A Power System Frequency Nadir Prediction Model Based on FNP-DBN Fusion-Driven Method.

Frequency nadir is an important indicator for frequency stability analysis and control in power system, so it is significant to calculate the frequency nadir quickly and accurately. To address the shortcomings of the current physical and data models, this paper proposes a physical-data fusion-driven model to calculate frequency nadir. Frequency nadir prediction (FNP) model is chosen to be the physical-driven link to quickly obtain the analytical expression of the frequency nadir as the initial results and ensure the model speed. Deep belief network (DBN) model is chosen to be the data-driven link to correct the error of the FNP model online and enhance the model accuracy. The two are fused by serial method, which can calculate frequency nadir quickly and accurately. Compared with current fusion-driven methods, the FNP model can substantially improve the model calculation speed and retain more critical factors, and the DBN model can greatly enhance the model prediction accuracy. Through the case studies in IEEE 39-bus system, the accuracy and validity of the proposed FNP-DBN fusion model has been demonstrated.