A learning automata local contribution sampling applied to hydropower production optimisation

Abstract

Learning Automata (LA) is a powerful approach for solving complex, non-linear and stochastic optimisation problems. However, existing solutions struggle with high-dimensional problems due to slow convergence, arguably caused by the global nature of feedback. In this paper we introduce a novel Learning Automata (LA) scheme to attack this challenge. The scheme is based on a parallel form of Local Contribution Sampling (LCS), which means that the LA receive individually directed feedback, designed to speed up convergence. Furthermore, our scheme is highly decentralized, allowing parallel execution on GPU architectures. To demonstrate the power of our scheme, the LA LCS is applied to hydropower production optimisation, involving several particularly challenging optimisation scenarios. The experimental results show that LA LCS is able to quickly find optimal solutions for a wide range of problem configurations. Our results also demonstrate that local directed feedback provides significantly faster convergence than global feedback. These results lead us to conclude that LA LCS holds great promise for solving complex, non-linear and stochastic optimisation problems, opening up for improved performance in a number of real-world applications.