State estimation and bad data detection in electrical power systems
The thesis studies the subjects of bad data detection and state estimation in electrical power systems which are the processes whereby voltage, power flow and switch status measurements gathered continuously in real-time are used in conjunction with a model of the system to calculate the voltage levels at every node in the system. Traditionally the state estimation process requires two stages. The first stage is the pre-processing of the measurements by a bad data detector in an attempt to remove all the measurements which are grossly in error. The second is the calculation of the voltage levels by a state estimator from the remaining measurements which are likely to contain small random errors. Conventional state estimation algorithms are very sensitive to measurement errors, especially switch status errors, and unfortunately it is not possible to ensure that all the measurement errors are removed by the bad data detector. The thesis presents a new algorithm for state estimation utilising linear programming which is able to function in the presence of not only bad analogue measurements but also switch status measurement errors, thus removing the need for a bad data detector. The proposed method of state estimation is also able to include in its model of the system the individual busbars and bus-couplers within a substation. This feature enables the state estimation algorithm to process and provide additional network information thus leading to a more useful and reliable data base.