from Chile to China
to keep you informed
In an interview with Solarplaza in the run-up to Solar Asset Management Asia 2018, the Programme Leader, Solar Power, Asia-Pacific of the Mott MacDonald Group, Phil Napier-Moore, explores the main differences in irradiation and yield estimation methods. He also touches upon how existing financial models navigate risks and limitations associated with these calculations. Additionally, Napier-Moore discusses the impact of system design on an operational PV plant with respect to performance and O&M.
Starting with solar resource, the two main factors affecting estimation of irradiation for a site are:
site-specific microclimate – mainly assessed through satellite data, given site-specific irradiation measurements are not available for most planned new projects
airborne aerosols – mainly assessed through impact on nearby ground-measured irradiation data, given relevant direct measurements are not generally available
The ground data provided by Japan Meteorological Agency (JMA) stations, with measurements from up to 61 stations, is generally very accurate. In addition, this data can be utilised to understand, with great accuracy, the impact of any aerosols in the atmosphere between space and the ground. Furthermore, there are some widely referenced irradiation databases (i.e. METPV or MONSOLA) which are based on sunshine hours from about 840 JMA stations. However, these databases can have unacceptably large error for application to utility-scale PV plants in our view, given both the conversion from sunshine hours to irradiance and use of older observation data. Therefore, the data obtained from JMA stations does not offer any significant insight into the microclimates at a given planned solar PV plant site, which is the primary challenge in Japan.
On the other hand, satellite data can be used in lieu of ground data to learn more about the microclimates at a specific part of the country to within 3 km resolution (which can also be further enhanced, to e.g. 250m). Therefore, the effect of microclimates on cloud patterns, the main driver of local irradiation conditions, can be identified. Nevertheless, satellite data should not be used alone; although it can supply a reading at any part of the country, it is not always accurate, often largely due to its lack of quality input data on aerosols.
One golden rule when dealing with satellite data is that it should always be validated with ground data in a representative location. It is noteworthy to mention that in numerous parts of the world, validation in a representative location could simply imply being situated in the same country depending on the complexity of the climate. For instance, the majority of PV plant sites in Chile and South Africa have fairly simple and plain topography with limited variation in irradiation, in which case a validation of data source 100 km away could be sufficient to prove its accuracy in another location. However, evidence from validations shows this does not apply to Japan due to the complexities associated with its microclimates as well as variations in pollution patterns over distance and through time. This results in a significant difference in the performance of satellite data across the Japanese archipelago.
In conclusion, using a combination of both satellite and ground data is the most plausible pairing in Japan as each method is limited in the nature of data it provides. This may spark the question of why using other database sources quoted in the market are not suggested. The rationale for this is the significant weaknesses linked to those methods, which has an impact on accuracy. Furthermore, there are not any reliable approaches to combine them with other data sources to enhance accuracy.
The yield estimation areas with most remarkable systematic differences amongst different players in the market are in shading and snow losses. Another point worth adding is that many still utilise a simplified approach under a JIA standard (C 8907:2005), which was primarily intended for rooftop solar and kilowatt projects when it was initially introduced. This standard entails some simplistic factors on matters including DC losses, shading and spectral performance. This approach is not recommended for utility-scale megawatt PV projects as the simplifications negatively affect the accuracy of the assessment.
“Generally, the best practice would be to use a method that combines both linear shading and non-linear shading effects also called electrical shading effect,” says Napier-Moore. He adds that this is considered industry standard internationally but not in Japan so far. Japan also has a number of very complex sites, for example golf course sites. The shading will vary on different parts of the site depending on slope steepness and direction of the underlying ground of the site, and on the way modules are mounted. It can very well be the case that some areas are best left empty due to the excessive amount of shading. This can prevent reducing the average ROI of the project.
For topographically complex sites, manual construction of a digital model of the plant is not only acutely time-consuming and labour-intensive but makes detailed layout optimisation impractical. Use of 3D layout simulation packages serves as an optimal alternative to increase efficiency and accuracy. These packages “autopopulate arrays based on a digital terrain profile; Helios 3D is the package that we use, as well as several EPCs that we work with, and with which we have been able to optimise large PV projects in Japan up to 258 MWp” explains Napier-Moore.
There are not widely-recognised industry standards to estimate the snow-related losses, which can have a decisive impact of the project’s performance. There are several reasonably plausible models available, however these are not used in a widespread manner. Nonetheless, these models have been successfully deployed in Japan. The main challenge is the insufficiency of snowfall data to represent the site and that snowfall significantly fluctuates from year-to-year. Thus, obtaining a definitive idea of long-term snow conditions at the site is difficult. Napier-Moore adds that “this area is relevant to yield estimation but also to design for successful operations. When we talk about design for snow loads, one critical input is maximum snow depth at the site”.
The design codes leave some room to select the maximum snow depth input depending on the location of the site; “In reviewing completed projects, we have often seen the least conservative option used rather than the most conservative value, to save costs on mounting structure design, and we have seen a significant number of failures.” adds Napier-Moore. Impact of soil heave when frozen can also often be neglected. The snow depth is additionally crucial regarding O&M, particularly in terms of snow clearing strategies and the site’s accessibility during snowy periods. This has an impact on a myriad of factors including the structure height, the site road layout, access to the inverter houses and so forth.
With respect to snow clearing strategies, the firm has seen limited progress on robotic applications for snow removal purposes. However, another potentially feasible technology is reverter technology, which is particularly useful for clearing snow from the PV modules rather than the inverter houses, access roads, etc. This technology injects DC currents into the modules in order to melt the snow. Furthermore, there are coatings available that help the snow slide off the modules. Another possible approach would be to utilise external heating grids, which can be used to remove snow from the site roads as well. These technologies are still in their infancy in terms of their application at utility scale.
The Mott MacDonald Group has worked on more than 200 PV projects in Japan with a capacity of just under 5 GW. One of the common issues observed has been the basic efficiency of the cables and transformers diverging from international standards. Napier-Moore attributes this to the fact that “There are no specific standards that are normally referenced in construction contracts for minimum efficiency or maximum losses of cables and transformers”. With respect to observed failures, it is more commonplace for structural or civil issues to occur, many of which are related to adequate drainage systems. This can lead to soil erosion, which may affect the stability of structures and downstream water quality.
With regards to risks in design, there is commonly a construction cost contingency allowed in the financial model. There are two scenarios depending on whether the issues are detected either prior to or after the plant is handed over. It is possible to ask the EPC to amend the issues within the lump sum fee, which is evidently only feasible if there is sufficient supervision of the EPC contractor beforehand. In the case that issues are not spotted and cause a failure during the operation, typical contingency levels are lower and do not allow for capital replacement or significant repairs.
All in all, lessons learned as the Japanese PV market has matured, and given the high early feed-in tariffs, offer major opportunities to revisit the early plant designs, for example with regards to cable and transformer selection and the plant monitoring system to identify possible improvement opportunities. While replacement of AC systems offers a low payback period, DC-AC ratio on plants can also be re-visited as a further optimisation opportunity.