This paper presents the results of application of sequential quadratic programming to the estimation of the unknown composite load model parameters. Traditionally applied estimation methods, such as nonlinear least squares or genetic algorithms, suffer from a number of issues. Genetic algorithms exhibit premature convergence and require high computational resources and nonlinear least squares method is very sensitive to the initial guess and can diverge easily. This paper provides a comparison of all three methods based on computer-generated signals serving as field measurements. Accuracy and precision are assessed as well as computational requirements.
Power system loads are one of its crucial elements to be modeled in stability studies. However their static and dynamic characteristics are very often unknown and usually changing in time (daily, weekly, monthly and seasonal variations). Taking this into account, a measurement-based approach for determining the load characteristics seems to be the best practice, as it updates the parameters of a load model directly from the system measurements. To achieve this, a Parameter Estimation tool is required, so a common approach is to incorporate the standard Nonlinear Least Squares, or Genetic Algorithms, as a method providing more global capabilities. In this paper a new solution is proposed -an Improved Particle Swarm Optimization method. This method is an Artificial Intelligence type technique similar to Genetic Algorithms, but easier for implementation and also computationally more efficient. The paper provides results of several experiments proving that the proposed method can achieve higher accuracy and show better generalization capabilities than the Nonlinear Least Squares method. The computer simulations were carried out using a one-bus and an IEEE 39-bus test system.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.