Research On Sequential Monte Carlo Simulation For Reliability Evaluation Of Bulk Power System Based On Cross Entropy

Because important reference from the angle of probability risk and reliability cost / benefit is provided for the planning and operation of power system, Power system reliability evaluation becomes one of the research highlights in power industry. In power system reliability evaluation, Sequential Monte Carlo simulation(sequential MCS) effectively simulate the time sequence characteristics of the system operation and fault recovery process. Therefore, not only the expected value of the reliability indexes is got, but also the probability density distribution is achieved. It is widely concerned. However, its applicability in engineering is severely limited severely by the contradiction between accuracy and computation cost. Hence, it is of great important academic and practical significance to study the convergence acceleration method of sequential MCS.The variance reduction technique is an effective method to speed up the convergence of Monte Carlo simulation, among which the important sampling method has been more studied. The Cross entropy algorithm, as a new important sampling method, has achieved good convergence effect in the non-sequential Monte Carlo simulation(non-sequential MCS) of power system reliability. But in the sequential MCS framework, it is worth studying how to combination the cross entropy algorithm and sequential MCS and accurately calculate the expected value and probability distribution of the reliability indexes. The main research contents include the following:The basic principle of Monte Carlo simulation method based on cross entropy is explained. In the non-sequential MCS, the components of the two state reliability model is submitted to the binomial distribution. The optimal importance sampling probability density function(IS-PDF) parameters can be effective estimated by the cross entropy algorithm, realizing the purpose of the variance reduction and convergence acceleration.On the basis of non-sequential MCS based on cross entropy, the sequential MCS based on cross entropy is put forward to calculate the optimal IS-PDF parameters. The detailed derivation of the optimal unavailability is given in this paper. Fatherly, three different calculation methods which is used to calculate failure rate and repair rate is also given. As a result, the failure rate and repair rate are used to evaluate power system reliability. Compared with other methods, the method in this paper makes full use of system state samples during parameters optimization to count reliability indexes. So the reliability assessment of the overall process is within the framework of the sequential MCS, which fatherly improve the efficiency of reliability assessment.The effect of the system state sequence by the sequential MCS based on cross entropy must be modified by the likelihood ratio, so four kinds of the likelihood ratios used to correct the system state duration and the likelihood ratios used to correct the system loss of load frequency are derived. Moreover, if the system state sequence is corrected only by the ratio of system state probability, the sequential Monte Carlo simulation is proved to be non-sequential Monte Carlo simulation. In order to obtain the accurate probability distribution of the reliability index, the basic principle of the parameter estimation of the optimal importance sampling probability density function based on cross entropy is proposed, which is that the system fault state frequency under the optimal reliability parameters stay less changed.