You are on page 1of 4

Light Metals 2009 Edited by: Geoff Bearne

TMS (The Minerals, Metals & Materials Society), 2009

Electrical power availability optimization at Alcoa Deschambaults smelter


Vincent Letellier1, Norman Plante2
Alcoa Canada Premire Fusion, Information Technologies, Deschambault, QC, G0A 1S0, Canada
2
Alcoa Aluminerie de Deschambault, Maintenance Departement, Deschambault, QC, G0A 1S0, Canada
1

Keywords: Utilization factor (F.U.), MW consumption, Energy demand, programmable logic controller (PLC)
Abstract
This document will detail what were the principal steps that
brought the team through the realization of the project. First, in
order to establish the bases for comprehension purposes, the
invoicing method of Hydro-Quebecs Large Customers will be
explained. Thereafter, differences between the available tools
related to the old and the new measurement will be detailed in
another section. Finally, a description of the results will be
described for each step of the project. It will be ended by the
conclusion section.

Todays energy scarcity affects technology and economic choices


in most industries. Even if Quebec is a huge producer of hydroelectricity, Alcoas plants located in this province, are facing the
issue. This situation forces optimization of available power in
order to increase plants productivity. One way to achieve this
goal is to increase power utilization directly at the electrical input
of the smelter, allowing a higher operating amperage in the
potline which is translated into more aluminum production.
Some modifications have been made to the electric control
system, at the main substation of the plant. Those modifications
allowed Alcoas Deschambault smelter to use 1.6 MW of power
over its previous average consumption, with the same contract
allowance. The rebuilt control system results in a utilization factor
(percentage of actual power usage over power contract) of up to
99.8% on a monthly basis. A significant gain for a very low
capital cost!

Invoicing Method
There are two different invoicing methods for the HydroQubecs Large Customers. Depending on the contract between
them and the customer, the invoicing is generally based on a 15
minutes period or on a 15 minutes sliding window broken into
three 5 minutes intervals. In this last case, for each 5 minutes
period, the average of the consumed power is calculated. Then,
the 15 minutes window is built with the average of these three
values. This is the method used for invoicing at Deschambault.

Introduction
Electrical energy is an important raw material for an aluminum
smelter. Alcoa Aluminerie of Deschambault Qubec (ADQ) is
fed in electricity by a state company called Hydro-Quebec. Two
electric high voltage lines (315 kV) are connected to the plant.
The total consumption of energy is of 3.660 GWh per year, which
represents the equivalent of the consumption of approximately
150,000 houses. It is not difficult to understand the efforts applied
to optimize this precious energy.

The Figure 1 represents this sliding window method. In this


example, the values retained for billing would be 420 MW,
417.67 MW and 414.67 MW.

In 2004, Hydro-Quebec has communicated the will to modify its


measuring equipment located at the entry of their Large
Customers companies. Through this phase of modernization, the
purpose of Hydro-Quebec was to measure each electric incoming
line independently. In the case of Deschambault, it represents 2
meters of energy connected to the two incoming high voltage
lines. Another goal of the state company was also to measure the
reactive power in a more accurate manner.
During the introduction of these new energy meters by HydroQuebec, Alcoa Deschambault gathered members in special team
to do this integration into the existing system. This team takes the
occasion to adapt its own measurement equipments with the same
generation of meters as Hydro-Qubecs choice. These new
models of meter also offered output signals representing the
instantaneous power; old meters did not. The availability of these
new signals, coming specifically from the invoicing meters,
brought us to think of a new method in controlling the power
delivered to the factory. The study of these new measurements
revealed that there was a significant quantity of energy not used
and worth to be recovered. The objective of the project was to
implement an automated control to adapt it to those new analog
signals and to this new situation.

Figure 1
Also, the maximum demand (PMAX) contracted with HydroQuebec is the value to which ADQ is committed to consume
during a month. The ratio of the power consumption over this
contractual demand corresponds to a measure of the efficiency of
the energy utilization which will be named the Utilization Factor
U.F.. Better is this ratio, better is our performance as the use of
this allowed energy.
The value that established the amount of the monthly invoice is
determined by the highest of the following values: the contractual
demand (PMAX) or the average power of the highest 15 minutes

333

period of the month. It should be noted that if the average power


of this 15 minutes period exceeds the contractual demand, a
monetary penalty applies. In the example of Figure 1, the value of
420 MW would be the peak demand of the month. If it is higher
than the contractual demand, it shall be used as the value to
establish the amount of invoicing.

First, using a signal reflecting the instantaneous variations helps


enormously the power controller to be proactive rather than
reactive. The core of the algorithm is a PID loop function and this
kind of control reacts much better to continuous signals than to
step variations (impulses). Also, the use of the impulse signals
has as disadvantage its inaccuracy in time. For example, an
impulse representing a quantity of energy consumed mainly at the
end of one 5 minutes period but also partly at the beginning of the
following period will be triggered only at the beginning of this
new period. This last period will thus be charged of an energy
which however would have been consumed mainly during the
previous one. With this type of measurement, considering a stable
consumption, there would have been low consumption in the first
case and possibly an exceeding demand on the other hand, even if
actually, the consumption of these two periods should have been
equal. The choice of using the analogue signals in the control
logic was to avoid this kind of chronological mismatch.

