AI-Driven Probabilistic Congestion Forecasting for Unbalanced Power Grids With Line Loss Calculations

The integration of distributed energy resources (DERs), including photovoltaics (PV) and electric vehicles (EVs), into unbalanced distribution grids has increased the likelihood of power congestion, affecting grid stability and reliability. Traditional congestion prediction methods often struggle with the spatially and temporally dynamic nature of these grids. This paper introduces an AI-driven framework for probabilistic congestion forecasting, incorporating machine learning and line loss estimation. By leveraging advanced graph-based wavelet transforms, the proposed method accounts for topological dependencies and non-stationary power flow variations across grid components. Unlike conventional approaches, this framework learns adaptive filter kernels and embeds topology explicitly, improving the model's ability to predict power congestion and estimate line losses under varying grid conditions. The results demonstrate that the proposed framework offers accurate probabilistic congestion forecasts and line loss calculations, with significant improvements in both speed and precision compared to traditional power flow models.