The aluminium industry is an important GHG producer due to its carbon dioxide emissions but also due to the perfluorocarbons (PFC) emissions emitted during a detrimental event known as anode effect (AE). The doctoral project presented in this thesis was realised to increase the understanding of the different mechanisms leading to the generation of PFC, in order to facilitate the quantification of PFC while facilitating a reduction of the total emissions.
Globally, a smelter’s PFC emissions are estimated using linear models based on monthly performance indicators. However, the precision of these methodologies is dependent on the total number of AE occurrence and new models are now necessary to assure adequate estimations of PFC emissions.
During this project, multiple measurement campaigns were performed to assign specific CF₄ and C₂F₆ amounts for each respective AE detected by the control system. Based on more than one thousand individual measurements, new models were proposed and compared to the already existing methodologies. The model considered with the best potential to be used widely across the industry, in terms of simplicity and efficiency, considers the PFC emission rate as a non-linear function of the polarised AE duration. Validation was performed based on data acquired in 7 different smelters to confirm an improved predictive efficiency. However, it also demonstrated that the line current has an important impact on the emission rate of PFC emissions. It was necessary to incorporate an additional variable into the equation to reach a higher level of precision. Finally, a generic model was developed with the ability to estimate the PFC emissions resulting from individual AE for cell technologies using prebaked anodes and line current higher than 440 kilo amperes.
The second aspect of the project is related to low voltage anode effect (LVAE) where a thorough study of the mechanism leading to their generation was performed. Based on gas composition measurements performed on individual cells, a first published model was established allowing quantification of PFC emissions resulting from LVAE. The measured accuracy of the model is ±25% for 2/3 of the studied scenarios. A sensitivity analysis was performed afterward on the model and the standard deviation among individual anode currents was found to be the variable having the best correlation with the presence of LVAE. It was also demonstrated that improvements in the gas extraction technique should lead to a better representativeness of the cell global condition, which is necessary in order to increase the predictive capability of the LVAE algorithm.
A transient mathematical model was developed to simulate the local alumina concentration and current density in an electrolysis cell for the 20 different anodic assemblies. Henceforth, it is possible to evaluate the homogeneity of the current distribution and predict if specific operation scenarios are more at risk to generate PFC emissions. Industrial measurements confirmed that a good correlation exists between the simulator and the reality for both the evolution of the alumina distribution and the LVAE predictive capability.
Finally, the knowledge acquired during this project and the proximity of the industrial partner allowed the development of a control algorithm to detect PFC generation while automatically launching a corrective action to eliminate the threat. Usage of this preventive treatment allowed a reduction of more than 50% on the AE frequency and a reduction of almost 50% related to the cell instability without any negative impact on other key performance indicators.