The ladder network parameter identification for transformer winding is crucial for the interpretation of the frequency response function data. The traditional identification method, mainly based on intelligent optimis...
详细信息
The ladder network parameter identification for transformer winding is crucial for the interpretation of the frequency response function data. The traditional identification method, mainly based on intelligent optimisation algorithm, is generally very time-consuming due to a large amount of computation. This study proposes to combine the intelligent algorithm and gauss-newton iteration algorithm (GNIA) to improve the optimisation efficiency notably with a sharply dropped calculation workload. These two methods are well-complementary since the intelligent algorithm holds excellent global search ability while the search of the GNIA is directional and quantitative. This study solves three key problems for the combined algorithms. The first problem is the calculation of the least-square correction solution to the network parameters in the iterationalgorithm. The treatment of the ill-conditioned Jacobian matrix in the iterationalgorithm is the second challenge. Another issue is the determination of the network parameter with zero sensitivity. The identification results on an isolated winding show that the combined algorithms can obtain a more precise solution with far less amount of computation.
Ladder network parameters identification of transformer winding based on frequency response analysis data is crucial to winding fault diagnosis. In the authors' previous study, the gauss-newton iteration algorithm...
详细信息
Ladder network parameters identification of transformer winding based on frequency response analysis data is crucial to winding fault diagnosis. In the authors' previous study, the gauss-newton iteration algorithm (GNIA) had been proposed to efficiently identify the network parameters. The core in GNIA is the Jacobian matrix, composed of derivative of frequency response function (FRF) to network components. This study proposes a generic algorithm to calculate this Jacobian matrix. Two key problems concerning the algorithm are solved elaborately. The first problem is the method for derivative calculation toward an arbitrary FRF based on adjoint network method (ANM). The other issue is the mathematical model construction of double ladder network and its adjoint network toward different FRFs to obtain all network branch voltages and currents, which are required in the derivative calculation by ANM. This generic algorithm can efficiently and effectively calculate the Jacobian matrix, which can be applied on ladder network parameters identification and winding fault diagnosis.
Determining the attitude using GNSS carrier signals is studied. It features an analytical approach to get an estimate as initial guess for iterative algorithms, in three steps. First, baseline vectors are estimated by...
详细信息
Determining the attitude using GNSS carrier signals is studied. It features an analytical approach to get an estimate as initial guess for iterative algorithms, in three steps. First, baseline vectors are estimated by least-squares method. Second, the constraint of the direction cosine matrix (DCM) is ignored and the least-squares estimates of its 9 elements are solved. Third, a mathematically feasible DCM estimate is extracted from the above estimated free matrix. An error attitude, formulated using the Gibbs vector, is introduced to relate the previously estimated and the true attitude, and the measurement model becomes a nonlinear function of the Gibbs vector. The gauss-newtoniteration is employed to solve the least-squares problem with this measurement model. The estimate of the roll-pitch-yaw angles and the variance covariance matrix of their estimation errors are extracted from the final solution. Numerical experiments are conducted. With 3 orthogonally mounted 3-meter baselines, 4 visible satellites, and 5-millimeter standard-deviation of the carrier measurements, the accuracy of the analytical solution can be less than 1 degrees in the root mean squared error (RMSE) sense. The convergence of the iteration is rather fast, the RMSE converges after only one iteration, with the converged RMSE less than 0.1 degrees.
暂无评论