Distribution reliability: Data calibration based on Monte Carlo simulation and evolutionary optimization

Power system failures are random events and their prevention can be managed by the regulatory agencies through pre-set performance targets. Once an acceptable target is established for a reliability index, it is assumed to be the limit value that must not be exceeded; otherwise the utility may receive a penalty that creates an extra operational cost. To maintain these targets under control, the utilities must regularly evaluate their system, and the reliability indices should be predicted using accurate system information. Most utilities do not have accurate information about their own system parameters (e.g., failure rates, etc.), and may use similar data for assessing reliability performance. Consequently, the predicted indices will not reflect the exact status performance of the distribution system. The solution can be reached by validating the database. This process includes detection, localization of the incorrect data, and, finally, the parameter correction. The validation methodology proposed in this paper uses Chronological Monte Carlo-based techniques to assess the probability distributions of system reliability indices, statistical evaluation of samples, and a metaheuristic-based optimization approach. Case studies on a simple distribution network and on a real system are presented and discussed.