The University of Queensland

THE UNIVERSITY OF QUEENSLAND

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Student Name: Phillip TREW

Course Code: MECH4500

Supervisor: Dr Peter Knights, Dr James Vaughan, Dr Alan Tomsett

Submission date: 25 October 2018

Faculty of Engineering, Architecture and Information Technology Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Abstract Demand for aluminium has increased over the last 60 years, this in conjunction with rising Australian energy prices has resulted in aluminium smelter electricity cost increasing. Additionally, within this overall energy price incline there are substantial demand and supply driven price variations for energy during each day. Smelters that rely on the electricity spot price can expect a larger variation in daily prices.

Due to these factors, this thesis aims to investigate the impact an amperage reduction has on the heat, electrical and mass balance of a conventional aluminium cell. A specific focus is placed on the relationship between the reduction of electrical inputs and aluminium production. Furthermore, this thesis seeks to understand the suitable economic environment required for the amperage reduction, and the feasibility of the reduction as an energy stabilisation strategy.

To achieve the aims of this project a quantitative strategy that involved the development of two models was adopted. One to simulate the heat, energy and mass balance of the cell, and the other to simulate to the economic impacts the reduction has on the cell. To analyse if amperage modulation is an effective energy stabilisation strategy, three types of reductions were tested; a step, staggered step and ramp reduction. These scenarios where tested for a reduction of 10, 15, and 20 kA. These scenarios were then compared by plotting the duration, opportunity cost and power saved. From this, the most viable reduction was selected with the use of weighted decision matrix.

From this analysis the 10kA ramp reduction was deemed as the most suitable energy stabilisation strategy. As it saves 115.92 MWhr over a 3 hour period when applied to all 840 cells. Furthermore, this reduction only incurs an opportunity cost of $-33.6 over the same period of time. However, an average peak energy price of 90.17 $/MWhr was used for this analysis. By increasing the energy price and plotting the opportunity cost it was found that this strategy becomes viable when an energy price of 90.47 $/MWhr is reached.

Finally, the following points where identified as areas of future study;

 Further optimisation of the model can be done by constructing a 3D model that simulates the cell dynamics in more detail.  A similar study could be conducted with the use of heat exchangers to control the heat loss from the cell.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

i. Table of Contents i. Table of Contents ...... 4 ii. List of Figures ...... 6 iii. List of Tables ...... 6 1. Abstract ...... 3 2. Introduction ...... 8 2.1. Scope ...... 8 2.2. Goals of the Project ...... 9 2.3. Assumptions ...... Error! Bookmark not defined. 3. Literature Review ...... 9 3.1. Hall-Héroult Aluminium Smelter ...... 9 3.1.1. Anode ...... 9 3.1.2. Cathode ...... 10 3.1.3. Electrolyte Bath ...... 11 3.1.4. Faradays laws ...... 12 3.1.5. Energy Balance of the Hall-Héroult Reduction Cell ...... 13 3.2. The Need for Energy Stabilisation ...... 13 3.3. Aluminium Smelter Cost Structure ...... 16 3.4. Current Power Modulation Research ...... 16 3.4.1. Shell Heat Exchanger ...... 17 3.4.2. Impact of Amperage Reduction on the Cell ...... 18 3.4.3. Energy stabilisation ...... 19 4. Methodology ...... 20 4.1. Technical Model Development ...... 20 4.2. Economic Model Development ...... 25 4.3. Strategy to Analysis the Results ...... 27 5. Results ...... 28 5.1. Technical Model ...... 28 5.1.1. Results: Step ...... 29 5.1.2. Results: Staggered Step ...... 30 5.1.3. Results Ramp ...... 31 5.2. Economic Model ...... 32 5.2.1. Results: Step ...... 32 5.2.2. Results Staggered Step ...... 33

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

5.2.3. Results: Ramp ...... 34 6. Analysis ...... 35 7. Sensitivity Analysis ...... 39 8. Conclusion ...... 41 9. Reference List ...... 42 10. Appendix ...... 44 10.1. Appendix A: Example Technical Model Calculation ...... 44 10.2. Appendix B: Example Economic Model Calculation ...... 45 10.3. Appendix C: Weighted Decision Matrix ...... 45

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

ii. List of Figures

Figure 1 Side section view of Pre-baked Aluminium Cell Figure 2 Retail price index of electricity in Australian Capital cities Figure 3 Electricity Demand and Price Fluctuations Snapshot Figure 4 Aluminium Smelter cost break down Figure 5 Shell Heat Exchanger Figure 6 Conventional Cell Response to an 80 kA Amperage Reduction Figure 7 Aluminium Smelter response with Shell Heat Exchanger Figure 8 Simulated Cell Response to 80 kA Reduction Figure 9 Technical Model with a Step Reduction for 10, 15 and 20 kA Figure 10 Technical Model with a Staggered Step Reduction for 10, 15 and 20 kA Figure 11 Technical model with a ramp reduction for 10, 15 and 20 kA Figure 12 Economic model with step reduction of 10 kA, 15 kA, and 20 kA Figure 13 Economic model with staggered step reduction of 10 kA, 15 kA, and 20 kA Figure 14 Economic model with ramp reduction of 10 kA, 15 kA, and 20 kA Figure 15 Duration of reduction for the three strategies at 10, 15 and 20 kA reduction Figure 16 Power Saved over 3 hour period for 1 cell Figure 17 Opportunity Cost of Cell Over 3 hour period for 1 cell Figure 18 Analysis of when Reduction becomes Profitable for Smelter Figure 19 Economic Model Sensitivity Analysis Figure 20 Technical Model Sensitivity Analysis

iii. List of Tables Table 1 Scope of the Project Table 2 Technical Model Inputs Table 3 Rate of Change of Liquidus point with Excess Fluoride Concentrations Table 4 Economic Model Inputs Table 5 Ramp, Step and Staggered Step Model outputs for 10, 15 and 20 kA reduction Table 6 Weighted Decision Matrix

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

iv. List of Definitions Opportunity Cost The benefits missed out on when selecting one alternative over another (the difference in profit) (Investopedia, 2018) Liquidus Temperature The lowest temperature at which an alloy is completely liquid, (LucasMihaupt Global Brazing Solutions , 2014) Electrolysis Process where electric current is passed through a substance to cause a chemical change (Encyclopaedia Britannica, 2014) Electrolyte A substance that conducts electricity as a result of the dissociation into positive and negative charged particles that migrate toward the positive and negative terminals of an electric circuit (Encyclopaedia Britannica, 2017) Electrode Electric conductor Anode Negatively charged electrode Cathode Positively charged electrode Coulomb Is the quantity of electrons transported in one second by a current of 1 A (Encyclopaedia Britannica, 2018) Sludge Undissolved alumina that is located on the bottom of the cell (Kai Grjotheim, 1993)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

1. Introduction The aluminium smelting industry is heavily reliant on the electric power supply available from the local grid. Where modern aluminium reduction cells require approximately 14.6 kW/hr to produce 1 kg of aluminium (World Aluminium, 2017). This dependency is further highlighted by the fact that the industry uses 12% of Australia’s National Electricity Market (NEM) (Perez, 2014). Due to this reliance an aluminium smelter’s profitability can be heavily dependent on energy prices that fluctuate with changes in energy demand.

Being such a heavy user of electricity, it may be feasible that under suitable economic and technical conditions, aluminium smelters can stabilise the level of electrical energy required to be supplied into the grid. This could be achieved by running the smelter at a reduced amperage during peaks in energy demand. This may increase the available energy supply for the local grid and reduce economic pressure caused by fluctuating energy prices. During this process, the heat balance of the cell would need to be managed to limit the growth of frozen cryolite

(Na3AlF6) as the cell will cool if the electric current is reduced. If the cell were to freeze, this would be highly detrimental and costly to the aluminium producer.

1.1. Scope This research project investigated the following aspects of using aluminium smelters for energy stabilisation.

In Scope Out of Scope Investigate the effects an amperage How the amperage reduction will be reduction has on the mass, heat and energy implemented to the smelter. balance of an aluminium reduction cell. Investigate the economic impacts of the Investigation of the reduction effects on the amperage reduction on an aluminium aluminium cell with a heat exchanger smelter. Investigate the feasibility of the amperage Investigation of other variables that reduction as a suitable energy stabilisation influence cell temperature, such as bath strategy. composition and anode cathode distance Investigate when the amperage reduction should be implemented and its duration. Table 1: Scope of the Project

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

1.2. Goals of the Project The main goal of this project is to investigate the heat, electrical, and mass balance of a conventional aluminium reduction cell. A specific focus is placed on the relationship between the reduction of electrical inputs and aluminium production. Furthermore, this thesis seeks to understand the suitable economic environment required for the amperage reduction, and the feasibility of the reduction as an energy stabilisation strategy.

