Abstract
Bio-indexing of hydro turbines has been identified as an important means to optimize passage conditions for fish by identifying operations for existing and new design turbines that minimize the probability of injury. Cost-effective implementation of bio-indexing requires the use of tools such as numerical and physical turbine models to generate hypotheses for turbine operations that can be tested at prototype scales using live fish. Blade strike has been proposed as an index variable for the biological performance of turbines. Here we report on an evaluation of the use of numerical blade-strike models as a means with which to predict the probability of blade strike and injury of juvenile salmon smolt passing through large Kaplan turbines on the mainstem Columbia River.
Numerical blade-strike models were developed for a 1:25-scale physical turbine model built by the U.S. Army Corps of Engineers for the original design turbine at McNary Dam and for prototype-scale original design and replacement minimum gap runner (MGR) design turbines at Bonneville Dam’s first powerhouse. The numerical blade-strike models were run in both deterministic and stochastic modes. Predictions of blade strike and injury probability made with the models were then compared to data available for the McNary 1:25-scale physical turbine model and the Bonneville Dam prototype-scale original and MGR turbine runner units.
Based on the comparisons of numerical blade-strike model predictions with 1) observations from both physical turbine models using beads and 2) prototype-scale live fish tests of turbine biological performance, we recommend the stochastic blade-strike models as the preferred method for prediction of blade-strike probability. This recommendation follows from much better agreement between numerical model and physical model predictions of blade-strike probability and prototype-scale observations of live fish mortality and injury assigned to blade strike. Our analysis clearly shows that consideration of the aspect that smaller fish present to the leading edges of turbine runner blades is a significant factor in assessment of blade strike.
Blade-strike probability predicted by the numerical blade-strike stochastic model agreed with the overall trends in blade-strike probability observed in both physical model predictions and prototype-scale live fish observations. However, the numerical model did not show the variability between operating conditions present in physical model bead strike data or in the trend in differences in fish injury and mortality estimates between Bonneville Dam original and MGR design runners. We conclude that blade-strike models, as implemented for juvenile salmon smolts, provide a good overview of the general blade-strike probabilities for prototype-scale turbines but lack the capability to contribute significantly to development of testable bio-indexing hypotheses. The requirements for numerical models that will prove useful for turbine bio-indexing are more rigorous than those provided by the simple blade-strike modeling done to date. The reason is that small differences in turbine biological performance, on the order of 1% or so in strike probability, are significant improvements or degradations in turbine biological performance because they may represent improvements or degradations of 25% to 50% due to the relative small percentage in strike probability or other turbine-blade-related injury. Numerical modeling, if it is to contribute to development of testable bio-indexing hypotheses, will likely need to move toward computational fluid dynamics models of moving machinery that may have a better chance of detecting and quantifying the differences critical to improved biological performance of operating hydro turbines.