A neural network approach to enhance blade element momentum theory performance for horizontal axis hydrokinetic turbine application

Abstract

Blade element momentum (BEM) theory is a commonly used tool to predict the performance of horizontal axis conversion systems, such as wind and water turbines. Moreover, BEM theory can be easily integrated into many optimization techniques to improve the turbine structure and performance reliability. BEM theory though conceptually simple has different sources of convergence issues. The main focus of this work was to introduce a computational intelligence technique, namely, multilayer perceptron (MLP) neural networks (NNs) to overcome the convergence issues regardless of their sources. To improve the BEM accuracy, NNs were also employed as a multivariate interpolation tool to calculate the lift and drag coefficients over an operational range of local Reynolds numbers. This technique was found to be easy to integrate into any modified BEM model such those account for blockage in channels. The BEM-NNs model was able to operate at a higher tip speed ratio, with no convergence problems, compared to other models. Integration of NNs as multivariate interpolation tool for hydrodynamic coefficient calculation further improved the power prediction compared to that when using a constant representative Reynolds number.