Optimal gas storage valuation and futures trading under a high-dimensional price process

We study the problem of optimal gas storage valuation under a high-dimensional multifactor price model. The problem is modeled as a Markov decision process which leads to a stochastic version of the popular rolling intrinsic value. We show that the rolling intrinsic solution is optimal for the case of extreme risk aversion when using the nested conditional value-at-risk. We solve the problem by combining optimal quantization with approximate dual dynamic programming. Our quantization method reduces the high-dimensional multifactor price model to a discrete lattice. We find that it is necessary to match price expectations on the lattice with price expectations of the continuous process. In a numerical study, we demonstrate that our approach yields 26%-55% higher values as compared with another state-of-the-art approach from the literature on the exact same problem instances.