Energy performance prediction of pump as turbine (PAT) based on PIWOA-BP neural network

Wenjin Yu,Peijian Zhou,Zhouqian Miao,Haoru Zhao,Jiegang Mou,Wenqiang Zhou

DOI: https://doi.org/10.1016/j.renene.2023.119873

IF: 8.7

2024-01-03

Renewable Energy

Abstract:The greenhouse effect and the depletion of fossil energy sources have emerged as significant challenges impacting development of society. The development of micro hydropower stations is crucial in reducing carbon emissions and achieving carbon neutrality. Among them, pump as turbine (PAT) technology is considered one of the most promising alternatives in small and micro hydropower. However, pump manufacturers don't provide performance curves in turbine conditions, leading to issues such as difficulty in selecting a model and inaccuracy in calculating the turbine's high-efficiency zone when employing PAT in practice. This study proposes a new solution to these problems by introducing a new PIWOA-BPNN model. The model integrates the enhanced WOA algorithm with the parameter-optimized multilayer BPNN and employs the innovative spiral guidance vector p to direct the random search within the WOA algorithm, significantly enhancing the prediction capability of the neural network while reducing time cost. Additionally, the internal parameters of the turbine are analyzed, and a network model incorporating the geometric parameters of the impeller as auxiliary input is proposed. Several experimental tests demonstrate that the method offers greater accuracy and reliability compared to current mainstream algorithms. The prediction accuracy of the proposed model is approximately 4 %, meeting engineering requirements.

energy & fuels,green & sustainable science & technology