基于熵产理论的PSO-BP优化PAT能量效率的改进方法

Abstract Pump as turbine (PAT) is a comprehensive equipment combining pump and turbine, and is one of the excellent pumped energy storage devices. The optimization of geometric diversion is an important means to improve the energy conversion efficiency of PAT. In order to study the effective energy conversion measures, this paper puts forward the method combining entropy production (EP) and PSO-BP to optimize the measures. Compared with the traditional PSO-BP method, this method first introduces the learning factor and the weight factor to improve the search efficiency of PSO, then describes the internal flow of PAT from the perspective of energy in terms of time-average and dissipative EP in the pulsating velocity field, and introduces it into the prediction of PSO-BP. To validate the effectiveness of the method, a cross Inducer is installed in the PAT, and both computational fluid dynamics (CFD) simulation and experimental testing are employed. The results show that the R 2 of EP-PSO-BP prediction model is increased to 0.95983 compared with PSO-BP of 0.77451, and predicted results are more consistent with the experimental results. Finally, the sensitivity analysis of several control parameters is also done to verify the influence of parameters on the prediction results. This method improves the optimization ability of PSO-BP and the prediction accuracy of PAT optimization measures, and is an effective method for PAT optimization.