EP0834601A1 - Process for controlling the bath temperature in an aluminium electrolysis cell - Google Patents

Abstract

Thermal regulation of individual electrolysis cells for the production of aluminium comprises directly measuring the temperature of the bath at regular time intervals and modifying the distance between the anode and the metal as a function of the measured values of the resistance R of the cell w.r.t. a consigned resistance (Ro). During each cycle of thermal regulation of duration (Tr) corresponding to a working sequence incorporated in the duration (T) of the work cycle of the cell the following steps are taken: (a) at least one measurement of the bath temperature (θ ) is effected; (b) from the n latest measures a corrected average temperature (θ mc) representative of the medium state of the cell assembly and free from the variations in the time and space due to the periodic operations of exploitation is determined; (c) an additional resistance (RTH) is determined, this may be positive or negative and comprises two terms: (i) a prior correction term (RTHa), calculated in a manner to neutralise by anticipation the irregular but known and quantified perturbations such as the addition of solids to the bath; and (ii) a posterior correction term (RTHb), calculated as a function of the corrected average temperature (θ mc) and a consigned temperature (θ o); and (d) the additional resistance (RTH) is then applied to the consigned resistance (Ro) of the cell in order to maintain or correct the temperature of the cell.

Description

TECHNICAL AREA

The invention relates to a method for regulating the temperature of the bath. an aluminum production tank by electrolysis of dissolved alumina in an electrolyte based on molten cryolite, according to the Hall-Héroult process.

STATE OF THE ART

The conduct of an electrolysis tank for the production of aluminum requires keeping its temperature as close as possible to its optimal operating temperature or equilibrium temperature. In practical the tank temperature is given by the maximum temperature at the heart of the tank, that is to say the temperature of the electrolysis bath. The operating conditions of a tank having been previously fixed and thereby the set temperature of the electrolysis bath is through an adjustment energy supplied to the tank compared to the energy consumed or dissipated by it, that it is possible to maintain the temperature of the bath at its setpoint. In this regard, we must remember the many advantages that there is, notably in terms of production costs, being able to regulate the as finely as possible the temperature of the electrolysis bath. So a increase of the electrolyte temperature by ten degrees Celsius lowers Faraday yield by about 2% while lower of the electrolyte temperature of ten degrees Celsius can reduce the already low solubility of alumina in the electrolyte and promote "the anode effect", that is to say the anode polarization, with abrupt rise of the voltage at the terminals of the tank and clearance in large quantities fluorinated and fluoro-carbon products.

By seeking to reduce the fluctuations of the thermal equilibrium and consequently of the chemical equilibrium of the bath which is intimately linked to it, for example by means of aluminum fluoride AlF _{3 additions} intended to adjust the acidity of the bath as well as its liquidus temperature or temperature at which solidification begins, the aim is to approach the optimal operating conditions, in particular for the equilibrium temperature. Faraday yields of around 95% can thus be reached, or even 96% in the case of acid baths, therefore containing a large excess of AlF ₃ which makes it possible to lower the equilibrium temperature to around 950 °. C or even below.

Another advantage of a very efficient thermal regulation is to promote the permanent maintenance of a sufficiently solidified bath slope thick on the sides of the tank and thus protect them against erosion, oxidation, chemical attack by liquids and aluminum. This protection of sides by the solidified bath slope obviously promotes the longevity of the pot lining and insofar as this solidified embankment is sufficiently thick, it causes a decrease in the lateral heat flow, hence a reduction in heat losses resulting in a reduction significant energy consumption.

In fact, even in the most recent state of the art, this thermal regulation is very delicate to implement industrially.

First of all because there are no efficient means to control the temperature of the sodium-fluoride electrolysis bath in a sufficiently reliable and frequent manner in the vicinity of 950 ° C. One cannot in fact have recourse to a temperature probe immersed continuously in the bath given its very great chemical aggressiveness. The use of a thermowell made of silicon nitride or titanium diboride placed in a side wall of the tank at the level of the bath and in which is housed a temperature probe according to FR 2104781 only makes it possible to measure the temperature of the bath 'in the vicinity of the wall and moreover with significant inertia, therefore without the possibility of rapidly detecting small temperature variations (2 to 3 ° C). Finally, indirect measurements of bath temperature and in particular electrical measurements based on variations in resistance of the bath with the temperature as recommended by SU 1236003 do not allow precise control of this temperature either because the resistivity of the bath varies locally due to that it is never perfectly homogeneous, but also over time because its composition changes with the additions of alumina and AlF ₃ .

Ultimately, the temperature measurements of the electrolysis bath are still very often performed manually and periodically by an operator who open the hood or the tank door and immerse a cane in the bath pyrometric. This procedure clearly presents many disadvantages: releases of fluorinated gases into the surrounding atmosphere, operator exposure to these harmful releases, low frequency of these measurements (typically 1 measurement every one or two days) difficult to perform and therefore not ensuring sufficiently monitored temperature control for perform precise and reliable regulation to meet new requirements of modern electrolytic cells.

