WO2017141134A1

WO2017141134A1 - Method for estimating dynamic state variables in an electrolytic cell suitable for the hall-héroult electrolysis process

Info

Publication number: WO2017141134A1
Application number: PCT/IB2017/050661
Authority: WO
Inventors: Yuchen YAO; Cheuk-Yi CHEUNG; Jie BAO; Barry Welch; Maria Skyllas-Kazacos; Sergey AKHMETOV; Ali Jasim BANJAB
Original assignee: Dubai Aluminium Pjsc; Newsouth Innovations Pty
Priority date: 2016-02-15
Filing date: 2017-02-08
Publication date: 2017-08-24
Also published as: EP3417095A1; GB201602613D0; CA3012166A1; CA3012166C; EP3417095A4

Abstract

Method of producing aluminium in an electrolytic cell using the Hall-Héroult electrolysis process, said method comprising the following steps: -dividing said cell into at least n subsystems, -for each subsystem, estimating the local value of at least one target or output parameter on the basis of the value of the measurement of at least one input parameter, -modifying at least one of said input parameters in the cell by exerting a physical action upon the cell, if at least one estimated value of said output parameter, for at least one subsystem, is substantially different from another estimated value of said output parameter for another subsystem.

Description

Method for estimating dynamic state variables in an electrolytic cell suitable for the

Hall-Heroult electrolysis process

Technical field of the invention

The invention relates to the field of fused salt electrolysis, and more precisely to the monitoring of Hall-Heroult process for making aluminium by fused salt electrolysis. In particular, the invention relates to a novel method of monitoring aluminium smelting process. This method is based on a particular approach of anode current measurement using digital communication for easy wiring, installation and maintenance, and applies a Kalman filter type state observer to a dynamic model of the process for the estimation of unmeasured process variables.

Prior art

The Hall-Heroult process is the only continuous industrial process for producing metallic aluminium from aluminium oxide. Aluminium oxide (Al₂0₃) is dissolved in molten cryolite (Na₃AIF₆), and the resulting mixture (typically at a temperature comprised between 940°C and 970°C) acts as a liquid electrolyte in an electrolytic cell. An electrolytic cell (also called "pot") used for the Hall-Heroult process typically comprises a steel shell, a lining usually made from refractory bricks, a cathode usually covering the whole bottom of the pot (and which is usually made from graphite, anthracite or a mixture of both), and a plurality of anodes (usually made from carbon) that plunge into the liquid electrolyte. Anodes and cathodes are connected to external busbars. An electrical current is passed through the cell (typically at a voltage between 3.7 V to 5 V) which splits the aluminium oxide in aluminium ions and oxygen ions. The oxide ions are reduced to oxygen at the anode, said oxygen reacting with the carbon of the anode. The aluminium ions move to the cathode where they accept electrons supplied by the cathode; the resulting metallic aluminium is not miscible with the liquid electrolyte, has a higher density than the liquid electrolyte and will thus accumulate as a liquid metal pad on the cathode surface from where it needs to be removed from time to time, usually by suction.

In order to decrease the capital cost per ton of production capacity of aluminum, the size of aluminium reduction cells, as well as the number of anodes per cell, tends to increase. Thus, keeping the cell in a balanced state has become increasingly important. One of the crucial process variables is the alumina content and its uniformity in the cell. The concentration of dissolved alumina needs to be regulated in a limited range to prevent anode effect and the formation of sludge, and its distribution affects the balance of the cell. As alumina is consumed during the electrolysis process, it must be added regularly, through so-called feeders. Feeders are most often associated with crust-breakers that provide a hole into the crust of solidified liquid bath (containing a mixture of cryolite and alumina) that forms on the top of the electrolyte layer; the feeder then dumps alumina powder through that hole into the liquid bath. However, the concentration of dissolved alumina in the electrolyte of an electrolytic cell does not lend itself to direct or continuous measurement during operation: a sample for chemical analysis is usually taken several times per weeks but not necessarily each day - this would be totally insufficient to identify for instance a drift of alumina concentration due to faulty equipment, such as malfunctioning of the crust breaker or feeder, or a leak of the feeder. What can be measured directly and continuously is cell resistance: the currently available control strategies of the process, such as the ones described in US 4,654, 129 and US 4,766,552, are largely dependent on the cell resistance. The cell resistance represents a combination of the average anode-cathode distance (ACD), global bath composition and bath properties in the cell, which is not able to reflect the spatial variations in the cell. As a consequence, approaches based on cell resistance would leave localised abnormalities undetected until they become severe and apparent on the cell resistance.

