Monday, 27 January 2014

What makes a WEC concept 'good'?



The topic of innovation in the design of WECs has been a bit of a recurring theme for me lately. It has cropped up in the form of an innovation competition for WECs, new concepts aired at this year's All Energy Opportunities conference, and in Weber's paper on the Performance/Readiness Matrix. All this leads to a conundrum that has been puzzling me for some time now:


How can you tell if a particular WEC concept is any good?”


Ultimately, the measure of fitness must be the cost of energy (CoE). This is often referred to as the levelised CoE, as the costs and income are averaged over the entire life-cycle of a project. The term CoE is used in many contexts. It can applied to particular projects (arrays); this is useful for plotting the trend in CoE on a project by project basis. It can also be averaged over all operational plant of a particular technology; this is useful for plotting learning curves - the trend in CoE with installed capacity. It can be used to describe the actual CoE calculated in retrospect, and it can also refer to the estimated value, calculated before all the project costs and incomes are known.

CoE is particularly hard to estimate for a new technology, where no projects have yet completed a full life-cycle. Uncertainties exist because some things are hard to know for sure, some things cannot be known in advance, and some things need to be discovered by doing.

CoE uncertainties – some things are hard to know for sure...


The devil is often in the details, and the details of how to calculate CoE, or what will impact on CoE, is still open to debate and interpretation. Some of this uncertainty could be reduced by agreement on standards (which will require both debate and consensus), further academic research, or commitment on a policy level. Examples of such types of unknown are:
  • Future social value attached to wave power, reflected in subsidies, grid costs, tariffs, or tax breaks.
  • Future installed capacity of wave power, which will determine cost savings due to increased operational experience, mass manufacture, specialist supply chains, site availability, grid capacity, etc
  • Whether finance costs and R&D costs (e.g. prototypes, working up programs, bench tests, certifications) should be included to any extent?
  • What discount rate should be applied to future costs (decommissioning/salvage) and incomes (power sales/value of salvage).
  • Drift in average resource characteristics due to climate change.
  • Resource availability: future vessel hire costs; bottlenecks in component (e.g. sub-sea cable) availability, or changes in raw material prices, due to competition from other sectors.

CoE uncertainties – God only knows!

Time machine

Some of the inputs to a CoE function contain inherent uncertainty. Barring a time machine or divine knowledge, there are some things that cannot be known with absolute certainty:
  • Annual variation in resource.
  • Future CoE of competing power technologies (particularly fossil fuel prices).
  • Availability and maintenance costs - there are several aspects that are outwith the control of the operator: the timing of faults requiring reactive maintenance, weather windows, component degradation and faults, accidents, and extreme weather events.
  • Less efficient power capture due to component degradation.
  • Project lifetime: how soon before the additional costs to extend plant lifetime are greater than the additional revenue gained as a result?
Although these things can never be known with absolute certainty, it is possible in some cases to reduce the levels of uncertainty.

CoE uncertainties – learn by doing


Barring annual resource variation, extreme events, fossil fuel prices, accidents and weather windows, operational experience can reduce the uncertainty bounds around the performance, availability, maintenance costs and project lifetime. In particular, sea trials can be very valuable. They can reduce risks in the following manner:

  • Provision of operational data to prove performance.
  • Provision of financial data to prove capital and operational expenditure.
  • Proof of system reliability in typical conditions, which narrows the uncertainty around availability and maintenance costs.
  • Proof that the engineering requirements have been met.
  • Identification of any non-linear behaviour not uncovered by simpler tests. Simpler tests, such as tank trials or smaller scale field trials, may not adequately identify such behaviour, due to simplified test conditions, scaling effects, or different design details resulting from the absence of challenges particular to a full scale prototype. Identification of non-linear behaviour reduces the uncertainties around performance and extreme loads.

CoE reductions due to tests at high TRLs


Furthermore, the information from sea trials can be useful for reassessing design specifications. Expenditure records from sea trials highlight actual costs of components and processes, allowing the identification of any that dominate CoE. This could suggest design changes leading to a lower CoE. Likewise, identification of non-linear behaviour and under-performing components is required in order for these problems to be addressed.

