A Machine Learning Framework With an Intelligent Algorithm for Predicting the Isentropic Efficiency of a Hydraulic Diaphragm Metering Pump in the Organic Rankine Cycle System

Abstract

The pump provides the necessary pressure and flow for the organic Rankine cycle (ORC) system. The traditional methods have obvious limitations when analyzing the time-varying characteristics of the key operating parameters of the pump. This study first introduces the scatter plot analysis method to analyze and evaluate the time-varying and coupling characteristics of the hydraulic diaphragm metering pump. Then, a machine learning-fitting algorithm hybrid model is constructed to solve and verify the actual matching correlation equation of the key operating parameters. In addition, the complicated non-linear relationship brings great challenges to obtaining the limit value of the pump isentropic efficiency. This study introduces the bilinear interpolation algorithm to systematically analyze the change trend between operating parameters and isentropic efficiency. Based on the wavelet neural network (WNN) with momentum term and particle swarm optimization-adaptive inertia weight adjusting (PSO-AIWA), a machine learning framework with an intelligent algorithm is constructed. Under this framework, the maximum isentropic efficiency of the pump can be stabilized at 70.22–74.67% under all working conditions. Through the theoretical analysis model, the effectiveness of this framework is evaluated. Finally, the optimal cycle parameters are evaluated. This study can provide direct significance for the analysis and optimization of the actual performance of the ORC system.