Studies On High-Risk Cascading Failure Screening For Large-Scale Hybrid AC/DC Grids

Due to the large scale of ultra-high voltage hybrid AC/DC power grids in our country,there is a risk that local failures may lead to global cascading failures and largescale blackouts.For hybrid AC/DC power grid,the local AC cascading failures in the receiving-end system will greatly reduce the voltage support capacity and easily lead to DC blocking failures.The HVDC transmission system transmits huge power.DC blocking failure will lead to a lot of load cutting of the power grid at the receiving-end,and seriously threaten the safe and stable operation of the power grid.Through the highrisk cascading failure screening,cascading failures with high probability and serious impact on power grid security can be identified,which can provide support for the development of security risk early warning and defense plan.It can effectively prevent the occurrence of major blackouts.In large-scale hybrid AC/DC power grid,there are many potential failure evolution paths.The access of a lot of renewable generation aggravates the uncertainty of power flow and significantly increase the complexity of cascading failure evolution,resulting in low efficiency of cascading failure screening and difficulty in timely warning and prevention of power grid security risks.Therefore,in-depth research on rapid and effective screening methods for high-risk cascading failure can provide sufficient time for the formulation of security risk warning and defense plans,which is an important basis for ensuring the safe operation of hybrid AC/DC power grid.In this dissertation,the risk of successive commutation failure caused by cascading failures is taken as the evaluation index to analyze the impact of different initial failures on AC/DC power grid and obtain high-risk initial failures.Considering the uncertainty of power flow caused by large-scale renewable generation and the defensive control actions adopted by the control system,high-risk cascading failures are screened out.By using machine learning techniques such as stacked denoised autoencoder and deep forest,and sequential importance sampling,stochastic response surface method,mathematical programming,the research work of hybrid AC/DC power grid cascading failure risk assessment,high-risk cascading failure screening considering the renewable generation uncertainty and the effects of defensive control actions is completed.The main contributions and innovations of this dissertation are summarized as follows:(1)An initial failure assessment method based on sequential importance sampling(SIS)and stacked denoised autoencoder(SDAE)is proposed to analyze the SCF risk caused by AC cascading failures with different initial failures.From the perspective of the AC cascading failure influence on DC system,the SCF risk index is established considering the cascading failure probability and SCF duration.In order to generate the cascading failure sample for calculating this index,a cascading failure sampling method based on SIS is proposed,which can effectively reduce the estimation variance and significantly reduce the sample size.In order to evaluate the SCF duration caused by each cascading failure sample,an evaluation network based on SDAE is established.The evaluation network selected topology features related to AC grid structure and short-circuit fault location as inputs,which could quickly evaluate the SCF duration caused by cascading failures.Simulation results of simple Shandong grid demonstrate that the cascading failure sampling method based on SIS can reduce sample size while ensuring sampling reliability,which saves the generation time of cascading failure samples.The estimation network based on SDAE can significantly improve the calculation efficiency of SCF duration.The proposed method can quickly evaluate the SCF risk caused by AC cascading failures,and screen out high-risk initial failures that affect the safe operation of AC/DC power grids.(2)Considering the power flow uncertainty of both power source and loads caused by large scale integration of renewable generation,a high risk cascading failure screening method is proposed.In order to quickly obtain the probability distribution of line load flow in the cascading failure process,an improved stochastic response surface method combining with distribution factor(SRSM-DF)is established.This method builds chaotic polynomial based on stochastic response surface method(SRSM)to characterize the mapping relationship between input variables and line load flow.In the subsequent process of cascading failure search,the polynomial coefficients are updated by distribution factor method to accelerate the calculation.Then the SRSM-DF calculation error judgment method based on deep forest is proposed.When the error is large,the SRSM based on AC power flow is used to update polynomial coefficients.Finally,a two-stage high-risk cascading failure screening strategy is established.The screening process is decomposed into high probability cascading failure screening and cascading failure risk calculation,which can rapidly screen high-risk cascading failures.Simulation results of east China power grid demonstrate that the proposed method can balance both the calculation speed and precision.The two-stage screening strategy reduces the complexity of cascading failure screening,and takes DC power loss into account when calculating cascading failure risk,which can rapidly screen high-risk cascading failures considering power flow uncertainty for hybrid AC/DC power grid.(3)Considering the influence of power grid defensive control actions in the development process of cascading failures,a screening and evolution mode analysis method for AC/DC high-risk cascading failures is proposed.In this dissertation,a mixed integer nonlinear programming model(MINLP)is established and linearized to obtain a mixed integer linear programming model,which simulates the defensive control actions.The two-stage screening strategy is improved.In the first stage,a game tree search strategy is established,in which failure selection and defensive control actions are carried out alternately.In the second stage,considering the cost of defensive control actions and the cost of emergency control actions,the risk of cascading failure is calculated.Based on the high risk cascading failure set,the weighted PrefixSpan algorithm is used to identify the key evolution patterns of failure set,and the weak links of power grid defensive control are analyzed.Simulation results of east China grid and west China grid demonstrate that compared with the existing methods,the established defensive control model has better defensive effect of cascading failure risk.However,the control ability of power grid is limited and unbalanced,which leads to some cascading failures still have high risk after the defensive control actions.The proposed evolution mode analysis method can identify the weakness of cascading failure defensive control.It can assist the dispatch and control system to formulate a more reasonable defense strategy,and improve the power grid security risk defense ability.(4)The cascading failure search software of hybrid AC/DC power grid is developed and applied in east China power grid.The software mainly includes power grid data conversion and cascading failure search.The data conversion function is to analyze the obtained power flow and dynamic model data,which is convenient for other modules to use.Based on the aforementioned research content,the cascading failure search function searches high-risk AC system cascading failures for power grid close to DC system,which provides a basis for power grid operation status evaluation and control decision making.