Multivariable Control of Aluminum Reduction Cells

This paper considers control of the aluminum reduction process, using a dynamic model developed from the literature and Virtpot, a model developed at Kaiser Aluminum. Analysis shows the process is controllable and observable, but not easily stabilizable using only one input, and that short-term changes in measured voltage result primarily from changes in alumina concentration rather than anode-to-cathode distance (ACD). Next, a multivariable control strategy is developed to regulate cell voltage by adjusting feed rate rather than beam movement. We introduce the idea of a feed voltage, obtained by subtracting expected voltage deviations due to ACD changes and beam moves from the filtered voltage. Feed rate is adjusted to compensate for deviations of feed voltage from its target. Simultaneously, beam movements are made to compensate for the difference in expected anode consumption and metal pad rise, based on changes in feed period. Simulations show the effectiveness of the proposed control strategy. Introduction The primary production of aluminum is a key process in the aluminum industry. A critical issue in this process is proper regulation of both the anode-to-cathode distance and the concentration of dissolved alumina. These quantities have a primary effect on the overall efficiency of the reduction process [1]. Unfortunately, it is difficult to directly measure either variable. Typical installations of aluminum reduction cells allow only the measurement of cell voltage, which is affected by both the anode-to-cathode distance and the dissolved alumina concentration. This fact makes the control of the aluminum reduction cell difficult [2]. In this paper we consider control of the aluminum reduction process. We take a model-based, control-theoretic approach, using a dynamic model structure to describe the process. This dynamic model is motivated by first principles arguments drawn from the literature and from results obtained from a detailed simulation Multivariable Control of Aluminum Reduction Cells Kevin L. Moore1, Nobuo Urata2 1Center for Self-Organizing and Intelligent Systems Utah State University 4160 Old Main Hill Logan, UT 84321 2Kaiser Aluminum & Chemical Corporation 2107 East Hawthorne Road Mead WA 99021