Spike-LSTM: A Hybrid Spiking Neural Architecture for Enhancing Renewable SCR Prediction

The Short-Circuit Ratio (SCR) is a critical metric for assessing power system stability under high renewable energy penetration. Accurate SCR prediction is essential for ensuring grid stability and optimizing renewable energy integration. However, traditional prediction methods, such as Long Short-Term Memory (LSTM) networks, struggle to capture the nonlinear and dynamic fluctuations of SCR, particularly in high-penetration scenarios. To address these challenges, this paper proposes Spike-LSTM, a hybrid architecture that integrates the gating mechanism of LSTM with the event-driven dynamics of Spiking Neural Networks (SNN). The Spike-LSTM model introduces a Gated-Spiking Unit (GSU), which replaces traditional continuous gating with spiking neuron dynamics, enabling faster response to abrupt changes and improved long-term dependency modeling. Additionally, a joint optimization framework is proposed to balance prediction accuracy and safety margin, ensuring conservative predictions near critical thresholds. The model is validated using real-world data from a provincial grid in China, with test sets ranging from 20% to 70% of the total data. Results demonstrate that Spike-LSTM consistently outperforms standard LSTM and conventional Spike-NN models across all test scenarios, achieving higher R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> values (up to 0.9694) and lower Mean Absolute Error (MAE) (as low as 0.785). The model's robustness is particularly evident in extreme cases, where it maintains high accuracy even with limited training data. These improvements highlight Spike-LSTM's ability to handle the dynamic and nonlinear characteristics of SCR, making it a reliable tool for power system stability analysis. The proposed architecture not only advances SCR prediction but also provides a foundation for future research in hybrid neural network applications for power systems.