Research On Line Parameter Identification In Distribution Network With PMU Measurement

The flexible,safe and reliable operation of distribution networks is of great significance.Compared with the existing supervisory control and data acquisition(SCADA)systems,synchronous phasor measurement unit(PMU)provides fast,accurate and reliable phasor measurement data for distribution networks,which can greatly improve the observability level of distribution networks.However,due to technical and economic constraints,it is difficult for the distribution networks to install PMU devices at all nodes in a short time.This leads to the differences in PMU placement conditions and technology requirements among different distribution networks.In order to comprehensively improve the observability,measurability and controllability of distribution networks,the researches of application technology based on PMU measurements have attracted extensive attentions.At the same time,the researches and developments of chip based and low-cost PMU devices will help to realize the wide deployments and applications of PMUs in distribution networks in the future.For the distribution networks with different PMU measurement placement conditions,the researches on line parameter identification with PMU measurements and state estimation with PMU measurements at all nodes are carried out in the dissertation.The main research findings are shown below:(1)For distribution networks in which PMU measurements do not meet the observability requirements,this dissertation proposes an identification method of distribution network line parameters based on impedance classification learning.Firstly,the electrical correlation relationships between line parameters and PMU phase measurements are considered when PMU measurements do not meet the observability requirements.The bases for classifying the line impedance are established.The data processing method which is helpful to extract the characteristics of measurement data is studied.An identification method of distribution network line parameters based on impedance classification learning is proposed.It solves the problem that the traditional methods cannot identify the line parameters when PMU measurements do not meet the observability requirements.The interpretability of the decision logic of the proposed method is given,which can be helpful to analyze the identification results.Secondly,the applicable conditions of whether the proposed method can identify the line parameters are given.This dissertation analyzes the interrelationships between the proposed method and PMU placement schemes.Finally,simulation experiments,hardware experiments and demonstration projects verify the effectiveness of the proposed method.(2)For distribution networks with time stamp mismatch of PMU and SCADA measurements,this dissertation analyzes the relationships between the measurement population and the measurement samples in the sampling technologies under the situation of time stamp mismatch,and proposes a PMU and SCADA hybrid data-driven detection and identification method of wrong line parameters in distribution grids.For line parameter wrong detection,a probability parameter identification index(PPII)of the conversion measurement mean value and the actual measurement mean value is proposed.The maximum value of PPII under the condition of the accurate line impedance is deduced to determine its threshold.Through the comparison between PPII and its threshold,the detection of wrong line parameters is realized.For line parameter identification,the solving process of standard deviation of power loss is simplified,which can be used for obtaining the power loss samples without time stamps.The probability density functions of these samples and PMU measurement data are analyzed.A power-loss chronological probability density function and a power-loss probability density function with the PMU time stamps are proposed to obtain the power-loss data with the chronological correlations and PMU time stamp characteristics.Using the power-loss data and PMU current measurements,the accuracy of line parameter identification is promoted.The effectiveness and practicability of the proposed method are verified by simulation experiments and hardware experiments when the time stamps of PMU and SCADA measurements do not match.(3)For distribution networks in which PMU measurements meet the observability requirements but may contain outliers,a distribution network line parameter identification method based on robust total least squares(RTLS)is proposed.In order to reduce the influence of the different accuracy of the coefficient matrix and observation data matrix on least trimmed squares(LTS)method,the objective function of the LTS method is improved.A RTLS method is proposed.Based on “concentration”-step of FAST-LTS algorithm,the solution flow of the RTLS method is given to reduce the calculation time of this method.The RTLS method is applied to line parameter identification,which improves the accuracy of line parameter identification when PMU measurement outliers and the different accuracy of the coefficient matrix and observation data matrix coexist.The effectiveness and stability of the proposed identification method are verified by IEEE33 and IEEE69 node system simulations.(4)For distribution networks in which PMU measurements are at all nodes but may contain outliers,a method of equivalence between long short-term memory and Metropolis-Hastings(E-LM)is proposed to reconstruct the measurement outliers.Then a distribution system state estimation scheme based on the E-LM method is built.Firstly,the target distribution based on time series measurement data is deduced,which can obtain the data with the chronological correlations to reconstruct the measurement outliers.Secondly,the long short-term memory(LSTM)network-based proposed distribution is established to be equivalent to target distribution of Metropolis-Hastings(MH)algorithm.The theory for the acceptance condition of MH algorithm can judge LSTM network prediction is revealed.The E-LM method is proposed.This method can decrease the number of the futile iterations of MH sampling algorithm and can realize the judgments of the prediction values of LSTM networks independent of the real values of measurements.This method reconstructs the measurement outliers.Then,a distribution system state estimation scheme based on the E-LM method is built to improve the accuracy of state estimation.Finally,simulation experiments verify the effectiveness of the proposed E-LM method for reconstructing outliers and the proposed scheme for improving the state estimation accuracy.