Hybrid model-driven short-term load forecasting and voltage stability assessment for energy storage optimization in coal mine power supply systems
Year: 2026
Type: article
Abstract: Coal mining power supply systems demand exceptionally high voltage stability, as even minor deviations may cause significant operational losses. To assess voltage stability in these systems, this study develops an integrated framework for energy storage optimization and scheduling. A hybrid model combining time series decomposition and Long Short-Term Memory (LSTM) networks is employed for short-term load forecasting. Through systematic sensitivity analysis, an optimal configuration with a 7-day input window and 128 hidden units is identified, reducing the RMSE by nearly 15 % compared to SARIMA and 8 % compared to GRU, making it significantly more accurate for load forecasting. Based on the IEEE 30-bus test system, combined with the load forecasting loads, Newton-Raphson power flow analysis and a parabolic voltage risk assessment model are implemented to accurately identify nodes experiencing voltage violations and quantify risk severity. Analysis reveals that certain nodes approach the voltage limit due to the sudden increase of load randomness. To address this, a mixed-integer second-order cone programming (MISOCP) model is constructed to optimize the configuration of energy storage, and the optimal installation locations and operation strategies of energy storage devices are solved. In the application of IEEE 30-bus system, the maximum voltage fluctuation is reduced from ±7 % to ±5 %, which significantly reduces the node voltage risk index. The research establishes an efficient and economically viable methodology for deploying energy storage in coal mining power supply systems. It provides a novel and practical solution for mitigating voltage quality issues, with critical implications for enhancing power supply reliability, reducing equipment degradation, and minimizing production interruption risks.
Source: Energy Reports
Authors: Tian Xia, Sheng Chen, Zhiqi Chen, Qiuhua Lin, Tongfu Fu
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,040
OpenAlex ID: W7118488661