Examining Different Input Formulations for Load Prediction in Regional Power Grids During N-1 Contingencies Utilizing Machine Learning Techniques

Accurate load prediction plays a pivotal role in effectively managing grids and ensuring the reliability of regional power networks. This study introduces an innovative input formulation for load prediction in regional power networks, which incorporates the topology of generating stations via principal component analysis (PCA). This proposed formulation offers a comprehensive understanding of system dynamics by considering the spatial distribution and connectivity of generating stations, as well as potential contingencies like generator outages. Such an approach facilitates better anticipation of load variations and system response across various scenarios, thereby strengthening the resilience and reliability of the power grid. The efficacy of this proposed formulation is substantiated through a case study, demonstrating its capacity to enhance load forecasting accuracy and support well-informed decision-making in grid operations. The mean absolute error of the prediction is 9.5% and 8.8% for normal operating conditions using linear regression LR and Gaussian processing regression GPR and 9.6% and 9.8% during N-l contingency. The PCA-based input formulation represents a valuable contribution to contingency load prediction by shedding light on the interaction between generating station topology and load dynamics within regional power system networks.