High-availabilty Evaluation For Large-scale System Using Error Back Propagation Neural Network

With the development of the times,the modern office and the automation of the enterprise are inseparable from the information system,and the information system becomes crucial to an enterprise.Not only for the Internet companies,traditional businesses are also as far as possible the traditional business are transferred to the information platform.The highly automated operation of the information platform has greatly accelerated the traditional business operations of traditional enterprises while saving a lot of labor costs.A large number of business was transferred to the information platform,then the maintenance of information system stability is a crucial issue.High availability is a measure of the stability of the system an important indicator of the system it represents the normal operation of the time ratio.With the continuous increase of the power grid enterprise system,the complexity of the internal information system is gradually increasing.The more complex the system,the more failures that may occur between the components,which also gives the entire system maintenance The availability poses a huge challenge.To maintain availability,you must first be able to calculate availability accurately,so grid companies urgently need a model that accurately estimates the availability of the entire system.This article is divided into two parts about the usability assessment model,and part of the usability assessment of a single component.Aiming at the usability assessment of a single component,this paper presents a fault rate estimation algorithm based on BP neural network.Through the study of historical fault data,it can be concluded that the failure rate function of a single component has a great influence on the real-time running status of a single component and the continuous running time The relationship between.Compared with the traditional algorithm,we can get only the relationship between the failure rate function and the running time.The failure rate function proposed in this paper can not only get the relationship between the failure rate function and the running time,but also introduce various parameters in real-time running.In the process of training BP neural network,the momentum method is introduced to dynamically modify the learning rate.Combined with the simulated annealing method,the algorithm avoids over fitting.Experiments show that,due to the introduction of more learning variables,the obtained failure rate function has a higher fitting degree on the test set than the traditional failure rate function.The other part is based on the usability assessment of the system network structure.Aiming at the evaluation of system network availability,this paper proposes a system disastrousness evaluation algorithm of set disjoint minimum path set and Monte Carlo simulation.Papers for a complex system network using search tree generation algorithm traversing the entire path,and draw the minimum point set of the starting point and the final point,and then proposed a disjoint minimum path set algorithm,the intersection of the road set decomposition is called not Intersected set of paths,making their probabilities independent,and then combined with the Monte Carlo simulation algorithm,calculates the availability of the entire network based on the disjoint sets of disjoint sets.Experiments show that compared with the traditional network availability algorithm based on Markov transition matrices,the system usability evaluation algorithm based on disjoint minimum path set and Monte Carlo simulation proves that the proposed method is faster and is not affected by the structure of the network Limit,the scope of application is high...