Research on Total Investment Estimation Method for Pumped Storage Power Stations Based on Hybrid Machine Learning Model

Le résumé fourni par la source

Pumped storage power station total investment estimation requires high accuracy for decision-making, but traditional methods fail to capture complex nonlinear and temporal characteristics. This paper proposes a hybrid model estimation method combining eXtreme Gradient Boosting(XGBoost) and Long Short-Term Memory (LSTM). This model first utilizes the powerful feature selection capability of XGBoost to screen key influencing factors from numerous technical and economic parameters; then, it adopts an Long Short-Term Memory-Back Propagation (LSTM-BP) combined model to deeply mine the long-term dependencies and dynamic temporal patterns implicit in the investment sequence, effectively integrating the static interpretation and dynamic prediction advantages of the models. Empirical research shows that compared to single models and traditional methods, the hybrid model constructed in this paper significantly improves estimation accuracy, with the mean absolute percentage error significantly reduced, providing a more accurate and reliable decision-making tool for the preliminary planning and investment risk management of pumped storage power stations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.