Decision Making for Low-Carbon Operation of Grid Mains Using Reinforcement Learning Algorithms

The realization of carbon neutrality goal, the power grid main network to promote low-carbon operation.To meet the needs of low-carbon operation of the main grid, this paper constructs an operation decision model based on reinforcement learning algorithm.By embedding algebraic equations with multi-source grid scheduling knowledge, knowledge such as trend calculation is transformed into machine language understandable by intelligences.At the same time, normative processing such as Max-min normalization is performed on the data to solve the interference caused by different unit magnitudes.Several microgrids are designed to conduct comparison experiments to judge the difference between the decision-making method of this paper and the traditional decisionmaking method.The results show that the extra amount of green power purchased by this paper's decision-making method in each time period is no more than 270kWh, which is much less than that purchased by the traditional method.The total cost of main grid operation is 29.09% lower than the traditional method.The total carbon carbon emission is decreased by 41.57% compared to the traditional method.The low carbon operation decision-making model of power grid main network based on reinforcement learning algorithm can reduce the economic cost of power grid main network operation and realize the low carbon goal at the same time.