State Estimation Of The Power System For Measurement Data Of Large Error

With the continuous development of economic life,production and living electricity load is increasing.The power system needs to be constantly adjusted to accommodate the ever-increasing amount of load.As a result,the grid structure is becoming more and more complicated.We need to keep abreast of the operational status of the power system,that is,to monitor the voltage amplitude and phase of each node of the grid.For the time being,there are two approaches to power system measurement:the Data Acquisition and Monitoring System(SCADA)and the Wide Area Measurement System(WAMS)based on the phasor measurement unit(PMU).Although WAMS has many advantages for SCADA,we still need SCADA due to technical and cost issues.Power system state estimation is an important part of SCADA.In order to understand the operating state of the whole power system,we usually need to use the weighted least squares power system state estimation algorithm,but the algorithm has some flaws,the most important one is the obvious error is that large measurement error data will affect the weighted least squares Estimation accuracy of the estimator.The commonly used weighted least squares weight selection method cannot effectively solve this obvious flaw.At present,the researchers only validate and eliminate the single large error measurement data,but when the number of large error data is redundant,these existing large error data recognition algorithms will not work effectively.This thesis mainly includes:First,several traditional weighted least squares state estimation algorithms and bad data identification algorithms are introduced in detail.And verify the feasibility of weighted least squares state estimation and bad data identification algorithm through simulation,and give the limitations of these algorithms.Second,a real-time weight adjustment weighted least squares power system state estimation algorithm is proposed.The algorithm is divided into two layers.In the first layer,the algorithm uses the inverse matrix as the weight of the state-based least-squares state estimation algorithm,which in turn estimates the entire power system so that we can obtain the estimated value of each measurement.For the second layer,the proposed algorithm uses The calculated value of the measured data obtained in the first layer is compared with the actual measured data to calculate a new weight value so as to calculate the second power system state estimation.The proposed algorithm benefits from the practical engineering characteristics of ’small error with large gain and large error with small gain’,and uses the better data as much as possible while effectively avoiding the impact of large errors on the accuracy of the state estimator Improve the estimation accuracy of power system state estimator.Third,a parallel adaptive linear neural network method is proposed.The method has the characteristics of two-layer structure and parallel processing.The left artificial neural network adopts a fixed,large step-size least mean squares(LMS)algorithm to adjust the weights and make the convergence faster.On the other hand,the right artificial neural network uses variable steps to achieve the minimum steady-state error.The feasibility of this method is verified on the IEEE 30-bus network and compared with the PSO(particle swarm optimization)and the simulation results of a single adaptive linear neural network.The simulation results show that the PSO algorithm has a faster convergence rate,but the estimation accuracy is poor,the parallel adaptive neural network convergence speed is in between,and the estimation accuracy is better than PSO.Therefore,parallel adaptive neural network improves the estimation accuracy and solves the convergence speed problem well.