Combining Forecasting and Multi-Agent Reinforcement Learning Techniques on Power Grid Scheduling Task

The continuous emission of greenhouse gases is causing global warming and the habitat of many animals is being affected. Efforts are constantly being made to mitigate this situation. In the field of electricity research, it is challenging to investigate how to operate and maintain the power grid economical, low-carbon and stable. In order to tackle the challenge, we propose grid scheduling techniques that combine forecasting and reinforcement learning techniques. Forecasting techniques can predict future electricity demand and solar power generation. Reinforcement learning uses the current state of the grid environment, future electricity demand and solar power generation to make charging and discharging decisions for energy storage devices. The result is economical, low-carbon electricity use and the grid is stable. We conduct a lot of experiments on a real dataset and the experimental results demonstrate the effectiveness of our approach.