2. Literature Review

2.1. Hall-Héroult Aluminium Smelter In nature aluminium does not occur in its pure elemental state, instead it is commonly found in combined forms such as oxides and silicates. Therefore smelters use an electrolysis method called the Hall-Héroult process to produce aluminium on an industrial scale. This process reduces liquid aluminium from dissolved alumina (Al2O3) in a molten cryolite (Na3AlF6) bath. Carbon anodes are placed in the bath and are gradually consumed, as oxygen from alumina

(Al2O3 ) is discharged onto the anode forming gaseous carbon dioxide (CO2). From this reaction pure liquid aluminium is formed and is deposited on top of the cathode. The overall chemical reaction for this process is shown below in Equation 1, (Grjotheim, Kvande, Foosnaes, Huglen, Lillebuen, Mellerud, Naterstad, 1993).

2퐴푙2푂3(푑𝑖푠푠표푙푣푒푑) + 3퐶 (푠) = 4퐴푙 (푙) + 3퐶푂2 (푔) (1) 2.1.1. Anode There are two anode designs that are used in the Hall-Héroult process; the pre-baked and the Soderberg anode. The pre-baked anode consists of a mixture of petroleum coke aggregate and a coal tar pitch binder, which are moulded into blocks and baked at 1120oC (Rivedal, 2018). In the final stage of producing pre-baked anodes, steel rods are cast into holes located on top of the blocks. In comparison, Soderberg anodes are continuously created by applying an anode paste of petroleum coke and tar pitch to the top of the anode. This paste slowly passes downwards and is baked into a solid by the heat generated in the bath (Zangiacomi, Garcia, Abreu, Kato, 2012).

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

The reaction that occurs at the anode is displayed below in Equation 2. As discussed previously, the anode is consumed during the electrolysis process and this reaction is shown below in Equation 3. Thus the pre-baked anodes need to be removed at regular intervals. This is usually after 22 to 26 days, depending on when the cell has consumed approximately 1/3 of the anode (Grjotheim et al, 1993).

2− − 2푂 − 4푒 → 푂2 (2)

퐶(푠) + 푂2(푔) → 퐶푂2 (푔) (3)

2.1.2. Cathode

As the oxygen from alumina (Al2O3) is discharged onto the carbon anode, liquid aluminium is formed on the cathode. Equation 4 displays the chemical reaction occurring at the cathode (Grjotheim et al, 1993).

퐴푙3+ + 3푒− → 퐴푙 (4) The theoretical cathode of the Hall-Héroult process is this liquid pool of aluminium. But it has become common practice within the industry to refer to the whole container of molten aluminium as the cathode (Grjotheim et al, 1993). The inner layer of this container is lined with pre-baked carbon blocks that are joined by a carbonaceous seam mix. Located on the bottom side of these blocks are grooves to accommodate for steel current collector bars. These blocks are insulated with bricks, which are positioned beneath and behind these pre-baked carbon blocks. This entire assembly is encased within a steel structure (Yurkov, 2017). Figure 1 displays a side section view of the completed assembly.

Figure 1: Side section view of Pre-baked Aluminium Cell, (Grjotheim et al, 1993)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.1.3. Electrolyte Bath

Molten cryolite (Na3AlF6) is used as an electrolyte due to its capacity as a solvent for alumina o (Al2O3). However one of its drawbacks is its high melting point of 1009 C (National Center for Biotechnology Information, 2018), therefore smelters add extra chemicals to reduce the melting point of the bath. These additives reduce the temperature of modern cells to an operation range of 920oC to 970oC, and their typical concentration ranges are displayed below (Prasad, 2000).

 10-12% aluminium fluoride (AlF3)

 2-4% alumina (Al2O3)

 4-6% calcium fluoride (CaF2)

It should be noted that these concentrations are constantly changing during the process. Fluorides are lost due to the vaporisation of the electrolyte and the hydrolysis reaction of dissolved aluminium fluoride with the water content of alumina, which may form gaseous hydrogen fluoride. Additionally fluorides are released in the form of CFC gases, this is especially the case when current density is high. These reactions result in the net removal of aluminium fluoride from the bath. But, with the use of dry-scrubbing most of these fluorides can be returned to the electrolyte. This results in a net deficit of 15-25kg of aluminium fluoride per metric tonne of aluminium produced (Grjotheim et al, 1993).

In addition, short term fluctuation to the electrolyte composition due to the continuous growth and melting of the frozen side ledge. This frozen ledge consist mainly of cryolite, with small amounts of calcium fluoride, alumina and aluminium fluoride, approximately less than 1% (Taylor & Chen, 2015).

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.1.4. Faradays laws A key component to the operation of aluminium smelter is determining the amount of materials consumed and aluminium produced. This is done by utilising two of Faraday’s laws of electrolysis;

1. The mass of a substance formed at each electrode is proportional to the number of coulomb passing through the cell (Bhattacharyya, 2015), and 2. For a given quantity of electric charge, the mass of a substance produced at each electrode is proportional to the molar mass divided by an integer that depends on the reaction in question (Bhattacharyya, 2015).

The equation derived from these two laws is displayed below in Equation 5, where ‘M’ is the molecular mass, ‘z’ the number of electrons involved in the reaction, ‘F’ is Faraday’s constant which is equivalent to 96,500 coulomb per equivalent, ‘I’ is current, ‘t’ is time, and ‘r o’ is the theoretical rate of production at the electrode.

푀퐼푡 (5) 푟 = 표 푧퐹 However, as with every processes there are losses involved, for the Hall-Héroult process this is accounted for by the current efficiency of the cell. Which is the ratio between measured rate of production and the theoretical rate of production. As seen in the below equation an estimate for the actual production rate can be determined by multiplying Equation 5 by the current efficiency, (Grjotheim et al, 1993).

푟 = 퐶퐸 ∗ 푟표 (6)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.1.5. Energy Balance of the Hall-Héroult Reduction Cell Due to the cell’s dependency on electricity, the energy balance of the cell is a major concern for all smelters. Therefore, over the years the Hall-Héroult cell has been modified to reduce the heat losses from the reduction cell. This has been done by optimising the thermal insulation of the cathode bottom and cell side walls. Due to this, the current dominant areas of heat loss are the cell walls, and the top of the cell. Where approximately 30 - 40% of the total heat loss is extracted from the cell walls, and 37-55% of total heat loss escapes via the top of the cell (Taylor, Etzion, Lavoie, Tang, 2014). However the heat balance of the cell is difficult, too much insulation and the protective layer of frozen electrolyte will begin to decrease. Exposing the cell walls to the electrolyte and causing rapid corrosion and intercalation of the carbon materials by sodium and its salts(Taylor et al 2014). Whereas, insufficient insulation will cause excessive frozen electrolyte to form on the cell walls and anode, which in turn will increase the cell voltage and thus power demand (Grjotheim et al, 1993).

2.2. The Need for Energy Stabilisation The use of a power modulation strategy by smelters to stabilise electricity grids has been identified as an opportunity within the industry. This is due to the variability of the aluminium market, and the continued increase and volatility of energy prices (Lavoie, Namboothiri, Dorreen, Chen, Zeigler, Taylor, 2011). Grid stabilisation is particularly important where a high proportion of renewable energy is used, for example South Australia and Germany.

Over the last 60 years the demand for aluminium has increased, with cells in 1948 producing 385 Al kg/day to modern cells producing 2475 Al kg/day. As a result two methods were used to satisfy this demand, either increase the amperage of existing cells, or design new high amperage cells with lower shell surface area per kW of heat generation (Namboothiri, Lavoie, Cotton, Taylor, 2009). Modern smelters operate cells in series under a high electrical current, with the largest cell operating at 600 kA (Gariépy, Couturier, Martin, Allano, Machado, Charmier, 2014). This in conjunction with the rising electricity prices seen in Figure 2, has resulted in smelter electricity costs increasing from 30% to approximately 40% of total production costs (Lavoie et al, 2011).

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Figure 2: Retail price index of electricity in Australian Capital cities, (Australian Competition & Consumer Commision , 2017)

Figure 2 highlights the overall energy price increase seen in Australia, where prices have increased by 61% (adjusted for inflation) in the last decade (Iggulden, 2017). Within this overall incline, there are substantial demand and supply driven price variations for energy during each day. Smelters that rely on the electricity spot price, can expect a larger variation in daily prices. This is exemplified in Figure 3, which is a snap shot of the NSW energy market on the 9th and 10th of May 2018. As seen in this figure the energy market experiences a 3 hour period of increased energy prices. Within this period the energy prices spike to a maximum value of 260 $/MWhr, before returning to normal at 70 $/MWhr.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

This pricing disparity can be attributed to two key factors. Firstly, the increased interconnection of larger urban centres creates a higher level of complexity which magnifies the daily, weekly and seasonal energy demand (Taylor & Chen, 2015). Additionally this complexity then increases the risk of supply problems occurring via network issues, as seen in South Australia in 2016. Secondly, the increase of renewable energy has magnified the variability of energy supplies. As renewable energy output is dependent on environmental conditions such as wind speed and solar irradiation, it has become more difficult to reliably predict the electrical output. This is amplified by the fact that energy cannot be easily stored and that the demand is price in- elastic, meaning that demand must be satisfied as it occurs (Ueckerdt, Brecha, Luderer, 2015).

