Power System State Estimation And Bad Data Detection And Identification Based On State-Forecasting

Power system state estimation is an important part of the energy management system of power system, and it plays an important role in intelligent analysis and decision-making of power grid dispatching. The traditional state estimation methods still face some difficulties. For example, if the measurement redundancy is inadequate, the result of state estimation of the system would not be ideal; or if there are bad data in the critical measurements of the system, the accuracy of the results of state estimation will be substantially reduced. That’s because the bad data in the critical measurements cannot be detected and corrected. Even worse, if the metering system or transmission system malfunction so that the estimator might be unable to obtain the critical measurements, it cannot work normally since the system is unobservable. State estimation based on state-forecasting, to some extents, can solve those problems above. The main work and results of this thesis are included below:1. Firstly, this thesis does in-depth research on the principle of the least squares support vector machine(LS-SVM) regression. Based on the least squares support vector machine regression model, an optimization scheme for the model parameters is proposed to improve the accuracy of the model:this thesis uses particle swarm optimization(PSO) algorithm, which is relatively mature at present, with the mean square error minimization as the objective, to optimized the related model parameters. And by adopting the system history state data to train the regression model, the system state-forecasting model is established.2. On the basis of system history state data, using the PSO-LS-SVR forecasting-model can obtain the forecasted state of the current system. After that, by using the measurement function the forecasted measurements of the current system can be obtained. For the problem of the result of state estimation is not ideal since the measurement redundancy deficiencies, and the system is unobservable owing to the critical measurements are not available, this thesis add the forecasted measurements as pseudo-measurements to the current system measurements. This method can perserve the measurements redundancy, so as to ensure the accuracy of the system state estimation. This method is proved to be effective through simulation experiments.3. In the context of the forecasted measurements is obtained by PSO-LS-SVM forecasting model, the anomalies is processed by innovation analysis. Due to the innovation analysis is performed before state estimation, it can avoid the bad data smearing effect and block identification efficiently. Also, it can detect the bad data in critical measurements correctly. Additionally, two kinds of combination of innovation analysis and residual analysis for the detection of sudden change are introduced and analyzed. These methods show their obvious superiority through simulation experiments.In this thesis, all the simulation experiments are performed on the platform of Matlab taking the standard transmission system of IEEE 14 nodes and IEEE 30 nodes as examples. Through the analysis and comparison with the traditional methods, the method this thesis demonstrated is proved its rationality, feasibility and effectiveness.