Research On The Bayesian Network Using Quantum Evolution Learning And Related Application

Quantum Evolutionary Algorithm (QEA) is one kind of new intelligent optimization algorithm, which is a combination of the quantum computing and evolutionary algorithm. In this thesis, an improved Quantum-inspired Evolutionary Algorithm (IQEA) is proposed. Compared with traditional QEA, it is not based on quantum-bits coding but based on real coding whilst the simulated annealing(SA) with roulette strategy is employed. Furthermore, since two new operators named optimized mutation and illegal figure modification are proposed and added in the proposed algorithm, it will be superior in solving the problem of Bayesian Network (BN) structure learning. In addition, a Bayesian Net based fault diagnosis model is proposed for electric meter fault diagnosis. The proposed algorithm will be utilized for optimizing and solving this model and a good accuracy of diagnosis is achieved.The main content of this thesis is as follows:Firstly, the principium and characteristics of QEA are deeply investigated while the principium and related research about Bayesian Network are discussed. In addition, the advantages of BN in fault diagnosis and the trend of related technology development are introduced.Secondly, based on the characteristics and current research about evolution algorithm (EA), a real coding based Quantum-inspired Evolutionary Algorithm (RQEA) is proposed. By using the advantages of other algorithms for reference, RQEA is significantly improved by using a synthesis of different algorithms. Furthermore, the logical derivation is introduced while the convergence of algorithm is verified mathematically.Thirdly, the principium of BN parameters learning and structure learning is introduced and an improved Quantum-inspired Evolutionary Algorithm for BN structure learning is proposed. The coding of proposed IQEA is firstly transformed from real coding to binary coding and a compact net structure is obtained by network learning operators such as genetic mutation, simulated annealing(SA) with roulette strategy and illegal figure modification. The proposed algorithm is verified at last by using Matlab software, which shows it can accelerate the convergence speed of BN learning and improve the training accuracy.Finally, the key feature of the fault of intelligent electric meter is introduced and a BN fault diagnosis model based on IQEA is proposed for diagnosing. For verifying the effectiveness of the proposed model, its advantages in solving system fault diagnosis are verified by MSBNx software.