Joint Topology/State Estimation For Active Distribution Systems Considering Multi-source Measurements

The growing penetration of renewable energy and participation of controllable loads in demand response introduces increasing source-network-load uncertainties to active distribution systems(ADSs).Under this circumstance,an accurate and efficient distribution system state estimation(DSSE)is essential to achieve the secure and optimal operation of ADSs.Considering the high cost,it is infeasible to install real-time measurements in overall ADSs.This will bring about low measurement redundancy or even unobservability of ADSs,thus reducing the estimation accuracy of DSSE.Moreover,aiming at economic and reliable operation,the network structures for ADSs are usually reconfigured.In this condition,their network topology may be erroneous due to both the insufficient topology identification devices with possible errors and the overdue calibration of incorrect topologies.This erroneous network topology will result in unreliable state estimates of DSSE that may influence advanced applications,thus challenging the security and economic operation of ADS.This dissertation investigates an accurate and efficient joint topology/state estimation method for ADS from the following three aspects:1)pseudo-measurement construction,2)estimation method proposing,and 3)extension of forecasting-aided state estimation(FASE)problem formulation.The estimation accuracy enhancement and calibration of network topology can be achieved by utilizing useful information lying in low-resolution measurements from smart meters(SMs),sequential operation states of ADSs,and the topology transition probability distributions.The main contributions and innovations of this dissertation are summarized as follows:(1)A high-precision pseudo-measurement construction method is proposed by utilizing low-resolution energy measurements provided by SMs,which ensures the system observability and provides measurement redundancy.Firstly,a hierarchical communication infrastructure is designed for ADSs to reduce the demand of massive SM data for the bandwidth of communication channels.Secondly,the nodal energy measurements provided by SMs are disaggregated by load/DG components via the online-learning dynamic fixed share(DFS)algorithm considering their different characteristics.Thirdly,based on the identified energy consumption of each component and their models considering temporal correlation,the detailed accurate pseudo-measurements in terms of nodal power injections are generated.Finally,the fusion method for multi-source measurements is proposed to address the inconsistency issue of measurement update resolutions.Simulation results validate that the generated pseudo-measurements can accurately track the variation of nodal power injections and thus improves the estimation accuracy of DSSE.(2)An adaptive particle filter(APF)method is proposed for forecasting-aided state estimation of ADS by using randomized QMC(RQMC)sampling,which enables the accurate and efficient state awareness of ADSs in case of low measurement redundancy.Firstly,a FASE-based problem is formulated based on the state space model of ADSs.This problem formulation can make use of temporal correlations in state evolution to reduce the sensitivity of DSSE on measurement redundancy.Secondly,an RQMC-PF estimator is proposed for nonlinear ADSs with non-Gaussian noises,which incorporates the RQMC sampling with high convergence rate based on scrambled low-discrepancy Sobol’sequence.The robustness of the RQMC-PF against measurement outliers is enhanced by applying anti-zero bias modification.Thirdly,to further make a trade-off between estimation accuracy and computational efficiency,the number of particles used in the RQMC-PF is adaptively determined with the root mean square of the estimated state covariance,yielding an APF.Finally,a multi-area state estimation scheme is designed to enhance the scalability of the proposed APF algorithm.Simulation results demonstrate that the proposed RQMC-APF method can efficiently provide accurate estimates for ADSs as well as achieve good robustness against measurement outliers and scalability for large-scale ADSs.(3)A forecasting-aided joint topology/state estimation method is proposed based on Bayesian non-parametric model for distribution systems considering topology changes in the presence of limited topology identification devices.The topology change process is modeled based on the Dirichlet-process hidden-Markov-model(DP-HMM),with which the joint topology/state estimation problem is formulated.By using probabilistic neural networks,the probabilities for ON/OFF states of each switch can be calculated,which enables establishing the base distribution for DP that makes DP applicable.Aiming at efficiency enhancement,a rule-based candidate set reduction method is proposed for the possible topologies considering the security and reliability constraints of ADSs to improve topology sampling validity.The RQMC-APF is utilized to resolve this joint estimation problem,whose particles are generated by steps based on the dependency of the augmented state variables(i.e.,the topology-related ON/OFF variables and continuous nodal voltages).In this way,the topology of ADSs can be calibrated while efficiently gaining their state awareness.Simulation results show that the proposed DP-HMM-PF-based joint estimation method can effectively calibrate the network topologies of ADSs while accurately gaining their state awareness.The proposed method also has good robustness against measurement outliers and scalability.In summary,this thesis focuses on the joint topology/state estimation for ADSs considering multi-source measurements,and has carried out comprehensive and in-depth research work.It can efficiently enables accurate and real-time state awareness for ADSs while calibrating the topology model in time,which provides data support for advanced applications in distribution management systems.