We calculated the IRR of >500,000 wind farm layouts in a 200km x 200km radius in the mountains of South Vietnam.
Local conglomerate in Vietnam securing sites for post FiT (feed-in-tariff) era beyond 2021.
Preliminary LCOE and IRR of 20 potential project on shore wind farm sites.
The customer is looking to quickly secure as many project sites as possible across 4 provinces in the south of Vietnam. They are a large regional conglomerate with diversified holdings looking to participate in Vietnam's onshore wind market beyond the FiT scheme of 2021. The team is a mix of investment professionals, project managers and engineers. So far they have been using a mix of GIS software together with Microsoft Excel to perform due diligence early stage acquisition opportunities as well as look for greenfield locations.
After 9 months of a rinse-repeat cycle performing origination through business development and then evaluating sites on a desktop basis, the customer was finding out that the asking acquisition price of early stage projects were at wildly varying valuations due to the euphoric nature of the Vietnam wind market. Even locations that have no existing development of wind, land prices and permitting risks rose immediately upon the customer revealing they are here to develop a wind farm.
Furthermore, the customer projected that there will no longer be feed in tariffs post 2021 and following the trend of global wind development, the PPA will either be a reverse auctioned or will be on a merchant basis. This means they had low margins of error for a) the acquisition price / development cost of sites, b) wind resource and energy output, c) civil costing and electrical losses and d) offtake curtailment.
Solution - Part 1 - Large area greenfield prospecting
The customer chose a prospecting area that straddled 4 provinces and encompassed an area approximately 200km x 200km in the south of Vietnam.
First we performed a down-scaling of mesoscale maps such as the Global Wind Atlas from a 3km resolution to 25 meters (at which it is a microscale map). The downscaled map provides the statistical distribution of windspeeds and wind directions over time at 25m intervals. The downscaling process for a 200km x 200km radius will produce a map with 64 million points x 12 directions x 3 Weilbull parameters = 2.3 billion values that describes the wind resource of this area. Traverse runs its own in house distributed servers comprising of over 240 CPU cores to run wind flow modelling software such as WasP.
Secondly we ran an "Inversed Nearest Roads" algorithm where for every 10m point (typical width of a road) in the 200km x 200km boundary, we performed an increasing radius search of the nearest access road. This is not a direct straight line distance as a straight line distance will pass under mountains, but instead the shortest road with an 8 degree slope with 45m turning radius using a modified variant of A* search. This produces a map with 400 million points where each point encodes a) the coordinates of the nearest major access road and b) the distance of this access road from that point. The same algorithm was applied to ask the question of where are the nearest substations ("Inversed Nearest Substations") but this was run at a much lower resolution.
Thirdly, the road and substation maps were combined with a map of protected areas, a map of urban zones and the previously produced wind resource map. An additional map called "terrain ruggedness" was created as a rough estimator of civil works. Altogether, we are then able to produce a multi-variate heatmap and divide it into 50 zones of 20km x 20km for 150MW sites and 50 zones of 10km x 10km for 75MW sites, for a total of 100 rectangular zones.
Fourth, these zones are then applied what is commonly known in the industry as "micro-siting". But we bring this to a whole new level. We perform micro-siting across t 100km x 100km area. For every rectangular zone of 10 to 20km, we generate seed layouts and calculate the gross energy and net energy. Here permutations on turbine type, hub height, number of turbines (which contributes to total installed capacity), position of turbine and other civil engineering considerations are applied.
After nudging the turbines into the highest energy positions, the candidate layout is snapshotted and road path + inter-array electrical loss optimization is performed. Then it goes back to energy optimization. Just from the energy and road costing we pass this into a simplified financial model and get a IRR + LCOE score. We call this a "tick-tock" optimization and for each rectangle, we perform between 5,000 to 25,000 such iterations. In the client case, a total of 500,000 layouts + configurations were permutated within the whole area of interest.
Solution - Part 2 - Due diligence credits
The first part (50 DD credits for early projects) performs the same exact methodology as the large scale search described above but for individual projects. The client can further upload project specific information such as no-go zones, substations and transmission lines, access points, weather station data etc.
The second part (25 DD credits for late stage projects) has a case study all on its own which you can read more about here. This is more of a traditional post measurement, project finance class WRA (wind resource assessment) and EPA (energy production assessment) combined with civil, electrical BOP cost and losses and IRR / LCOE simulations.
Click the button below to request for the sample reports that is a representative example of what we delivered in this case study.Download Sample Reports