Executive Summary

1        Introduction

This report’s primary aim is to provide yield projections for the proposed Linear Fresnel Reflector (LFR) technology plant at Collinsville, Queensland, Australia.  However, the techniques developed in this report to overcome inadequate datasets at Collinsville to produce the yield projections are of interest to a wider audience because inadequate datasets for renewable energy projects are commonplace.  Our subsequent report called ‘Energy economics and dispatch forecasting’ (Bell, Wild & Foster 2014a) uses the yield projections from this report to produce long-term wholesale market price and dispatch forecasts for the plant. 

2        Literature review

The literature review discusses the four drivers for yield for LFR technology:

  • DNI (Direct Normal Irradiance)
  • Temperature
  • Humidity
  • Pressure

Collinsville lacks complete historical datasets of the four drivers to develop yield projections but its three nearby neighbours possess complete datasets, so could act as proxies for Collinsville.  However, analysing the four drivers for Collinsville and its three nearby sites shows that there is considerable difference in their climates.  This difference makes them unsuitable to act as proxies for yield calculations.  Therefore, the review investigates modelling the four drivers for Collinsville.

We introduce the term “effective” DNI to help clarify and ameliorate concerns over the dust and dew effects on terrestrial DNI measurement and LFR technology.

We also introduce a modified Typical Metrological Year (TMY) technique to overcome technology specific TMYs.  We discuss the effect of climate change and the El Niño Southern Oscillation (ENSO) on yield and their implications for a TMY.

2.1     Research questions

Research questions arising from the literature review include:

The overarching research question:

Can modelling the weather with limited datasets produce greater yield predictive power than using the historically more complete datasets from nearby sites?

This overarching question has a number of smaller supporting research questions:

  • Does BoM adequately adjust its DNI satellite dataset for cloud cover at Collinsville?
  • Given the dust and dew effects, is using raw satellite data sufficient to model yield?
  • Does elevation between Collinsville and nearby sites affect yield?
  • How does the ENSO cycle affect yield?
  • Given the 2007-12 electricity demand data constraint, will the 2007-13 based TMY provide a “Typical” year over the ENSO cycle?
  • How does climate change affect yield?
  • Is the method to use raw satellite DNI data to calculate yield and retrospectively adjusting the calculated yield with an effective to satellite DNI energy per area ratio suitable?
  • How has climate change affected the ENSO cycle?

A further research question arises in the methodology but is included here for completeness.

  • What is the expected frequency of oversupply from the Linear Fresnel Novatec Solar Boiler?

3        Methodology

In the methodology section, we discuss the data preparation and the model selection process for the four drivers of yield.  We also discuss the development of the technology specific TMY and sensitivity analysis to address the research questions on climate change and elevation.

4        Results and analysis

In the results section we present the selection process for the four driver models.  We also present the effective to satellite DNI ratio, the annual variation in gross yield, the selection of TMMs for the TMY based on monthly yield, the sensitivity analysis results on climate change and elevation, and the frequency of gross yield exceeding 30 MW.

5        Discussion

We analyse the results within a wider context, in particular, we make a comparison with the yield calculations for Rockhampton to address the overarching research question.  We find that the modelling of weather at Collinsville using incomplete weather data has higher predictive performance that using the complete weather data at Rockhampton but recommend using the BoM’s one-minute solar data to improve the comparative test.  Other findings include the requirement to increase the current TMM’s selection period 2007-13 to incorporate more of the ENSO cycle.  There is less than 0.3% change in gross yield from the plant in the most likely case of climate change but there is a requirement to determine the effect of climate change on electricity demand and the ensuing change in wholesale electricity prices.

6        Conclusion

In this report, we have addressed the key research questions, produced the yield projections for our subsequent report ‘Energy economics and dispatch forecasting’ (Bell, Wild & Foster 2014a) and made recommendations for further research.

Publication Details