Event-Driven Power Outage Prediction using Collaborative Neural Networks

Abstract

An increase in the weather events can result in an increase in the frequency of grid faults, thereby leading to a decrease in electricity supply reliability. To develop a better performance model for outage forecasting, it is essential to consider various types of events for different complex weather conditions and constraints. A conventional event-driven forecasting algorithm is only appropriate for a single type of event scenario and unable to adapt effectively to multiple types of event scenarios concurrently. A multiple-constraints event-driven outage model that depends on a maximum entropy function, which transforms a complex event-driven outage problem into two continuous differential single-objective subproblems using the collaborative neural network (CONN) algorithm is proposed in this article. The CONN event-driven forecasting algorithm, validated by many experiments, successfully resolves the problem of the difficulty in obtaining very large complex weather events and outages data.