This repository displays the implementation and results of my master's thesis. The implementation is in the directory src/, the experiments can be found in exp/ including a notebook results.ipynb showing the reproducible results.

Building on Jiang and Powell, I model the problem of bidding into the NYISO real-time market as an energy storage operator. I use a simpler backward approximate dynamic programming approach with a scenario lattice to determine a near-optimal policy, i.e. a decision rule for placing bids. In stylized experiments testing the approximation quality, the method performs just as well as in Jiang and Powell while leading to a speedup of 500-1000x in computation time on rather weak hardware. However, it was not yet compared to a state of the art method in a more realistic setting.