But it is above all the difficulty of controlling the thermal equilibrium of the tank because of its inertia which makes it very difficult to implement a regulation of tank temperature, especially since the tank is of high capacity. Indeed, the drifts can be long to appear but, when they appear, they are difficult to contain and to correct. Some disturbances are part of the normal operation of the tank. Among them some return at regular time intervals (change of anode others) are irregular and of variable magnitude (addition of solid bath for example). We can therefore predict these disturbances and keep them matters but it is not the same with unpredictable disruptions (anode effect, sudden change in temperature due to an anomaly in operation).

In practice, one acts in a punctual manner on various parameters which have an indirect corrective effect on the temperature and in particular the excess of AlF ₃ compared to the composition of the cryolite, determined by sampling and chemical analysis in the laboratory. This regulation which implements corrective additions of AlF ₃ is generally qualified as thermal in the sense that it takes account of the excess of AlF ₃ and of the temperature and that it ultimately acts on the temperature because of the relationship between the chemistry and the thermal of the electrolyte, but this thermal effect is obtained with a significant delay. This traditional mode of regulation does not take into account the differences in reaction time between thermal and bath chemistry under transient conditions, whereas the role of regulation is precisely to intervene as soon as the tank tends to move away from its point of equilibrium. The thermal of the tank (the temperature of the bath) reacts quickly to a thermal stress. For example, the tank reacts very quickly to an increase in power, even if the reaction takes its full extent only after a few hours or ten hours due to the thermal inertia of the tank. On the contrary, the chemistry of the bath, in particular the excess of AlF ₃ , changes only with a significant delay, the effect of an addition of AlF ₃ appearing only several tens of hours at several days after the time of the addition.

Furthermore, it should be remembered that the higher the excess of AlF ₃ , the more the electrical resistivity of the bath increases, which results, if the resistance at the terminals of the tank is kept constant, by a reduction in the anode distance- metal (DAM) which can be detrimental to Faraday performance. Conversely, a lack of AlF ₃ leads to a reduction in the resistivity of the bath which, if the resistance of the tank is kept constant, results in an increase in the unnecessary anode-metal distance which is detrimental to energy efficiency.

On a similar principle EP 0671488A describes a thermal regulation process according to which a theoretical calculation of the energy dissipated in and by the electrolysis cell in its various forms is periodically carried out: energy necessary for the reduction of alumina but also energy absorbed by the various additives, such as alumina and AlF ₃ , as well as by operating operations (changes of anode for example). This dissipated energy is compared to the energy supplied to the tank for a predefined operating speed. The deviations are then corrected by acting on the set resistance, which is increased by increasing the anode-metal distance (DAM), if there is a deficit in the energy supplied, or which is reduced by decreasing the anode-metal distance if there is an excess of energy. However, if we only consider the return of heat by the re-oxidation of aluminum corresponding to the one hundred lack of Faraday yield, very unstable over time and depending on the state of the tank, or the fluctuating mass of cover product based on alumina and solid bath which falls into the tank when the anode is changed, it is obvious to a person skilled in the art that the accuracy of such a theoretical calculation can be at best 5%, which corresponds to an indeterminacy of several tens of degrees. Such a method is therefore inapplicable for finely regulating the temperature of the bath of an electrolysis tank to within a few degrees.

In addition, the author's certificate SU 1 183 565 describes a method of regulation temperature according to which the temperature of the bath of the tank and the anode-metal distance is directly and only modified proportionally, on the one hand, to the difference between the last temperature measured and the set temperature, and on the other hand, the difference between the last measured temperature and the previous one. This approach does not hold account of the various disturbances that are part of industrial operations normal electrolytic cells, such as anode changes and solid bath additions, which disturbances cause variations in temperature up to several tens of degrees. For example, after the installation of a new anode, the temperature of the bath drops very quickly and very strongly, especially in the vicinity of this anode. The in accordance with SU 1 183 565 would in this case impose a large increase in the anode-metal distance which would cause, due to the thermal inertia of the tank, an over-adjustment and, consequently, an abnormal heating of the tank and a thermal imbalance prejudicial in particular to the energy consumption and Faraday efficiency.

Thus, no known method of thermal regulation of an electrolytic cell makes it possible to directly detect, and a fortiori to instantly correct, a small thermal imbalance of the bath, and the subsequent corrective actions of the temperature carried out indirectly by regulating the amount of AlF _{3 is} insufficient to avoid thermal and chemical fluctuations.

PROBLEM

With the search for very high levels of performance on the tanks modern high capacity, it has become essential to regulate very precise and reliable the temperature of the electrolysis bath compared to a target equilibrium temperature or setpoint temperature, this in particular to obtain a Faraday yield of at least 95%, or even 96% with acid baths, simultaneously improving the yield energy of the tanks, very sensitive as previously indicated to thermal equilibrium fluctuations and as a result of the stabilization of the bath slope solidified on the sides of the tank.

