在現有的冰水系統上,對於冰水主機負載幾乎未特別考量其個別效率,而是採用平均負載法,所以近年來已有研究提出相關演算法來做冰水主機負載分配的最佳化,但是在研究中對於冰水主機耗電模型皆採用線性迴歸分析方式建立,其對於非線性問題尚有準確度較低缺點。 因此,本研究利用類神經網路(Neural Networks;NN)來建立冰水主機運轉耗電量模型,並且利用粒子群演算法(Particle Swarm Optimization;PSO)來進行負載分配最佳化,進而可以達到節約能源目的。 本論文使用一高耗電量案例來做分析,以類神經網路結合粒子群演算法之結果與採用迴歸分析之平均負載法做比較,結果顯示,類神經網路結合粒子群演算法在冰水主機負載最佳分配上,相較平均負載法於不同負載有較佳結果,最低改善節能效率在55%負載下為2.63%,而最高改善節能效率50%負載下為7.8%。因此,本研究結果發現應用類神經網路在建立冰水主機耗電模型上收斂速度快,再結合粒子群演算法於冰水主機負載分配最佳化上準確度又高,因此可以運用在空調系統與其它相關最佳化問題上。
On the existing chiller water system, has not considered its individual efficiencies of chiller then uses the mean load method. Therefore, the recent years had the research to propose that the related calculating method to make the optimization of chiller loading, but power consumed model in the research regarding to use the linear regression analysis mode establishment, it still had the accuracy low shortcoming in regarding the non-linear problem. This dissertation used neural networks (NN) to build the model of power consumption of the chiller and use particle swarm optimization (PSO) algorithm to optimize the chiller loading for minimizing power consumption. We obtained 2.63% power saving on 55% chiller partial load rate (PLR) and 7.8% power saving on 50% PLR after analysis and comparison with the linear regression (LR) and equal loading distribution (ELD) methods. Therefore, the NNPSO method solved the problem with fast convergence on optimal chiller load (OCL), and produced highly accurate results within a short timeframe. The proposed approach can be applied to air-conditioning systems and other related optimization problems.