Figure 3: Electricity Demand and Price Fluctuations Snapshot, (Australian Energy Market Operator, 2018)

This changing dynamic in the external forces impacting the local aluminium market is applying economic challenges for energy intensive industries such aluminium smelters. These key industry players are significantly motivated to lessen the current impacts of energy price variation to ensure they remain profitable. Formulating a cost-effective method to reduce smelter energy demands during periods of increased spot pricing, has the potential to stabilise energy supply as the grid becomes more reliant on renewable energies.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.3. Aluminium Smelter Cost Structure As highlighted above the aluminium industry is highly dependent on the electricity supply and demand. This along with the alumina and carbon anodes represent the three major costs of the smelter, As seen in Figure 4, alumina and energy prices consist of 38% of the cost structure each. Whereas raw materials consist of 11% of the costs, with cost of replacing carbon anodes taking up a significant portion of this figure. As the anodes consist of petroleum coke and a coal tar pitch binder, and requires a significant quantity of heat to bake the anodes at 1120OC. These two factors make the anode manufactures reliant on the global oil and coal prices. Alumina and carbon anode costs remain fairly consistent for most smelters, whereas electricity and labour costs vary depending on the area the smelter operates in.

Figure 4: Aluminium Smelter cost break down, (Phillips, Bristo, Tian, 2015)

2.4. Current Power Modulation Research Recently, research into power modulation of aluminium smelters has predominately focused on how the amperage reduction affects the temperature of the cell, and the freezing of the bath. The research has primarily investigated how Shell Heat Exchangers (SHE) minimise the temperature drop and reduce bath freezing. Further detail about these topics are discussed below and are used later in this thesis to help develop the technical model of this thesis.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.4.1. Shell Heat Exchanger Currently compressed air jets are used to regulate the cell sidewall temperature. This system is difficult to regulate as it wastes potential heat recovery and increases the dust dispersion within the pot room. Given the limitations of this method The Light Metals Research Centre at the University of Auckland has developed the SHE, which has the ability to provide controlled cooling to the side wall of aluminium smelter cells. As seen in Figure 5, the SHE consists of 3 components; the exchanger body, the outlet, and the air knife assembly which increases the air flow into the exchanger. When tested on a 350kA potline the maximum sidewall temperature was reduced by 150oC (Namboothiri et al, 2009).

Figure 5: Shell Heat Exchanger (SHE) (Namboothiri et al, 2009)

However, with power modulation the heat balance needs to be altered so that either heat is generated or heat loss is reduced to prevent ledge freeze. A study by Lavoie et al (2011) investigated the use of SHE technology to increase the duration of the power modulation window. The heat exchanger was again tested on a 350kA potline resulting in the SHE acting as an insulator when operated without forced air extraction. Thereby causing an increase in the average cell temperature by approximately 75oC (Lavoie et al, 2011).

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

2.4.2. Impact of Amperage Reduction on the Cell A recent study conducted by Taylor and Chen (2015) constructed models to predict and control the response of a cell to reductions in amperage. The study was focused on two main scenarios; the reduction applied to a smelter using current technology and the reduction applied to a smelter using a SHE (Taylor & Chen, 2015). These models were developed using a simplified energy balance, and the model without the SHE has been recreated in Section 3.

It was found by Taylor and Chen (2015) that when the amperage reduction was applied to the model without the SHE technology, the energy imbalance immediately reduced the bath temperature to near liquidus temperature. Freezing of the bath then occurs at the rate of the energy imbalance across the bath and frozen bath interface. Bath temperature then decreases at approximately the same rate as the liquidus point reduction rate. Furthermore, the study found that the reduction could only be maintained for approximately half an hour before the bath’s liquid mass and temperature began to approach a critical point of 2500kg and 900oC respectfully (Taylor & Chen, 2015). However the cell can only be reduced to a temperature of 1218 -1223 K before the cell is at risk of significant damage, as sludge can form on surface of the cathode if the temperature falls below this limit for extended periods of time (Lavoie et al, 2011).

Figure 6 Conventional Cell Response to an 80 kA Amperage Reduction, (Taylor & Chen, 2015).

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

The second model estimated that the SHE would reduce the heat lost via sidewalls by approximately 60%, while the top heat lost would decrease by 30%. Under these assumptions it was observed that the initial reduction was similar to the no SHE case. Following this, the rate of temperature reduction is slower and begins to level out at 930oC after 3 hours with the bath mass levelling out at 2950kg. The temperature and the mass starts to recover after the 3 and a half hour mark. Although this is a large improvement compared to the no SHE case, this reduction could not be maintained due to risk of sludge forming in the cathode and low alumina

(Al2O3) dissolution rate. Nevertheless it is suitable for power modulation where the reduction could be applied for 1 to 2 hours and then returned to normal operating conditions (Taylor & Chen, 2015).

2.4.3. Energy stabilisation Aluminium smelters can stabilise the energy supplied into the power grid by reducing energy consumption during periods of high demand and price thus effectively returning capacity to the grid. Preventing the need for equivalent capacity produced by conventional means that would otherwise be added to the grid (Depree, Düssel, Patel, Reek, 2016). A study by N. Depree et al (2016) investigated this by using SHE to control the temperature reduction caused by downward power modulation. Tests were conducted on a group of 12 pots, with the ability to increase current by up to 25kA above the rest of the line (Depree et al, 2016). Short term energy modulation were conducted to simulate the rapid fluctuation of the energy market, as seen in Figure 7. The results suggested that the frequency of these modulations were too high, as the pots were not reaching steady state before the following reduction was implemented (Depree et al, 2016). A possible scenario for further investigation would be to stagger the downward power modulation of the smelter potline. Thus giving each potline sufficient time to recover to steady state operation before the next power reduction is implemented. Additionally it should be investigated if these reductions would have a significant economic impact on the smelter, and if these reductions ‘return’ enough capacity back to the grid to be deemed as an effective energy stabilisation method.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Figure 7: Aluminium Smelter response with Shell Heat Exchanger (Depree, 2016)

3. Methodology

3.1. Technical Model Development The technical model is derived from the work done by Taylor and Chen (2015), and outputs a temperature time graph that shows the rate of heat loss once the amperage reduction is applied. The model reproduces this behaviour by using the variables outlined in Table 2 and the following assumptions.

 At time 0 the system operates under steady state conditions  The concentration of calcium fluoride and alumina remain constant throughout the amperage reduction  60% of the heat loss via the top of the cell is due to heat loss through the anodes  The energy demand spike has a duration of 3 hours  The amperage reduction does not cause a decrease in production’s efficiency

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Variable Value Source

Initial Bath Temperature (푻풃풂풕풉) 1233 K (Taylor & Chen, 2015)

Initial Liquidus Temperature (푻푳) 1223 k (Taylor & Chen, 2015)

Decomposition Voltage (푽풅풆풄풐풎풑풐풔풊풕풊풐풏) 1.2 V (Taylor & Chen, 2015) Over-potential (anode + cathode) 0.5 V (Taylor & Chen, 2015)

(푽풐풗풆풓풑풐풕풆풏풕풊풂풍) Ohmic resistance (R) 8.75e-6 Ω (Taylor & Chen, 2015)

Potential due to reaction (푽풓풆풂풄풕풊풐풏) 2.05 V (Taylor & Chen, 2015)

Initial amperage (푰풄풆풍풍) 240 kA (Taylor & Chen, 2015)

Enthalpy change due to state change (∆푯풇풖풔풊풐풏) 520 kJ/kg (Taylor & Chen, 2015) Bath Mass (푴) 4500 kg (Taylor & Chen, 2015)

Specific heat for constant pressure (푪풑) 1.7 kJ/kgK (Taylor & Chen, 2015) Heat transfer coefficient from bath/metal to 0.8 kW/m2 (Taylor & Chen, 2015)

ledge (풉풔) K

Heat transfer coefficient from bath to anode (풉푨) 1.13 (Severo & Gusberti, kW/m2K 2009)

Initial excess aluminium fluoride (푨풍푭ퟑ풆풙) 10% (Taylor & Chen, 2015) 2 Cell wall contact surface area (푨풔) 13 m (Taylor & Chen, 2015) 2 Anode contact surface area (푨푨) 21.6 m (R&D Carbon, 2018) Time step (풅풕) 72 s Table 2: Technical Model Inputs

The main equations used to construct this model are displayed below, with more detailed calculations shown in Appendix A. One of the key assumptions for this model was that when operating at normal amperage, the system is at steady state. Thus the steady state heat loss of the cell is;

(7) 푄표푟𝑖푔𝑖푛푎푙(푘푊) = (푉푑푒푐표푚푝표푠𝑖푡𝑖표푛(푉) + 푉표푣푒푟푝표푡푒푛푡𝑖푎푙(푉) + 푅(휇훺)

∗ 퐼푐푒푙푙(푘퐴) − 푉푟푒푎푐푡𝑖표푛(푉)) ∗ 퐼푐푒푙푙(푘퐴)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Equation 7 how conductive heat lost via the cell side walls and anode can be calculated by utilising the percentages of heat lost from previous thermodynamic cell analysis. Taylor et al

(2014) found that 38% of 푄표푟𝑖푔𝑖푛푎푙 exited the cell via the cell walls, and 48% via the top of the cell. An additional equation needs to be developed for the heat lost via the anode, which assumes that approximately 60% of heat is lost via the top of the cell. This value is based on previous thermodynamic simulations conducted by Taylor et al (2014). These findings indicated that the proportion of heat lost via the anode varied between 57.4 and 61.7, depending on the anode immersion depth and anode cover thickness (Taylor et al, 2014).

