While large-scale wind farms and solar power stations have been used widely as supplement to the nuclear, fossil fuels, hydro and geothermal power generation, at smaller scales these resources are not reliable to be used independently and may result in load rejection or an over size design which is not cost effective. A possible solution to solve this issue is using them as part of a hybrid power system. Complexity in design and analysis of hybrid renewable energy systems (HRES) has attracted the attention of many researchers to find better solutions by using various optimisation methods. Majority of the reported researches on optimal sizing of HRES in the literature are either only considering one objective to the optimisation problem or if more than one objective is considered the effect of uncertainties are ignored. This dissertation work investigates deterministic and stochastic approach in design of HRES. In deterministic approach it shows how adding a battery bank to a grid connected HRES might result in more cost effective design depending on different grid electricity prices. This work also investigates the reliability of HRES designed by conventional deterministic design approach and shows the weakness of common reliability analysis. To perform the stochastic approach the renewable resources variation are modelled using time series analysis and statistical analysis of their available historical meteorological data and the results are compared in this work. Chance constrained programming (CCP) approach is used to design a standalone HRES and it is shown that the common CCP approach which solves the problem based on the assumption on the joint distribution of the uncertain variables limits the design space of problem. This work then proposes a new method to solve CCP to improve the size of design space. This dissertation comprises multi-objective optimisation method based on Non-dominated Sorting Genetic Algorithm (NSGA-II) with an innovative method to use CCP as a tool in estimating the expected value of the objective function instead of Monte-Carlo simulation to decrease the computational time.
|Publication status||Accepted/In press - Feb 2015|