Integrated learning model for water intake capacity of Tyrolean weirs under supercritical flow

ABSTRACT Tyrolean weir can be used as an effective solution to address floatation and sediment deposition in runoff hydropower stations. To improve the efficiency and accuracy of calculating this structure's water intake capacity. The integrated learning algorithm random forest (RF), the firefly algorithm (FA), and the exponential distribution algorithm (EDO) are utilized to develop the algorithm that can be used for the Tyrolean weir Cd and (qw)i/(qw)T prediction models. Sobol's method and SHAP theory are introduced to analyze the above parameters quantitatively and qualitatively. It is shown that EDO-RF is the optimal prediction model for the Tyrolean weir's discharge coefficient and the Froude number Fr has the greatest influence on the Cd prediction results; when Fr < 30, the greater the negative influence of Fr on the model prediction results. When Fr > 30, the greater the positive influence of Fr on the model prediction results. FA-RF is the optimal prediction model for the Tyrolean weir water capture capacity (qw)i/(qw)T, with the ratio of bar length to bar spacing L/e being the largest; When L/e < 20, the greater the negative influence of L/e on the model prediction results. When L/e > 20, the more significant the positive impact of L/e on the model prediction results.