Research On Application Of Classification Algorithm In Hydropower Station Equipment Failure Prediction

Electromechanical systems of the hydropower station consists of large number,variety,high integration of electromechanical equipment,is the core of the hydropower station normal operation,the working condition of equipment directly affects the production operation of hydropower station.With the development of science and technology,there are more and more unmanned and few manned hydropower station.Intelligent development bring lots of conveniences to the people,also bring new problems—more equipment failures,how to effectively analyze and deal with the equipment failure has become an important issue.At present,for improve the reliability and stability of the system,using computer related technology to diagnose some key equipment has gradually become an effective method for the maintenance of hydropower stations,but this maintenance method has limitations and hysteresis.Data mining technology can extract hidden,unknown and useful information from a large number of incomplete and noisy data,it used to hydropower station in electromechanical equipment failure prediction can strong support for equipment maintenance work.According to the characteristics of equipment failure data in Caijiazhou Hydropower Station,this paper studies the classification problem of unbalanced dataset,analyzes the limitations of traditional single classification algorithm,two kinds of combined classification algorithms based on Microsoft Decision Tree classifier and Microsoft Naive Bayes classifier are proposed—MDB and WMDB,and the two algorithms are evaluated with common performance indicators.Use Microsoft Visual Studio 2008 and Microsoft SQL Server 2008 R2 as the development tool and the data platform,design and realize the failure prediction system of hydropower station equipment.The multidimensional analysis of failure data is carried out,achieves multi-directional and multi-angle browsing of the data.The integration of two kinds of combined algorithms is realized by using the algorithm extension function of SQL Server Analysis Server(SSAS).Take the failure data of Caijiazhou hydropower station in the first half of 2016 as the test set,the practical application of the model based on the two algorithms is completed by window operation and command operation.Finally,the performance of the model is verified by lift chart and classification matrix,the verification results of the model are consistent with the evaluation results of the algorithms,it is further proved that the WMDB algorithm and the MDB algorithm are effective improvements to the single classification algorithm,the WMDB algorithm among them has the best comprehensive performance,and can meet the needs of the practical application of the hydropower station equipment failure prediction system.