In order to improve the cell control, and, in particular, to identify defective feeders or crust breakers early, it would be desirable to gain localized information on operating parameters within an electrolysis cell. As an example, it is known (see K. Rye et al., "Current redistribution among individual anode carbons in a Hall-Heroult prebake cell at low alumina concentration" , TMS Light Metals 1998, p. 241-246) that anodes located in areas where the concentration of dissolved alumina is low take less current than anodes in areas where the alumina concentration is higher. Various methods and devices for measurement of the current through each individual anode have been proposed. These methods include the calculation of the anode currents from voltage drop measurement across a length of conductor (see US 4,786,379 and WO 94/002859), and the use of Hall Effect sensors (see J.W. Evans and N. Urata ("Wireless and Non-Contacting Measurement of Individual Anode Currents in Hal-Heroult Pots; Experience and Benefits", Light Metals 2012, TMS, p. 939-942).

The measurement of anode current distribution in a cell provides a general guide to the cell conditions in the vicinity of each anode, but a quantitative analysis allowing to obtain localized information on the electrolysis process in a Hall-Heroult cell is very difficult. This is because the variation and interactions of anode currents are complicated, for example, through a number of ways as follows: (i) While the anode current affects local cell conditions, it is also affected by these conditions.

(ii) Variation in one anode current will affect others due to the controlled line current. (iii) Anode current represents a combination of localized variables. As they are coupled, the separation of one variable from others is not easy,

(iv) There are mass and energy transfers within the cell that will cause spatial variations, which, in turn, can alter the anode current distribution. Therefore, anode current needs to be treated as part of a dynamic system, where the variations of process variables are taken into account.

State observer is a mathematical tool in system science that is used to estimate the internal states of a dynamic system based on the measured inputs and outputs of the system. It provides the basic structure to the process monitoring and control methods in a wide range of industrial applications.

Kalman filter-type state observer has been disclosed in a number of publications, such as US 4,814,050 and WO 2009/067019. In these approaches, the state estimation is based on the whole cell, thus the spatial variations are not accounted for. The use of Kalman filter-type state observer with anode currents in Hall-Heroult cells has been discussed in a certain number of papers: Jakobsen et.al., "Estimating alumina concentration distribution in aluminium electrolysis cells", Proc. 10^th IFAC Symposium on Automation in Mining, Mineral and Metal Processing, p.253-258 (2001); K. Hestetun, M. Hovd, "Dectecting abnormal feed rate in aluminium electrolysis using extended Kalman filter", IFAC World Congress, Praha 2005; and K. Hestetun, M. Hovd, "Detection of abnormal alumina feed rate in aluminium electrolysis cells using state and parameter estimation", 16^th European Symposium on Computer Aided Process Engineering and 9^th International Symposium on Process Systems Engineering, Elsevier 2006, p.1557-1562. These papers attempted to estimate local cell conditions using the measured individual anode current. However, as pointed out by the authors, the ACD change cannot be accommodated in these methods. In fact, the initial conditions need to be carefully chosen in the mentioned papers, and are assumed to be constant, which deviates the purpose of state estimation. The problem that the present invention endeavors to resolve is therefore to propose an improved method for monitoring of aluminium reduction cells which can deal with the interactions between spatial process variables and interaction between current in different anodes. Moreover, the invention wishes to provide such a method, which allows for different initial estimation of states and can be used effectively online. And eventually, the method for monitoring aluminium reduction cells should be simple, reliable and robust in view of its use in an industrial environment.

Object of the invention

The present invention is devoted to the estimation of spatial alumina concentration and local ACD using a multi-stage state observer. In each stage, the cell is discretized successively into subsystems, with each containing a number of anodes. The states in each subsystem are estimated based on a dynamic model using the measurement of anode currents associated with the subsystem, measurement of cell voltage, feeding information and the estimation from the previous stage.