OBJECT OF THE INVENTION

The method according to the invention provides a solution to the problem of individual thermal regulation of electrolytic cells. It consists of acting on the temperature of the tank via the setpoint resistance Ro, which is modulated so as to correct the temperature both in advance and by feedback. On the one hand, the correction in advance, known as "a priori", takes into account known and quantified disturbances and allows compensate in advance for effects on tank temperature. On the other hand, the correction by feedback, known as "a posteriori", consists, starting from the direct measurement of the bath temperature at regular time intervals electrolysis, to determine an average temperature corrected according periodic operations and to compensate for variations and deviations of this temperature from a set temperature. The corrections are made by regularly adjusting a so-called value of additional resistance, positive or negative, added to the resistance of tank setpoint, so that it warms the tank temperature towards the setpoint and limits variations over time.

• on effectue au moins une mesure de température  du bain ;
• on détermine à partir des n dernières mesures, une température moyenne corrigée mc, représentative de l'état moyen de l'ensemble de la cuve et affranchie des variations dans le temps et l'espace dues aux opérations périodiques d'exploitation ;
• on détermine une résistance additionnelle corrective RTH, positive ou négative, constituée de 2 termes :
  • un terme RTHa de correction a priori, calculé de manière à neutraliser par anticipation les perturbations irrégulières mais connues et quantifiées comme les additions de bain solide,
  • un terme RTHb de correction a posteriori, calculé en fonction de la température moyenne corrigée mc et de la température de consigne o, de manière à faire tendre la température moyenne corrigée de la cuve mc vers la valeur de consigne o et à en limiter les variations dans le temps ;
• on applique la résistance RTH à la résistance Ro de consigne de la cuve, pour maintenir ou corriger la température de la cuve.

More specifically, the subject of the invention is a process for thermal regulation of an aluminum production tank by electrolysis of alumina dissolved in an electrolyte based on molten cryolite according to the Hall-Héroult process comprising direct measurement and at intervals of regular times of the bath temperature, and comprising modifications of the anode-metal distance as a function of the measured values of the resistance of the tank R with respect to a reference resistance Ro, characterized in that, during each cycle of thermal regulation of duration Tr, corresponding to a work sequence included in the operating cycle of the tank of duration T:

• at least one temperature measurement  of the bath is carried out;
• a corrected average temperature mc, representative of the average state of the entire tank and free from variations in time and space due to periodic operating operations, is determined from the n last measurements;
• an additional corrective resistance RTH, positive or negative, consisting of 2 terms is determined:
  • an a priori correction term RTHa, calculated so as to neutralize in advance irregular disturbances which are known and quantified as solid bath additions,
  • an a posteriori correction term RTHb, calculated as a function of the corrected mean temperature mc and the set temperature o, so as to cause the corrected mean temperature of the tank mc to tend towards the setpoint o and at limit variations over time;
• the resistance RTH is applied to the setpoint resistance Ro of the tank, to maintain or correct the temperature of the tank.

The term RTHb is advantageously calculated using a regulator, preferably according to an algorithm comprising a proportional action, integral and derivative.

Generally, the calculation of RTHb is performed such that, if the corrected average temperature of the bath is lower than the temperature of setpoint, i.e. if mc <o, this additional resistance is increased consequently, if the corrected average temperature mc is on the way to decrease we also increase this additional resistance in consequence, if the corrected average temperature is higher than the set temperature, i.e. if mc> o, this resistance is reduced accordingly and if the corrected average temperature mc is in the process of increasing, this additional resistance is also reduced Consequently.

Preferably, the values of RTHb are limited so as to keep them at within an allowable range, including a lower safety threshold (RTHb min) and an upper safety threshold (RTHb max). In practice, the values calculated from RTHb which go outside the admissible range are brought back to the closest threshold value. Such a limitation of the values allowed for RTHb make it possible in particular to avoid over-corrections that could cause abnormal temperature values.

The bath temperature measurement is a point measurement in space (at a given point in the tank) and over time (at a given point in time a periodic measurement cycle). The temperature of the bath varies at the same time according to the place of the tank where we are placed (at a given time) and according to the time of the measurement (at a given location). If we consider the effect of change of an anode for example, at a given time, the temperature the lower the measured anode is, the closer it is to the point of measure, and over time, the measured temperature is lower as the change of anode is recent. The temperature measurement is therefore not directly usable, even when the tank is in normal and fixed operating conditions, i.e. correctly regulated, stable and avoiding, by appropriate waiting, the direct impact of disturbing operations of operation or adjustment such as change anode, metal casting or specific regulation procedure.

It is therefore necessary to perform an average over time permettantm allowing to overcome short-term temperature fluctuations, especially variations due to known periodic disturbances and in particular to periodic operations, but a correction must also be made spatial Δ to obtain a value representative of the entire tank, i.e. mc = m + Δ. This spatial temperature correction determined experimentally can reach 10 ° C depending on the operations considered and the position of the measuring point.

In practice, the bath temperature must be measured at least once per cycle of thermal regulation Tr corresponding to a work sequence. This measurement can be done manually discontinuously but much more efficiently using a special semi-continuously submerged sensor in the bath and allowing for much greater temperature measurements frequency for example every hour.

