-Henri Poincare
Taking the spirit of the quote above Vulcan is now able to predict and foresee US coal stocks using our advanced satellite imagery.
In the previous article releases, we mentioned that US coal stocks are reported but on a 2 to 3-month lag. To solve this, we predicted that if one has a sufficient sample of coal plants, one could predict all other coal plant volumes. Using our satellite data on the selected coal plants in the US, we then set out to test it.
Our selected coal plants focused on the plants that had the largest propensity to react to market conditions rather than the ones that run fully regulated in a constant state. With these plants, we were able to run a historical regression analysis against the total fleet and produce a high statistical connection (R^2 ~0.9). Then after applying it to our selected coal stocks, we were able to start forecasting US coal stock 2 to 3 months out. In the graph below you can see a back test conducted the results of which are a direct result of our regression analysis.
The lowest point of the stock level was identified to be September with US coal stockpiles climbing back up in October. We anticipate the EIA to report the stock levels to continue to rise in November and for some inventories to even approach must burn levels before a steep drop in December. Compared to the spring shoulder months, gas as a percentage of total coal and gas generation has been lower so perhaps, we did see some force burns. This does highlight one value of being able to see coal stocks more up to date so one can favor down the gas burns when coal inventories get to the high end.
All this data will be available on the Vulcan dashboard for subscribers to Vulcan.
