Improved uncertainty analysis for tidal energy project development

Abstract

High investment risk is a key barrier to the commercialisation of the nascent tidal energy sector. An increase in investor confidence can unlock funding for early arrays, the lessons from which can provide further de-risking, leading to further investment. This thesis focussed on increasing investor confidence by improving the uncertainty analysis methods used to quantify the overall uncertainty in key investment decision metrics; energy yield, levelised cost of energy (LCOE) and internal rate of return (IRR). A Monte Carlo Analysis (MCA) framework for tidal energy annual yield uncertainty analysis was developed and compared to the currently recommended ISO-GUM method. It was shown that key assumptions implicit in ISO-GUM are inaccurate for most realistic projects. Crucially, the resultant error provides an overly optimistic view of a project's P90 energy yield. By modelling a range of realistic projects, it was shown that the ISO-GUM P90 yield overestimate exceeds 2% for a maximum resource uncertainty between 4% and 11%, depending on the project, with increasing uncertainty leading to larger errors. It is difficult to judge accurately where within that range a given case crosses the 2% error threshold, as it is a complex function of numerous project specific variables. This undermines confidence in ISO-GUM results, even in cases where the method may be acceptable, because it is not possible to deduce the validity for a particular project a priori. MCA does not make the same assumptions and provides consistently accurate results. A modification to the standard ISO-GUM process was also proposed as a simpler alternative to MCA, with an improvement in results compared to the standard method, but the residual error would still remain unquantified. A generic cost modelling tool for probabilistic discounted cashflow analysis using MCA was also developed. The tool accepts user specified uncertainty distributions in a multitude of flexibly defined input variables defining a project's CapEx, OpEx, yield and finances to produce distributions representing uncertainty in LCOE and IRR. It was compared to commonly used deterministic methods for a realistic tidal energy project. MCA provides highly resolved results compared to the point estimates from deterministic methods. The improved decision support provided by MCA was demonstrated and the scope for misinterpreting the deterministic outputs was highlighted. The significance of several common cost modelling assumptions was tested and the difference between probabilistic and deterministic sensitivity analysis was highlighted. A probability weighted deterministic method was suggested and shown to provide useful indicative results at a reduced effort compared to MCA. Finally, the impact of the ISO-GUM P90 yield error on the P90 LCOE and IRR was quantified for several cases by propagating the ISO-GUM and MCA yield uncertainty distributions through the cost model. MCA propagates input distributions through the functional relationship between the inputs and outputs. For any application, this reduces the unquantified approximations in the results compared to the simpler methods considered. This leads to not only more accurate results, but also a higher confidence in the results. The use of MCA is therefore recommended for annual yield and financial performance uncertainty analysis for tidal energy projects.