Application of machine learning into organic Rankine cycle for prediction and optimization of thermal and exergy efficiency

Organic Rankine cycle (ORC) is a promising technology to recovery and utilization of low grade thermal energy. In recent years, there are few researches on ORC performance prediction based on Machine Learning, mainly due to a lack of reasonable methodology and case demonstration. This paper presented a comprehensive method to achieve a reasonable application of Machine Learning into ORC research for prediction and optimization of ORC's parameter and performance. Firstly, a cycle database was established by thermodynamic modeling, including four ORC configurations and seven working fluids. Then, for Machine Learning, the Back Propagation Neural Network (BPNN) and Support Vector Regression (SVR) prediction models for ORC were built by predicting error analysis with part of the database which can determine the best parameters of BPNN and SVR. Finally, taking RORC as example, cycle parameter analysis and multi-objective optimization of ORC were conducted based on the thermodynamic model and prediction model to maximize the thermal and exergy efficiency simultaneously. By the prediction and optimization results, it can be deserved that the accurate and fast prediction of the thermal efficiency and exergy efficiency of ORC with multi-parameter, multi-configuration and multi-working fluid was realized, and the optimization results based on the prediction model as the proxy model were also greatly close to the traditional optimization results based on the thermodynamic model. It should be noted that the comprehensive performance of prediction and optimization will be better with more data input. In conclusion, considering accuracy, calculation time, economic cost and safety, the ORC prediction and optimization method proposed in this paper is a promising technology combining Machine Learning and energy utilization, which could provide a new perspective for research in this field.