Taking into account the corrections in time and space, we then calculate the average temperature corrected from bath temperature measurements thermal regulation cycles of duration Tr included in the cycle operating anode and casting change whose duration T is generally 24, 30, 32, 36, 40, 42 or 48 hours, this gives the corrected average temperature mc which is used for regulation. In practice, this temperature is recalculated as a sliding average corrected after each new bath temperature measurement made at at least once per thermal regulation cycle of duration Tr corresponding to a work sequence generally of 4, 6, 8 or 12 hours.

Figures 1a to 1c illustrate the calculation of the corrected average temperature, which is used to determine the correction term RTHb at item j, in the case where an anode change was made after measuring the temperature at station d - 4 and where the calculation of the average temperature is performed using the temperature values measured at stations j - 3 to j. The figure la corresponds to the case where the changed anode is at a so-called position intermediate with respect to the measurement point, hence the fact that Δ is zero. The Figure 1b corresponds to the case where the changed anode is relatively close to the measuring point, hence a positive Δ. Figure 1c corresponds to the case where the anode changed is relatively far from the measurement point, resulting in a negative Δ.

• soit sous forme de température moyenne corrigée mb obtenue directement à partir des mesures de température du bain dont les valeurs sont généralement comprises entre 930°C et 980°C, cette température moyenne corrigée mb étant comparée à la température de consigne o de la cuve par exemple 950°C,
• soit sous forme de température moyenne corrigée différentielle md représentant l'écart de température entre la température moyenne corrigée mb précédemment définie et la température de liquidus l du bain, sachant qu'à une composition chimique donnée du bain d'électrolyse correspond une température de liquidus donnée. On connaít sous le nom de surchauffe cet écart de température entre la température du bain et la température de liquidus, il s'ensuit dans le cas présent que la température moyenne corrigée différentielle md n'est autre que la surchauffe moyenne corrigée. Celle-ci est comparée à la température différentielle de consigne od ou encore surchauffe de consigne fixée par les paramètres d'exploitation de la cuve tenant compte notamment du flux thermique latéral (proportionnel au coefficient d'échange moyen entre le bain et le talus multiplié par la surchauffe) lié à l'épaisseur du talus de bain solidifié latéral.

It should also be specified that the corrected average temperature mc can be formulated in 2 ways:

• either in the form of corrected average temperature mb obtained directly from bath temperature measurements, the values of which are generally between 930 ° C and 980 ° C, this corrected average temperature mb being compared to the set temperature températureo of the tank for example 950 ° C,
• either in the form of a corrected differential average temperature md representing the temperature difference between the corrected mean temperature mb previously defined and the liquidus temperature l of the bath, knowing that to a given chemical composition of the electrolysis bath corresponds a given liquidus temperature. We know under the name of overheating this temperature difference between the bath temperature and the liquidus temperature, it follows in this case that the corrected differential average temperature md is none other than the corrected mean overheating. This is compared to the differential setpoint temperature od or setpoint overheating set by the operating parameters of the tank, taking into account in particular the lateral heat flux (proportional to the average exchange coefficient between the bath and the slope increased by overheating) related to the thickness of the lateral solidified embankment.

Thus, as an additional resistance adjustment parameter is used RTHb, either the corrected average temperature mb, or the temperature corrected mean differential md usually called overheating corrected mean, that is to say the 2 parameters at the same time, for example as it is described in the implementation of the invention (example e), where the temperature corrected mean mb is chosen as the basic setting parameter for the additional resistance and where the corrected average superheat md is taken into account if it exceeds a fixed threshold.

If the corrected average superheat md is used as the setting, the corresponding temperature l of the liquidus, traditionally calculated from the chemical composition of bath which should therefore be determined simultaneously during the work sequence considered. Liquidus temperature and overheating can also be obtained by direct measurement on the tank electrolysis using an appropriate device.

• d'une part des perturbations irrégulières mais connues et quantifiées comme par exemple les additions de bain solide dont on neutralise a priori et par anticipation l'action de refroidissement par une augmentation de la résistance Ro de consigne de la cuve à l'aide d'une résistance additionnelle positive RTHa dont la valeur est calculée en fonction du débit d'addition de bain broyé, cette augmentation de la résistance de consigne étant en pratique mise en oeuvre par une légère augmentation de la DAM dans la cuve,
• d'autre part des perturbations imprévisibles (incidents ou anomalies de fonctionnement) qu'il convient de détecter le plus tôt possible pour les contenir puis les corriger rapidement et retrouver la température de consigne o ou od si l'on considère la surchauffe de consigne et cela par l'application d'une seconde résistance additionnelle positive ou négative RTHb à la résistance Ro de consigne de la cuve.

