A Novel Solar Power Interval Prediction Based on Combined Multiple Quantile Regression Models

To address the challenges posed by the stochastic nature of solar power, this paper proposes an interval forecasting model designed to enhance the accuracy of solar power predictions and improve the stability of power infrastructure. The model integrates multiple quantile regression approaches to overcome the limitations of traditional quantile regression, which typically relies solely on a single quantile loss function. Key metrics, including Prediction Interval Coverage Probability (PICP) and Prediction Interval Normalized Average Width (PINAW), are introduced into the optimization criteria and optimized using a particle swarm optimization algorithm. Additionally, the upper and lower bounds of the prediction intervals are derived by aggregating results from different models through upper and lower bound weights, which are continually updated based on model performance. Experimental results demonstrate that the proposed method outperforms other approaches in terms of generalization ability and robustness when tested on a real-world solar power dataset, highlighting its effectiveness and reliability.