14 August 2023 in IEEE Transactions on Power Systems
This work considers the European Resource Adequacy Assessment, which is a pan-European resource adequacy process that is being developed by the European Networks of Transmission System Operators for Electricity (ENTSO-E). A critical part of this process is the so-called Economic Viability Assessment model, which aims at determining future expansion and retirement capacity opportunities for the entire European network. As such, the problem is stochastic. Nevertheless, due to computational constraints, simplified approaches have been followed by ENTSO-E. Our work formulates the problem as a two-stage stochastic problem and proposes two decomposition algorithms for solving the problem which are implemented in a high-performance computing infrastructure. The first is a subgradient-based algorithm, and the second uses a relaxation of the second stage (the economic dispatch) in order to speed up the subgradient calculation thus achieving a considerable reduction in solution time. We compare our schemes against the commonly used Bender's decomposition. We compare the obtained stochastic solution against the deterministic solution proposed by ENTSO-E for their 2021 study and analyze the impact of the stochastic solution on various adequacy indicators.
20 February 2023 in IEEE Transactions on Power Systems
We consider the stochastic dual dynamic programming (SDDP) algorithm - a widely employed algorithm applied to multistage stochastic programming - and propose a variant using experience replay - a batch learning technique from reinforcement learning. To connect SDDP with reinforcement learning, we cast SDDP as a Q-learning algorithm and describe its application in both risk-neutral and risk-averse settings. We demonstrate the superiority of the algorithm over conventional SDDP by benchmarking it against PSR's SDDP software using a large-scale instance of the long-term planning problem of inter-connected hydropower plants in Colombia. We find that SDDP with batch learning is able to produce tighter optimality gaps in a shorter amount of time than conventional SDDP. We also find that batch learning improves the parallel efficiency of SDDP backward passes.
24 August 2021 in Computational Management Science
We study different parallelization schemes for the stochastic dual dynamic programming (SDDP) algorithm. We propose a taxonomy for these parallel algorithms, which is based on the concept of parallelizing by scenario and parallelizing by node of the underlying stochastic process. We develop a synchronous and asynchronous version for each configuration. The parallelization strategy in the parallelscenario configuration aims at parallelizing the Monte Carlo sampling procedure in the forward pass of the SDDP algorithm, and thus generates a large number of supporting hyperplanes in parallel. On the other hand, the parallel-node strategy aims at building a single hyperplane of the dynamic programming value function in parallel. The considered algorithms are implemented using Julia and JuMP on a high performance computing cluster. We study the effectiveness of the methods in terms of achieving tight optimality gaps, as well as the scalability properties of the algorithms with respect to an increasing number of CPUs. In particular, we study the effects of the different parallelization strategies on performance when increasing the number of Monte Carlo samples in the forward pass, and demonstrate through numerical experiments that such an increase may be harmful. Our results indicate that a parallel-node strategy presents certain benefits as compared to a parallel-scenario configuration.
6 April 2021 in IEEE Transactions on Automatic Control
We consider a Markov control model in discrete time with countable both state space and action space. Using the value function of a suitable long-run average reward problem, we study various reachability/controllability problems. First, we characterize the domain of attraction and escape set of the system, and a generalization called $p$ -domain of attraction, using the aforementioned value function. Next, we solve the problem of maximizing the probability of reaching a set $A$ while avoiding a set $B$ . Finally, we consider a constrained version of previous problem, where we ask for the probability of reaching the set $B$ to be bounded. In the finite case, we use linear programming formulations to solve these problems.
1 November 2026 in IEEE Transactions on Power Systems
23 February 2022 in IEEE Transactions on Smart Grid
Priority service and multilevel demand subscription have been proposed as two alternative methods for the mobilization of residential demand response. Whereas priority service relies on the differentiation of electricity service according to reliability, multilevel demand subscription further differentiates electricity service according to duration. Despite its increased complexity, multilevel demand subscription promises increased operational efficiency, as it permits a finer differentiation of consumer classes by the utility. It also allows households to reduce their electricity bills relative to priority service. This paper proposes a framework for quantifying these effects. We design a modeling approach for evaluating the performance of these different aggregator service offerings in a system with utility-scale renewable supply, residential renewable supply, and residential storage. We compare priority service to multilevel demand subscription, and discuss the implications of these different residential demand response options on operational efficiency and consumer expenditures for electricity service on a realistic model of the Belgian power market. We show how the comparison between the two schemes is affected by the adoption of a different time resolution in a detailed case study.