If the determination of a corrected average temperature mc (that is to say mb or md) is representative of the average condition of the entire tank and freed by a corrective term for variations due to operations operating periods such as anode changes, it does not take into account the effects on the bath temperature:

• on the one hand, irregular but known and quantified disturbances such as for example the solid bath additions, the cooling action of which is a priori and anticipated neutralized by an increase in the resistance Ro of the setpoint of the tank using an additional positive resistance RTHa the value of which is calculated as a function of the flow rate of addition of ground bath, this increase in the set resistance being in practice implemented by a slight increase in the DAM in the tank,
• on the other hand unpredictable disturbances (incidents or malfunctions) that should be detected as soon as possible to contain them then correct them quickly and find the set temperature o or od if we consider the overheating of setpoint and this by applying a second additional positive or negative resistance RTHb to the setpoint resistance Ro of the tank.

Thus, the additional resistance includes a term RTHa, for which it is required account at certain items, intended to compensate in advance for irregular disturbances but known and quantified as the additions of solid bath, and an RTHb term calculated according to the values of mb and md compared to the set values, as well as their evolution.

It is therefore from a reference resistance Ro periodically corrected of a value RTH = RTHa + RTHb that the regulation of the tank takes place. AT from Ro, which may include other terms (e.g. terms intended to ensure the electrical stability of the tank), the regulation generally modify the anode-metal distance (DAM) so that if the resistance R measured regularly across the cell (with R = (U-E) / lc, U terminal voltage, E electrolysis voltage and lc intensity of the electrolysis current) remains below the set resistance, the regulation gives an order to raise the anode frame to increase the anode-metal distance (DAM) so as to increase the resistance of the bath and to approach the set resistance. Conversely, if the resistance measured becomes greater than the set resistance, the regulation gives an anode frame descent order to decrease the anode-metal distance (DAM), so as to decrease the resistance of the bath and to approach the set resistance.

The process according to the invention will be better understood from the description detailed of its implementation based on Figures 1 to 4 corresponding to typical profiles of temperature evolution during the cycles of thermal regulation.

IMPLEMENTATION OF THE INVENTION

• composition du bain : cryolithe AlF₃, 3 NaF + 12 % excès AlF₃
• température de consigne o = 950°C
• température liquidus l = 938°C
• surchauffe de consigne od = 12°C
• durée du cycle de régulation thermique Tr = 1 poste de 8 heures
• durée du cycle d'exploitation T = 32 heures
• nombre de mesure de température par poste = 1
• moyenne corrigée calculée sur les 4 dernières mesures de température
• résistance de consigne Ro = 5,930 µΩ
• plage admissible pour RTHb fixée à RTHb = - 0,100 µΩ et RTHb max = + 0,200 µΩ
• résistance R aux bornes de la cuve calculée périodiquement à partir de la relation R [ohm] = (U-E) / lc , où U est la tension aux bornes de la cuve en volts, lc l'intensité du courant d'électrolyse en ampères et E la tension d'électrolyse avec par exemple E=1,65 volts dans le cas présent.

The method according to the invention was implemented for several months on prototypes of electrolytic cells with precooked anodes supplied at 400,000 amperes. The alumina is introduced directly into the molten electrolysis in successive doses of substantially constant mass through several introduction orifices kept permanently open by a crust breaker. The bath additions in the form of a ground bath or of cryolite and the AlF ₃ additions intended respectively for adjusting the volume and the acidity of the bath are carried out in a similar manner:

• composition of the bath: cryolite AlF ₃ , 3 NaF + 12% excess AlF ₃
• set temperature o = 950 ° C
• liquidus temperature l = 938 ° C
• setpoint overheating od = 12 ° C
• duration of the thermal regulation cycle Tr = 1 8-hour shift
• duration of the operating cycle T = 32 hours
• number of temperature measurements per station = 1
• corrected average calculated on the last 4 temperature measurements
• setpoint resistance Ro = 5,930 µΩ
• admissible range for RTHb fixed at RTHb = - 0.100 µΩ and RTHb max = + 0.200 µΩ
• resistance R at the terminals of the cell calculated periodically from the relation R [ohm] = (UE) / lc, where U is the voltage at the terminals of the cell in volts, lc the intensity of the electrolysis current in amperes and E the electrolysis voltage with for example E = 1.65 volts in the present case.

Bath temperature measurements made at least once per shift station 8 hours on stable tank, adjusted and out of process disturbing operation or adjustment are carried out in very good conditions with the temperature and bath level measurement device electrolysis as described in patent FR-2727985 (= EP-A-0716165). This device indeed allows with the same probe many and frequent bath temperature measurements with an accuracy of ± 2 ° C for each unit measurement, without manual intervention therefore without safety risks and operator health.

