Dynamic state estimation and control of power systems

This thesis reports on an observer based approach, i.e., deterministic functional observer approach and probabilistic unscented Kalman lter and particle lter approaches to the problem of dynamic state estimation (DSE) of generators in power systems. The proposed DSE algorithms presented in this thesis are all decentralized in which only the measurement of the local phasor measurement units (PMUs) installed on the generator busbars are used. Consequently, any change in the power system topology does not aect the DSE process. The performance of the above mentioned probabilistic approaches have been tested on dierent IEEE test systems and are compared when the covariances of the noise in the PMU measurements are altered.

Furthermore, in this thesis, application of estimation techniques in load frequency control (LFC) of highly interconnected power systems is investigated. The proposed techniques for LFC are quasi-decentralised since the DSE based controllers need the measurement of tie-line powers (power in the transmission line that connects two dierent power system areas). The analysis and design of LFC controllers presented in this thesis is dierent to traditional methods previously reported in the literature, the method proposed in this thesis considers the entire network topology. The proposed schemes have been implemented on the IEEE 39-bus 10-generator 3-area test system considering noise in the PMU measurements and also generator parameter deviations from nominal values. The proposed probabilistic unscented transform based method, proposed in this thesis for LFC, takes into consideration noise in PMU measurements. Moreover, the advantages of functional observer based method in comparison to traditional state observer based method in LFC are demonstrated in this thesis.

The application of DSE to detect faults in dynamical systems is also investigated in this thesis. A functional observer based fault detection technique for dynamical systems with application to wind turbines is proposed. The proposed scheme has the ability to detect faults independent of chosen observer parameters. The theoretical development of the fault detection algorithm and its application to wind turbines is presented in this thesis.

The theoretical development presented in this thesis on DSE algorithms and DSE based LFC controllers hinges on the state of the art latest technology available in PMUs to provide synchronized measurements using global positioning system data. The thesis presents a contribution in decentralized estimation of generator states and its utilization in LFC and fault detection.