Construction of investment impact index and LASSO regres-sion prediction for pumped storage power stations

Pumped storage power stations (PSPS), as a form of energy storage technology, are deployed extensively in power systems dominated by renewable energy due to their flexible energy storage and regulation capabilities. Investment decisions for new power stations require com-prehensive consideration of cost-driving factors and estimation of total project investment. However, current cost management methods in this area remain immature. In this study, we propose a cost impact factor analysis and prediction model for PSPS. Firstly, descriptive statistics and potential relationship construction were conducted on data from 33 PSPS projects in China to preliminarily identify potential cost-driving factors. Secondly, the Grey Relational Analysis (GRA) method was employed to weight the impact factors and select significant ones. Subsequently, utilizing these significant impact factors as labels, a LASSO regression model was established to construct the relationship between the factors and unit cost, and to accomplish predictive tasks based on the constructed relationship. Finally, the predictive model was validated for its effectiveness and sensitivity analysis. The results indicate that factors such as dam volume and rated head are significant cost-driving factors for PSPS. The established LASSO regression model achieves high-quality regression prediction within an accuracy deviation of less than 5%, and the regression results show high sensitivity to factors like reservoir capacity and major material prices.