The term RTHb was calculated by a regulator including an action proportional, integral and derived, and in some cases including a term of correction of overheating. The proportional correction term P has been calculated with a correction coefficient fixed at p = - 0.0400 µΩ / ° C, this coefficient corrector preferably being in the range - 0.5000 µΩ / ° C ≤ p ≤ - 0.0002 µΩ / ° C; the integral corrective term I has been calculated with a coefficient corrector fixed at i = - 0.00005 µΩ / ° C, this corrector coefficient being preferably in the range - 0.10000µΩ / ° C ≤i≤0.00000µΩ / ° C; the derivative corrective term D was calculated with a correction coefficient fixed at d = - 0.0200 µΩ / ° C, this correction coefficient preferably being included in the range - 0.5000 µΩ / ° C ≤ d ≤ 0.0000 µΩ / ° C. The correction coefficient of overheating s was - 0.0150µΩ / ° C in the cases described, this coefficient corrector s preferably being in the range - 0.5000 µΩ / ° C ≤ s ≤ 0.0000 µΩ / ° C.

In addition to the value of RTHb, the term has also been taken into account RTHa correction, which term was equal to + 0.058 µΩ in the cases presented (in proportion of the bath addition flow crushed by the automatic device power supply).

Cases a) to e) presented below correspond to different situations observed during the months of implementation of the method according to the invention. These cases correspond respectively to Figures 2 to 5, in which the evolution of the values between two successive values is shown in line thin for m and thick line for mc.

a) Case where mc was increasing and where the term RTHb was in the range admissible (according to figure 2)

The average values m obtained were:
m (d) = 943.5 ° C and m (d-1) = 942.5 ° C.

An anode change was made during station j - 4, before the temperature measurement, and during station j, also before the temperature measurement. The temperature correction Δ determined by the regulator according to the correction tables stored and applied to the average temperature was + 4.2 ° C for station j, which corresponds to the fact that the anode changed at station j was very close to the temperature measurement point, and - 0.9 ° C for station j - 1, which corresponds to the fact that the anode changed at station j - 4 was relatively far from the point of temperature measurement. Thus, the corrected average temperatures were as follows:
mc (j) = mb (j) = 943.5 + 4.2 = 947.7 ° C;
mc (d-1) = mb (d-1) = 942.5 - 0.9 = 941.6 ° C.

Corrected average temperatures actually reveal a trend pronounced at the increase in the temperature of the tank that does not reveal only partially the uncorrected average temperature.

• terme correctif proportionnel P = p x (mb(j) - o) = - 0,0400 x [947,7-950] = + 0,092 µΩ
• terme correctif intégral I = l(j - 1) - i x (mb(j) - o) = 0,00005 - 0,00005 x [947,7 - 950] = 0,00017 µΩ arrondi à 0,000 µΩ pour le calcul de RTHb
• terme correctif dérivé D = d x (mb(j) - mb(j-1)) = - 0,0200 x (947,7 - 941,6) =-0,122 µΩ

donc RTHb = 0,092 + 0,000 - 0,122 = - 0,030 µΩ.These values were then used to calculate the PID regulation parameters of the term RTHb of station j:

• proportional corrective term P = px (mb (j) - o) = - 0.0400 x [947.7-950] = + 0.092 µΩ
• integral corrective term I = l (j - 1) - ix (mb (j) - o) = 0.00005 - 0.00005 x [947.7 - 950] = 0.00017 µΩ rounded to 0.000 µΩ for RTHb calculation
• derived corrective term D = dx (mb (d) - mb (d-1)) = - 0.0200 x (947.7 - 941.6) = -0.122 µΩ

therefore RTHb = 0.092 + 0.000 - 0.122 = - 0.030 µΩ.

Although the temperature mb (j) is less than o, the rapid growth of the temperature makes the term derivative preponderant and leads to introduce an additional negative resistance RTHb = - 0.030 µΩ which remains in the admissible range for RTHb.

The correction term RTH for item j was therefore equal to:
RTH (j) = RTHa + RTHb = + 0.058 µΩ - 0.030 µΩ = + 0.028 µΩ.

Thus, despite a fairly marked tendency to increase the tank temperature, RTH correction is actually slightly positive because the a priori correction term RTHa, which counterbalances the term of RTHb a posteriori regulation, anticipates cooling.

b) Case where mc was decreasing and where RTHb was in the admissible range (according to figure 3)

The average values m obtained were:
m (d) = 951.3 ° C and m (d-1) = 954.9 ° C

In this case, an anode change was made during station j - 3. The temperature correction applied was + 1.5 ° C for stations j and j - 1, which corresponds to the fact that the changed anode was relatively close to the temperature measurement point. The corrected average temperatures were therefore:
mc (j) = mb (j) = 951.3 + 1.5 = 952.8 ° C
mc (d-1) = mb (d-1) = 954.9 + 1.5 = 956.4 ° C

For the parameters of the PID regulation at station j, we obtain:
P = - 0.0400 x (952.8 - 950) = - 0.1 12 µΩ
I = 0.0001 1 - 0.00005 x [952.8 - 950] = - 0.00003 µΩ rounded to 0.000 µΩ
D = - 0.0200 x (952.8 - 956.4) = + 0.072 µΩ
therefore RTHb = - 0.112 + 0.000 + 0.072 = - 0.040 µΩ

The proportional term prevails over the derivative term and leads to introduce an additional negative resistance RTHb = - 0.040 µΩ which remains within the admissible range and which aims to reduce the temperature of the tank.

