Steady Upstream for Loose Control
Custom-built proportional integral loops overcome PLC software limitations
By Saed Hussain, Mark Cusac, and James Muri
The Norumbega Reservoir used to be an essential part of the water transmission system for the metropolitan Boston area. Now the Norumbega Covered Storage Tank (NCST) facility replaces it as one of the largest drinking water storage tanks in the world.
Outflow from the NCST fluctuates widely throughout the day. The cause stems from a 405-MGD water transmission system supplying the Boston area that suffers a large diurnal demand variation. To keep the transmission system and the upstream treatment processes running steadily, the system should make as few changes as possible in the upstream supply valve’s position. To accomplish this, the level of the 115-million-gallon water storage tank becomes the process variable in a custom-built timed, reverse-acting, proportional/integral closed-loop level controller. This controller modulates remote upstream water supply valves through a minimal amount of movements, allowing the loose control needed to satisfy the hydraulic system’s varied needs.
Almost all programmable logic controllers (PLCs) offer traditional proportional/integral/derivative (PID) controllers. While this facilitates many automation processes, it does not overcome existing PLC software limitations on ready-to-use PID blocks, tuning parameters, and loop update time when you need loose control in certain applications or when the characteristic process time parameters are very large. There is a custom-built ladder logic PID loop that overcomes existing PLC software limitations to achieve its goal.
Water transmission system
The tank hydraulic gradeline floats on the hydraulic gradeline of the Metrowest Tunnel, whose Western segment supplies water to the NCST. Upstream water supply valves control the entire flow from the Metrowest Tunnel. The tank supplies water to the eastern segment of the Metrowest Tunnel, feeding the entire metropolitan Boston area and compensating for the variation in demand through different times of the day.
The 115 million-gallon reinforced concrete tank consists of three cells. Operators indicate the level in each cell and record it in a supervisory control and data acquisition system using two redundant level sensors. These are the basis of control to send an automatic position set point to the remote upstream water supply valves. The goal of the automatic system is to allow the NCST to absorb variations in the demand downstream, while letting the tank level move up to four ft above or below its normal level set point. Although the Norumbega tank is a flow through, other tanks in the system are not. The fluctuation in level keeps the water fresh in this and other connected water tanks in the system.
To improve the reliability of the level signal, operators calculate the average level signal for each cell from the reading of the two-level sensors. If the operator chooses to use both level sensors, the two sensors provide two different level readings for each cell; however, if both instruments are working correctly, these two readings are within an allowed error deadband. Add the two level signals and divide by two to produce an average level reading to use as a feedback signal for the automatic control of the remote upstream water supply valves.
If one of the two level readings drifts excessively or gives a zero or full-scale reading, the error exceeds the error deadband allowed. Determine the level sensor that has drifted excessively or failed by comparing each of the two level readings to the last recorded “good” value, and remove it from the averaging process. An alarm brings the failure to the operator’s attention. Until the faulty instrument returns to service, only one level sensor determines the level reading.
When either one of the two remote upstream water supply valves is in remote and available for automatic control, the selected Norumbega level signal will send a remote auto position set point to this valve via radio link. The control is possible with timed, reverse acting, PI closed-loop level controllers associated with one of the three tank cell levels. When a custom-built PI controller switches to manual, operators may make manual position adjustments of the remote upstream water supply valves. The logic provides smooth transfer between auto and manual by recalculating the accumulated error required to prevent any major fluctuations in the control output signal when the remote valve switches to automatic control.
