A Research On Distributed Optimal Power Flow Calculation Of Large Scale Distribution Network In Cloud Environment

The smart grid improves the collection and real-time supervision ability of related data by installing amounts of sensors and data-collecting devices in order to realize the smart power transmission and distribution,which makes it become an inevitable trend of power grid in the future.However,the operation of smart grid has led an explosive growth of those collected data and has been equipped with characteristics of Big Data.Among the optimal power flow calculation of current large-scale power networks,when the traditional calculation methods meet the power-system data equipped with big data,some disadvantages may appear which makes it difficult to meet the real-time calculation of the smart grid,such as the slow calculation and low operational efficiency;However,the most existing parallel computing methods are used in professional parallel machines and have low cost performance.Therefore,how to realize the optimal power flow calculation with high speed and high cost performance has become an important issue which needs to be solved during the development of the smart grid.This paper studies the parallel computing methods of the optimal power flow for large-scale power-distribution networks under the cloud environment.With the help of the Map-Reduce distributed parallel programing framework,the proposed methods can be used in the Hadoop group with higher cost performance.Specifically,this paper firstly puts forward the performance model of the optimal power flow calculation which faces the Map-Reduce framework.This model can analyze and quantify the execution time of the algorithms under different groups,and guide the decomposition and computing granularity of the power grid.Based on this performance model,this paper puts forward the load balancing algorithm for the optimal power flow calculation.Under the given group resources,it will determine the optimal algorithm decomposition method and computing granularity through the stimulated annealing method;and it will realize the load balance through the feeder restructuring method,which may optimize the computing speed and efficiency of the optimal power flow under cloud environment.For experiment,this paper has compared the proposed method and the serial optimal power flow calculation.The results shows that the former method is better than the traditional one which will save 68.3% of time.At the same time,this paper has also verified the computing time of the optimal power flow algorithm under the situation of load balance and load imbalance.The experimental data shows that the load balancing algorithm proposed in this paper can save 43.7% of the computing time for the optimal power flow compared with the load imbalance algorithm.