푄푠𝑖푑푒(푘푊) = 0.38 ∗ 푄표푟𝑖푔𝑖푛푎푙(푘푊) (8)

푄푡표푝(푘푊) = 0.48 ∗ 푄표푟𝑖푔𝑖푛푎푙(푘푊) (9)

푄푎푛표푑푒(푘푊) = 0.6 ∗ 푄푡표푝(푘푊) (10) By modifying Equation 7, the heat lost under reduced amperage conditions can be determined as shown below in Equation 11. Where ‘R’ is the ohmic resistance of the cell at normal operating conditions, and is calculated by using, 푅 = 푉표ℎ푚𝑖푐/퐼푐푒푙푙

푄푟푒푑푢푐푒푑(푘푊) = (푉푑푒푐표푚푝표푠𝑖푡𝑖표푛(푉) + 푉표푣푒푟푝표푡푒푛푡𝑖푎푙(푉) + 퐼푟푒푑푢푐푒푑 (푘푊) (11)

∗ 푅(휇훺) − 푉푟푒푎푐푡𝑖표푛(푉)) ∗ 퐼푟푒푑푢푐푒푑(푘퐴) From Equations 7 and 11 the global instantaneous energy imbalance when the amperage reduction is applied is;

푄∆ (푘푊) = 푄표푟𝑖푔𝑖푛푎푙(푘푊) − 푄푟푒푑푢푐푒푑(푘푊) (12) Once the global energy imbalance of the cell is determined the local energy imbalance that occurs at the liquid-solid interface needs to be calculated. This interface is where the freezing of the bath occurs on both the cell walls and anode, and is displayed below in Equation 12.

Where ∆퐻푓푢푠𝑖표푛 is the change in enthalpy of the electrolyte to convert its state from a liquid to 푑푀 a solid at constant pressure. While 푓푟푒푒푧푒 is the rate the bath freezes with respect to time. 푑푡

푄푠𝑖푑푒(푘푊) + 푄푎푛표푑푒(푘푊) (13) 푘푊 푘푊 = ℎ ( ) ∗ 퐴 (푚2) ∗ (푇 (퐾) − 푇 (퐾)) + ℎ ( ) 푠 푚2퐾 푠 푏푎푡ℎ 푙 퐴 푚2퐾 푘퐽 ∗ 퐴 (푚2) ∗ (푇 (퐾) − 푇 (퐾)) + ∆퐻 ( ) 퐴 푏푎푡ℎ 푙 푓푢푠𝑖표푛 푘푔 푑푀 (푘푔) ∗ 푓푟푒푒푧푒 푑푡 (푠)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

As stated previously, at time zero it is assumed that the system operates at steady state. By applying this assumption to Equation 12 it can be assumed that no freezing occurs (∆퐻푓푢푠𝑖표푛 ∗ 푑푀 푓푟푒푒푧푒 = 0) and thus the conductive heat losses (left side of equation) equal the convection 푑푡 heat losses (right side of equation).

푑푀 By rearranging Equation 13 for ∆퐻 ∗ 푓푟푒푒푧푒, the energy lost due to the electrolyte 푓푢푠𝑖표푛 푑푡 freezing can be derived. Where the rate of freezing depends on the side wall and anode surface area available for the electrolyte to freeze onto. In this case it is the full liquid height, including the aluminium metal height, and the surface area of the anode that has been immersed in the electrolyte. By combining this with Equation 12, the instantaneous energy imbalance of the cell can be solved using Equation 14.

푘푊 (14) ∆퐸 (푘푊) = 푄∆(푘푊) − [푄 (푘푊) + 푄 (푘푊) − ℎ ( ) ∗ 퐴 (푚2) 푠𝑖푑푒 푎푛표푑푒 퐴 푚2퐾 퐴 푘푊 ∗ (푇 (퐾) − 푇 (퐾)) − ℎ ( ) ∗ 퐴 (푚2) ∗ (푇 (퐾) 푏푎푡ℎ 퐿 푠 푚2퐾 푠 푏푎푡ℎ

− 푇퐿(퐾))] The change in bath temperature over a time step 푑푡 can be determined by rearranging Equation

15 for 푇푏푎푡ℎ, and substituting the result from Equation 14 into 15.

푘퐽 (15) 푑 (푀(푘푔) ∗ 퐶 ( ) ∗ ∆푇 (퐾)) 푝 푘푔퐾 푏푎푡ℎ ∆퐸 (푘푊) = 푑푡 (푠) However as the temperature of the bath drops the electrolyte begins to freeze, as a result the mass of the bath needs to be recalculated after each time step. This is done by rearranging

Equation 13 for 푑푀푓푟푒푒푧푒 and integrating for time step, 푑푡 the resulting mass after the time step is displayed in Equation 16. Where 푀𝑖 is the electrolyte bath mass from the current iteration, and 푀𝑖+1 is the new electrolyte bath mass after time step, 푑푡.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

푀𝑖+1(푘푔) (16)

= 푀𝑖(푘푔)

푄 (푘푊) + 푄 (푘푤) − ∫ [ 푠𝑖푑푒 푎푛표푑푒 푘퐽 ∆퐻 ( ) 푓푢푠𝑖표푛 푘푔

푘푊 2 푘푊 2 [ℎ푠 ( 2 ) ∗ 퐴푠(푚 ) ∗ (푇푏푎푡ℎ − 푇푙) + ℎ퐴 ( 2 ) ∗ 퐴퐴(푚 ) ∗ (푇푏푎푡ℎ − 푇푙) − 푚 퐾 푚 퐾 ] 푘퐽 ∆퐻 ( ) 푓푢푠𝑖표푛 푘푔 ∗ 푑푡

The result of the electrolyte freezing is that the mass of excess aluminium fluoride is concentrated within a smaller quantity of liquid. This increase corresponds to a change in the bath liquidus temperature, according to Equation 17. Here it has been assumed alumina feed strategy varies so that the levels of calcium fluoride and alumina remain at constant concentrations of 4.5% and 2% respectfully.

푑푇 (퐾) (17) 푇 (퐾) = 푇 (퐾) + 푙 ∗ ∆퐴푙퐹 (%퐴푙퐹 ) 퐿,𝑖+1 퐿,𝑖 ( ) 3 3 푑퐴푙퐹3 %퐴푙퐹3 Where; 푑푇푙 is the rate of change of liquidus temperature with excess aluminium fluoride 푑퐴푙퐹3

푑푇푙 concentration, and ∆퐴푙퐹3, is the increase in excess aluminium fluoride as a percentage. is 푑퐴푙퐹3 computed by using Table 3 shown below, which is the rate of change of liquidus point between excess fluoride concentrations of 8% and 20%.

Excess 푨풍푭 풅푻 ퟑ 풍 풅푨풍푭ퟑ 8 -3.24 10 -4.33 12 -5.47 14 -6.64 16 -7.84 18 -9.07 20 -10.32 Table 3: Rate of Change of Liquidus point with Excess Fluoride Concentrations, (Taylor & Chen, 2015) 24

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

3.2. Economic Model Development The economic model outputs the individual cells profit, variable cost, revenue and fixed cost per hour for both the normal operating conditions and reduced amperage. The model simulates this by utilising the variables outlined in Table 4. Where the energy price is the average peak price over an 8 month period. Similarly the aluminium price is averaged over an 8 month period, excluding anomalies. It should be noted that this model assumes that the production efficiency (current efficiency (CE)) remains constant throughout the reduction.

