Deep learning based active loss estimation in power grid with N-1 reactive support contingencies

The efficient and optimal operation of power systems is critically dependent on accurate estimation and reduction of network losses. Traditional techniques such as Gauss-Seidel (GS) and Newton-Raphson (NR) are widely used to compute power system operating conditions and associated power losses. However, these methods are inherently iterative and computationally intensive, which limits their suitability for real-time applications in modern power system operations. In practical power systems, contingencies such as the tripping of generators, synchronous condensers, or capacitor banks can significantly affect the reactive power balance of the system which in turn affect the voltage profile of the grid. This paper presents a data-driven approach for real-time estimation of active power losses under different reactive power support contingency scenarios. An Artificial Neural Network (ANN) model is developed and benchmarked against other regression-based methods. The ANN model demonstrates superior performance in terms of accuracy and generalization under all operating scenarios. The proposed ANN estimator is implemented using Python libraries in conjunction with the Panda Power simulation environment. The effectiveness of the proposed approach is validated using the IEEE 14-bus test system. Simulation results demonstrate that proposed ANN-based loss estimator provides superior prediction accuracy, faster response and better generalization capability as compared to traditional methods, making it a promising and robust tool in predicting power system losses.