Abstract: The service restoration of distributed system is an important and complex task.The ability of service restoration has great effect on the reliability and safety of the distributed system.The restoration operation requires massive analysis for data from the system information,which can lead to high complexity and long operating period.In order to solve this problem,a service restoration algorithm based on ESN and MOA is presented.This algorithm combines ESN and NSGA-Ⅱ to accomplish the unsupervised learning for the information input which has no need for the output of the training sample.The experiment based on Matlab/Simulink has showed that this algorithm can make the system learn the restoration method effectively and make immediate response to the input information.