Variable Value Source Number of cells 840 (Taylor & Chen, 2015) Current Efficiency (CE) 0.94 (Taylor & Chen, 2015) Faraday’s Constant 96500 coulomb/equivalent (Grjotheim et al, 1993) Number of electrons 3 (Grjotheim et al, 1993) Molecular mass 26.98 g/mol (Grjotheim et al, 1993) Price of Alumina 0.48 AUD/kg (Australian Goverment - Department of Industry, Innovation and Science , 2018) Price of Aluminium 2.92 AUD/kg (Investing.com, 2018) Price of Anode 0.55 AUD/kg (Made-in-China.com, 2018) Price of Energy during 90.17 AUD/MWhr (Australian Energy Market peak demand Operator, 2018) Rate of anode consumed 0.4 kg/kg of Al (R&D Carbon, 2018) per kg of aluminium produced Rate of alumina consumed 1.91 kg/kg of Al (Kvande & Drabløs, 2014) per kg of aluminium produced Table 4 Economic Model Inputs

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

The main equations used to construct this model are displayed below. By combining two of Faraday’s laws, an equation for the amount of aluminium produced over time ‘t’ can be developed. Multiplying the result from this equation with the current efficiency of the cell can produce the actual production rate for the cell, as shown in Equation 18. Current efficiency is the ratio of measured production rate and theoretical production.

푔 (18) 푘푔 푀 ( ) ∗ 퐼(푘퐴) ∗ 푡(푠) 푟 ( ) = 퐶퐸(%) ∗ 푚표푙 퐴푙푢푚𝑖푛𝑖푢푚 퐶 푠 푧퐹 ( ) 푚표푙 The variable cost for the aluminium smelter can be broken down into three main pricing components these being; alumina, anode and electricity costs. The electrolyte is not considered in the variable cost as it is not consumed. Furthermore, aluminium fluoride is also not considered as its consumption during this electrolysis process is negligible.

The power consumed by the cell can be derived from the simple Power Law shown below in Equation 19. By multiplying this value by the energy rate, the energy component of the variable cost can be calculated. Whereas, the alumina and anode components are determined by multiplying the rate of component consumed (kg/kg of aluminium produced) by the rate of aluminium production (kg/hr). The aluminium production rate is determined by Equation 18 and the price per kilogram of component. By combining these three components, an estimate for the variable cost of one aluminium smelter cell can be determined as seen in Equation 20.

푃푒푛푒푟푔푦(푘푊) = 퐼 (푘퐴) ∗ 푉(푉) (19)

$ $ 푘푔푎푙푢푚𝑖푛푎 (20) 퐶푣푎푟𝑖푎푏푙푒 ( ) = [푝푎푙푢푚𝑖푛푎 ( ) ∗ 푟푎푙푢푚𝑖푛푎 ( ) ℎ푟 푘푔푎푙푢푚𝑖푛푎 푘푔푎푙푢푚𝑖푛𝑖푢푚

$ 푘푔푎푛표푑푒 + 푝푎푛표푑푒 ( ) ∗ 푟푎푛표푑푒 ( )] 푘푔푎푛표푑푒 푘푔푎푙푢푚𝑖푛𝑖푢푚 푘푔 $ ∗ 푟 ( 푎푙푢푚𝑖푛𝑖푢푚) + 푝 ( ) 푎푙푢푚𝑖푛𝑖푢푚 ℎ푟 푒푛푒푟푔푦 푀푊ℎ푟

∗ 푃푒푛푒푟푔푦(푀푊ℎ푟)

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Due to the lack of smelter financial data available to the public, an estimate of the fixed cost was used instead. The estimation was evaluated by first calculating the total cost of the cell at 240kA. This was done by using the cost structure seen in Figure 4 and the variable cost calculated in Equation 20. The fixed cost was then derived as a proportion of total costs at full capacity, as seen in Equations 21 and 22.

$ $ (21) 퐶 ( ) = 0.87 ∗ 퐶 ( ) 푣푎푟𝑖푎푏푙푒 ℎ푟 푡표푡푎푙@240푘퐴 ℎ푟 $ $ (22) 퐶 ( ) = 0.13 ∗ 퐶 ( ) 푓𝑖푥푒푑 ℎ푟 푡표푡푎푙@240푘퐴 ℎ푟 Finally the profit of one aluminium smelter is determined by first calculating the revenue of the smelter, which is the price of aluminium per kilo multiplied by the rate of production, shown in Equation 23. The profit of the cell is then deduced by subtracting the variable and fixed cost calculated in Equation 20 and 22 from revenue, as seen in Equation 24 below.

푅푒푣푒푛푢푒 = 푝퐴푙푢푚𝑖푛𝑖푢푚 ∗ 푟푎푙푢푚𝑖푛𝑖푢푚 (23)

푃푟표푓𝑖푡 = 푅푒푣푒푛푢푒 − 퐶푣푎푟𝑖푎푏푙푒 − 퐶푓𝑖푥푒푑 (24)

3.3. Method to Analysis the Results To determine if amperage modulation is an effective energy stabilisation strategy, three scenarios were tested. The first scenario being an amperage step function, where the amperage is stepped down to the desired level followed by a step up once the limit has been reached. Secondly a staggered step, which consists of two step downs, with the 1st lasting for 1hr, followed by a 2nd which last until the limit has been reached. Finally the ramp scenario, which utilises a sine function to gradually decrease the amperage to the desired level, where it is held until the limit has been reached. These 3 scenarios were tested for an amperage reduction of 10, 15 and 20 kA.

These scenarios were then compared by plotting the variables of interest, and selecting the most viable solution with the use of the weighted decision matrix shown in Appendix B. The variable of interest are; duration of the reduction, opportunity cost of performing the reduction, and finally the power saved.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4. Results

4.1. Technical Model Figure 8 displays how the cell’s temperature reacts to an 80 kA reduction in amperage. After the reduction is applied, the bath temperature experiences a rapid decrease until it nearly reaches the liquidus temperature. At this point the bath temperature decreases approximately at the liquidus temperature rate. This behaviour aligns with the expected temperature drop that is derived from heat transfer and mass balance principles. As the heat source is reduced due to the amperage reduction the bath temperature approaches the temperature of its surroundings. In this case that is the temperature of frozen electrolyte, which is equivalent to the liquidus temperature. Thus the rapid drop in bath temperature to near liquidus temperature. From this point the electrolyte begins to freeze which dictates the rate of liquidus temperature reduction and thus the bath temperature reduction.

This model is supported by the results of previous studies conducted by Mark Taylor and John Chen (2015) which used a model with similar conditions. Where the time to decrease the bath temperature to 1173K is 1.5 hours at which time the bath mass is approximately 2500 kg. Additionally, at a time of 0.5 hours the bath temperature is 1213K. To validate this model these values were compared to results derived from the model under similar conditions. From Figure 8, it can be seen that at the times 0.5 and 1.5 the model produces the 1217.57 K and 1173.51K respectfully, thus validating the model.

Figure 8: Simulated Cell Response to 80 kA Reduction

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.1.1. Results: Step As discussed previously any power modulation strategy is only effective if the cathode surface temperature remains above 1218 -1223 K. As sludge can form on surface of the cathode if the temperature falls below this limit for extended periods of time, causing severe damage to the cathodes (Lavoie et al, 2011). Therefore to minimise the damage done to the cathode, the bath temperature of the model is limited to 1223K before initial operating conditions are applied.

Figure 9 displays the results from the technical model with an amperage step reduction of 10 kA (left), 15 kA (right) and 20 kA (bottom). By inspection of the graphs it can be seen that the smaller amperage reduction causes less severe initial shock to the system. As the bath temperature decreases from 1233 K to approximately 1230 K, compared to the 10 K drop experienced by the 80 kA reduction. The bath then begins to freeze at the rate derived from Equation 13, with the bath temperature decreasing according to the change in liquidus temperature, which is also related to the bath mass. Once the bath temperature reaches approximately 1223 K the initial operating conditions are once again applied. This occurs at different points within the 3 hour period depending on the magnitude of the reduction. As expected the 10 kA reduction takes the longest, with the limit being reached at approximately 2.96 hours, followed by the 15 kA reduction with 1.94 hours, and lastly the 20 kA reduction with 1.41 hours. After the limit has been reached and initial operating conditions are applied, another shock to the system occurs. Where the bath temperature sharply increases, followed by a gradual increase according to the change in liquidus temperature.

Figure 9: Technical Model with a Step Reduction for 10, 15 and 20 kA

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.1.2. Results: Staggered Step Figure 10 displays the results of the staggered step amperage reduction for the 10 kA (left), 15 kA (right), and 20 kA (bottom) scenarios. This strategy consists of 2 stages with the first being an intermediary reduction that is initiated at time 0 and has a duration of 1 hour. During the second stage the amperage is reduced to the desired level and is maintained at this level until the limit of 1223K is reached. These scenarios follow a similar trend to the step reduction described previously. However when the second reduction occurs, the cell experiences another shock to the system where the bath temperature decreases depending on the magnitude of the reduction. With the 10 kA reduction experiencing slight drop in bath temperature followed by increased temperature rate of change. A similar behaviour occurs in the 15 kA but the 20 kA reduction has a more noticeable temperature drop and change in the rate of reduction. As seen below the result of this staggered step is that the cell can operate at the reduced amperage value for longer periods of time. With the 10 kA reduction, the bath temperature limit of 1223 K is not reached until 3 hours after the initial reduction while the 15 kA takes 2.92 hours and the 20 kA takes 2.24 hours. At which point the initial operating conditions are applied, and the cell follows the same pattern as discussed in the step reductions.

