An Algorithm for Automatically Tuning the Weights of Performance Indices for Monitoring Power System Loading or Security

This paper presents an algorithm for automatically updating the coefficients of performance indices (PI) used to monitor power system's security, loading or network adequacy, as the system operating conditions change. The objective of the algorithm is to maximize the effectiveness of the PI in terms of predicting operating limit violations. This is accomplished by tuning the PI coefficients so that the PI measures the robustness of the operating point to system changes in terms of how far the point is from the operating limits. The approach taken is to formulate the coefficient optimization problem as a constrained volume maximization problem, in which the PI coefficients are used to maximize the volume of an approximate convex set (defined by the PI) inside the set of all feasible operating points. By exploiting the salient properties of the volume maximization problem, an efficient sequential gradient minimization algorithm has been developed to tune the PI coefficients. This algorithm was implemented to tune the coefficients of the PI used to rank contingencies of the EPRI 39-bus system in accordance to the severity of voltage problems.