The dynamic model used in each stage of the state observer has the following features: 1. It includes the dynamics of alumina addition, dissolution and consumption. The addition of the alumina is referred to the dumps from the feeders. The dissolution of the alumina follows a specific rate equation and the consumption of the dissolved alumina can be obtained from Faraday's equation. In addition, the model also deals with the mass transfer between subsystems due to the induced bath flow.

2. It includes the dynamics of ACD variation, which is associated with the consumption of carbon anodes, the beam movement, and the accumulation of liquid aluminium at the cathode.

3. It calculates cell voltage from a semi-empirical equation, which is a function of anode current, alumina concentration, ACD and other parameters. Cell voltage equations are known by a person skilled in the art and can be found in the standard textbook of Grjotheim, K. & Kvande, H. 1993, "Introduction to aluminium electrolysis: understanding the Hall-Heroult process", Aluminium-Verlag, Dusseldorf, or in Haupin's paper, "Interpreting the components of cell voltage", TMS Light Metals 1998, p. 531-537.

The model can be represented as the following discrete-time state-space form:

ACD^N+1 = f₃(ACD^N ,BM^N )

(equation 1) where N represents the time step, c_un is the concentration of the undissolved alumina, k_un is the effective rate constant for the dissolution of alumina, c_d is the concentration of the dissolved alumina, M is the mass of bath in the subsystem, BM is the variation in beam movement, /^' is the total anode current flowing into the subsystem impacting the consumption of dissolved alumina and the rate of change of ACD, d-ι and d₂ represent the mass transfer between relevant subsystems, v is the cell voltage, h represents the voltage equation, Θ represents the relevant process parameters.

The estimation of spatial alumina concentration and ACD is achieved by a state observer. In an advantageous embodiment the observer is of Kalman filter type, which estimates the state variables as well as their uncertainties, and sequentially updates the estimation when the next measurement is available, based on a weighted average. According to an essential feature of the present invention, the estimation result from one stage is used an input in the next stage.

A first object of the invention is a method of producing aluminium in an electrolytic cell using the Hall-Heroult electrolysis process, said cell comprising

- a cathode forming the bottom of said electrolytic cell and comprising a plurality of parallel cathode blocks, each cathode block carrying at least one current collector bar and two electrical connections points,

- a lateral lining defining together with the cathode a volume containing the liquid electrolyte and the liquid metal resulting from the Hall-Heroult electrolysis process,

said cathode and lateral lining being contained in an outer metallic shell,

- a plurality of anode assemblies suspended above the cathode, each anode assembly comprising at least one anode and a metallic anode rod connected to an anode busbar (so-called "anode beam"),

- a plurality of alumina feeders by which alumina powder is fed into the liquid bath,

said method comprising the following steps:

- dividing said cell into at least n subsystems,

- for each subsystem, estimating the local value of at least one target or output parameter on the basis of the value of the measurement of at least one input parameter, - modifying at least one of said input parameters in the cell by exerting a physical action upon the cell, if at least one estimated value of said output parameter, for at least one subsystem, is substantially different from another estimated value of said output parameter for another subsystem.

Said output parameter is the local alumina concentration in the subsystem and/or the alumina dissolution rate and/or the local anode-cathode distance in the subsystem.

Said input parameter is selected from the group formed by: currents of anodes in the subsystem, anode beam movement in the subsystem, cell voltage in the subsystem, alumina dumps in the subsystem, ACD change rate, and/or said input parameter is the output of a previous estimate of one or more of said local variables.

In an advantageous embodiment said input parameter is the individual anode current determined for each anode in the subsystem.

The number of subsystems (n) is advantageously chosen between 2 and 16, and is preferably equal to 4. Said subsystems correspond to sectors of substantially same length, divided along the main dimension of the cell. In an advantageous embodiment said output parameter is the local concentration of alumina in the bath, and said physical action comprises increasing or decreasing the alumina dump in the subsystem in which the estimated concentration of alumina in the bath is inferior or superior, respectively, to a predetermined target value.

In another advantageous embodiment, which can be combined with the previous one, said output parameter is the anode - cathode distance, and said physical action comprises increasing or decreasing the anode - cathode distance, when the estimated anode - cathode distance is inferior or superior, respectively, to a predetermined target value.

