Integrating high dimensional quadratic regression with penalties based predictive modeling for hydro power plants accurate tariff prediction

In order to optimize the financial and operational cost of an hydropower plant in a micro-grid operation, it is required to accurately forecast the per unit generation cost and per unit selling price in a competitive energy trading market. This study presents a novel high dimensional quadratic regression with penalty based predictive model for forecasting generation cost and selling price per unit in hydroelectric energy systems. The proposed model addresses the limitations of conventional method such as SVR, SARIMA and LSTM by integrating polynomial interaction terms with L2 regularization to balance model complexity and generalization. A total of 12 features including operational variables and nonlinear combinations are pre-processed using outlier detection normalization and interpolation techniques. The model is benchmarked across multiple time intervals using a comprehensive set of key performance indicators. Compared to benchmarking models, the proposed approach consistently achieves the lowest forecast error. Computational complexity is minimized with 100 parameters and training time of 2 s. Graphical and numerical evaluations confirm the model's accuracy and suitability for spot market forecasting within hydro-DISCOM integration. The study concludes with recommendations for real time deployment and extension into hybrid intelligent forecasting framework. In future the work can also be integrated to different policy & weather dynamics and impact analysis in predicting the per unit selling price of the energy.