An Online Economic Scheduling Method Based on Lagrangian Relaxation Reinforcement Algorithm

Online economic dispatch is one of the important means to ensure the safe and stable operation of power grids in uncertain environments. At present, reinforcement learning has shown good results in the field of economic dispatch, but its practical application in power grids still needs to address the lack of interpretability in the learning process. To this end, a Lagrangian relaxation reinforcement algorithm for online economic scheduling is proposed. Firstly, transform the mathematical model of online economic scheduling into a Markov decision process model suitable for deep reinforcement learning. Then, taking the deep deterministic policy gradient algorithm as an example, the fusion of constraint conditions and policy neural networks is achieved based on the Lagrange relaxation principle, and the constrained Markov decision process model is transformed into an unconstrained Markov decision process model. Finally, the effectiveness of the proposed algorithm was verified through actual power grid examples, which can ensure that the decision results strictly meet the constraints of power grid regulation while improving the training efficiency of the intelligent agent.