Accelerating Dual Dynamic Programming for Stochastic Hydrothermal Coordination Problems

Mid-term hydrothermal coordination problems for large scale systems are difficult to be solved. As the complexity of the model increases the results become more accurate, however more expensive is to solve the problem. We propose new strategies to accelerate Dual Dynamic Programming (DDP), which is one of the leading techniques in stochastic optimization. The proposed strategies consist in local convergence tests in scenario sub-trees, as well as analysis of the stability in the values of state variables along the nodes, in order to avoid unnecessary forward and backward passes and therefore saving CPU time and memory requirements. We also introduce an asynchronous parallel scheme which makes the best use of the available resources. The efficiency of our approach is shown for the real large-scale Brazilian system.