Figure 10: Technical Model with a Staggered Step Reduction for 10, 15 and 20 kA

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.1.3. Results Ramp Figure 11, displays the results of the ramp amperage reduction for the 10 kA (left), 15 kA (right), and 20 kA (bottom) scenarios. Here, a sine function is used to gradually decrease the amperage to the desired level where it is held until the limit has been reached. Again, the reductions follow a similar patter as the previous scenarios where the bath temperature rapidly decreases followed by a gradual decrease. While the previous scenarios experienced a sudden shock to the system when stepped down, the ramp reduction has a smoother transition. As seen in Figure 11 below, the ramp reduction has no sudden decrease in bath temperature when transitioning from the ramped amperage to a constant amperage. This amperage is held constant until the bath temperature limit of 1223 K is reached, where the original operational conditions are applied. This point occurs at different positions, for the 10 kA the limit isn’t reach until 3 hours after the initial reduction, the 15kA’s limit occurs at 2.58 hours and the 20 kA’s limit occurring at 2.06 hours. Once the initial operating conditions are applied the cell follows the same pattern as discussed in the previous reductions

Figure 11: Technical model with a ramp reduction for 10, 15 and 20 kA

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.2. Economic Model The economic model uses the basic profit equation outlined in Equation 24, as such there are multiple variables that influence the profitability of the cell. The first and most important variable is the revenue of the cell, this is proportional to the price of aluminium and the production rate. Secondly, the variable cost of the cell is proportional to the alumina, anode, and energy prices. Additionally, the power consumed by the cell and the aluminium rate have a direct impact on the variable costs. Finally, the Fixed Cost is proportional to labour cost and other expenses such as overheads and maintenance.

4.2.1. Results: Step According to Faraday’s laws, the rate of aluminum production is directly related to the amperage of the cell. Consequently, the revenue, variable cost and profit are all dependant on the amperage of the cell. This patteren is present in Figure 12, where the revenue, variable cost and profit all vary with respect to the change in cell amperage. As seen below the revenue of the cell and the variable cost decrease for when each respective amperage reduction is applied. Resulting in a profit decrease of the difference between the revenue and variable costs as fixed cost remian constant.

10 kA Step Reduction 15 kA Step Reduction 1400.00 1500.00 1200.00 1000.00 1000.00

800.00 $/hr 600.00 $/hr 400.00 500.00 200.00 0.00 0.00 With Reduction Without With Reduction Without Reduction Reduction

Revenue Variable Cost Fixed Cost Profit Revenue Variable Cost Fixed Cost Profit

20 kA Step Reduction 1500.00

1000.00

$/hr 500.00

0.00 With Reduction Without Reduction

Revenue Variable Cost Fixed Cost Profit Figure 12: Economic model with step reduction of 10 kA, 15 kA, and 20 kA 32

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.2.2. Results Staggered Step As discussed previously the variable cost and revenue are dependent on the amperage of the cell, as each variable mimics the amperage reductions. As discussed previously it takes approximately 3, 2.92 and 2.24 hours before the limit is reached for each respective amperage reduction. As seen below the revenue of the cell and the variable cost decrease for when each respective amperage reduction is applied. Resulting in a profit decrease of the difference between the revenue and variable costs as fixed cost remian constant.

10 kA Staggered Step Reduction 15 kA Staggered Step Reduction 1400.00 1400.00 1200.00 1200.00 1000.00 1000.00

800.00 800.00 $/hr $/hr 600.00 600.00 400.00 400.00 200.00 200.00 0.00 0.00 With Reduction Without With Reduction Without Reduction Reduction Revenue Variable Cost Fixed Cost Profit Revenue Variable Cost Fixed Cost Profit

20 kA Staggered Step Reduction 1400.00 1200.00 1000.00 800.00

$/hr 600.00 400.00 200.00 0.00 With Reduction Without Reduction

Revenue Variable Cost Fixed Cost Profit

Figure 13: Economic model with staggered step reduction of 10 kA, 15 kA, and 20 kA

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

4.2.3. Results: Ramp As seen below the revenue of the cell and the variable cost decrease for when each respective amperage reduction is applied. Resulting in a profit decrease of the difference between the revenue and variable costs as fixed cost remian constant. Additionally, the period of the reduction approximately equal to 3, 2.58, and 2.06 hours for each respective amperage reduction. However, unlike the staggered step and step reductions the ramps’ amperage, and thus revenue, variable cost and profit reduces gradually over time. To compare this to the previous two scenarios the duration of reduction, opportunity cost and energy saved were compared for 10kA, 15kA and 20kA to determine which strategy and amperage reduction is the most effective.

10 kA Ramp Reduction 15 kA Ramp Reduction 1400.00 1400.00

1200.00 1200.00

1000.00 1000.00

800.00 800.00 $/hr 600.00 $/hr 600.00

400.00 400.00

200.00 200.00

0.00 0.00 With Reduction Without Reduction With Reduction Without Reduction

Revenue Variable Cost Fixed Cost Profit Revenue Variable Cost Fixed Cost Profit

20 kA Ramp Reduction 1400.00

1200.00

1000.00

800.00

$/hr 600.00

400.00

200.00

0.00 With Reduction Without Reduction

Revenue Variable Cost Fixed Cost Profit

Figure 14: Economic model with ramp reduction of 10 kA, 15 kA, and 20 kA

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

5. Analysis To determine which reduction strategies are a viable solution, the following variables were plotted effectively to compare and contrast the results. The most viable solution was then selected using the decision matrix shown in Appendix C. The variables of interest are the following:

 The duration of the amperage reduction  The opportunity cost of performing the reduction (profit difference between cell with the reduction and cell without the reduction)  The power saved

These values were determined by integrating the corresponding function with respect to the duration of the amperage reduction, and can be seen below in Table 5

Ramp Step SS

10 15 20 10 15 20 10 15 20 Duration 3 2.58 2.06 2.96 1.94 1.41 3 2.92 2.24 Opportunity -0.04 -0.15 -0.23 -0.09 -0.19 -0.31 -0.05 -0.19 -0.3 cost Power Saved 0.138 0.169 0.164 0.172 0.168 0.161 0.146 0.21 0.2 Table 5: Ramp, Step and Staggered Step Model outputs for 10, 15 and 20 kA reduction

Inspection of Table 5 and Figure 15 below, showcases that only the 10kA reduction for the ramp and staggered step scenarios achieve the desired 3 hour duration of reduced amperage. However the other scenarios are still considered as although the energy market does experience increased energy prices for approximately 3 hours, it also experiences smaller peaks for a shorter period of time. It is these smaller periods of increased energy prices where the other scenarios and reduction would be useful to implement, such as the 10 kA step and 15 kA staggered step.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Duration of Reduction 3.5

2.5

2 Ramp 1.5

Staggered Step Duration (hrs) Duration 1 Step

0.5

0 10 15 20 Amperage Reduction (kA)

Figure 15: Duration of reduction for the three strategies at 10, 15 and 20 kA reduction

To determine if these reductions are a suitable stabilisation strategy, the amount of power saved needs to be analysed. As seen below in Figure 16, the staggered step with a 15 kA reduction reduces the cell power consumption by approximately 0.21 MWhr over a 3 hour period. Although relatively small when compared to operational maximum demanded by the grid over the same 3 hour period of 42333 MWhr (Australian Energy Market Operator, 2017). When the reduction is applied to all of the cells of a generic smelter the amount of power saved is increased to 176.4 MWhr which is 0.42% of the operational maximum demanded by the grid.

Power Saved Over 3 Hour Period 0.25

0.2

0.15 Ramp 0.1 Staggered Step Step Power Saved (MWhr) Saved Power 0.05

0 10 15 20 Amperage Reduction (kA)

Figure 16: Power Saved over 3 hour period for 1 cell

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

While the power saved due to the reduction is important, it is redundant unless the smelter has an incentive to reduce the amperage during periods of peak energy prices. Thus the opportunity cost of the reductions are compared to determine which is most profitable for the smelter. As seen in Table 5 and Figure 17, all of the strategies reduce the profitability of the cell. Where the 10kA ramp reduction has the least impact on the cell’s profitability, with a value of -$0.04. When this reduction is applied to all of the smelter cells an opportunity cost of -$33.6 for the 3 hours of reduction is produced.

