Heterogeneous particle swarm optimization with an application of unit commitment in power system

To meet increasing electricity demand and reduce tremendous greenhouse gas emission at the same time, the power infrastructure is integrated with renewable resources. Due to intermitted and unpredictable nature of renewable resources, the amount of generated power generated from renewable resources is uncertain and unreliable. The power scheduling problem becomes so complex that traditional optimization methods are no longer suitable for the current power systems. Computational intelligence (CI) methods have the adaptive ability to address this kind of complex and uncertain optimization problems from changing environments and have been demonstrated in many power system problems such as decision-making, forecasting, unit commitment, optimal power flow, renewable energy power systems, etc. Therefore, the particle swarm optimization algorithm, a variant of CI technique, is investigated and deployed to tackle unit commitment power problem with/without considering uncertainty in renewable resources. The first part of the thesis is focused on the development of new particle swarm optimization (PSO) algorithms. PSO belongs to a class of evolutionary algorithms as well as swarm intelligence. PSO is known to be easy to implement and effective in solving large-scale non-linear optimization problems. In this thesis, a new heterogeneous comprehensive learning particle swarm optimization (HCLPSO) algorithm is proposed to improve the performance of PSO by balancing exploration and exploitation. In addition, Ensemble of particle swarm optimization algorithms called (EPSO) algorithm that combines various PSO algorithms is also proposed. The EPSO algorithm is focused on combining various PSO algorithms to improve optimization capability. Before addressing unit commitment problem in power system, the proposed algorithms are evaluated on CEC2005, CEC2013 and CEC2014 benchmark problems. A comparative study is conducted between the proposed PSO algorithms and recent algorithms. The experimental results show the performance of newly proposed PSO algorithms is better than other state-of-the-art PSO algorithms. In the second part, we focus on the unit commitment (UC) problem, an important power optimization problem. A successful practical solution for this problem can benefit the industry in a huge operational cost saving. The problem determines start-up and shut-down schedule of power generating units over a scheduled period while meeting the system demand and spinning reserve requirements at minimum production cost. In this thesis, a new priority listing method called time-ahead priority listing is introduced to generate a feasible unit schedule and the proposed HCLPSO is used to address economic dispatch of the UC problems of 10-, 20-, 40-, 60-, 80- and 100-bus power systems over 24 hour scheduling horizon. The performance of a hybrid solution of combining time-ahead priority listing and HCLPSO is compared with other hybrid methods. The proposed hybrid model outperforms other hybrid models on all small and large power systems and provides the lowest minimum production cost. In the third part of the thesis, an independent system operator of 10-bus power system is integrated with renewable resources of solar and wind energies. A set of the scenarios are considered for uncertainty in solar, wind energies and power demands. Binary HCLPSO algorithm is proposed to solve cost-emission optimization of UC problem, considering reducing the greenhouse gas emission. The performance of binary HCLPSO algorithm is compared with integer-coded GA, improved binary PSO and hybrid approach of binary PSO and real-coded PSO. The results showed that binary HCLPSO algorithm is able to handle UC problem under certainty in renewable resources.DOCTOR OF PHILOSOPHY (EEE