Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/83483
DC FieldValueLanguage
dc.contributorDepartment of Building Services Engineering-
dc.creatorLi, Ning-
dc.identifier.urihttps://theses.lib.polyu.edu.hk/handle/200/6826-
dc.language.isoEnglish-
dc.titleArtificial neural network based modelling and control of a direct expansion air conditioning system-
dc.typeThesis-
dcterms.abstractDirect expansion (DX) air conditioning (A/C) systems have been increasingly used over the recent decades in buildings, especially in small to medium scaled buildings. This is because they are more energy efficient and more flexible in installation, but cost less to own and to maintain, as compared to large chilled water based central A/C systems. Conventional DX A/C units equipped with single-speed compressor and fan rely on on-off cycling of compressor to maintain the indoor dry-bulb temperature, leading to either a space overcooling or an uncontrolled equilibrium indoor air humidity, and resulting in a reduced level of thermal comfort for occupants and low energy efficiency. With the development of variable-speed drive technology, it becomes possible for DX A/C units to have the speeds of their compressors and supply fans varied, so as to achieve simultaneous control over both indoor air temperature and relative humidity (RH). On the other hand, artificial neural network (ANN) has been proven to be powerful in modeling the dynamic operating performance of a nonlinear multivariable system, because ANN has a powerful ability in recognizing accurately the inherent relationship between any set of inputs and outputs without requiring a physical model. This ability is essentially independent of the system complexity such as nonlinearity, multiple variables, coupling, with noise and uncertainty. An ANN-based control strategy which could deal with a nonlinear multivariable complex system, such as a DX A/C system, can then be developed. As an intelligent nonlinear dynamic control method, an ANN-based control strategy offers a viable solution to the control over complex systems. This Thesis reports on a study of developing a multi-input multi-output (MIMO) control strategy that can simultaneously control indoor air temperature and humidity by varying speeds of both compressor and supply fan in a DX A/C system, using ANN-based modeling and control approaches. The Thesis starts with reporting the development of a two-in two-out ANN-based steady-state model for an experimental variable speed DX A/C system. The model can be used for simulating the steady-state total output cooling capacity (TCC) and Equipment Sensible Heat Ratio (SHR) of the DX A/C system under different combinations of compressor and supply fan speeds. Extensive experiments were carried out to collect data for ANN training and testing, as well as for validating the ANN-based steady-state model developed. The ANN-based steady-state model has been validated experimentally by comparing the measured results of TCC and SHR using the experimental DX A/C system with the predicted results using the ANN-based steady-state model developed. The ANN-based model developed can be used to predict the steady-state operating performance of the experimental DX A/C system with a higher accuracy.-
dcterms.abstractSecondly, the Thesis presents the development of an ANN-based dynamic model for the experimental DX A/C system, linking the indoor air temperature and humidity controlled by the DX A/C system with the variations of compressor and supply fan speeds. The ANN-based dynamic model has been validated experimentally by comparing the measured results of indoor air dry-bulb and wet-bulb temperatures under different compressor speed and/or supply fan speed using the experimental DX A/C system, with the predicted results using the ANN-based dynamic model developed. The calculated values of average relative error (ARE) and maximum relative error (MRE) when experimentally validating the ANN-based dynamic model developed indicated the high accuracy of the ANN-based dynamic model developed. Thirdly, using the ANN-based dynamic model developed, an ANN-based controller for controlling simultaneously the indoor air temperature and humidity by varying the compressor speed and supply fan speed in a space served by the experimental DX A/C system was developed. This ANN-based controller was designed using the direct inverse control (DIC) strategy. The controllability tests including command following test and disturbance rejection test were carried out using the experimental DX A/C system, and the test results showed that the ANN-based controller developed was able to track the changes in setpoints and to resist the disturbances, with adequate control accuracy and sensitivity. Finally, to further address the problem of limited controllable range for the ANN-based controller, which is common to all controllers developed based on system identification, an ANN-based on-line adaptive controller has been developed and is presented. The ANN-based on-line adaptive controller was able to control indoor air temperature and humidity simultaneously within the entire expected operational range by varying compressor and supply fan speeds. The controllability tests for the controller were carried out using also the experimental DX A/C system. The test results showed that the ANN-based on-line adaptive controller developed was able to control indoor air dry-bulb and wet-bulb temperatures both near and away from the operating condition at which the ANN-based dynamic model in the ANN-based on-line adaptive controller was initially trained, but within the entire range of operating conditions, with a high control accuracy.-
dcterms.accessRightsopen access-
dcterms.educationLevelPh.D.-
dcterms.extentxxiii, 195 leaves : ill. ; 30 cm.-
dcterms.issued2012-
dcterms.LCSHAir conditioning.-
dcterms.LCSHAir conditioning -- Control.-
dcterms.LCSHNeural networks (Computer science)-
dcterms.LCSHHong Kong Polytechnic University -- Dissertations-
Appears in Collections:Thesis
Show simple item record

Page views

34
Last Week
2
Last month
Citations as of Mar 24, 2024

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.