Real-Time ANN-Based Multi-Indicator Voltage Stability Assessment for Mesh Distribution Systems with High SPV-EV Penetration

• Proposes ANN-based method for rapid voltage stability assessment in Mesh distribution networks. • Utilizes VSI, LAM, and GAM to quantify system voltage stability under stress. • evenberg-Marquardt algorithm achieved best accuracy for ANN model training. • Demonstrates superior performance over CPF and JBSA in speed and efficiency. • Enables real-time stability analysis amid high EV and SPV system penetration. Simultaneous integration of solar photovoltaic (SPV) generation and electric vehicle (EV) charging creates competing overvoltage and undervoltage conditions that stress meshed distribution systems and complicate real-time voltage stability assessment. Conventional techniques such as Predictor–Corrector Continuation Power Flow (CPF) and Jacobian-Based Sensitivity Analysis (JBSA) provide accurate stability margins but are computationally intensive and less suitable for rapid operational monitoring. This study proposes an artificial neural network (ANN)-based framework for real-time voltage stability assessment in a meshed distribution system under combined SPV and EV penetration. The model simultaneously estimates the Voltage Stability Index (VSI), Load Availability Margin (LAM), and Generation Acceptability Margin (GAM) for 1,698 operating scenarios derived from a modified IEEE 33-bus meshed system. The optimized ANN achieves a mean squared error of 0.00039729 and a regression coefficient of approximately 0.999, indicating high predictive accuracy. In terms of computational performance, the ANN requires only 0.15 minutes of execution time and 2.99 MB of memory, whereas CPF requires 0.728 minutes and 28.64 MB, and JBSA requires 1.185 minutes and 7.76 MB. This corresponds to the ANN being approximately 79% faster than CPF and 87% faster than JBSA, while reducing memory usage by 89.56% compared to CPF (using 10.44% of CPF’s memory) and by 61.47% compared to JBSA (using 38.53% of JBSA’s memory). The ANN maintains comparable accuracy to CPF and JBSA, with mean absolute percentage errors (MAPE) of 4.282% and 2.822%, respectively. Furthermore, the framework reliably identifies Bus-18 as a critical weak bus due to multiple loop power redistributions, a vulnerability often overlooked by conventional approaches. These results demonstrate that the proposed ANN method enables accurate, lightweight, and near real-time voltage stability monitoring suitable for practical operational deployment in meshed distribution networks.