Power System Resilience Enhancement Using Artificial Intelligence

Extreme weather events and natural disasters are the major cause of power outages in the United States. An accurate forecast of component outages and the resultant load curtailment in response to extreme events is an essential task in pre- and post-event planning, recovery and hardening of power systems. Power system resilience improvement is investigated in this work from component outage prediction to identifying the potential power outages in the system to estimating probable load curtailment due to these outages and offering methods for grid hardening. Initially, two machine learning based prediction methods are proposed to determine the potential outage of power grid components in response to an imminent hurricane, namely a second order logistic regression model and a three-dimensional Support Vector Machine (SVM). The logistic regression model defines the decision boundary, which partitions the components\u27 states into two sets of damaged and operational. Two metrics are examined to validate the performance of the obtained decision boundary in efficiently predicting component outages. The proposed three-dimensional SVM furthermore leverages its accuracy-uncertainty tradeoff to achieve highly accurate results, which can be further used to schedule system resources in a predictive manner with the objective of maximizing its resilience. The performance of the model is tested through numerical simulations and validated based on well-defined and commonly-used performance measures. After training the outage estimation model, the predicted component outages are plugged into a load curtailment minimization model to estimate the nodal load curtailments in the system. The standard IEEE 30-bus system with a combination of hurricane path and intensity scenarios are used to study the model where the results demonstrate that the proposed modelling framework is capable of effectively capturing the dynamics of load curtailment estimation in response to extreme events. Furthermore, a machine learning based grid hardening model is proposed with the objective of improving power grid resilience. The predictions from previous stages are fed into the proposed grid hardening model, which determines strategic locations for placement of distributed generation (DG) units. In contrast to existing literature in hardening and resilience enhancement, this work co-optimizes grid economic and resilience objectives by considering the intricate dependencies of the two. The numerical simulations on the standard IEEE 118-bus test system illustrate the merits and applicability of the proposed model. The results further indicate that the proposed hardening model through decentralized and distributed local energy resources can produce a more robust solution that can protect the system significantly against multiple component outages. Finally, a probabilistic load curtailment estimation model is proposed through a three-step sequential method. At first, to determine a deterministic outage state of the grid components in response to a forecasted hurricane, a machine learning model based on TWSVM is proposed. Then, to convert the deterministic results into probabilistic outage states, a posterior probability sigmoid model is trained on the obtained results from the previous step. Finally, the obtained component outages are integrated into a load curtailment estimation model to determine the potential load curtailments in the system. The simulation results on a standard test system illustrate the high accuracy performance of the proposed method