Multi-Objective Optimal Operation System For Reservoirs Based On Spark

The reservoir operation can reasonably control the water storage and emission of the reservoir by their regulating capacity on the runoff,which could make them have the function on flood-and-drought resisting.The issue of the optimal operation of reservoirs is a multi-objective optimization problem with complex constraints.The data scale of the reservoir operation becoming larger with the speeding up of the progress of the water information,which made the calculation of the reservoir operation is more complicated and need the technologies supporting of the parallel computing and big data.Therefore,this research will focus on the handing techniques of the big data that revolve around the multi-objective optimal operation of the reservoirs,and applied it on the parallelization of the multi-objective optimal operation of the reservoirs.The authors investigated the development state of the multi-objective optimal operation of the reservoirs;research the data processing and parallel computing relevant technologic,model of the multi-objective optimal operation of reservoirs and the optimal operation method based on genetic algorithm,propose the parallel methods of genetic algorithm based on spark.Based on above basis,we use Linux operating system,Hadoop2.7.1+Spark1.5.2+Spark on Yarn to build Spark big data supporting environment,then designed and developed the multi-objective optimal operation of reservoirs based on spark using Scala language.The software mainly have three function modules,including the data collect module,data saving module and the reservoir operation computation module.The data collect module mainly achieve the source data reception and integration of the reservoir group,and convert it into the format that support by the large-scale database Hbase.The data storage module mainly achieve the data storage of the reservoir group schedule by HDFS,the collecting database stored by the Hbase and the storage of the inner data and the computation results that required for the Spark cluster are used by the HDFS.The reservoir operation computation module is the hard core of the whole system,which involve a series of complex operation.This would lead to the characteristics of large data volume and the complex calculation,the calculation can achieve by the parallel operation of the Spark that based on RDD and major steps including the RDD data cluster creation,the population initialization,the parallelization fitness calculation and so on.System operation results show that the multi-objective optimization optimal operation system for reservoirs based on Spark not only accelerated the speed of operation and the average annual electric energy production of the reservoirs also improved.