Dam Seepage Analysis Based on Causal Testing and Regression Analysis

Dam seepage is a critical issue affecting the safe operation of reservoir dams, making the monitoring and early warning of abnormal seepage conditions particularly important. Currently, analyses of dam seepage primarily focus on using finite element methods to invert seepage conditions and employ regression analysis and classical machine learning methods to predict seepage. However, there has been limited analysis of the relationships among various influencing factors. The subjectivity of input factors in seepage safety monitoring models, the imprecision of factor relationships, and the randomness of parameter selection can all lead to uncertainty in model predictions. Therefore, to identify the primary factors influencing reservoir seepage issues, we took a specific reservoir project as an example and employed stepwise regression analysis and Granger causality tests to comprehensively examine the relationships between reservoir water level, rainfall, seepage pressure at various locations, and seepage pressure around the dam. Based on this analysis, the key influencing factors for seepage pressure around the dam were identified. The results indicate that reservoir water level and seepage pressure influence the seepage pressure around the dam. The stepwise regression method can comprehensively screen for potential influencing factors. Meanwhile, GCT utilizes time lag characteristics to further narrow the range of influencing factors. These two methods significantly narrow the scope of screening for factors affecting seepage pressure around the dam. These two methods can be used to narrow down the range of factors influencing seepage pressure around the dam, reduce interference from these factors, scientifically eliminate spurious correlations and redundant variables, and efficiently and reliably detect and provide early warnings of abnormal dam seepage.