This thesis is structured as follows:
Chapter 2 introduces a novel optimisation algorithm, Function Optimisation by Learning Automata (FOLA). It mainly includes the following aspects em- ployed in FOLA: dimensional search, path value, action selection, reinforce- ment signal, cell value and state memory. FOLA has been compared with a number of classical Evolutionary Algorithms (EAs) on 13 high-dimensional benchmark functions, which represent a wide range of challenging optimisa- tion problems. The simulation results presented in this chapter have shown that FOLA is able to undertake search in continuous states and achieve ac- curate solutions efficiently. The analysis of FOLA, in terms of convergence characteristics, computation time and parameters, is also carried out in the chapter. To further evaluate FOLA’s performance, it is also compared with two popularly used EAs and four newly-proposed EAs, on 9 complex multi- model benchmark functions. The experimental results demonstrate the supe- riority of FOLA over other algorithms for most benchmark functions, in both the convergence rate and the accuracy of finding optimal solutions. FOLA is able to reduce computation time greatly, especially for high-dimensional functions.
Chapter 3 presents a multi-objective optimisation by learning automata (MOLA). The approaches employed in MOLA, including reinforcement scheme, form- ing of Pareto set, and the process of searching and learning, are introduced in detail in this chapter. MOLA has been compared with two popular weighted- sum based multi-objective EAs on four multi-objective benchmark functions that comprise low and high-dimensional models, convex and non-convex mod- els, and continuous and discontinuous models respectively. The simulation results of these algorithms have been analysed with respect to the accuracy of the experimental results and the range of the Pareto front. MOLA is also compared with multi-objective evolutionary algorithm based on decomposi- tion (MOEA/D) and non-dominated sorting genetic algorithm II (NSGA-II),
1.4 Overview of this Thesis 26
on thirteen multi-objective benchmark functions, which are difficult to resolve due to complicated Pareto set. The simulation results have presented the great superiority of MOLA over MOEA/D and NSGA-II, as MOLA can obtain more accurate Pareto optimal solutions, and find wider range of the Pareto front.
Chapter 4 describes two applications of FOLA. The first concerns with the opti- mal power system dispatch and voltage stability problem. The problem is to reduce the fuel cost whilst enhancing the voltage stability of the power sys- tem. Simulation studies are undertaken on the standard IEEE 30-bus and 57- bus power systems respectively. The advantages of FOLA have been demon- strated by comparing its performance with that of improved PSO and GA. The second case concerns the renewable energy of wind power. Wind power pen- etrated power systems bring new challenges to power system operation since the dynamic nature of wind power. In the thesis, FOLA is applied to tackle the optimal power flow problem which aims to achieve economic power dis- patch and voltage stability enhancement in dynamic wind power penetrated systems. FOLA is compared with the improved PSO and GA, on the modi- fied IEEE 30-bus and 57-bus power systems respectively, which are penetrated with time-varying wind power. The experimental results have demonstrated that FOLA outperforms PSO and GA, as it tracks the changing system config- uration more rapidly and accurately than the improved PSO and GA. In addi- tion, FOLA is compared with CLPSO and CPSO based on the modified IEEE 118-bus power system. FOLA is able to minimise the fuel cost and enhances the voltage stability of the power system more efficiently in comparison with the other two algorithms.
Chapter 5 presents three applications of MOLA. The first two are concerned with the increasing attention on pollutant emission and the use of wind power. The chapter applies MOLA to solve the problem of economic emission dispatch and voltage stability enhancement in wind power penetrated power systems, which are incorporated with fixed-speed and variable-speed wind generators
respectively. MOLA is compared with MOEA/D and NSGA-II on the mod- ified IEEE 30-bus power system and new England test power system respec- tively. The simulation results have demonstrated the superiority of MOLA over NAGA-II and MOEA/D. The third application of this chapter concerns the deregulated market. The deregulation of the power market creates more competition and more trading mechanisms for market players. MOLA is applied to maximise social benefit and enhance voltage stability simultane- ously in a deregulated electricity market. MOLA has been compared with MOEA/D and NSGA-II, based on the challenging optimisation problems on the IEEE 30-bus power system. In this application, the simulation results have again demonstrated the superiority of MOLA over MOEA/D and NSGA-II, as MOLA can find wider and evenly distributed Pareto fronts, and obtain more accurate non-dominated solutions.
Chapter 6 concludes this thesis based on the experimental results obtained in this study. The merits of FOLA and MOLA are further discussed, as well as their application capabilities. Additionally, this chapter includes suggestions for future work.
Appendix gives the details of single objective and multi-objective benchmark func- tions employed in the thesis, as well as the notations used in the thesis.