Research Of Fault Diagnosis Method For Building Electrical

Electrical system is an extremely crucial part in any buildings,guaranteeing the normal function of the whole building.If electrical system is broken down,it could apparently cause a series of adverse consequences,e.g.the stop of electrical equipment and the data loss of computer.Moreover,the failure of electrical system would lead to more serious results in some special organizations such as financial institution,hospital and television station,where there are higher requirements in terms of electrical reliability.As we know,fault diagnosis is an efficient way to enhance the security of electrical system.Consequently,it is highly necessary and essential to design a reliable and stable fault diagnosis equipment for building electrical system.Motivated by the above problem,in this paper we research and design building electrical fault diagnosis model based on data.The contributions and novelties of this paper are summarized as follows:Firstly we design a new data acquisition system which is built and optimized according to Fault Simulating Platform of Building Electrical System.It is highly suitable for collecting the building electrical fault diagnosis data.Secondly we establish a model for building electrical fault diagnosis based on bayesian method.And naive bayes theory is used to determine the conditional probability distribution of bayesian network.The simulation test shows that fault diagnosis model based on bayesian can update conditional probability distribution through parameter learning process to improve the diagnostic accuracy,which can satisfy the requirements of large scale data fault diagnosis.Then a building electrical fault diagnosis model based on relevance vector machine is established,which not only has realized the good classification performance of small sample data,but also can reveal the uncertainty of the fault classification results through the form of the probability output.As simulation experiment results show that the building electrical fault diagnosis model based on RVM is suitable for small sample data classification problem.Finally,on account of noise has a marked impact on building electrical fault diagnosis model,we explore the application of sparse representation theory in building electrical fault diagnosis,and put forward a fault diagnosis model based on redundant dictionary.It is highly useful to provide the efficient and real-time fault diagnosis for building electrical system.