Said method can comprise the following steps,

- measuring at least one value, called here "level 1 measured value", of said input parameter for the whole cell,

- obtaining at least one value, called here "level 1 estimated value", of said output parameter, on the basis of said level 1 measured value, using a mathematical observer;

- dividing said cell into at least two subsystems called "level 2 subsystems", - for each level 2 subsystem, measuring at least one value, called "level 2 measured value", of said input parameter,

- for each level 2 subsystem, obtaining at least one value, called "level 2 estimated value", of said output parameter, on the basis both of said level 2 measured value and of said level 1 estimated value, using said mathematical observer.

It can further comprise the following steps, carried out in a recursive (cascaded) way:

- dividing said cell into at least 2" subsystems of level n,

- for each subsystem of level n: measuring at least one value, called "level n measured value", of said input parameter

- for each subsystem of level n: obtaining at least one value, called "level n estimated value", of said output parameter, on the basis both of said level n measured value and of said level estimated value, using said mathematical observer.

In an advantageous embodiment said mathematical observer is of the Kalman filter-type.

Another object of the invention is an electrolytic cell suitable for the Hall-Heroult

electrolysis process, said cell comprising

a lateral lining defining together with the cathode a volume containing the liquid electrolyte and the liquid metal resulting from the Hall-Heroult electrolysis process, said cathode and lateral lining being contained in an outer metallic shell,

a plurality of anode assemblies suspended above the cathode, each anode assembly comprising at least one anode and a metallic anode rod connected to an anode busbar (so-called "anode beam"),

a plurality of alumina feeders by which alumina powder is fed into the liquid bath, and said electrolytic cell being characterized in that it comprises specific means for carrying out a method according to the invention.

Said specific means comprise a specifically programmed microprocessor. In an advantageous embodiment said cell comprises means to determine individual anode currents for each anode. The method according to the present invention has several advantages in view of prior art, amongst which:

1. The observer produces quantitative spatial estimation of state variables based on a dynamic model, which provides a more systematic way of utilizing anode current measurements.

2. The observer has a flexible structure. The exact discretization of the cell can depend on the level of details required and the area of interests.

3. The multi-stage configuration ensures system observability (without which the system is not observable, which means that all the state variables cannot be estimated) and reduces modelling errors and uncertainties as estimations from the previous stage are served as additional information as well as constraints.

4. The observer is tuned properly so that it produces converging results at different starting conditions of reasonable ranges.

Figures

Figure 1 is a schematic view, showing an electrolytic cell for carrying a monitoring process according to the invention.

Figure 2 is a schematic diagram of the multi-stage state observer.

Figure 3 shows the estimated spatial alumina concentration during a time when a feeder is blocked on purpose.

Figure 4 shows estimated spatial alumina concentration during a time when a feeder or crust breaker problem has occurred. Description

1. General presentation

An aluminium smelter plant comprises a plurality of electrolytic cells arranged the one behind the other (and side by side), typically along two parallel lines. These cells are electrically connected in series by means of conductors, so that electrolysis current passes from one cell to the next. The number of cells in a series is typically comprised between 50 and over 400, but this figure is not substantial for the present invention. The cells are arranged transversally in reference of main direction of the line they constitute. In other words the main dimension, or length, of each cell is substantially orthogonal to the main direction of a respective line, i.e. the circulation direction of current.

Referring to figure 1 , a Hall-Heroult electrolytic cell 1 , the general structure of which is known per se, first comprises a cathode 2 forming the bottom of said electrolytic cell and comprising a plurality of parallel cathode blocks, each cathode block being provided with at least one current collector bar and two electrical connection ends. A lateral lining 3 defines together with the cathode a volume V containing the liquid electrolyte 8 and the liquid metal 9 resulting from the Hall-Heroult electrolysis process, said cathode and lateral lining being contained in an outer metallic potshell 4. Said electrolytic cell further comprises a plurality of anode assemblies suspended above the cathode, each anode assembly comprising at least one anode 11 -18 and a metallic anode rod 6 connected to an anode busbar 7 (so-called anode beam). In the shown example, there are two rows of eight anodes 11 -18, only one row being illustrated. Moreover, the cell includes four aluminium feeders 21 -24 (linked to an outside alumina supply, not shown on the figure) regularly provided along the main dimension of the cell, between its tap end T and its duct end D.