The electrical power-factor of the factory is also a parameter that


affects the invoicing. This parameter will not be treated in this
document but just assume that a minimum has been fixed by
Hydro-Quebec.
Control Algorithm
The total incoming power of the plant is managed by a
programmable logic controller (PLC). This PLC receives as
inputs the signals of the two energy meters measuring the power
of their corresponding high voltage lines.
An algorithm,
programmed in this PLC, supervises the power consumption. Its
output controls the current delivered to the potline. Obviously,
this operation on the potline current modifies the total
consumption of the plant and this variation is received in feedback
by the controller. The two independent factors causing variation
on the electrical load of the factory are the impedance in the
potline (replacement of anode, starting of pot, anode effects) as
well as the auxiliary power of the plant (big motors, electrical
heating furnaces, large conveyors).

Energy Management vs Invoicing Mode


As explained previously, the invoicing is based on 15 minutes
sliding windows. It is thus very important to make sure that the
controller takes this factor in account in order not to cause
overshoot of the contractual demand throughout this period. The
controller always reacts on the power over the current 5 minutes
period. This period is by definition also the last period of the
current 15 minutes window. Then, to respect the setpoint over
this period, built of three 5 minutes intervals, it is necessary to
keep in memory the value of the last two periods. The method
used to store these two 5 minutes periods had been to implement a
10 minutes moving average built from 600 values computed every
second. These values are themselves based on averages of the
power analogue signal sampled at the scan time of the PLC. Each
second, this moving average is then refreshed.

Three faces of the automated control were analyzed and modified


: The transmission method of the power readings towards the
PLC, the energy management vs invoicing mode and the
validation of the instantaneous power signals. These three steps
of modifications to the algorithm of control are described bellow.
Electrical Power Readings, Transmission Mode
Before the modernization phase of the meters, the PLC received
impulses as inputs of the energy measurement. Each pulses
represented a consumption of 100 kWh of energy. The invoicing
meters of Hydro-Quebec offered only these output signals. More,
these impulses were equivalent to the summation of the energy of
the two incoming lines.

In order to provide to the PID loop a measurement that will


represents the average power over all the 15 minutes window
under its control, the solution was to add the 600 moving average
values to the multiplication of the instantaneous value by a
constant of 300. This measurement is then built up with 900
values divided in two parts: The past 10 minutes period (600
values moving average) and the next (or current) 5 minutes period
(300 values). The goal is always to obtain the highest power
average over every 15 minutes sliding windows without any
overshoot of the contractual demand. Thats why the controller
has to be aware of the past and has to be able to foreseen the
future. The action of multiplying the instantaneous value by 300
brings a predictive effect on the PID loop over the 5 minutes
period in progress. This operation is equivalent to apply the
instantaneous value over all the 5 minutes period, thus predicting
the final power average value of this period. PID function then
reacts consequently, foreseeing to be lower or higher than the
contractual demand. At the same time, this manipulation adds a
derivative effect on the PID loop.

The new energy meters provide various types of output signals:


impulses of energy as well as analogue signals representing the
instantaneous power. These values come independently from the
two meters and do not reflect anymore the sum of the two lines.
There is an enormous advantage to use the output signals of the
invoicing meters: the same elements (current & voltage
transformers as well as energy meters) are used for invoicing and
within the control system. This eliminates errors between the
value transmitted to the control system and the value for
invoicing. This is why these new signals are now connected and
used by the PLC.
In addition, after analysis, it was decided that the analogue signals
would be used for control instead of using impulse signals. This
choice can seem to be doubtful at first sight. Indeed, the analogue
signal is less accurate for power consumption measurement on a
certain period of time than the impulse signal which comes from
the integration of the power, directly in the instrument. However,
it will be shown here that the advantages associated to the
analogue signals will decrease this negative effect.

Figure 2

334

Validation of Instantaneous Power Signal

matter what the configuration is applied to the high voltage


incoming lines.

In the section Electrical Power Readings, Transmission Mode, it


was discussed the choice of the analogue signal instead of the
energy impulses mode. It was explained that the analogue mode
has precision inaccuracy. Indeed, this effect was observed during
the analysis of data regarding the project. According to certain
conditions of operation or configuration of the equipments of the
main sub-station, the response of total electrical power of the
plant differed from one configuration to another. After analysis, it
was found that these differences came from the load variations
applied to each incoming line. According to the range of the
power measured by a meter, the analogue output did not
correspond to the awaited value due to this conversion between
the instantaneous value and the analogue output which has not a
perfect linear response.

Results
All the changes brought to the control system were described in
the preceding section. The implementation of this set of
modifications obtained excellent results. They are presented
below.
Initially, Graph 1 represents the total electrical power
consumption of the plant BEFORE modifications were applied.
The blue line is expressed in 15 minutes averages on a 6 days
period. The Y axis is in MW. Data is from the end of year 2003.
It should be noticed that the contractual demand (red line) is at
418.8 MW and the consumption average is 417.0 MW. It is
important to note that the average does not take in account the
values lower than 416.8 MW. The reason is that calculation
should take in account only the data in normal operation of the
plant. It will be the same for the other results.

