Contribution a l'exploitation robuste des reseaux electriques par la methode de Taguchi.

摘要

Today, power system analysis algorithms are becoming increasingly available. However, most existing algorithms still have some deficiencies that limit their scope of application and practical value. The solution is not always satisfactory and the adjustments of the analytical solution to a practical one are often required. In OPF, variables as shunt controls are rounded off to their discrete steps and the control actions to correct constraints violations, if any, or to enforce power system in a secure operating region are made by a suboptimal procedure. The same procedure is applied in power system stability to adjust parameters of stabilizers and coordinate regulators in the presence of power system variability or interaction between machines and sub-systems. An arbitrary adjustment procedure leads to a high objective function cost and constraint violations. Fast and rigorous algorithms that handle a mix of continuous/discrete variables and power system parameter variability require sophisticated mathematical tools and are computationally complex and not attractive for real time applications. A best compromise is to develop a logic or a procedure to improve the performance of the fast standard algorithms to be acceptable for practical purposes.;Here, we use a quality control based method to improve the performances of the solution in terms of robustness and desirability. The goal is to obtain a best setting of the input which results in a performance characteristic of the output close to target with minimum variability. A post-optimality procedure which incorporates the philosophy and the method of Taguchi is applied. Desirability functions are used to assess how close system responses are to their targets. Power system variability is included in the analysis by means of the expected value of a loss function which is estimated by Taguchi orthogonal arrays approach. The best and less sensitive solution over uncontrollable power system variability is selected by means of signal-to-noise ratios.;The procedure used here couples statistical analysis with a well-defined problem optimization solving strategy. It can be helpful in all power system problems where parameter variability is incorporated in the analysis and the exact solution or mathematical model are not available. Decision making under uncertainty or sensitivity analysis in the presence of interactions between system variables may be evaluated. Based on orthogonal arrays, the approach handles discrete variables as shunt controls. Operators in VAR planning and/or contingency analysis can quickly and efficiently select through a large number of factors those that actively affect a response variable.