The Hall-Heroult process as such, the way to operate the latter, as well as the cell arrangement are known to a person skilled in the art and will not be described here in more detail. It is sufficient to explain that the current is fed into the anode beam 7, flows from the anode beam 7 to the plurality of anode rods 6 and to the anodes 11 -18 in contact with the liquid electrolyte 8 where the electrolytic reaction takes place. Then the current crosses the liquid metal pad 9 resulting from the process and eventually will be collected at the cathode blocks forming the cathode 2. In the present description, the terms "upper" and "lower" refer to mechanical elements in use, with respect to a horizontal ground surface. Moreover, unless otherwise specifically mentioned, "conductive" means "electrically conductive". In the present invention embodiment, attempt is made to estimate the local alumina concentration, alumina dissolution rate and average anode cathode distance (ACD), which are referred to as "state variables", in the above mentioned electrolyte. This estimation is carried out on the basis of the determination of four parameters, i.e. P1 : anode currents, P2: alumina dumps,

P3: anode beam movement, P4: cell voltage

In view of this determination, the cell is provided with four series of sensors or actuators, each being relative to one respective of the above listed parameters:

S11 to S18: means to sensors determine anode current. In the framework of this invention anode currents can be determined using any method. In an advantageous embodiment the number of these sensors is equal to the number of anode rods and they can be positioned on the beam between two adjacent anode rods.

A21 to A24: alumina dumps actuators. Each of these actuators is positioned at the outlet of its respective feeder 21 -24, and is controlling the opening and closing of said feeder. Each opening of the outlet is releasing a fixed volume of powdered alumina into the electrolytic bath, and the quantity of alumina fed through each feeder is calculated from said fixed volume and the density of alumina. We will refer here to a "measurement" in relation with the determination of the number of alumina dumps, although in practice it is usually not the mass of dumped alumina that is measured but the number alumina dumps of a volume of powder predetermined by the volume of the feeder is counted by the actuator.

S3: one single beam movement sensor for the whole cell.

S4: one single cell voltage sensor for the whole cell. The method according to the invention does not depend upon the means and methods used for the determination of the input parameters, provided that said means are capable of determining said input parameters.

All the sensors and actuators are connected to a central unit U; this connection is symbolized as dotted lines on figure 1 . Said central unit U is provided in particular with a calculator configured to use an extended Kalman Filter. On figure 1 the voltage beam sensor (measuring the anode beam height) and voltage beam actuator (modifying the abode beam height) is symbolized by the letter B, and the sensor measuring the cell voltage by the letter Z.

We describe here an embodiment of the method according to the invention, in relation with figure 2, where:

u^w ₍y is the input to the y^'-th observer in the /^'-th stage at time step N,

x^Ni_j is the estimation result from the y^'-th observer in the i-th stage at time step N.

Time step 0

In step 0 the system is initialised. The number of stages in the cascaded state observer is decided. Reasonable initial estimations of the global (average) state variables and estimation of error covariance are made.

Time step 1

Stage 1 The four input parameters (line current, ACD change rate (determined from anode consumption rate and aluminium accumulation rate), beam movement, alumina dumps) are determined (measured) for the whole cell. More precisely, one global value is used for each parameter, even if more than one value is measured; this means, for example, that the alumina dumps for the four feeders are summed up to become one global value, and the current value used here is the line current, which is the summation of all individual anode currents. These values serve as u¹ n to the observer. The estimation of states at time step 1 (x¹ n) is produced using the standard extended Kalman filter algorithm with the initial estimation of states and error covariance. x¹ n contains the estimation of average alumina concentration in the bath, alumina dissolution rate and average ACD at time step 1 .

In step 0, initial estimations of these variables are made, they are used to calculate an estimated cell voltage using equation (1). The difference between this estimated cell voltage and the measured cell voltage is then used to adjust the estimated variables. After some time (normally one feeding cycle), the estimated variables will be within reasonable range of the real variables.

- Stage 2

At this stage, the cell is divided into two identical subsystems, i.e. SUB1 1 and SUB12 (shown on figure 1 ).

