The Role of ML and AI in Managing Forced Outages in Hybrid Energy Systems
04 HYB24-99
Presented by: Mladen Kezunovic
Power systems may be experiencing more frequent forced outages in the future due to increased complexity of hybrid energy systems combined with more frequent occurrence of extreme weather events. To prevent such a trend from happening new approaches to outage management focused on outage risk prediction and early prevention through optimized mitigation measures are being investigated. The basis for this innovative approach is a plethora of Machine Learning (ML) and Artificial Intelligence (AI) techniques that rely on historical outage data and variety of environmental condition causing outages.
This paper summarizes results of recent projects demonstrating usefulness of such approaches in predicting outage State of Risk (SoR) in transmission, distribution and Inverter-based Resources (IBRs) interfaced to the grid. The implementations are based on the use of field-recorded power system data coming from synchrophasor measurements and outage records in transmission and distribution systems, and internal resource management systems in the IBR-interfaced third party owned hybrid systems.
Besides the data coming from the utility or IBR-interfaced systems, data from variety of environmental databases not owned by utilities is utilized is to correlate the causes of the forced outages.
Once the causes of forced outages are identified, the next step is to predict the State of Risk (SoR) of outages using ML and AI. Applying the framework from Figure 2 in clockwise direction, ML and AI may be used to predict weather events that create hazard for power system operation, and then the grid vulnerability to such hazards may be determined based on historical record, eventually resulting in the State of Risk (SoR) determination as a product of hazard and vulnerability. Such an SoR prediction is then used to manage and mitigate the risk impacts leading to an optimized operation of a hybrid system to improve its resilience to outages. The paper illustrates such uses of ML and AI through variety of use case implemented using utility data.
This paper summarizes results of recent projects demonstrating usefulness of such approaches in predicting outage State of Risk (SoR) in transmission, distribution and Inverter-based Resources (IBRs) interfaced to the grid. The implementations are based on the use of field-recorded power system data coming from synchrophasor measurements and outage records in transmission and distribution systems, and internal resource management systems in the IBR-interfaced third party owned hybrid systems.
Besides the data coming from the utility or IBR-interfaced systems, data from variety of environmental databases not owned by utilities is utilized is to correlate the causes of the forced outages.
Once the causes of forced outages are identified, the next step is to predict the State of Risk (SoR) of outages using ML and AI. Applying the framework from Figure 2 in clockwise direction, ML and AI may be used to predict weather events that create hazard for power system operation, and then the grid vulnerability to such hazards may be determined based on historical record, eventually resulting in the State of Risk (SoR) determination as a product of hazard and vulnerability. Such an SoR prediction is then used to manage and mitigate the risk impacts leading to an optimized operation of a hybrid system to improve its resilience to outages. The paper illustrates such uses of ML and AI through variety of use case implemented using utility data.