Deep Reinforcement Learning Modeling for Power Carbon Emission Monitoring and Prediction

Carbon dioxide emission reduction relies on an accurate carbon emission monitoring system, and China's electric power industry accounts for 40% of the country's total carbon dioxide emissions, so the study of carbon emission monitoring methods for the electric power industry is of great significance.This paper proposes the monitoring direction of grid carbon emission data from the perspectives of pressure, flow, humidity, temperature, etc., associates the calculation formula of total carbon emission, establishes the monitoring model of grid carbon emission, and implements real-time monitoring of carbon emission in the power industry.Based on the current monitoring framework of the power system, the carbon flow calculation of the system is carried out, Q learning algorithm and VMD are introduced, and a combined prediction model based on VMD and DL is proposed for predicting the carbon emissions on the power generation side.The difference between the real value and the predicted value of the carbon emissions prediction of the VMD-DL model is not large, and the mean value of the error is at -1.77910-4 megatonnes, and the combined model obtains the prediction of the optimal carbon flow Data.Using the VMD-DL model in the regional scenario, the out-of-sample 2018-2022 five-year minimum percentage error of 0.005% and the average absolute error of 0.01% show that the model error is small and the measured values fit the growth trend of total energy consumption in five years better.