A Flood Season Division Model Considering Uncertainty and New Information Priority

The seasonal temporal characteristics of runoff are crucial in regional water resources scheduling and administration, flood warning and forecasting, and efficient utilization of water resources through large-scale hub engineering scheduling. Traditional flood season division models are based primarily on consistency assumptions and deterministic indicators and can provide only a single division scheme. However, data consistency is greatly influenced by climate change and human activities, and indicators are becoming more uncertain. As a result, the consistency assumption of traditional flood season division models is unsatisfactory, and deterministic indicators cannot fully reflect the uncertainty of division indicators. Ultimately, the traditional flood season division model cannot adapt to the new form. In response to the above issues, the Flood Season Division Model based on the New Information Priority Principle - Interval Indicators (FSDMBNIP-II) is proposed. Following interval number theory, the IV-FSDI is calculated for both the training and validation sets. The IV-FSDI can represent the fluctuation of flood season division indicators through the interval radius to reflect the uncertainty of flood season division indicators. Optimization models are developed based on IV-FSDI for the two sets. NSGA-II is employed for solving optimization models, resulting in two Pareto solution sets. Two final flood season division schemes are selected for the two Pareto solution sets using the TOPSIS method. According to the new information priority principle, based on the matrix comparison method, the training and validation set are weighted, with more weight given to newer data. This allows final model to differentiate and learn from the information in new and old data to handle inconsistent data. Based on the weighting results, the two final flood season division schemes are combined. The performance of FSDMBNIP-II was validated using the control basin of the Wudongde hydropower station as the research area and compared with the performances of a model without considering the new information priority principle (MWCNIPP), a model without considering the interval of indicators (MWCII), and a set pair analysis method (SPA). The comparative results showed that, in the predetermined scenarios, when compared to MWCNIPP, FSDMBNIP-II was the optimal model for 66.7