Back to results list
Please use this identifier to cite or link to this item:
|Title:||Hybrid intelligent optimization techniques and its industrial applications||Authors:||Lai, Chung Yee Johnny||Keywords:||Industrial engineering.
Hong Kong Polytechnic University -- Dissertations
|Issue Date:||2014||Publisher:||The Hong Kong Polytechnic University||Abstract:||This thesis focuses on developing efficient methods to solve different real-world optimisation problems. The proposed methods are based on Evolutionary Computation (EC) to perform the optimisation. Results in the following areas will be reported. (1) An improved Differential Evolution with Double Wavelet Mutations (DWM-DE) is proposed as a general-purpose evolutionary algorithm. (2) An improved intelligent optimiser that integrates two optimisation engines is proposed to solve high-dimension and complex optimisation problems. (3) The industrial optimisation problem of Economic Load Dispatch with Valve-Point Loading (ELD-VPL) is used to test the performance of the proposed methods in (1) and (2). (4) The biomedical application of hypoglycaemia detection is employed as a real-world complex classification platform for testing the performance of the proposed methods in (1) and (2). In this thesis, an improved optimisation algorithm and an intelligent optimiser are proposed for high-dimension complex optimisation problems. The algorithm is an improved version of Differential Evolution (DE) called DE with double wavelet mutations (DWM-DE). By introducing the double wavelet mutations in DE, the searching process is enhanced by offering an effective balance between the exploration and exploitation of the solution space for better solution reliability and quality. In the DE mutation operation, a wavelet function is employed to control the mutation factor F. In the DE crossover operation, a wavelet-based second mutation mechanism is proposed to modify the trial vectors within the population. A suite of 29 benchmark test functions is employed to test the performance of the proposed DWM-DE. The experiment results show that the proposed DWM-DE is a useful tool for solving optimisation problems, and it offers better results in terms of solution reliability, solution quality and convergence rate. The experiment results reflect that DWM-DE is particularly suitable for complex problems with a high dimension.
The intelligent optimiser embeds two DE engines into one single system. Through sharing the population information of the two DE engines, the optimiser offers better searching performance. The user of the intelligent optimiser is not required to set the parameter values of the optimiser. The two DE engines operate in parallel and an internal fuzzy controller is employed to adjust the parameter values adaptively in real time during the iteration process. The fuzzy controller takes the searching process information of the population as input. The Student T-Test method is employed to obtain the difference of the population information in the two engines. The resulting intelligent optimiser is capable of dealing with different high-dimension complex optimisation problems efficiently. A suite of 29 benchmark test functions is employed to test the performance of the proposed intelligent optimiser. The experiment results show that the proposed intelligent optimiser a useful tool for solving optimisation problems, and it offers better results in terms of solution reliability, solution quality and convergence rate. In particular, the experiment results show that the intelligent optimiser could offer much better results when the problem is complex and the problem dimension is high (>30). The ELD-VPL problem concerns the process of sharing the power demand among online generators in a power system for the minimum fuel cost. The proposed DWM-DE algorithm and intelligent optimiser are employed to solve the ELD-VPL problem. Two different ELD-VPL problems of different scales have been tested. It is observed that the proposed methods give better optimal costs when compared with other techniques in the literature. A fuzzy inference system (FIS) is employed as a classifier to classify the presence of hypoglycaemic episodes for Type 1 diabetes mellitus (TIDM) patients by measuring some physiological signals continuously from human body. It captures the relationship between the presence of hypoglycaemic episodes and the physiological signals of corrected QT interval of the electrocardiogram (ECG) signal and heart rate. The proposed DWM-DE algorithm and intelligent optimiser are employed to optimise the FIS parameter values that formulate the fuzzy rules and fuzzy membership functions. Data of 15 children with TIDM are studied and used in the training and testing process for the proposed FIS. The experiment results show that the two proposed optimisation methods could offer good performance on training the FIS. The resulting FIS can offer good performance on doing hypoglycaemia detection.
|Description:||xvi, 183 leaves : ill. (some col.) ; 30 cm.
PolyU Library Call No.: [THS] LG51 .H577P EIE 2014 Lai
|URI:||http://hdl.handle.net/10397/6858||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b26960734_link.htm||For PolyU Users||203 B||HTML||View/Open|
|b26960734_ir.pdf||For All Users (Non-printable)||3.4 MB||Adobe PDF||View/Open|
Citations as of Mar 18, 2018
Citations as of Mar 18, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.