ANN-based procedure to obtain the optimal design and operation of the compression chiller network – Energy, economic and environmental analysis

In a large-scale chiller plant, a network of chillers is a suitable solution to respond to the variable cooling demand. The energy consumption of a chiller network depends on the configuration and the control strategy of the chiller network in different conditions. This study presents a general procedure for designing a chiller network for a building with an arbitrary annual cooling demand distribution. The procedure determines the optimal configuration considering the quantity, the size ratio, and the energy performance of chillers. The particle swarm optimization (PSO) algorithm is used for each configuration to find the optimal chiller loading distribution. Then, the optimal configuration is selected through a life cycle cost analysis. In order to predict a general chiller performance curve with an arbitrary nominal capacity, an artificial neural network model is developed based on 20 available commercial chillers in the market. The chiller performance prediction includes determination of COP and actual capacity of a chiller in terms of nominal capacity, chilled water temperature, cooling water temperature, and partial load ratio. The simulation is carried out in TRNSYS, which linked to MATLAB to implement the PSO optimization strategy. The results show that for networks with two, three, and four chillers, the optimal selection of chiller network configuration under the PSO strategy reduces the energy consumption by 26.30, 26.06, and 26.18%, respectively, compared to the conventional configuration under the baseline strategy. The life cycle cost for these configurations is also reduced by 17.93, 17.69, and 18.56%, respectively.