The four input parameters are determined for the each subsystem SUB1 1 and SUB12, plus cell voltage and possibly outputs from previous state estimation stages. ACD change rate (determined from anode consumption rate and aluminium accumulation rate) and beam movement are determined (measured) for the whole cell. Local variables (i.e. variables that are determined for each subsystem SU B1 1 and SUB12) that are used are anode current and alumina dumps. More precisely, one global value is used for each subsystem, even if more than one value is measured. As a consequence, anode current in stage 2 means the summation of all individual anode current in the subsystem, and alumina dump in stage 2 means the summation of alumina dumps in each subsystem.

At this stage, the filter receives two types of input data, i.e. measured parameters of stage 2 (u¹ ₂i and u ₂), as well as estimated parameters of stage 1 (x¹ n).

u¹2i contains the respective anode currents, cell voltage, beam movement and alumina dumps measured in the first subsystem, while u ₂ contains the respective anode currents, cell voltage, beam movement and alumina dumps measured in the second subsystem. The estimation of states in the two subsystems is obtained using the standard extended Kalman filter, subject to the constraint that the average of x¹ ₂i and x ₂ is x¹ n , the estimated parameters of stage 1. x¹ ₂i and x ₂ contain the estimated alumina concentration, alumina dissolution, rate and ACD in the two subsystems; these are the outputs.

It should be noted that the operation of this filter at stage 2 is different from the one at stage 1. - Stage 3 and beyond

The cell is divided into further subsystems (SUB21 to SUB24 on figure 1). The cell is subsequently divided into subsystems and the estimation results in the previous stage are used in the current stage. The estimation stops when the desired number of stages has been reached. As in step 2, local variables that are used in each subsystem are anode current and alumina dumps.

Time step 2 and beyond

The standard Kalman filter recursion is applied in each stage, using the estimation results from previous time step as the new starting point.

In the outlined method, the estimated alumina concentration and ACD are produced after n^th stage state observer at each time step. It is an approach which provides online monitoring of spatial variables of a smelter cell. Upon the basis of the estimated values, for each parameter, appropriate regulation of the electrolytic cell can be carried out.

Thus, let us consider that the estimated value of alumina concentration is lower in subsystem SUB21 than in the other subsystems. The operator will then increase the flow rate of alumina (or the number of dumps of a predetermined volume of alumina powder), delivered by feeder 21.

Let us consider that the estimated value of anode cathode distance is lower in subsystem SUB21 than in the other subsystems. The operator will then adjust the anode- cathode distance of the anodes in the said subsystem by raising or lowering the anode Examples

These examples present results obtained through the method according to the invention. Example 1

The cell has been divided into four subsystems, where zone 1 is at the tap end, zone 2 is in between zone 1 and zone 3, zone 4 is at the duct end and adjacent to zone 3. In this example, a feeder in zone 2 is blocked on purpose while the total alumina feed rate to the cell is maintained as usual, i.e. the other feeders dump more to compensate the blocked feeder. The goal was to estimate the spatial variables in the four subsystems.

The cascaded state observer therefore contained three stages. Following the procedure described above, the spatial distribution of alumina concentration in the four zones was estimated and figure 3 depicts the estimated values of spatial alumina concentration in response to the blocked feeder. As can be seen, following the feeder blocking, there is a significant difference in the spatial alumina concentration in the four zones depending on the location. The estimated alumina concentrations in zone 1 and zone 2 decrease, while the concentrations in zone 3 and zone 4 increase. The concentration in zone 2 is the lowest since the feeder in this area is blocked, but due to the flow of electrolyte, the concentration in zone 1 is also relatively low. The same applies to the concentration in zone 3. As for zone 4, it is the furthest away from zone 2 and therefore it is the least affected by the depletion of alumina in zone 2. It causes the alumina content in this zone to increase, and to become the highest in the cell.

It should be noted that apart from the blocked feeder, the alumina addition rate in the other feeders was the same in this example. Even so, the cascaded state observer is able to produce the reasonable and logical estimation of spatial variables based on the information available.

