Research On Optimal Control Method Of Cigarette Factory Packing Workshop Based On Neural Network

As the economy grows and carbon emissions rise,digitalization,intelligence and energy efficiency have become inevitable trends in the development of the cigarette manufacturing field.The electric cooling system is a key component in controlling the indoor environment of the cigarette factory packing workshop.It undertakes the important task of stabilizing the temperature and humidity and improving the quality of cigarette production.However,the electric cooling system represents a significant contributor to overall energy consumption during production due to the high amount of energy it requires for its operation.Therefore,reducing its energy consumption through a series of energysaving measures can significantly reduce the cost of cigarette production.Despite the non-linearity and strong coupling that are characteristic of electric cooling systems which traditional optimal control methods tend to overlook,it is crucial to select methods that appropriately accounts for these system characteristics to accurately study and optimize the energy-saving potential of these systems.This study aims to optimize the parameters and energy efficiency of the electric cooling system of a cigarette factory,based on feature modeling and time-series data prediction to reduce the emissions and consumption of the electric cooling system.The main research contents of the thesis are as follows.(1)The means of energy saving and consumption reduction of the electric cooling system are analyzed,and the method of variable flow rate and temperature difference control of the electric cooling system is determined.Based on the physical laws and empirical formulas to establish the equipment’s energy consumption and indoor temperature and humidity evolution model,and the LMA method is used to fit the parameters to be determined to construct the equipment energy consumption and temperature and humidity mechanism-data model.(2)For the characteristics of electric cooling system operation parameters and indoor and outdoor temperature and humidity data,data preprocessing is performed utilizing data cleaning and transformation,feature down-scaling,and sliding windowing.According to the characteristics of different data sets,the ATT-TCN-Bi GRU algorithm is selected for outdoor temperature and humidity sequence prediction,and the PSO-XGBoost algorithm is used for indoor temperature and humidity and electric cooling system operation parameters sequence prediction.To illustrate the superiority of the algorithms proposed in the thesis,other deep neural network models are also constructed and the model prediction performance is evaluated cross-sectionally using metrics such as average loss.(3)A single-objective optimization task for system energy minimization and a multiobjective optimization task for system energy minimization and room temperature and humidity stability is established.With the mechanism-data model as the objective function and the production environment requirements and expert experience as the bundle conditions,the single-objective and multi-objective optimization objectives of the electric cooling system are optimized using the SEGA and NSGA-III to obtain the point-by-point optimal energy consumption of the electric cooling system and its corresponding operating parameters.To reflect the generalization of the model,three load conditions,such as high,medium,and low,and models such as PSO algorithm and NSGA-II are set in the thesis,and evaluation criteria such as energy saving rate and quality loss coefficient are used for cross-sectional comparison.The above modeling,prediction,and optimization simulation results illustrate the feasibility of the neural network-based optimal control method for the electric cooling system in packing workshop and provide some reference for the research of energy saving and consumption reduction of the factory cooling equipment.