Combined learning and analytical model based early warning algorithm for real-time congestion management
Fast and accurate estimation of branch flows is essential for power system security assessment. Reliable operation of the electric power system highly depends on a transmission network that is immune to miscellaneous contingencies and immediate actions against those contingencies. Operators run optimal power flow (OPF) in every 5-10 minutes resolution to determine transmission flow and to investigate any prospective overload in the system. However, for large scale systems, it becomes computationally costly and intractable. This paper presents a combined machine learning and analytical model-based scheme to make an early prediction of line congestion. The proposed method does not require solving OPF to determine branch loading that makes it suitable for real-time management of overload in transmission systems. A trained model predicts optimal generations based on demand information. Then, shift factors are used to calculate line flows. The proposed algorithm is based solely on net nodal injection regardless of the source of injection. As a result, it can easily capture uncertainties. Also, the scheme uses line outage distribution factors to forecast branch loading after N-1 contingency. Numerical results on the EPRI 39-bus system, IEEE 57-bus system, and the IEEE 118-bus system show the effectiveness of the proposed algorithm for early prediction of transmission congestion and overload before and after the occurrence of contingencies.
Publication Source (Journal or Book title)
2020 IEEE Texas Power and Energy Conference, TPEC 2020
Hasan, F., & Kargarian, A. (2020). Combined learning and analytical model based early warning algorithm for real-time congestion management. 2020 IEEE Texas Power and Energy Conference, TPEC 2020 https://doi.org/10.1109/TPEC48276.2020.9042548