Energy Efficiency Optimization Of Industrial Boiler Combustion Field Based On Improved Deep Reinforcement Learning

The energy consumption of China is usually based on coal,and the diversified energy consumption structure will not undergo fundamental changes at least in the short term.Driven by the goal of "peaking carbon dioxide emissions by 2030 and achieving carbon neutrality by 2060",it will face increasing challenges.Resource and environmental constraints,and the energy saving and emission reduction of coal-fired industrial boilers are still an important part of the country’s sustainable energy development.The combustion field of the boiler is the most direct and important state quantity that reflects the combustion process and the state of the equipment.Energy efficiency optimization research is of great significance for improving the energy saving and emission reduction capabilities of industrial boilers and reducing carbon emissions.However,it is difficult to establish an effective combustion mechanism model using traditional analytical mathematical model methods,and it is difficult to find the best parameters for boiler operation.However,the data-driven deep reinforcement learning model method has the significant advantage of not relying on the system model,which is the energy efficiency in the field of industrial boiler combustion.The optimization problem provides new research ideas.This dissertation studies the energy efficiency optimization of the industrial boiler combustion field based on deep reinforcement learning,comprehensively considers the various index parameters of the combustion field,and proposes an energy efficiency online optimization algorithm to improve the energy efficiency of the combustion field.The main research contents include:(1)A Boiler Combustion Field Power Prediction Algorithm Based on Deep Reinforcement Learning.The energy efficiency index during boiler combustion is the key to accurately grasping the coal utilization rate of the boiler.Predicting the output power of the industrial boiler can not only better understand the working status of the boiler,but also provide a reference for subsequent energy efficiency optimization control.This paper combines the environmental information of the industrial boiler combustion field,such as: coal feed rate,steam value,water feed rate,etc.,to design a boiler combustion field output power prediction model based on Deep Q-Network(DQN).The simulation results show that the prediction results of the combustion field power prediction model using the DQN model have higher accuracy.(2)On-line optimization algorithm of combustion field energy efficiency based on improved DQN algorithm.In order to make full use of the existing combustion field environment information,realize rapid response adjustment of control strategies,adjust the boiler combustion state to the best in time,and maximize the fuel utilization of the combustion field,an online optimization control algorithm is designed.The online optimization control algorithm combines the combustion field power prediction model with the energy efficiency optimization model.The overall operation process is divided into two phases,namely the offline training phase and the online optimization control phase.In addition,in order to improve the applicability of the DQN model in the combustion field energy efficiency optimization scenario,optimization and improvement are made on the basis of the model: by improving the training process of the DQN algorithm model,the stability of the algorithm during the training process is improved;through the training process Optimize the experience playback mechanism to make the convergence direction of the training process more accurate and faster.