Following this observation, an algorithm of validation of the


analogue signal was developed in order to correct the problem.
To achieve this goal, it was necessary to have access to a very
accurate value of the power consumption. The solution to this
problem was the development of a computerised system of
measurement. The core of this system is a server executing a
program which communicates with the energy meters and stores
the data in a central database. Figure 3 explains interconnections
and dataflow between equipments. The work team takes the
occasion to connect the meters of Hydro-Quebec to this new
server. The data in the internal memory registers of invoicing
meters were therefore stored and easily accessible by the PLC
controller. The accuracy of these values is very high and fulfills
the need for this project.

Performance prior changes, november 2003


420.0
Instantaneous Power

Power

419.0

Power (MW)

418.0
417.0
416.0
PMAX : 418.8 MW
Pmoy : 417 MW
Delta : 1.8 MW

415.0
414.0
24-nov

25-nov

26-nov

27-nov

28-nov

29-nov

30-nov

Time

Graph 1
In 2004, a few weeks after the changes, the results were already
visible. At this stage, only the first two modifications were
applied (Electrical Power Readings, Transmission Mode and
Energy Management vs Invoicing Mode).
On Graph 2, we can observe that the power consumption curve (in
blue) nearly approached the setpoint of the controller (in red).
There was at that time an increase of 1.4 MW of the consumed
power average. This increase went up our Utilization Factor
ratio from 99.50% to 99.85%.

From now on, the data of high degree of accuracy is now


accessible by the control system. The method used for making
possible the transfer between one entity to another was to develop
an interface in the existing SCADA system that gathers the data
from the database and transmits them towards the PLC. This
process occurs over the Ethernet network. Two informations are
transmitted at the end of each 5 minutes period. They represent
the value of the power consumption for the last period for each of
the two lines. It is important to remember that the values are
coming from the internal memory registers of the invoicing
meters of Hydro-Quebec. It means that the control is always
online with the billing values. At the same time, the program
integrates the instantaneous power analogue signal for the same
period; it is the same signal used by the power controller.
Thereafter, the algorithm compares these two values of energy
and thus determines the error between the analogue signal and the
invoicing value. By software programming, this error is then
applied respectively to each input signal. Thus, the program
compensates the nonlinear error brought by the selected signal
type. This ensures the constancy of the error on the signal no

Performance AFTER 1st phase, march 2004


420.0
Instantaneous Power

Power

Power (MW)

419.0
418.0
417.0
416.0

PMAX : 418.5 MW
Pmoy : 418.1 MW
Delta : 0.4 MW

415.0
414.0
23-mars

24-mars

25-mars

Time

Graph 2

335

26-mars

27-mars

28-mars

manager in place in 2004 for the confidence that she had put in
our project.

More benefits were obtained after the last modification described


in the section called Validation of Instantaneous Signal. This
time, 0.2 MW were added to the total consumption of the plant.
This brought the monthly Utilization Factor up to a level of
99.92%. The average power is found to be then at only 0.2 MW
of the contractual demand.
Performance today, august 2008
420.0
Instantaneous Power

Power (MW)

419.0

Pmax

418.0
417.0
416.0

PMAX : 417.8 MW
Pavg : 417.6 MW
Delta : 0.2 MW

415.0
414.0
01-aot

06-aot

11-aot

16-aot

21-aot

26-aot

31-aot

Time

Graph 3
The rise up of our Utilization Factor from 99.5% to more than
99.9% enabled us to recover 1600 kilowatts of continuous power
which generated an increase in production over than 900 metric
tons of aluminum per year. This represents additional incomes
estimated at 2 500 000$ per year based on the price of aluminum
on the London Metal Exchange in 2008. A total investment of 75
000$ was necessary to achieve the project. A return on
investment of approximately 11 days (0.03 year).
Conclusion
As we mentioned in the introduction, electrical energy is a
precious raw material for aluminum smelters and has a great
impact on their production costs. This is why we invested time to
optimize the availability of electrical energy supplied to the
smelter. At the implementation step of the project, we were
expecting good results and they finally exceeded our expectations.
The objective was subsequent to the analysis of new data
according to modifications our energy supplier made to their
equipments.
In the Results section, looking at the Graph 3, we have shown that
we still have a good performance today. These results improved
our Utilization Factor, improve the accuracy of the controller and
brought a significant increase to our production of aluminum.
Also, through out the realization of this project, we have
demonstrated that the company uses energy in a responsible
manner and this way broadcasts a good corporative citizen image.
These results will not stop our willingness to improve our
efficiency. Still today, we try to increase even more our
Utilization Factor, although each additional kilowatt represents a
real challenge.
Thanks
Thanks to Alcoa Deschambault for having allowed us to
participate in this event. Special thanks to the maintenance

336

You might also like