Distributed Reinforcement Learning SARSA-λ for Optimal Power Flow Calculation of Distribution Network

With the increasing of renewable energy to the distribution network will lead to new uncertainties in the operation of the power system. Therefore, the traditional optimal power flow algorithm is difficult to adapt to complex multi-objective functions, to meet the real-time requirements of large-scale power grid calculation. In this paper, an optimization algorithm based on distributed reinforcement learning SARSA-λ is proposed. In the algorithm, each agent uses the standard multi-step SARSA-λ algorithm to independently assume the learning tasks of the system, and forms the overall sense of the optimal through unified cooperation. The IEEE 57-bus system is used as a simulation case, as well as comparing the two traditional reinforcement learning methods and optimal power flow solution methods, the results show that the algorithm proposed in this paper has better robustness to complex multi-objective functions, while the traditional optimization algorithms are unable to compute real-time large-scale power grids. This paper provides a feasible and effective new method for solving the multi-objective optimal power flow problem for complex power grids.