Power System Repair Scheduling Technique Considering Extreme Weather Events And Its Application In Hardening Planning

Nowadays,ensuring the normal operation of power systems under natural disasters has become a top priority for all countries.China is a country with vast territory and distinct topography,where numerous extreme events have been reported over the past few decades.Thus,it is urgent for electric utility companies to study the restoration strategies under extreme weather events.Power system infrastructures usually suffers from large-scale damages under extreme weather events.Given limited resources,it is impossible to swiftly restore the load and contain the social-economic loss without efficient repair scheduling strategies.Unfortunately,few research efforts have been devoted to this field as of today.Sponsored by the National Science Fund for Distinguished Young Scholars of China(No.51725701),this dissertation carries out a series of studies on power system repair scheduling to improve the power system resilience.Key issues studied in this dissertation include the sequence-dependency and uncertainty of repair time,as well as the incompleteness of failure information.The repair crews are usually exposed to varying weather conditions during the repair of power infrastructures.Besides,the transportation of repair resources must be considered in the repair process.Hence,the repair time of failure components is dependent on the sequence of repair tasks in which they are executed.To minimize the accumulated load pick-up deficit during post-disaster recovery,a repair scheduling model considering sequence-dependent repair time is established.The model of task lead time considering the routing of repair units and model of task execution time considering time-varying repair rate are proposed.The time-varying repair time function is then linearized and embedded in the model in the form of constraints.To realize the model application in large-scale system,a two-stage optimization algorithm is proposed.The method decouples the search of task sequences and the optimization of components activation,and solves them by a bi-pheromone search method and an iterative right shift technique,respectively.The model is validated on IEEE 30-bus and 118-bus systems.The results show that the efficiency of the proposed method is higher than directly solving the mathematical programming formulation.The consideration of repair time sequence dependency has a nonnegligible impact on the system functionality loss after the occurrence of disasters.Due to the subjectivity and inaccuracy of field investigation,there exists great uncertainties in the repair time estimation.To tackle the issue,a dynamic repair scheduling model based on stochastic Markovian process is established.First,the concept of decision point is proposed,together with the definitions of scheduling states,the state transfer function,as well as the cost function considering the uncertainty of state transition paths.The model cannot be directly solved by enumerating the state transition paths.Therefore,a solution method based on look-ahead strategy approximation is tailor-made.First,a simplified model assuming equal time interval between decision points is established,which yield a priority list with only trivial computation time.The priority list is later combined with the stage-dependent balancing coefficients to approximate the cost-to-go function.Finally,the repair task assigned at each decision point is derived according to the approximated function.The accuracy and efficiency of this method can be further improved by increasing the look-ahead depth and creating a lookup table of system topology.The results on IEEE-14 and IEEE 118-bus systems show that the proposed method result in close-to-optimal loss with uncertain repair time.Compared with the traditional stochastic optimization method,the proposed model can be solved within a minute,which proved its applicability in online scheduling.The communication/transportation systems may be disrupted in extreme weather conditions.In this situation,it is hard to assess the damage severity of system components in an ad-hoc manner through field investigation.A global repair rule set(GRRS)is designed to handle the problem when only partial failure information is available.GRRS can be conveniently applied to any failure cases.Taking the historical/simulated failure cases as input,each individual global repair rule was obtained by Ada Rank learning-to-rank training algorithm,which regards the outputs of the dynamic repair scheduling model as training inputs.Then,through the clustering of training cases,the global repair rule is split and retrained to form a rule set for dealing with different types of failure scenarios.Finally,a method based on multi-label K-nearest neighbor is proposed to select the optimal rule from the rule set and realize the repair scheduling given incomplete failure information.The results on IEEE 118-bus system show that the proposed global repair rule set results in only small precision loss when less failure information is available.Besides,the method does not rely on the mathematical optimization model when used in an actual failure case.This brings huge advantage in computation efficiency and enables its application in the pre-disaster resource allocation problem,in which the analysis of numerous post-disaster failure cases is required.The existing power system planning models did not consider the post-disaster restoration.To bridge the research gap,a four-level data-driven power system hardening planning model is proposed along with its solution method.The model determines the reliability parameter enhancement scheme for each power system component given the total planning budget.On the top level,the optimal combination of hardening scheme is obtained.On the second level,the component survivability is used to represent the spatial distribution of weather intensity under different extreme events,whose probability distribution is characterized by an ambiguity set to deal with the insufficiency of historical data.On the third level,an ambiguity set for failure state probability is established.The survivability delivered from the upper level is used to define the marginal probability of the ambiguity set,thereby eliminating the highly nonlinear terms of reliability index.The fourth level solves the repair scheduling problem given a deterministic failure state.Furthermore,a cutting-plane constraint method and a failure state sparsification technique are developed to solve the model.The results on IEEE RTS-79 and IEEE 118-bus systems indicate that,compared with the existing models,the proposed model can effectively reduce the total workload of power system repair and improve the recovery rate following the occurrence of extreme weather events.Meanwhile,the accuracy of planning decisions increases with the amount of available data.