Neural Network Predictive Control of a Vapor Compression Cycle

Abstract

This study investigates the Neural Network Predictive Control of a vapor compression cycle (VCC). VCC consists of four components, namely the compressor, electronic expansion valve (EEV), evaporator and condenser. Modeling of the compressor and EEV is carried out with the static relationships, while modeling of the evaporator and condenser is performed with the lumped parameter moving boundary method. The established thermodynamic model is validated against the ASPEN model with the same design specifications. The neural network is trained off-line with the input and output signal data of the established model. The solution of the optimization problem for the each time step is achieved with the metaheuristic method called Whale Optimization Algorithm in the predictive controller. Ultimately, performances of the four different controllers, namely the cooling load, first law efficiency, entropy generation and second law efficiency, are compared with each other. The results show that the entropy generation controller achieves the most favorable exergy destruction performance with 0.2% lower than the worst performer cooling load controller. It is also observed that the second law efficiency controller is the best performer in terms of the overall second law efficiency through the simulation time.