Opportunity Cost Incurred Over 3 Hour Period 0 10 15 20 -0.05

-0.1

-0.15 Ramp

-0.2 Staggered Step Step

-0.25 Opportunity cost ($) cost Opportunity

-0.3

-0.35 Amperage Reduction (kA)

Figure 17: Opportunity Cost of Cell Over 3 Hour Period for 1 cell

As seen in Table 6, the duration was allocated a weighting of 0.3, as the reduction is an ineffective energy stabilisation strategy unless it can last for the full duration of the peak energy demand. The same reasoning is used for the construction of the duration scale where the highest scale is allocated to the reductions which have a duration of 3 hours. While the others are spaced 0.5 hours apart. Additionally as stated previously, smelters will not implement the reduction unless an incentive is provided. Thus the opportunity cost is allocated a weighting of 0.45 to reflect this. Finally, power saved is allocated a 0.25 weighting to reflect the power “effectively” supplied back into the grid. Based on these weightings the 10 kA ramp and staggered step reductions tied for first rank, followed by the 15 kA staggered step for 3rd rank.

As opportunity cost has the highest weighting it is used to differentiate between the scenarios tied for 1st place. Therefore the 10 kA ramp reduction is select as the most viable solution as it has an opportunity cost of -$0.04 for one cell over a 3 hour period.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

Scale 1 2 3 4 5 Duration 1-1.5 1.51-2 2.01-2.5 2.51-2.99 3 Power Saved 0.12-0.14 0.141-0.16 0.161-0.18 0.181-0.20 0.201-0.22 Opportunity Cost 0.35-28 0.279-0.21 0.209-0.14 0.139-0.07 0.06-0 Ramp Step Staggered Step Criteria Weighting 10 15 20 10 15 20 10 15 20 Duration 0.3 5 4 3 4 2 1 5 4 3 Power Saved 0.25 2 3 3 3 3 3 2 5 5 Opportunity Cost 0.45 5 3 2 4 3 1 5 3 1 2.5 Total 4.25 3.3 3.75 2.7 1.5 4.25 3.8 2.6 5 Ranking 1 5 8 4 6 9 1 3 7 Table 6: Weighted Decision Matrix

From the previous analysis none of the scenarios provide an economic incentive to the smelter to reduce the amperage of the cell during periods of peak energy prices. However this analysis is calculated using an average peak energy price of 90.17$/MWhr where in fact the energy price regularly spikes over this value. Therefore the price when the reduction becomes profitable needs to be determined. This was done by increasing the price of the 10kA ramp reduction and plotting the opportunity cost. The price where the reduction becomes profitable can be deduced by taking the point where the curve intersects with the x-axis. As this is the point where the difference between the cell’s profit with the reduction, and cell profit without the reduction is 0. It can be seen in Figure 18, that this point occurs at an energy price of approximately 90.47 $/MWhr.

Opportunity Cost vs Energy Price 0.02

0.015

0.01

0.005

0 90.5

-0.005 90.4

90.55 90.41 90.42 90.43 90.44 90.45 90.46 90.47 90.48 90.49 90.51 90.52 90.53 90.54 90.56 90.57 90.58 90.59 -0.01

-0.015

Energy Price ($/MWhr) Opporunity Cost over 3hr period ($) period 3hr over Cost Opporunity Figure 18: Analysis of when Reduction becomes Profitable for Smelter 38

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

This analysis demonstrates that an amperage reduction during periods of peak energy demand is a sufficient energy stabilisation strategy. Specifically the 10 kA ramp reduction when applied to all 840 cells can be sustained over a 3 hour period, whilst still providing 115.92 MWhr or 0.27% of the total energy demanded during this period. However from the analysis it can be seen that this reduction can only be applied when the energy price is above 90.47 $/MWhr.

6. Sensitivity Analysis A sensitivity analysis was conducted to highlight how volatile the technical and economic models are to uncertainties and errors present in the input variables. This was done by varying the input variables of the 10kA ramp reduction by ±10% and recording the change in the model output for one cell. The variables were then ranked from most volatile to least to determine which variables are the most sensitive, as seen in Figure 19 and 20.

Economic Sensitivity Analysis

Price of Aluminium

Price of Energy

Price of Alumina

Fixed Cost

Price of Anode

-80 -60 -40 -20 0 20 40 60 80 Change in Profit due to ±10% Change in Variable ($)

10% + 10% -

Figure 19: Economic Model Sensitivity Analysis

Based off Figure 19 the uncertainties that would have the most impact on the model would be the ones that influence the price of aluminium, price of energy and price of alumina. These variables produced a total change in the simulation out of $128.26 for price of aluminium, $49.29 for price of energy, and $40.27 for price of alumina. From this analysis the following factors were identified that could influence the uncertainty of the input variables;

 It is difficult to vary raw material consumption over the short periods of modulation that is being investigated. This could increase the costs incurred by the cell over the time period of interest.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

 Due to the volatility of the input prices to outside forces, averages were used in the economic model. As such there may be some uncertainty present within these variables.  This model simulates the ideal case where production efficiency does not decrease when the reduction is applied. However for an actual cell there would be a change in efficiency when the reduction is applied. Thus introducing uncertainties into the revenue and variable cost of the model.

Technical Sensitivity Analysis

Bath Mass

Excess AlF

Anode Area

Anode Heat Loss

Wall area

-6 -4 -2 0 2 4 6 8 Change in Final Bath Temperature due to ±10% Change in Variable

10% + 10% - Figure 20: Technical Model Sensitivity Analysis

From Figure 20, the uncertainties that have the most influence on the technical model are the ones that affect the mass of the electrolyte and the percentage of excess AlF3. These two variables produced a temperature change of 10.18K for the mass of the electrolyte and 5.24K for the percentage of excess AlF3. This analysis identified the following factors that could influence the uncertainty of the input variables;

 The technical model does not account for the temperature fluctuations that occur at multiple positions throughout the cell. These fluctuations could push the bath temperature to below 1223K and thus reducing the duration of the reduction.

 The technical model assumed that only percentage of excess AlF3 changed with the reduction and this was done to simplify the model. However as the temperature drops the bath composition will change thus influencing the liquidus and bath temperatures.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

7. Conclusion As stated in the analysis the 10kA ramp reduction is the most suitable strategy to stabilise the local energy market. It provides 115.92 MWhr back into the grid over 3 hours, whilst incurring an opportunity cost of -$33.6 over the 3 hour period. By using this scenario it was establish that the smelter will gain a financial incentive when the energy price increases above 90.47 $/MWhr.

However due to the assumptions that were required to perform this analysis, uncertainties have been built into the models. These uncertainties where identified from the sensitivity analysis, and are displayed below;

For the economic model;

 Fixed cost could be pushed up to 20-25% of total cost due to difficulty of varying raw material consumption.  Averages where used for the input prices, due to the volatility of outside forces.  The model simulates the ideal scenario where production efficiency remains constant when reduction is applied. Where as in an actual cell there would be some variance.

For the technical model;

 The technical model simulates the average bath temperature and does not account for temperature fluctuations that occur at multiple positions throughout the cell. These fluctuations could push the bath temperature below 1223K thus reducing the duration of the reduction.

 This model assumed that only percentage of excess AlF3 changed with the reduction, this was done to simplify the model. However as the temperature drops the bath composition will change hence influencing the liquidus and bath temperatures.

From this analysis the following were identified as areas of future study;

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

8. Reference List Australian Competition & Consumer Commision . (2017). Retail Electricity Pricing Inquiry, Preliminary Report . Australian Competition & Consumer Commision . Australian Energy Market Operator. (2017). AEMO. Retrieved from Maximum and Minimum Demand : https://www.aemo.com.au/Electricity/National-Electricity-Market-NEM/Planning- and-forecasting/Electricity-Forecasting-Insights/2017-Electricity-Forecasting-Insights Australian Energy Market Operator. (2018, May 10). Data Dashboard. Retrieved from AEMO: https://www.aemo.com.au/Electricity/National-Electricity-Market-NEM/Data-dashboard Australian Goverment - Department of Industry, Innovation and Science . (2018, July 1). Office of Chief Economist . Retrieved from Resources and Energy Quarterly June 2018: https://archive.industry.gov.au/Office-of-the-Chief- Economist/Publications/ResourcesandEnergyQuarterlyMarch2018/documents/Resources-and- Energy-Quarterly-March-2018-Aluminium-alumina-and-bauxite.pdf Bhattacharyya, B. (2015). Electrochemical Micromachining for Nanofabrication, MEMS and Nanotechnology. Kolkata: William Andrew Applied Science Publishers . Depree, N. Dussel, R. Patel, P. Reek, T. (2016). The 'Virtual Battery' - Operating an Aluminum Smelter with Flexiable Energy Input . Light Metals, 571-576. Encyclopaedia Britannica. (2014, March 5). Encyclopaedia Britannica. Retrieved from Electrolysis: https://www.britannica.com/science/electrolysis Encyclopaedia Britannica. (2017, June 07). Encyclopaedia Britannica. Retrieved from Electrolyte: https://www.britannica.com/science/electrolyte Encyclopaedia Britannica. (2018, March 28). Encyclopaedia Britannica. Retrieved from Coulomb: https://www.britannica.com/science/coulomb Gariépy, R. Couturier, A. Martin, O. Allano, B. Machado, A. Charmier F. (2014). Preparation and start-up of Arvida Smelter, AP60 Technological Center. Light Metals, 797-801. Grjotheim, K. Kvande, H. Foosnaes, T. Huglen, R. Lillebuen, B. Mellerud, T. Naterstad, T. (1993). Introduction to Aluminium Electrolysis. Dusseldorf: Aluminium-Verlag. Iggulden, O. (2017, October 16). ABC News. Retrieved from Power prices up 63pc since 2007, ACCC report shows, as watchdog casts doubt on Clean Energy Target: http://www.abc.net.au/news/2017-10-16/clean-energy-target-not-certain-to-lower-power- prices-accc/9052094 Investing.com. (2018, September 16). Investing.com. Retrieved from Aluminium Futures : https://au.investing.com/commodities/aluminum-historical-data Investopedia. (2018, October 20). Investopedia. Retrieved from Opportunity Cost : https://www.investopedia.com/terms/o/opportunitycost.asp Kvande, H. & Drabløs, A. (2014). The Aluminum Smelting Process and Innovative Alternative Technologies. Journal of Occupational and Environmental Medicine, 23-32. Lavoie, P. Namboothiri, S. Dorreen, M. Chen, J. Zeigler, D.P. Taylor, M.P. (2011). Increasing the Power Modulation Window of Aluminium Smelter Pots with Sheel Heat Exchanger Technology. Light Metals, 369-374.

