Optimal Prosumer Decision-Making Using Factored MDPs

Authors: Angelos Angelidakis, Georgios Chalkiadakis

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental simulations verify the effectiveness of our approach; and show that our exact value iteration solution matches that of a state-of-the-art method for stochastic planning in very large environments, while outperforming it in terms of computation time.
Researcher Affiliation Academia Angelos Angelidakis Electronic and Computer Engineering Technical University of Crete Greece, GR73100 aggelos@intelligence.tuc.gr Georgios Chalkiadakis Electronic and Computer Engineering Technical University of Crete Greece, GR773100 gehalk@intelligence.tuc.gr
Pseudocode Yes Algorithm 1: Value iteration for solving the FMDP
Open Source Code No The paper does not provide a specific repository link or explicit statement that the source code for the described methodology is publicly available.
Open Datasets Yes The data used in our prediction of residential load consumption for the area, comes from the Public Service Company of New Hampshire, and is freely available in their website (http://www.psnh.com/).
Dataset Splits No The paper mentions using data from the Public Service Company of New Hampshire but does not specify any training, validation, or test dataset splits or a methodology for creating them.
Hardware Specification Yes All experiments were conducted on a 2.10 GHz x 4 Intel Core i3-2310M processor, with 8GB of memory.
Software Dependencies No The paper mentions using 'Gaussian Processes' and 'Bayesian Linear Regression' for prediction, and refers to 'SPUDD' and 'RENES', but it does not provide specific software names along with their version numbers required for replication.
Experiment Setup Yes Our value iteration method, operating over a problem horizon corresponding to 24 hours... By setting the horizon T to be equal to the number of time steps at which the prosumer is required to act, we can incorporate the tms factored state into the problem s horizon, thus effectively reducing the size of the state space. Our problem is naturally a finite-horizon problem, thus we employed a finite-horizon VI method. [and in Table 1] Horizon |S A| Bounded Region Size Value Iteration (hours) SPUDD (hours) Script Runtime 24 664290 15 1.76 13.4992 0.184 90 15.84 46.9188 1.19 2624490 15 8.7603 36.98 0.73975 48 664290 15 3.5 16.8221 0.4271