Multi-Lag Benders Decomposition for Power Generation Planning with Nonanticipativity Constraints on the Dispatch of LNG Thermal Plants

Multi-stage Benders decomposition has long been applied to mid and long term power generation planning, which is usually conceived as a stochastic problem, where the dispatch of the plants at each node of the scenario tree is determined based on current system conditions and on a number of possible future realizations of the random variables. However, recent logistic issues have forced the dispatch of Liquefied Natural Gas (LNG)-based thermal plants to be determined some months in advance to its actual generation, which leads to nonanticipativity constraints on the dispatch of these plants. This paper proposes a multi-lag variant of Benders decomposition, where additional state variables corresponding to the pre-order dispatch of LNG plants are introduced to the recourse function of each stage. Artificial variables are defined so that the number of state variables does not depend directly on the number of LNG plants in the system. Numerical results show the efficiency of the application of the proposed approach in the stochastic dual dynamic programming and L-shaped solving methodologies used to solve the long and mid-term generation planning problems for the large-scale Brazilian system.