Research On Power Generation Dispatching And System Design Of Cascade Hydropower Station Based On Parallel Computing Technology

With the establishment of the 13 hydropower bases,China forms the world’s largest connected hydropower system.Allocating water resources and optimizing the operation and management of reservoir group operation have gradually become a focus of research.With the increasing scale of the hydropower systems and complex scheduling requirements,the modeling and solving matter of reservoirs dispatching has become incredibly difficult.The traditional mathematical programming method and heuristic intelligence algorithm are inadequate to meet the realistic needs of engineering.The improvement and innovation of algorithms as well as the introduction of the latest computing technology are urgent for research.Therefore,this dissertation takes the four reservoirs Xiluodu,Xiangjiaba,Three Gorges and Gezhouba in the Lower Reaches of the Jinsha River-Three Gorges Cascade as the research object,combines the parallel and distributed computing technology with the optimal operation of cascade hydropower stations,and proposes the Fork/Join based Parallel-Ant Colony Optimization(FP-ACO)algorithm adapted to the single-machine multi-core environment.Further,this dissertation comes up with a big data computing framework,Spark,based Distributed-Ant Colony Optimization(SD-ACO)algorithm.Based on all the above,this dissertation designs and develops a unified data management platform suitable for the distributed water resources management system.The contents of this work and the related innovative results are as followed:(1)Focusing on the problem of a long calculating time for the cascade hydropower station optimal dispatching,this dissertation constructs a mid-and-long term power generation optimal dispatching model for cascade hydropower stations to maximize the power generated.This dissertation also proposes the FP-ACO algorithm based on the Fork/Join framework through analyzing the parallel computing technology and the Fork/Join parallel framework with an example of the Ant Colony Optimization algorithm.The results of the research of the four reservoirs in the Lower Reaches of the Jinsha River-Three Gorges Cascade show that the FP-ACO algorithm can maximize the utility of the computing resources of parallel computer,which effectively reduce the calculating time while improving the quality of the optimization results,and provide an efficient algorithm supporting the optimal dispatching of cascade hydropower stations.(2)For the concerns of the FP-ACO algorithm’s limited efficiency improvement and insufficiency of single-machine memory,this dissertation proposes the SD-ACO algorithm based on Spark,a big data computing framework and the parallelly design strategy of the FP-ACO algorithm.The experimental results of the four reservoirs show that under a single-machine environment,the optimization result and calculating time of the SD-ACO algorithm are superior to the FP-ACO algorithm.Under a cluster environment,the SD-ACO algorithm has an excellent parallel speedup ratio and scalability with an efficient implement of the cluster storage and computing resources,reduces the operation and system response time,and improves the applied synchronization.The SD-ACO algorithm has the practical application value in engineering.(3)Focusing on the issues of the highly coupled model algorithm and data management as well as the slow interactive response in the traditional water resources management system,this dissertation constructs a unified data management platform with Representational State Transfer(REST)architecture based on the database technology and Spring application framework.The effect of "The Decision Support System of Water Resources Management for the Lower Reaches of the Jinsha River-Three Gorges Cascade Hydropower Plants" applied in the Changjiang River Three Gorges Cascade Dispatch Center shows that through the unified,standardized,and simplified data management toward the massive heterogeneous water resources data,the platform not just reduces the development cycle of scheduling model but also effectively improves the operation of the business model.This platform provides a new data management integration design idea for the water resources management system.