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

LucasMihaupt Global Brazing Solutions . (2014, March 26). LucasMihaupt Global Brazing Solutions . Retrieved from Liquidus vs. Solidus: https://www.lucasmilhaupt.com/en- US/about/blog/2014/3/liquidus-vs-solidus Made-in-China.com. (2018, June 1). Made-in-China.com. Retrieved from Prebaked Carbon Anode T1-T3 with Best Quality: https://jj-coke.en.made-in- china.com/product/gvRQnDTAuVch/China-Prebaked-Carbon-Anode-T1-T3-with-Best- Quality.html Maroney, T. (2013). System Restart and Ancillary Services Draft Report . Baulkham Hills: Tomago Aluminium . Namboothiri, S. Lavoie, P. Cotton, D. Taylor. M. P. (2009). Controlled Cooling of Aluminium Smelting Cell Sidewalls using Heat Exchanger Supplied with Air. Light Metals , 317 - 322. National Center for Biotechnology Information. (2018, May 26). PubChem Compound Database. Retrieved from Cryolite: https://pubchem.ncbi.nlm.nih.gov/compound/cryolite#section=Top Perez, E. (2014). Australian Aluminium Council Submission to the Energy White Paper - Issues Paper. Canberra: AUSTRALIAN ALUMINIUM COUNCIL. Phillips, H. F. Bristo, S. Tian, W. (2015). Metal Prospects-Aluminium Market Outlook – Fourth Quarter 2015. RBC Capital Markets . Prasad, S. (2000). Studies on the Hall-Heroult Aluminum Electrowinning Process. Journal of the Brazilian Chemical Society, 245-251. R&D Carbon. (2018, May 25). Prebaked Anodes for Aluminium Electrolysis. Retrieved from R&D Carbon: https://www.rd- carbon.com/data/documents/publications/general/Prebaked_Anodes_Aluminium_Electrolysis _2014.pdf Rivedal, O. (2018, May 29). Aluminium for Furture Fenerations . Retrieved from Anode Production: http://primary.world-aluminium.org/processes/anode-production/ Severo, D.S. Gusberti, V. (2009). A MODELLING APPROACH TO ESTIMATE BATH AND METAL HEAT TRANSFER COEFFICIENTS . Light Metals, 557-562. Taylor, M. Etzion, R. Lavoie, P. Tang, J.(2014). Energy Balance Regulation and Flexible Production: A New Frontier for Aluminium Smelting . Metallurgical and Materials Transactions , 292- 302. Taylor, M. Chen, J.(2015). Technique for Low Amperage Potline Operatin for Electricity Grid Storage. Metallurgical and Materials Transactions, 87-98. Ueckerdt, F. Brecha, R. Luderer, G. (2015). Analyzing major challenges of wind and solar variability in power systems. Renewable Energy, 1-10. World Aluminium. (2017, June 30). PRIMARY ALUMINIUM SMELTING ENERGY INTENSITY. Retrieved from World Aluminium: http://www.world-aluminium.org/statistics/primary- aluminium-smelting-energy-intensity/ Yurkov, A. (2017). Refractories for Aluminum: Electrolysis and the Cast House. Moscow: Springer International Publishing. Zangiacomi, C. E. Garcia, G.J.L. Abreu, A.L.T. Kato, C.R. (2012). EXPERIENCES ON ANODE RECONSTRUCTION PROCESS IN S0DERBERG TECHNOLOGY . Light Metals , 1235- 1239. 43

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

9. Appendix

9.1. Appendix A: Example Technical Model Calculation 푄표푟𝑖푔𝑖푛푎푙 = (1.2 + 0.5 + 0.00000875 ∗ 240 000 − 2.05) ∗ 240 000

푄표푟푔𝑖푛푎푙 = 420 000 푊

푄푠𝑖푑푒 = 0.38 ∗ 푄표푟𝑖푔𝑖푛푎푙 = 159600 푊

푄푡표푝 = 0.48 ∗ 푄표푟𝑖푔𝑖푛푎푙 = 201600 푊

푄푎푛표푑푒 = 0.6 ∗ 푄푡표푝 = 120960 푊

푄푟푒푑푢푐푒푑 = (1.2 + 0.5 + 160000 ∗ 0.00000875 − 2.05) ∗ 160 000

푄푟푒푑푢푐푒푑 = 168000 푊

푄∆ = 푄표푟𝑖푔𝑖푛푎푙 − 푄푟푒푑푢푐푒푑 = 252 000 푊

∆퐸 = 252 000 − [(159 600 + 120 960) − 1.13 ∗ 21.6 ∗ (1233 − 1223) − 0.8 ∗ 13 ∗ (1233 − 1223)]

∆퐸 = 319 520 푊

319.520 ∗ 72 = ∆푇 = 3.01 4500 ∗ 1.7 푏푎푡ℎ

푇퐵푎푡ℎ,𝑖+1 = 푇퐵푎푡ℎ,𝑖 − ∆푇푏푎푡ℎ = 1229.99퐾

푀𝑖+1 = 4500 159.6 − 120.96 − [1.13 ∗ 21.6 ∗ (1233 − 1223) − 0.8 ∗ 13 ∗ (1233 − 1223)] − ∫ ∗ 푑푡 520

푀𝑖+1 = 4457.15 푘푔

∆퐴푙퐹3 = %퐴푙퐹3,𝑖+1 − %퐴푙퐹3,푛 = 0.1133%

푑푇푙 푇퐿,𝑖+1 = 푇퐿,𝑖 + ∗ 푑푡 ∗ ∆퐴푙퐹3 푑퐴푙퐹3

푇퐿,𝑖+1 = 1223 − 4.33 ∗ 72 ∗ 0.001133

Using Aluminium Smelters for Energy Stabilisation: A study of Demand Management for Aluminium Reduction Cells

푇퐿,𝑖+1 = 1222.65 퐾

9.2. Appendix B: Example Economic Model Calculation

26.98 ∗ 240 ∗ 3600 푟 = 0.94 ∗ 퐴푙푢푚𝑖푛𝑖푢푚 3 ∗ 96500

푟퐴푙푢푚𝑖푛𝑖푢푚 = 75.69 푘푔/ℎ푟

푃푒푛푒푟푔푦 = 240 ∗ 4 = 960 푘푊

$ 퐶 = [0.48 ∗ 1.91 + 0.55 ∗ 0.4] ∗ 75.69 + 0.96 ∗ 91.17 = 173.57 푣푎푟𝑖푎푏푙푒 ℎ푟

퐶 $ 푣푎푟𝑖푎푏푙푒 = 퐶 = 199.5 0.87 푡표푡푎푙@240푘퐴 ℎ푟

$ 퐶 = 0.13 ∗ 퐶 = 25.94 푓𝑖푥푒푑 푡표푡푎푙@240푘퐴 ℎ푟

$ 푅푒푣푒푛푢푒 = 푝 ∗ 푟 = 221.01 퐴푙푢푚𝑖푛𝑖푢푚 푎푙푢푚𝑖푛𝑖푢푚 ℎ푟

9.3. Appendix C: Weighted Decision Matrix Scale 1 2 3 4 5 Duration 1-1.5 1.5-2 2-2.5 2.5-2.99 3 Power 0.12-0.14 0.14-0.16 0.16-0.18 0.18-0.20 0.2-0.22 Saved Opportunit 0.35-28 0.28-0.21 0.0.21-0.14 0.14-0.07 0.07-0 y Cost Ramp Reduction Step Reduction Staggered Step Reduction Criteria Weighting 1 15 20 10 15 20 10 15 20 0 Duration 0.3 Power 0.25 Saved Opportunit 0.45 y Cost Total Ranking