Identification of synchronous machine parameters from field flashing and load rejection tests with field voltage variations

- Identification models for field flashing and load rejection tests are derived.
- Optimization problem formulated as orthogonal regression to take noise into account.
- Constrained optimization solved by interior point method.
- Detailed analysis of method performance using synthetic data.
- Application to real data from the commissioning of a hydro generator.

In this paper, a method for the identification of the synchronous machine parameters using data from field flashing (FF) and load rejection (LR) tests performed during commissioning, is proposed. The identification using data acquired during field flashing, a commissioning standard procedure, allows an initial estimation of the d-axis open-circuit transient time constant. A model dependent on that time constant, which relates the applied field voltage and the field current, the model inputs, and the generator terminal voltage, the model output, is derived. In the conventional load rejection, the applied field voltage is constant. This poses a practical restriction, since a constant voltage source must be available. In the load rejection identification method proposed in this paper, a model is derived which relates variable field voltage and field current, the model inputs to the generator terminal voltage, the model output, following load rejection. From the field flashing and load rejection models, the synchronous machine parameters are determined solving an optimization problem, formulated as a nonlinear least-squares problem or as an orthogonal distance regression. The orthogonal distance regression is suited to take into account noise in the model input. The proposed method is applied to synthetic data generated by simulation of a generator with known parameters. Statistical analysis shows a maximum error with relation to the true values of 35.8% for the nonlinear least-squares problem and 5.8% for the orthogonal distance regression. The method is also tested using real data acquired during commissioning of a 140 MVA hydro powerplant. The Normalized Sum of Squared Errors (NSSE), a metric to evaluate the deviation of the simulated response with relation to the measurements, gives a value less than 1% in most of the cases and a maximum of 2.3%, corroborating the accuracy of the proposed method.