Multi-objective Parameter Optimization Of Grate Cooler Based On Improved RNN

Grate cooler is the key equipment in cement clinker production process.The adjustment of relevant control parameters of grate cooler has a significant impact on energy consumption,product quality and stable operation of related equipment.Because the control process of cement production is relatively complex and belongs to slow process parameter control,there is serious time lag in each control parameter of grate cooler,and the lag time of different control parameters is different,which brings great difficulties for energy saving and consumption reduction of grate cooler.This project takes grate cooler,the key equipment in cement production,as the research object.It studies how to establish a high-precision prediction model of grate cooler parameters and how to optimize the relevant control parameters,and completes the modeling of key parameters of grate cooler and optimization of control parameters.According to the production process of cement clinker,this paper systematically analyzes the process flow and working mechanism of grate cooler,summarizes the target variables that can reflect clinker quality and energy consumption in grate cooler,and summarizes the key parameters that affect the target variables.The delay time of key parameters is analyzed in detail,and the delay characteristics of each parameter are summarized.By analyzing the working mechanism of grate cooler,the grate down pressure,secondary air temperature,clinker outlet temperature and power consumption are determined as the optimization objectives,and the corresponding constraints are established,which lays the foundation for the subsequent modeling and optimization work.Based on the delay characteristics of grate cooler parameters,this paper proposes a long-term and short-term memory network(ACT-LSTM)based on adaptive calculation time,establishes a prediction model of grate cooler parameters,and solves the problem of low accuracy of prediction model caused by inconsistent delay time of key parameters of grate cooler.Compared with the prediction accuracy of other classical time series networks,the prediction model established in this paper has outstanding advantages in grate cooler parameter prediction.Aiming at optimizing the grate cooler parameters,we improved the fast non dominated sorting genetic algorithm(NSGA-II)and proposed a multi-objective bounded stable optimization algorithm(BS-NSGA-?).The algorithm integrates stability optimization and emergency stop strategy optimization to reduce the power consumption of grate cooler,ensure the stability of grate pressure,improve the quality of clinker products and reduce the loss of emergency stop,so that the grate cooler can operate more stably on the basis of energy saving.After testing,when BS-NSGA-? algorithm is used to optimize the grate cooler parameters,the average power consumption per minute is reduced by 13.2%,and the clinker temperature at the outlet is more stable.We also designed the optimization software of grate cooler control parameters,including reading real-time operation data,real-time optimization of control parameters,display of optimization results,database storage and other functions.Based on the ACT-LSTM parameter prediction model and BS-NSGA-? optimization algorithm proposed in this paper,the software optimizes the control objective of grate cooler in real time,guides the adjustment of controllable parameters of grate cooler,and provides a favorable guarantee for the automatic operation of the whole grate cooler control system.