Optimal Temperature Control Of Hydrometallurgical Zinc Electrolyte Based On Deep Reinforcement Learning

The electrowinning process is the most energy-intensive process in zinc hydrometallurgy,and the electrolyte temperature is a key parameter for process control.By adjusting the cooling tower fan current to ensure the optimal matching of the electrolyte temperature with the ion concentrations and current density,the current efficiency can be effectively reduced.However,there exist many factors that affect the electrolyte temperature,including the current density and the meteorological temperature.More than that,those variables fluctuate frequently and have obvious uncertainties,resulting in the optimal matching of the electrolyte temperature,the ion concentrations,and the current density is difficult to establish,and the stable control of the electrolyte temperature runs into difficulties.Therefore,to improve the current efficiency of the zinc electrowinning process,this thesis took the current efficiency as the optimization objective and studied the optimization of electrolyte temperature for multiple working conditions.By establishing the optimal matching of electrolyte temperature and process variables,a stable control strategy for electrolyte temperature based on deep reinforcement learning is proposed.The main work of this thesis is as follows:(1)For the problem that the current efficiency is difficult to be estimated accurately caused by the complex coupling between the process variables and the current efficiency,a current efficiency estimation method based on causal convolutional network is proposed.Based on the hydrogen-zinc competition reaction,the input features of the model are selected,and the causal dependence between the input features and the current efficiency on the time scale is incorporated into the machine learning to estimate the current efficiency accurately.(2)Considering the data distribution of multiple working conditions and the small adjustment space of electrolyte temperature,this thesis took the working conditions as the matching object.First,the acid ion concentration,zinc ion concentration,and current density were used as the input parameters to divide the working conditions,and the working conditions are used as the matching unit between the electrolyte temperature and the process variables to simplify the matching mode;then,based on the current efficiency estimation model,taking the current efficiency as the optimization objective,the optimal temperature under different working conditions is solved to provide the control goal for the subsequent stable control.(3)To solve the difficulty of stable control of electrolyte temperature caused by numerous influencing factors and its uncertainties,an electrolyte temperature stable control framework based on multi-agent integration was proposed,and the sub-models for electrolyte temperature control based on deep Q network were built.By designing the action and state space,an interactive learning method is established to eliminate the interference of the uncertainty of process variables on the performance of the controller.In accordance with the control goal solved by the current efficiency optimization model,the stable control of the electrolyte temperature is realized.