Service Restoration Algorithm For Energy Network Based On Machine Learning Technology

Smart grid is becoming the trend in development of power system development and smart distributed network is an important part of smart grid technology.The smart distributed network has fully functional service restoration system to improve the stability and reliability of the power system,distributed energy sources are also supported to be linked to the distributed network.Brilliant service restoration ability can help distributed network system can reduce the impact of the client to minimum degree when faults occur and improve user experience and the stability of the network.Service restoration is the task that the network system transfers loads from shortage areas to other feeders to restore loads through operating switches of the circuits and tie switches of different nodes when faults occur.The main task of service restoration is to find the best route to restore loads from shortage areas,this is a typical multi optimizing nonlinear problem.The main content and thesis of this article is listed below:(1): Introducing the structure and classes of distributed network.Summarizing and analyzing current research situation and technical difficulties of service restoration technology,introducing several frequently used methods for service restoration.(2): Introducing theory about machine learning,and relative technologies used on service restoration.Presenting a service restoration algorithm based on Echo-State Network/ESN.Making the distributed system to learn service restoration plan through brilliant dynamic characteristic and designing Tree Node Visiting method to keep radial structure of the network.(3): Combining Non-dominate sort genetic algorithm-ii/NSGA-II with the service restoration algorithm based on ESN,using the brilliant global searching and multi optimizing ability of NSGA-II to help the system carrying out unsupervised learning.This learning method doesn't need to collect training sample from historical data,therefore can save much time in the long term.(4): Using IEEE Standard 16 node distributed network as experiment platform to put the service restoration algorithm presented in this article into implementation.After learning different cases,the system could quickly respond properly to different fault information.