Self-organized Critical State Identification In UHVAC/UHVDCSending-side System With High Wind/Photovoltaic Power Penetration

New induction factors of large-scale blackouts emerge in UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration.The power flow distribution,voltage,frequency,dynamic behavior after failure and power angle of sending-side system are changed compared to traditional AC power system.Each blackout in the cascading failures has relevance and acts like a chain.Dispatcher nearly has no time and current scheduling control is hard to stop cascading failures after the trigger fault.Self-organized critical state is in a boundary situation before cascading failures.SOC identification can help to prevent and control the cascading failures.Against the above background,SOC identification in UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration is presented in this paper.The major works are summarized as follows:Impacts mechanism of UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration on SOC is presented based on their characteristics.On the one hand,wind/photovoltaic power have characteristics of fluctuation,high/low voltage cascading tripping and low-inertia.They can bring about voltage fluctuation which can't meet system's operational requirements.They can also increase the risk of wind turbine generators/pv module cascading tripping and low frequency oscillation.On the other hand,fault occurs in AC power system near inverter side can bring about DC commutation failure and DC block.It can lead to AC bus transient overvoltage and large-scale power flow transferring,after which could be cascading trips.The interaction of the above deteriorate normal running state and dynamic behavior after failure occurred in sending-side system.It weakens the ability against blackouts and promote evolving toward SOC state.Impact indexes of SOC considering characteristics of high wind/photovoltaic power penetration and UHVAC/UHVDC are extracted.Impact indexes of SOC in traditional power system are expended based on characteristics of high wind/photovoltaic power penetration and UHVAC/UHVDC.Partial correlation computational method is presented to compute the correlation between impact indexes and SOC.And impact indexes proves to be available in representing the evolution degree of SOC.Based on critical phase transformation and critical slowing down theory,Mann-Kendall test is used to compute the threshold values of power system's SOC indexes.And computing method of indexes'threshold values of power system SOC is presented.Identification method of SOC state in UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration based on semi supervised learning-radial basis function neural network is proposed.This method can both use unlabeled samples to improve learning performance and has optimal approximation and global optimization of traditional RBF neural network.So it takes less computation time and gets higher rate of correctness.SOC indexes and threshold values can create a small number of 1st kind of sample,2nd kind of sample and a large number of 3rd kind of sample.The cluster center,weight and width of SSL-RBF neural network can be computed with those training sample.SOC state identification is simulated in Jiuquan UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration.And it can provide theoretical and technical basis for real time prevention of large-scale blackouts.Using 2017 winter peak load running state of Gansu grid as initial state,some typical evolutionary branches are designed to make it evolve to SOC state.In the process a large number of 3rd kind of sample are created.And a small number of 1st kind of sample,2nd kind of sample are created with Mann-Kendall test and K-S test.With those training sample a neural network model can be built to identify SOC state in Jiuquan UHVAC/UHVDC sending-side system with high wind/photovoltaic power penetration.The result of SOC state and non-SOC state identification is verified 93%and 92%correct.The method has outstanding performance.