Energy efficiency prediction of refrigerating station system based on BP neural network and optimization of operating strategy

The energy consumption of central air-conditioning systems accounts for a significant proportion of building energy consumption. As the "heart" of a central air conditioning system, the refrigerating station system is complex and consumes a large amount of energy, making its energy efficiency improving crucial for building energy savings. This study focused on the refrigerating station system of an office building, which includes four chillers. The key issue addressed was how to improve the overall energy efficiency of the chillers during operation and optimize the associated high-dimensional operational parameters. A data-driven approach was adopted, using a back-propagation neural network to construct an energy efficiency prediction model for the cold station, covering six different operating modes of the chillers to explore the impact of related operational parameters on energy efficiency. Subsequently, with the goal of maximizing overall energy efficiency, a genetic algorithm was employed to iteratively optimize five parameters in single-chiller operating modes and nine parameters in dual-chiller operating modes simultaneously. Three typical days were selected for validation, and the results showed that compared to the original operating conditions, energy efficiency improved by approximately 4.3 %, 5.2 %, and 5.6%, respectively, achieving the goal of energy-efficient operation. This study provides valuable reference for improving the energy efficiency of refrigerating station systems.