Custom-built PI level controller
When the PI controller switches to automatic PI control mode, the output control sent to the remote upstream water supply valves to modulate their position stems from the following equation:
Calculate the accumulated error by the summation of the error between the process variable (cell level) and the desired tank level set point each time the PI loop updates. Multiply it by the loop update time:
PI controller deadband
One of the goals of the system is to make as few changes as possible in the upstream controlling valve’s position. This helps keep the transmission system and the upstream treatment processes running steadily. For this reason, the system tries not to let these valves change their position unless there is a considerable change in the water level at Norumbega Storage Tank. To achieve this, the control signal currently sent to remote upstream modulating valves store in a temporary register inside the PLC memory each time the PI controller loop executes. This stored value calculates the change in the control output signal sent to the modulating valves. If the change in the PI control output is less than the deadband the operator enters, the operator ignores the change in the PI controller output, and the same control signal goes to the remote upstream water supply valves to maintain their position. Only when the change in the PI control output is greater than a deadband entered by the operator will the new control output serve as the control signal for the upstream water supply valves. It is this deadband that allows the loose control of the storage tank level to satisfy the system hydraulic needs. If the water level in the tanks fluctuate only a small amount away from the set point, the system makes no valve changes.
Auto/manual smooth transfer
To prevent sudden changes in valve position when switching the remote upstream water supply valves from manual to automatic control mode, use the position feedback of the valve switched to automatic, along with the current error term, to manually calculate the accumulated error required to prevent any fluctuations in the control output signal when the valve switches to automatic control. Calculate the accumulated error using:
There are some known pitfalls and common limitations for setting up PID control using a PLC. All ready-to-use PID controller blocks in a PLC tightly control any process to maintain a steady-state system and shorten a transient time response. Therefore, using any of these PID controllers to control the 115 million-gallon water storage tank level would result in maintaining a steady water level inside the tank, while drastically changing the flow through the water transmission system. This would cause major shocks to the treatment processes upstream.
The output deadband of this controller is the key point of how the system works. This custom-built PI controller allows the process variable (the level in the tank cells) to fluctuate during the day instead of trying to maintain a stable level set point. Looking back at the goals for this system, this feature satisfies the need to allow the tanks to fluctuate. The deadband allows a loose control of the process variable, so water in the tanks cycle through and the level stays within acceptable limits to satisfy the downstream demand while minimizing the movement of the upstream water supply valves. Equally important, minimizing the remote upstream supply valves’ movements avoids excessive wear in these 72 in valves since replacement is costly and difficult.
Given the huge storage volume of this tank, it is a matter of several minutes (not seconds) to expect any significant change in the tank level (process variable) that is a direct result of the downstream demand, not noise generated by the level sensors. This large response time demonstrates another common limitation of traditional ready-to-use PID function in a PLC: its ability to only control processes with reasonably fast response. Its application to processes with large time delay, where cause and effect are not instantaneous, is less common and shows poor stability. Therefore, the PLC programming software manufacturers limit the value at which you can set the PID loop update in order to avoid excessive overshoot and large settling time. But to satisfy the system process, we had to set the loop update time (Dt) of the PI controller to minutes, not seconds, which exceeded the loop update time limits of the ready-to-use PID blocks in the existing PLC.
More importantly, for this level controller to react to the downstream demand trend over a period that may expand over a couple of days or more, we had to set the integral time constant (Ti) of the integral part to about six hours, which exceeded the limits on setting this variable for ready-to-use PID block in the existing controller.
This is when the need for a custom-built PI controller first materialized, since its loop update time or its reset time are not limited. Rather than integer-type variables, the system uses floating point-type variables that accept values as high as 120 seconds of loop update time and two days of reset time.
About the Authors
Saed Hussain and Mark Cusac are control systems engineers at Camp Dresser & McKee, Inc. in Cambridge, Mass. James Muri is a senior program manager in T&T operations process engineering at Massachusetts Water Resources Authority in Marlborough, Mass
Industrial Pressure and Level Measurement Engineering (EI05) http://www.isa.org/training/ei05
The Consumer Guide to Non-Contact Level Gauges Flowmeters (EI12PC6) http://www.isa.org/noncontact
Industrial Pressure, Level, and Density Measurement by D. Gillum http://www.isa.org/pressureleveldensity