Abstract: Reliable operation of modern power systems demands accurate forecasting. Errors can reduce resilience, especially in mission-critical military, defense, and strategic infrastructures. While hybrid forecasting models perform well under normal conditions, their robustness to noisy data still needs thorough evaluation. In this work, a sensitivity and robustness analysis of a novel adaptive hybrid forecasting framework is examined using two scenarios of correlated Gaussian noise. These scenarios were selected to represent measurement degradation, communication failures, and physical disturbances commonly encountered in resource-limited defense operational environments. The proposed hybrid method integrates the Multi Model Partitioning Filter, Nonlinear Autoregressive Exogenous models, and a Genetic Algorithm for Resource Allocation, with performance evaluated using real commercial data. The results highlight the successful performance of the proposed method and underline the importance of the sensitivity analysis as a validation process for forecasting methods intended for defense energy systems that need to remain highly reliable at all times.

Keywords: Defense Energy Systems, Hybrid Forecasting, Mission Critical Infrastructure, Sensitivity Analysis, Robustness Assessment, Power System reliability, Gaussian Noise, Defense Energy Systems

Pages: 595-606