The correction term RTH at position j was therefore equal to:
RTH (j) = RTHa + RTHb = + 0.058 µΩ - 0.040 µΩ = + 0.018 µΩ.

This slightly positive term, which reflects a compensating effect mutual a priori and a posteriori correction terms, leads to a relatively low setpoint correction.

c) Case where mc was substantially constant, with mb> o, and where RTHb left of the admissible range (according to figure 4)

The average temperature values obtained were:
m (d) = 955.0 ° C
m (d-1) = 955.6 ° C.

In this case, an anode change was made during station j - 2. The temperature correction applied was + 1.2 ° C for stations j and j - 1, which corresponds to the fact that the changed anode was relatively close to the temperature measurement point. The corresponding corrected mean temperature values were:
mc (j) = mb (j) = 955.0 + 1.2 = 956.2 ° C
mc (d-1) = mb (d-1) = 955.6 + 1.2 = 956.8 ° C.

It will be noted that the difference between the corrected mean temperatures mb (j) and mb (j-1) is less than 1 ° C therefore the precision of the unit measurements of temperature that one would expect from the most efficient devices.

For the parameters of the PID regulation of station j, we obtain:
P = - 0.0400 x (956.2 - 950) = - 0.248 µΩ
I = - 0.00008 - 0.00005 x [956.2 - 950] = - 0.00039 µΩ rounded to 0.000 µΩ
D = - 0.0200 x (956.2 - 956.8) = + 0.012 µΩ
therefore RTHb = - 0.248 + 0.000 + 0.012 = - 0.236 µΩ, which is limited to - 0.100 µΩ, since it is located below the lower safety threshold.

The correction term RTH at position j was therefore equal to:
RTH (j) = RTHa + RTHb = + 0.058 µΩ - 0.100 µΩ = - 0.042 µΩ.

The proportional term here becomes preponderant compared to the term derivative and the significantly high level of temperature leads to introduce an additional negative resistance RTHb, certainly limited to - 0.100 µΩ (low limit), but important and which counterbalances the term early correction RTHa.

d) Case where mc was substantially constant, with mb <o, and where RTHb was within the permissible range (according to figure 5)

The average temperature values obtained were:
m (d) = 944.1 ° C
m (d-1) = 945.7 ° C

An anode change was made during station j - 4, before the temperature measurement, and during station j, also before the temperature measurement. The temperature correction applied was + 1.5 ° C for stations j, which corresponds to the fact that the changed anode was relatively close to the temperature measurement point and - 0.9 ° C for station j - 1, which corresponds to the fact that the changed anode was relatively far from the measurement point. The corresponding corrected mean temperature values were:
mc (j) = mb (j) = 944.1 + 1.5 = 945.6 ° C
mc (d-1) = mb (d-1) = 945.7 - 0.9 = 944.8 ° C

The correction of the average temperature reveals that the tendency to the increase is actually in the opposite direction from what the uncorrected mean temperature, which leads to a change in sign of the action derived from the term RTHb.

For the parameters of the PID regulation at station j, we obtain:
P = - 0.0400 x (945.6 - 950) = + 0.176 µΩ
I = - 0.00018 - 0.00005 x [945.6 - 950] = + 0.00004 µΩ rounded to 0.000 µΩ
D = - 0.0200 x (945.6 - 944.8) = - 0.016 µΩ
therefore RTHb = + 0.176 + 0.000 - 0.016 = + 0.160 µΩ

The proportional term takes precedence over the derivative term and the significantly low temperature level leads to the introduction of strong additional positive resistance RTHb = + 0.160 µΩ which remains in the admissible range from - 0.100 µΩ to + 0.200 µΩ.

The correction term RTH at position j was therefore equal to:
RTH (j) = RTHa + RTHb = + 0.058 µΩ + 0.160 µΩ = + 0.218 µΩ.

The combined effect of the a posteriori correction term and the a priori correction can largely compensate for a negative deviation, and significant, compared to the set point combined with a tendency to predictable cooling.

e) Case where the calculation of RTHb took into account the correction of overheating

This consideration of overheating may be subject to certain conditions, namely in this case: RTHb value greater than zero and superheat value greater than the set superheat.

The overheating correction can be applied to RTHb in example d).

Thus we found RTHb = + 0.160 µΩ and an overheating md (j) = 15.7 ° C at from the liquidus temperature calculated from the composition bath chemical.

We are targeting an operating regime at 12.0% excess AlF ₃ , 938 ° C liquidus temperature, 950 ° C set temperature and 12 ° C overheating.

As the 15.7 ° C overheat is above 12 ° C, a term is obtained overheating correction S of - 0.0150 x (15.7 -12) = - 0.056 µΩ, i.e. RTHb corrected = + 0.160 - 0.056 = + 0.104 µΩ.

The correction term RTH was therefore equal to:
RTHa + RTHb = + 0.058 µΩ + 0.104 µΩ = + 0.162 µΩ.

