophysics and Geomatics, China University of Geosciences, Wuhan 430074, China;
2. Headquarters of Prevention and Control of Geo-Hazards in Area of TGR, Yichang 443000, China)
Abstract:Prediction of ground water level is ignificant in evaluation of landslide stability. The evolution process of the ground water level in landslides is a nonlinear dynamic system which is controlled by the hydrogeology condition and to suffers from comprehensive effects by multiple influential factors such as rainfall, reservoir water level, temperature and so on. There is the nonlinear response between ground water level and its influencing factors. According to ground water level data of Baijiabao landslide in the Three Gorges reservoir area, the response relationship between influential factors and ground water level variation was analysed, and the characteristics of ground water level in the landslide were discussed. Using a nonlinear genetic algorithm and support vector regression (GA-SVR) model, the values of ground water level in landslides is predicted. Predicted values of the GA-SVR model are consistent with the measured values. The mean squared error of the GA-SVR model is only 0.013, which is less than those of radial basis function artificial neural network (RBF-ANN) model by 154%. And the squared correlation coefficient of the GA-SVR model reaches 0.929, which is more than those of RBF-ANN model by 10%. It is indicated that the GA-SVR model has a great fitting and generalization ability. It is an effective method for prediction of ground water level in landslides.
Key words:ground water level; prediction; landslides; support vector machine; genetic algorithm
