Power Flow Out of Limit Correction Control for Grid Branch based on Reinforcement Learning and Sensitivity

At present, based on sensitivity and artificial intelligence methods, the prevention and correction of finite power flow violations in power grid branches are prone to the phenomenon of multiple branch cross-limit adjustments and falling into local optimum. This paper proposes a method based on reinforcement learning and sensitivity Based on the power grid branch active power flow out-of-limit correction control method, this method firstly proposes a branch power flow heavy-load/over-limit unit adjustment combination division strategy based on the sensitivity calculation method, selects the key unit groups participating in power adjustment and the adjustment properties, and A power grid branch active power flow over-limit correction control method based on a deep deterministic strategy gradient algorithm is proposed, and the dimensionality of the agent's continuous output action space is reduced through the key unit adjustment properties, which speeds up the convergence speed of the agent's training. The final simulation example shows that The effectiveness and feasibility of the method.