It should also be noted that the correction coefficients p, i, d and s as well as their ranges of variation were first determined by theoretical calculations using the formulas and calculation tools of the Research Laboratory of Aluminum Fabrications Pechiney. They were then refined experimentally on the basis of the results obtained during the implementation of temperature regulation on test vessels, knowing that the configuration is all the more suitable as it allows bath temperatures to be obtained. more stable and tighter around the target temperature. These correction coefficients p, i, d and s determined in the present case for tanks of intensity lc = 400,000 amperes are easily transposable to tanks of different intensity lc '<lc or lc'> lc knowing that the previous values can be defined in relative value with respect to the intensity l 'so that:
p '= px lc / lc' = px (4 x 10 ⁵ A) / lc '
i '= ix lc / lc' = ix (4 x 10 ⁵ A) / lc '
d '= dx lc / lc' = dx (4 x 10 ⁵ A) / lc '
s' = sx lc / lc '= sx (4 x 10 ⁵ A) / lc'

INDUSTRIAL APPLICATION

In the table below are grouped the most characteristic values obtained during several months of operation with 400,000 amp tanks working first without regulating the temperature of the bath (A) then with regulating the temperature according to the invention (B). AT B AlF ₃ excess targeted % 11.8 13 Total standard deviation σ% 1.5 0.8 AlF3 excess at +/- 2 σ% 8.8 to 14.8 11.4 to 14.6 Target temperature ° C 953 947 Total standard deviation σ ° C 7 3 Temperature at +/- 2 σ ° C 939 to 967 941 to 953 Faraday yield % 94.9 96.2 Tank voltage volts 4.25 4.14 Specific energy kWh / t (ton Al) 13350 12830

With the process according to the invention, there is both a tightening of the temperature adjustment ranges and of the AlF ₃ contents around the set values and thereby the possibility of working at a lower temperature with a more acid bath without risking problems associated with walking too cold, such as poor dissolution of the alumina and bogging of the cathode bottoms since the minimum temperature of the bath remains above 940 ° C. The result is an improved Faraday yield of 1.3% and a specific energy per tonne of metal reduced by almost 500 kWh / t Al.

Claims (11)

Process for the thermal regulation of an aluminum production tank by electrolysis of alumina dissolved in an electrolyte based on molten cryolite according to the Hall-Héroult process comprising direct measurement and at regular time intervals of the bath temperature, and comprising modifications of the anode-metal distance as a function of the measured values of the resistance of the tank R with respect to a reference resistance Ro, characterized in that, during each thermal regulation cycle of duration Tr corresponding to a sequence of work included in the operating cycle of the tank of duration T: at least one temperature measurement  of the bath is carried out; a corrected average temperature mc representative of the average state of the whole of the tank is determined from the last n measurements, free of variations in time and space due to periodic operating operations; an additional resistance RTH, positive or negative, consisting of 2 terms is determined: an a priori correction term RTHa, calculated so as to neutralize in advance irregular disturbances which are known and quantified as solid bath additions, an a posteriori correction term RTHb, calculated as a function of the corrected mean temperature mc and the set temperature o, so as to cause the corrected mean temperature of the tank mc to tend towards the setpoint o and at limit variations over time; the additional resistance RTH is applied to the setpoint resistance Ro of the tank to maintain or correct the temperature of the tank. Method according to claim 1 characterized in that the term RTHb is calculated by a regulator. Method according to claim 1 or 2, characterized in that the calculation of the RTHb term includes a proportional, integral action algorithm and derived. Method according to any one of Claims 1 to 3, characterized in what the spatial temperature correction determined experimentally can reach 10 ° C depending on the operations considered and the position of the measuring point. Method according to any one of claims 1 to 4, characterized in what the corrected average temperature mc is calculated from bath temperature measurements of thermal regulation cycles Tr included in the operating cycle of anode change and casting whose duration T is conventionally 24, 30, 32, 36, 40, 42 or 48 hours. Method according to any one of claims 1 to 5, characterized in what the thermal regulation cycle corresponds to a sequence of work whose duration Tr is conventionally 4, 6, 8 or 12 hours. Process according to any one of Claims 1 to 6, characterized in what the corrected average temperature mc is expressed as of a temperature mb deduced directly from the temperature measurements of the bath and compared to the set temperature o. Process according to any one of Claims 1 to 6, characterized in what the corrected average temperature mc is expressed as a differential temperature md corresponding to the difference between the direct corrected mean temperature mb previously defined and the liquidus temperature l of the bath, also called overheating corrected mean, which is compared to the differential temperature of setpoint or setpoint overheating od. Method according to claims 1, 7 and 8, characterized in that one uses as an adjustment parameter for the additional resistance RTHb la corrected average temperature mb or corrected average superheat md or a combination of these 2 quantities. Method according to claim 8, characterized in that the temperature of liquidus l of the bath is calculated from the chemical composition of bath. Method according to claim 8, characterized in that the temperature of bath liquidus and overheating are obtained by direct measurement on the electrolytic cell using an appropriate device.

Images