An AGC Dynamic Optimization Method Based on Proximal Policy Optimization

The increasing penetration of renewable energy introduces more uncertainties and creates fluctuations in power systems than ever before, which brings great challenges for automatic generation control (AGC). It is necessary grid operators to develop an advanced AGC strategy handle uncertainties. dynamic optimization a sequential decision problem that can be formulated as discrete-time Markov process. Therefore, this article proposes novel framework based on proximal policy (PPO) reinforcement learning algorithm optimize regulation among each generator advance. Then, the detailed modeling process reward functions state action space designing presented. application proposed PPO-based simulated modified IEEE 39-bus system compared with classical proportional−integral (PI) other algorithms. results case study show make frequency characteristic better satisfy performance standard (CPS) under scenario large systems.