This is one reason why there has been such a strong focus on reaching high technology readiness levels (TRLs) in the wave power industry. Not only do prototype sea trials reduce the uncertainty bounds around many CoE inputs (performance, availability, capital expenditure, operational expenditure, and the full scale engineering requirements), but they are a way of learning by doing. There are some cost reductions that can be achieved just by testing at full scale, and learning from the mistakes.

CoE reductions due to tests at low TRLs


However, there is another route for cost reductions: innovation. Development and testing of innovative ideas is more expensive at high TRLs (appending existing designs, or going straight to sea trials) than at low TRLs (starting from scratch). If we are going to consider searching for innovative WEC concepts, it is worth considering the problems particular to working at low TRLs, and how to address them.

Sources of uncertainty at low TRLs


As already discussed, working at low TRLs (numerical modelling and tank testing) is an economically efficient way of conducting an innovative design process. However, the boundaries of uncertainty on estimates of CoE are wide.

Unconscious bias – It is human nature to treat uncertainty in an optimistic manner. It should be expected that at low TRLs, in the absence of evidence otherwise, unconscious (non-blind) bias would result in an overestimation of performance, and an underestimation of costs and engineering requirements.

Non-linearities – Furthermore, any behaviour not yet identified due to limits of investigations at low TRLs, will result in over-optimistic CoE estimations. This is because such behaviours are likely to be non-linear. Non-linearities, like taxes, result in you taking home less than you expected.

Operating principles – If the basic operating principles - the physics of typical dynamic behaviour - are not fully understood, this could result in either an underestimation or an overestimation of performance. Neither case is helpful, as it could lead to the rejection of potentially viable concepts (in the case of performance overestimation, viable concepts could be rejected in favour of the misunderstood concept). An understanding of the basic operating principles is crucial for innovative design. This is particularly the case at low TRLs: it is possible to test a design that 'feels right' with a sea-going prototype, and from the results to know within acceptable uncertainty bounds whether it was a viable device, without any understanding of the operating principles. However, the less risky approach of weighing up alternative concepts at lower TRLs is highly dependant on an understanding of the operating principles.

Addressing uncertainties at low TRLs

Unconscious bias – The problem of unconscious bias can be addressed if one team tests several very different concepts, and is aware of the importance of impartiality. In a research context, this might be an academic institution that has not previously had in-house WEC device development. In an industrial context, this might happen if a larger company were to work in partnership with several device developers simultaneously.

Non-linearities – the problem of potential unseen show-stoppers that only become apparent at higher TRLs can be mitigated by ensuring that several concepts are allowed to progress to higher TRLs. Market forces have already had the impact of supporting concepts that have very different niches in terms of location and operating principle. Genetic diversity makes for a stronger gene pool.

Operating principles – it is vital that anyone working in technology development of WECs understands the operating principles of WECs, particularly if decisions are based on comparisons of data that has high levels of uncertainty (associated with low TRLs). I would recommend this as a worthy topic for debate and research, and plan to contribute to this discussion in the near future.

Performance as a metric – At early stages of development, CoE might not be the best criteria for weighing up alternatives. The uncertainty limits on CoE are wide because it is made up of several inputs which each have wide uncertainty limits. Simpler metrics such as performance to volume ratio, or performance to weight ratio, might be more suitable, as they contain only a sub-set of the uncertainty contained in CoE. Note they also contain only a subset of the information; good performance in a numerical model or tank test does not guarantee good CoE in sea trials. At low TRLs, good relative performance can be a useful proxy for CoE. At higher TRLs, an estimation of CoE is a better metric.




Image credit:
Blogger's own photo of a mural in Bristol, corner of Hillgrove and Jamaica Str. There's a slight possibility it was inspired by Hokusai's Great Wave.

No comments:

Post a Comment

Comments