Research On Impact Of Load-sharing Effect On Reliability Of Converter Valve And Probabilistic Load Flow Simulation Algorithm

In recent years, high-power electronic devices become widely applied to the field of power system, especially for FACTS and HVDC technology. In long-distance large-capacity transmission, different frequencies interconnection, enhancement of the controllability of the AC power grid and maximize the power grid transmission capacity, etc., high-power electronic devices play an irreplaceable role. As a new generation of HVDC Transmission technology, VSC-HVDC introduced all-controlling power electronics device, voltage source converters and PWM control technology, solved many problems which exist in conventional HVDC. However, due to the immaturity in their own development, complex structure and the tie in closely with the grid, reliability of high-power electronic devices and the impact of the power system on its reliability should never be neglected. Existing VSC-HVDC reliability model, especially for converter valve, is not perfect and requires to be improved. It is difficult to reflect the impact of the changing work environment on the reliability of power electronic devices by the model based on fixed failure rate and analytic method. This paper introduces the load-sharing effect for reliability modeling of the VSC-HVDC converter valve, proposes the correspond disposal methods according to the characteristic between voltage and current load-sharing effect.Meanwhile, equivalent model proposed for current load-sharing effect takes the varying power flow into account. Therefore, probabilistic load flow (PLF) based on simulation is need for equivalent current load-sharing model. Due to large calculation quantities in Monte Carlo simulation (MCS) method, extended Latin hypercube sampling (ELHS) is proposed in this paper for PLF evaluation to improve computation efficiency and reduce calculation time.This paper conducts the following research based on the above:(1) Based on the theoretical basis of the existing research on reliability power electronic equipment, the reliability model of VSC-HVDC converter valve using the k-out-of-n:G model are presented, and the quantitative relationship between redundant set and reliability indicators is discussed.(2) Due to the disadvantages of the existing fixed failure rate reliability modeling, the proposed converter valve reliability model based on the load effect is presented. Instead of analytical algorithm, Monte monteCarlo simulation is used to solve the reliability model to explore the influence on the reliability indexes with the voltage load-sharing effect through the aspect of the unrepairable and repairable models. Calculation and study on the spare parts reserves strategy of the converter valve is presented.(3) Due to the disadvantages of the existing steady-state VSC-HVDC converter model, the converter equations are derived in the most general format and a generalized loss formula with the converter losses quadratically dependent on the converter current. This paper proposed a unified method for power flow calculation in the AC power systems with embedded VSC-HVDC system.(4) Based on the steady power flow model of VSC-HVDC, PLF evaluation with VSC-HVDC is proposed. According to the results, equivalent calculation of varying current is proposed to take varying power flow into account in reliability model of valve module.(5) In view of existing Latin Hypercube sampling (LHS) technique based on fixed sample sizes and the disadvantage that cannot be extended on the existing sampling points, the extended Latin hypercube sampling is presented. On the basis of the study of variance reduction mechanism of LHS, the existing error analysis and convergence criterion tool, this paper points out that the coefficient of variation cannot be used as convergence criterion for LHS, and proposes the relative change of the output variable estimator used as the convergence criterion for LHS.(6) ELHS is applied to the calculation of PLF. The numerical example has shown that the method is valid and effective. By comparing the probability of the absolute error in MCS, CLHS and ELHS, the efficiency of the proposed convergence criterion is validated. On the basis of a single expansion program, this paper presents the idea of multiple expansion programs for obtaining more accurate and detailed convergence trend to improve the accuracy of the proposed convergence criterion.