A Machine Learning Approach to Weak Grid Identification for Large Scale Electrical Power Systems
MetadataShow full item record
This research proposes a weak grid detection method that is a steady state screening method to identify clusters of buses where potential coordinated voltage oscillations could occur. The objectives of this method are to identify any potential weak grid areas in transmission planning cases with future projected renewable energy under a planning horizon. So that necessary improvements such as transmission projects, voltage support devices or conventional generation are identified for the scenarios in the cases that are being studied. Furthermore, after the necessary improvements have been identified and made, the change in system strength can be quantified. The benefit of this method is that it is a steady state screening method and does not require computationally and manually intensive dynamic simulations. Several different scenarios in the form of cases can quickly be analyzed, with the resulting areas identified and quantified in each scenario. The contribution of this method is that it is robust and can be applied to any large scale electrical power system under varying operating conditions. In the proposed method, a random forest algorithm is used to cluster the buses into the weak grid areas based on features extracted from the short circuit current and electrical distance. This method was applied to the ERCOT Current Trends Long Term System Assessment (LTSA) Transmission Planning Case for the year 2031 to benchmark the capability of the tool and see if it could identify predicted weak grid areas. It was also applied to a transmission planning case representing a Synthetic Texas Network. To demonstrate the robustness of the tool and ability to identify weak grid areas under different operating conditions and for different systems. The system conditions that were studied in the 2031 ERCOT Current Trends Long Term System Assessment (LTSA) Transmission Planning Case that was analyzed do not reflect actual ERCOT operating conditions. The conclusions in this thesis are only the author’s opinion and do not reflect ERCOT’s official position.
Clark, Angelica (2017). A Machine Learning Approach to Weak Grid Identification for Large Scale Electrical Power Systems. Master's thesis, Texas A & M University. Available electronically from