Research On Adaptive Identification Method For Abnormal Data Of Dam Monitoring Based On Association Rules And Machine Learning

The monitoring system of a dam is crucial for ensuring its safety and stability.However,when it comes to safety monitoring,the generation of abnormal data is inevitable.These abnormal data can significantly affect the accuracy and reliability of the monitoring system,resulting in unnecessary risks and losses in production.Therefore,identifying abnormal data in dam monitoring is a challenging problem both domestically and internationally.Currently,traditional methods for cleaning abnormal data mainly analyze single monitoring effects without considering the correlations between dam monitoring sequences.Moreover,these methods lack adaptability and are unable to process complex data,which can lead to inefficiencies caused by factors such as the distribution of monitoring data.As a result,this thesis conducts research on an adaptive recognition method of dam monitoring anomaly data based on association rules and machine learning.The main achievements are as follows:(1)The correlation between environmental quantity and effect quantity of dam monitoring is analyzed using the Apriori algorithm based on the principle of association rules.The results show a strong correlation between the water level sequence of the environmental quantity monitored by Longgang Reservoir Dam and the horizontal displacement sequence of dam monitoring effect quantity.Similarly,a strong correlation was found between the water level sequence and P1 sequence of osmometer.However,there is no correlation between P2 sequence of osmometer and P3 sequence of osmometer.(2)A method for identifying abnormal data in dam monitoring sequences based on association rules and the DBSCAN algorithm is proposed.The algorithm uses DBSCAN to identify irrelevant sequence anomaly data and output washable anomaly data.For monitoring data of strong-correlated sequences,the two sequences are combined according to monitoring time in DBSCAN algorithm,which realizes anomaly data recognition of dam monitoring based on Association rules.(3)For irrelevant sequence anomaly data,the random forest regression model is used to reconstruct single anomaly data.The accuracy of the random forest regression model for the reconstruction of irrelevant sequence anomaly value is validated using BP neural network,support vector machine model(SVM)and random forest regression to forecast and compare.(4)For strong-correlated sequence anomaly data,LSTM neural network model is used to reconstruct it,and the goodness of fit and mean square error of strong-correlated sequences are validated using the network model.(5)Based on the above research outcomes,a self-adaptive recognition software for dam monitoring anomaly data is developed using Python programming language,Py Charm integrated development environment,Python standard library Tkinter,and other third-party libraries.The software’s feasibility and applicability are verified using actual monitoring data of Longgang Reservoir.