Total real and reactive power losses in electrical power systems are an inevitable phenomenon and occur due to several factors, such as conductor resistance, transformer impedance, line reactance, equipment losses, and phase unbalance. Minimizing them is crucial to the system’s efficiency. In this study, an artificial neural network, specifically a Multi-layer Perceptron, was employed to predict total real and reactive power losses in electrical systems. The network is composed of three layers: an input layer consisting of the variables loading factor, real and reactive power generated on the slack bus, a hidden layer, and an output layer representing the total real and reactive power losses. The training method used was backpropagation, adjusting the weights based on the desired output. The results obtained, using datasets from IEEE systems with 14, 30, and 57 buses, showed satisfactory performance, with a mean squared error of around 10⁻⁴ and a coefficient of determination (R²) of 0.998. In validation with 20% of the data that was not part of the training, the network demonstrated effectiveness, with a mean squared error around 10⁻³. This indicates that the network was able to accurately predict total power losses based on loads, generating estimates close to the desired values.

Technical power losses in power systems are unavoidable, caused by factors such as transformer impedance, conductor resistance, equipment inefficiencies, line reactance, and phase imbalances. Reducing these losses is essential for improving system efficiency. This study introduces an innovative approach using Artificial Neural Networks (ANN) combined with the graphical interface to predict complete curves of real and reactive power losses in power systems under various contingencies. The key advantage of this methodology is its speed, allowing quick estimation of power loss curves both in normal and contingency conditions, whether mild or severe. ANN models excel at capturing the nonlinear behavior of power systems, eliminating the need for iterative methods commonly used in traditional approaches. The results showed that the ANN performed effectively, with a mean squared error during training below the specified threshold. For samples not included in the training set, the network accurately estimated 99% of the real and reactive power losses within the specified range, with residuals around 10⁻³ and an overall accuracy rate of 99% between the desired and obtained outputs. Additionally, a Graphical User Interface (GUI) was implemented to facilitate user interaction, allowing for easy visualization of power-loss predictions and real-time adjustments.

The estimation of nodal voltage angles is essential for the efficient and secure operation of electrical power systems. Traditionally, this estimation depends on the prior solution of the power flow problem, a procedure that may become computationally expensive in studies involving multiple operating scenarios. The present research demonstrates relevant potential impact by employing artificial neural networks for the direct estimation of nodal voltage angles. Although the power flow is used during the database generation stage, it is no longer required during the model application phase, resulting in a significant reduction in computational effort of approximately 80%. The proposed model was trained using the error backpropagation algorithm, achieving a mean squared error (MSE) on the order of 10⁻³ in only four iterations, with a training time of approximately 3 s and a correlation coefficient (R) of 0.99. Validation using unseen samples (10% of the samples) also demonstrated high accuracy, with an error on the order of 10⁻³ between the estimated and expected values. Furthermore, the P–θ curves were successfully obtained for the IEEE 14-, 30-, and 57-bus systems, enabling the identification of the maximum loading point with low estimation error. These results confirm the neural network’s ability to generalize and operate as a reliable estimator under different system loading conditions. Another relevant aspect of this work was the sensitivity analysis of the input variables, which made it possible to identify the variables that exerted the greatest influence on the neural network response. In addition, a normalization step was proposed to overcome distortions associated with the magnitudes of the input data, demonstrating the efficiency and robustness of the proposed method.