The variability of both wind and solar energy means that energy production predictions are fundamental both for managing the generation-demand curve and for assessing power generation projects. In order to be able to integrate thousands of GW of variable renewables into the system, we will have to further develop prediction systems to be able to anticipate future generation as accurately as possible. And this is where the problem comes in: the renewable sector is still very optimistic in its predictions of future production.
Types of predictions
We differentiate the predictions between those oriented to the management of the network and those aimed at assessing a renewable project
- Grid management: basically these are necessary for the grid operator to be able to anticipate minutes, hours and even days to the renewable generation curve and thus be able to adjust the demand and generation curve. NREL explains very well the types of predictions in this field:
It is a field where great advances are being made, hand in hand with advances in weather forecasting and pushed by grid operators, who, in order to reach the levels of penetration of renewables in the system, will need not only storage capacity (hydro, batteries, etc), but also medium and long term forecasting capacity.
- Project assessment: these are the predictions associated with a specific project and are used in the development phase of the project to finance it or assess it with a view to its sale. They are based on the extrapolation of real measurements on site and are carried out by specialised independent engineering firms with software designed for this purpose. In principle we have all the ingredients for a good prediction, but nevertheless, historically it has been found that the production of the plants is often overestimated. Let’s see in this article why this is happening
– How accurate are the predictions?
Well, this is not an easy question to answer because both prediction and actual production data are usually not public. Park operators do not make these data public either as they are very sensitive and could affect their business, but there have been some independent studies that have provided very valuable information
- NREL study on wind projects: this is the most ambitious study I know of, as they studied 62 wind projects in the USA from operators such as EDF, EDPR, Enel or Iberdola and with turbines from the main manufacturers such as Vestas, SGRE or GE. Its main conclusions:
- There is an important tendency to overestimate production:
Depending on the filters applied, the deviation varies, but it can be concluded that for projects less than 10 years old, the overestimation of energy is in the order of 5%.
- The most recent parks have less estimation error
This is good news because it seems to be getting better
- KWh Analytics study on solar projects: Given that the solar industry is more recent than wind power, one might think that solar projects have learned from the mistakes of their wind power counterpart and there are no forecasting errors, but surprisingly, the KWh Analytics report concludes that the average deviation between actual and estimated production is 6.3% in favour of the estimate. Surprisingly, this is an even greater error than in the case of wind power and may have to do with the greater immaturity of the industry and the products (irradiance studies, panel degradation, losses in the inverters, bifaciality, etc).
– Why is production overestimated?
Let’s see possible causes:
- Calculation tools: it is clear that calculation tools can be improved, but I do not think that this is the reason because if there were uncertainty, it would be both over- and underestimation and as we can see, this is not the case.
- Quality of measurements: in the case of wind power, wind measurement campaigns last a minimum of 12 months and are carried out by specialised companies, so this does not seem to be the cause
- Manufacturers’ data: the engineering companies that carry out the calculations are based on both the measurements and the operational parameters provided by the manufacturers. And since it is such a competitive sector, these parameters are a sales argument which often “twist” the data to make the most favourable case.
- Developer pressure: although the previous studies are carried out by an independent engineering company, the one who pays is the developer who, in turn, will use the study to seek funding or to sell the project. This is why, explicitly or implicitly, the pressure to make the study “benevolent” is sometimes great, since the profitability or even the viability of the project often depends on its outcome.
- Climate change: although it seems like science fiction, some say that the general rise in temperatures will reduce the wind resource and make it more variable. Solar panels also work less well with temperature. I think that, even if this is the case, it cannot explain the great deviation and furthermore in recent projects.
– And does this have a solution?
- It is clear that experience is improving the calculations and I am sure that we continue to see development in the tools
- Sharing data on both predictions and actual production would be a major advance. Like any statistical process, the more data in the sample, the more reliable the result. And it would also be easier to detect and audit the over-optimistic parameters used.
In conclusion, in view of the new phase where wind and solar will be the main sources of generation, we have to be a credible and reliable sector. To do this, we must move towards a scenario where we work on a mixed public-private data scheme: onsite measurements and financial calculations in the private part of the project while the yield, losses and extrapolation parameters should be public and part of a large database that is fed by each project studied and in operation…today this sounds like science fiction but I believe that in a few years we will see steps towards this scenario