Example 2:

In this example, the configuration of the discretised cell was the same as that in Example 1. Figure 4 shows the estimated spatial alumina concentration in the four zones with an unexpected feeder or crust breaker problem, which was later discovered to be in zone 2. Similar to what is shown in Example 1 , there is a significant difference in the spatial alumina concentration in the four zones towards the end of the period. The estimated alumina concentrations in zone 1 and zone 2 are relatively lower than the concentrations in zone 3 and zone 4. It indicates that there is a depletion of alumina concentration in zone 1 and zone 2, which may be due to the feeder/breaker problem. It should be noted that in this example, the cascaded state observer does not know the feeder/breaker problem a priori, instead it assumes the regular alumina addition rate. Even so, the estimation of spatial variables is able to pinpoint the problematic area in the cell. This allows the quick identification of the defective feeder or crust breaker device, and eventually speeds up the work of the maintenance team in the plant.

Claims

1. Method of producing aluminium in an electrolytic cell using the Hall-Heroult

electrolysis process, said cell comprising

said cathode and lateral lining being contained in an outer metallic shell,

said method comprising the following steps:

- dividing said cell into at least n subsystems,

- for each subsystem, estimating the local value of at least one target or output parameter on the basis of the value of the measurement of at least one input parameter,

- modifying at least one of said input parameters in the cell by exerting a physical action upon the cell, if at least one estimated value of said output parameter, for at least one subsystem, is substantially different from another estimated value of said output parameter for another subsystem.

2. Method according to claim 1 , wherein said output parameter is the local alumina concentration in the subsystem and/or the alumina dissolution rate and/or the local anode- cathode distance in the subsystem.

3. Method according to claim 1 or 2, wherein

- said input parameter is selected from the group formed by: currents of anodes in the subsystem, anode beam movement in the subsystem, cell voltage in the subsystem, alumina dumps in the subsystem, ACD change rate,

- and/or said input parameter is the output of a previous estimate of one or more of said local variables.

4. Method according to any of claims 1 to 3, wherein said input parameter is the individual anode current determined for each anode in the subsystem.

5. Method according to any of claims 1 to 4, wherein n is chosen between 2 and 16, and is preferably equal to 4.

6. Method according to any of claims 1 to 5, characterized in that said subsystems correspond to sectors of substantially same length, divided along the main dimension of the cell.

7. Method according to any of claims 1 to 6, wherein said output parameter is the local concentration of alumina in the bath, and wherein said physical action comprises increasing or decreasing the alumina dump in the subsystem in which the estimated concentration of alumina in the bath is inferior or superior, respectively, to a predetermined target value.

8. Method according to any of claims 1 to 7, wherein said output parameter is the anode - cathode distance, and wherein said physical action comprises increasing or decreasing the anode - cathode distance, when the estimated anode - cathode distance is inferior or superior, respectively, to a predetermined target value.

9. Method according to any of claims 1 to 8, characterized in that it comprises the following stages,

- dividing said cell into at least two subsystems called "level 2 subsystems",

- for each level 2 subsystem, measuring at least one value, called "level 2 measured value", of said input parameter,

10. Method according to any of claims 1 to 9, characterized in that it further comprises the following steps, carried out in a recursive or cascaded way:

- dividing said cell into at least 2" subsystems of level n,

- for each subsystem of level n: measuring at least one value, called "level n measured value", of said input parameter - for each subsystem of level n: obtaining at least one value, called "level n estimated value", of said output parameter, on the basis both of said level n measured value and of said level (n-1) estimated value, using said mathematical observer.

1 1. Method according to claim 10, characterized in that said mathematical observer is of the Kalman filter-type.

12. Electrolytic cell suitable for the Hall-Heroult electrolysis process, said cell comprising - a cathode forming the bottom of said electrolytic cell and comprising a plurality of parallel cathode blocks, each cathode block carrying at least one current collector bar and two electrical connections points,

a plurality of alumina feeders by which alumina powder is fed into the liquid bath, and said electrolytic cell being characterized in that it comprises specific means for carrying out a method according to any of claims 1 to 11.

13. Electrolytic cell according to claim 12, characterized in that said specific means comprise a specifically programmed microprocessor.

14. Electrolytic cell according to claim 12 or 13, characterized in that said cell comprises means to determine individual anode currents for each anode.