Development of a Novel Artificial Intelligent Coupling (ANN+PSO) Method for On-line Energy Optimization of Building HVAC System

Abstract

Since the turn of the century, commercial buildings started to install Building Management System (BMS) for the Heating Ventilating and Air Conditioning (HVAC) system to provide continuous thermal comfort for the occupants inside the building. The emergence of advanced BMS has led to the promise of maintaining thermal comfort for occupants. However, the function of BMS does not possess optimize technique to maximize the energy-saving potentials. As a result, a significant amount of energy is wasted every day. The objective of this research project is to establish an Artificial Neural Network (ANN) coupling with Particle Swarm Optimization (PSO) algorithm to model and optimize the water side of an HVAC system. A commercial hotel in Hong Kong is selected as a case study to demonstrate the capability of the proposed coupling algorithm, which expects to yield a significant energy-saving without lower the comfort level of its occupants.

In the research study, on-site experiments have been conducted to determine the required time for the water-side air-conditioning system to take effect in a hotel bedroom was around 15 minutes. Subsequent experiment results had indicated that by focusing on intraday (intra-hour) period, an accurate cooling load prediction could be achieved by using historical cooling load data only. A coupling method that included using ANN to model the chiller performance in different ambient situations and cooling load demands, and then followed by PSO to perform the optimal calculation, was developed for this research project. By focusing on the chiller section only, it can be demonstrated that such ANN-PSO coupling method can lead to a significant energy reduction result.

In general, the performance of an ANN model depends on the quality of data used and the nature of data preprocessing in its training algorithm. Because of the robustness issue, a hybrid data selection algorithm was proposed, which aimed to preprocess the data that was going to be adopted for ANN training and to predict the power output of a water-side air-conditioning system. The research data, which was directly collected from the hotel building management system of a commercial hotel and used to model the air-conditioning system. Several different modelling techniques including typical ANN with no data tampering, ANN with Predictive Data Selection (PDS) method, ANN with Ensemble (ENS) method and ANN with ensemble and hybrid data selection (ENS-PDS) techniques were developed using the same data collected. The performances of these ANN models were then compared, which allowed an on-line method to be constructed. Besides, trials were also conducted to validate and to provide a benchmarking difference between the proposed ANN(ENS-PDS)-PSO coupling method and the traditional fixed set-point method. The verification case study reported 21.5% of energy-saving, which can validate the success of the proposed algorithm.

Date of Award	3 Aug 2020
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Wai Ming LEE (Supervisor)