Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/144572
Title: Predicting chiller system parameters using artificial neural networks : an experimental study
Authors: Venkatesalu Ishwariya
Keywords: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
Issue Date: 2020
Publisher: Nanyang Technological University
Abstract: Air-conditioning is vital for Industrial, Manufacturing, and Commercial applications. Chillers are the core cooling source to provide air-conditioning to the required area which is a part of the HVAC system. Chiller units consume more energy compared to other systems in HVAC systems. Energy conservation provides the opportunity for better innovation and development both economically and environmentally. This dissertation involves the prediction of chiller’s two main output parameters. Chilled water output and Cooling load are parameters considered in this dissertation, with which the energy optimization could be done more effectively. The focus is to Predict the chiller parameters through Multi-input and Multi-Output Regression model using Neural Networks in Python. The uprising technology of Artificial Neural Networks is considered as the new normal for parameter prediction and optimization in HVAC - air conditioning systems. The availability of the excess amount of data through various sources is the main reason for the growth of ANN in the Machine Learning and Artificial Intelligence field. The accuracy of the ANN model relies on data and neural network structure. It has extensively been used in HVAC systems for optimization, including Chiller System which is the high energy-consuming equipment. The chiller system analyzed in this dissertation is of Vapour Compressed, air-cooled condenser, piston compressor and direct expansion evaporator type. Temperature, Pressure and Flow rate of Water and Refrigerant is the key parameters collected through sensors. The virtual integration of the chiller system in LabVIEW software helped in data acquisition of the key parameters. The Neural Network model of Chiller System targets to train and analyze the accuracy of the output predicting terms. This research aims to initiate the use of predicted target variables to adjust the other key parameters in the system to optimize the chiller to be energy efficient. The research work is directed to implement the algorithm to predict the Chiller parameters of Chilled Water Output and Cooling load, using the Artificial Neural Networks (ANN). The proposed chiller energy consumption model was evaluated for accuracy metrics in terms of R^2 multiple-output regression score function on the given datasets. A comparative study is also performed to check the performance of the modelled Neural Network using regression analysis and Ridge regression model. The best curve fit between the actual and predicted parameters graph with higher R^2 score is considered and results were explained to aid the scope of this dissertation. The best R^2 accuracy model is obtained with 29 input features, 9 hidden layer neurons, Adam optimizer with an initial learning rate set to 0.1. The dissertation study results are obtained as R^2 the score of 0.896 on the Training dataset and 0.884 on the testing dataset for the implemented Neural Network. Comparing with the sklearn ridge regression model with 0.8807 Training R^2 score and 0.875 Testing R^2 score. From the results obtained the modelled Linear regression with a neural network is having significantly improved performance in terms of R^2 score comparison.
URI: https://hdl.handle.net/10356/144572
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
Final Version_Amended_DISSERTATION REPORT_PREDICTING CHILLER SYSTEM PARAMETER USING ARTIFICIAL NEURAL NETWORKS_ AN EXPERIMENTAL STUDY_MSc.EEE_VENKATESALU ISHWARIYA.pdf
  Restricted Access
Master Dissertation_Venkatesalu Ishwariya1.65 MBAdobe PDFView/Open

Page view(s)

267
Updated on Jun 21, 2022

Download(s)

4
Updated on Jun 21, 2022

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.