Gas storage valuation: Comparison of recombining trees and Least Squares Monte-Carlo simulation

In this paper two approaches for natural gas storage valuation are compared. Whereas the first approach uses recombining trees and the dynamic programming principle the second one uses Least Squares Monte Carlo simulation to compute the optimal storage value and strategy. Both methodologies rely on simulated spot prices. In a first step, therefore the deterministic impacts on the historic natural gas price are determined. The results are then used to simulate appropriate natural gas prices. Whereas the results of these simulation runs are clustered to be applied by the recombining tree valuation, the unclustered simulation results are used to calculate the optimal storage schedule and value via Least Squares Monte Carlo simulation. Applying both approaches on a German case study, the methodologies are compared concerning valuation results, valuation speed and discretization sensitivity.