In this study, a novel application of neural networks that predict thermal comfort states of occupants is proposed with accuracy over 95%, and two optimization algorithms are proposed and evaluated under two real cases (general offices and lecture theatres/conference rooms scenarios) in Singapore. The two optimization algorithms are Bayesian Gaussian process optimization (BGPO) and augmented firefly algorithm (AFA). Based on our earlier studies, the models of energy consumption were developed and well-trained through neural networks. This study focuses on using novel active approaches to evaluate thermal comfort of occupants and so as to solves a multiple-objective problem that aims to balance energy-efficiency of centralized air-conditioning systems and thermal comfort of occupants. The study results show that both BGPO and AFA are feasible to resolve this no prior knowledge-based optimization problem effectively. However, the optimal solutions of AFA are more consistent than those of BGPO at given sample sizes. The best energy saving rates (ESR) of BGPO and AFA are around -21% and -10% respectively at energy-efficient user preference for both Case 1 and Case 2. As a result, an potential benefit of S$1219.1 can be achieved annually for this experimental laboratory level in Singapore.
翻译:在本研究中,提议对神经网络进行新的应用,以预测居住者的热舒适状态,精确度超过95%,并在新加坡的两个实际案例(一般办公室和讲座剧院/会议室设想方案)下提出并评价两种优化算法,两种优化算法是巴伊西亚高山工艺优化(BGPO)和增强的消防飞行算法(AFA)。根据我们先前的研究,通过神经网络开发了能源消费模式,并对其进行了良好的培训。本研究的重点是使用新的积极方法,评价居住者的热舒适状态,以解决一个旨在平衡中央空调系统的能源效率和居住者的热舒适的多重目标问题。研究结果表明,BGPO和AFA是可行的,可以有效解决以前没有以知识为基础的优化问题。但是,根据我们先前的研究,AFA的最佳解决办法比BGPO在特定样本规模上的最佳解决办法更加一致。BGPO和AFA分别采用21%和-10%的最佳节能率方法,在以节能方式优先选择第1号案件和2号案件案例2,新加坡每年实现S12.1级实验室的潜在利益。