Development of modeling and optimization technologies for vapor compression refrigeration cycle

Abstract

This thesis presents the research results on the new modeling for vapor compression cycle (VCC) based on single layer feed-forward neural network (SLFN) trained with extreme learning machine (ELM), and the development and implementation of a novel optimization approach based on SLFN model. Furthermore, the centralized and decentralized optimization problems of VCC are formulated, the corresponding solution methodology are then developed. The details of these methodologies are as follows: 1. A single hidden layer feed-forward neural network (SLFN) is used to model the VCC, and an optimal system model, associated with cooling load, environment and system operating states, is developed. The SLFN model is then combined with a grid search algorithm to compute the system optimal set points for different cooling loads. Simulation and experiment results show that the SLFN trained with ELM achieves higher prediction accuracy than the ones with the back-propagation and support vector learning machines. 2. The model-based centralized optimization for VCC is developed based on a modified genetic algorithm, which is capable of minimizing the system energy consumption under cooling load, hybrid models and physical constraints, achieving the accurate estimates of the operating states and the global optimal set point. 3. The decentralized methodology is proposed with application to the VCC, by decomposing the complex global optimization of VCC into the local evaporator, condenser and compressor optimizations based on component hybrid models and interactive constraints. The decentralized optimization method is then further simplified to the unconstrained subsystem optimization with gradient based search methods with the result that the time cost can be efficiently reduced compared with the centralized optimization. The main contribution of this thesis is to propose three systematic approaches in optimizing the vapor compression refrigeration cycle. In practice, these approaches can be chosen accordingly to balance the energy saving and speed requirements. These approaches offer significant advantages versus traditional VCC control and optimization methods.