SMALL PLC REVAMPS PRODUCTION BATTERY CONTROL

Michael A. Argo
ARCO Oil & Gas Co.
Midland

For a West Texas oil production facility, installation of a PLC (programmable logic controller) increased operator information, improved control system reliability, and enhanced throughput.

The new controls also reduced nuisance alarms, repair time, and the possibility of spills.

The H.T. Boyd production battery, located 45 miles north of Denver City, Tex., was built in 1958. Today, the battery serves 68 wells and handles 9,600 b/d of fluid.

The original control system consisted of relay panels dedicated to individual pieces of equipment or processes, i.e., each of the engine-driven pumps had a dedicated control panel. Over time, as the production of the field changed and equipment was replaced or added, the control system became difficult to operate, maintain, and modify.

In 1989, the existing control system was revamped through the installation of a small PLC. At the start of this project, the battery consisted of four storage tanks, three engine-driven main pumps, a charge pump, skim tank pump, and LACT (lease automatic custody transfer) unit.

During this project, a fourth main pump, electric-motor driven, was added. A schematic of the facility is shown in Fig. 1.

PUMPS

The main pumps were instrumented for the following alarms and shutdowns:

High pump discharge pressure
Low pump discharge pressure
Low engine lubricant level
Low pump lubricant level
Engine overspeed
High engine coolant temperature.

Individual motor instruments were switch gauge-style units with oil-immersed contacts. The alarm/shutdown circuits were normally open with the contacts closing on an alarm or shutdown condition.

This circuit arrangement is often used in the belief that it produces fewer nuisance alarms. The prime shortcoming of this contact arrangement is the lack of a self-check of the circuit integrity between the end device and the controller.

The preferable arrangement is for the contacts to close on normal process values. Then, when the process reaches an alarm value, the contacts open to alarm or shutdown the equipment.

Likewise, failures interrupting the control circuit cause the control system to operate and result in operator intervention.

In contrast, systems with the contacts open on normal conditions can suffer unintentional circuit breaks, such as accidental excavations or loose terminations, without any direct indication of a problem.

Each pump's control panel was located immediately adjacent to its respective unit in the pump building. This exposed the controls to high levels of vibration and possible flooding of the control boxes in the event of a leak.

TANKS AND CONTROLS

The four tanks at the battery are: two produced water tanks, one freshwater tank, and one skim tank.

Level instruments on the produced water tanks are switches with local indication. The produced water tank serving as the pump supply was set up with low alarm and low shutdown set points. The skim tank and freshwater tank each had high level set points.

NEW SYSTEM

The new control system performed the engine and pump protection functions listed previously as well as several new functions.

To avoid adding additional tanks, more importance was assigned to uninterrupted operation of the gas engine-powered pumps. Nuisance alarms and call-outs were to be minimized. The control system was designed for ease of understanding and operating. Included in the design were the following:

Engines shut down sequentially on high discharge pressure.
All engines shut down immediately on low discharge pressure.
Charge pump shuts down when all main pumps are down.
Electric drive pump starts when any other main pump shuts down, unless the shutdown was for high or low discharge pressure.
Radio alarm activates when all main pumps shut down and on high tank level.
Individual pump control panels are consolidated into a single panel.
Pump and battery alarms are consolidated into a single panel.

PLC HARDWARE

Standard construction methods were used for wiring between the end devices and the controller. Rigid metal conduit was run except where limited flexibility was needed.

For example, the main instrument wiring runs in rigid conduit and the individual drops to the engine are flexible conduit.

A single NEMA 4 (watertight, dust-tight, sleet resistant indoor and outdoor) box housed the PLC system and the status display consisting of a General Electric Series One Plus controller with:

1,724 words of memory
Base I/O (Input/Output) rack and two expansion I/O racks with a total of 6-8 channel digital input modules
7-8 channel digital output modules
Two spare slots.

To the greatest extent possible, the physical wiring of the inputs and outputs for a single device, such as Main Pump 2, were consolidated into one input or output module as required. This approach allowed the person troubleshooting the unit to concentrate on a single group of wires or single module.

A light-box annunciator was used instead of a series of individual lamps to indicate equipment and alarm status. No on-board logic was purchased with the light-box.

Portions of the PLC program were dedicated to "first out," lamp test, and other functions. Approximately 30% of the memory and 70% of the output channels were dedicated to driving the annunciator.

Pump operating modes for the engine-driven units were selected by standard two position (run/stop) selector switches. A three position (hand/off/auto) selector switch determined the single electric drive unit's operation.

Reset, alarm acknowledge, emergency stop, and lamp test push buttons were normally open, momentary contact units.

Because the control panel was outdoors, all components exposed to weather were industrial heavy duty, weather roof items.

PLC SOFTWARE

The program for controlling the equipment, local alarm panel, and remote alarm was developed on the General Electric portable programming unit, the midpoint programming device for the GE Series One PLC.

The other programming devices available were the hand-held programmer and the GE computer-based programmer.

The portable incorporates a liquid crystal display capable of displaying multiple rungs of ladder logic while indicating the real-time status of the rung components such as input contacts and output relays. The portable can also drive a printer directly. This allows for rapidly creating hard-copy documentation.

After developing the logic for the first engine, the second and third unit's logic only required substituting the appropriate I/O reference numbers.

Unfortunately, the portable programming unit did not allow for copying lines of ladder logic, therefore each module, although essentially identical, had to be individually keyed in.

As in the hardware installation, logic modules were written for each pump as a self-contained package to minimize hunting through widely separated areas of the program while troubleshooting.

Although a programming and documentation package was available for use on a personal computer, annotation and operational notes were handwritten. The GE software was originally written for a dedicated programming unit and was extremely frustrating to try and operate on a standard personal computer.

Since this project, General Electric has revised its software.

Now it is easier to use on a personal computer. Also, its capabilities were made comparable to the programming and documentation packages available from other PLC vendors and third party companies.

Other documentation, such as input/output listings, internal coil (relay) and retentive relay records, and timer/counter tables, were all prepared on a personal computer using a form generation program available at the time.

Tables 1 and 2 are examples of timer/counter and input/out forms.

The program was relatively simple. The control schemes were all simple on or off logical decisions. No math or analog values were handled in this application although the PLC used does have those capabilities.

The "first out" alarm capability, as illustrated in Fig. 2, was a requirement from the operations department to expedite troubleshooting.

Each alarm input was placed into a latching circuit, i.e., a circuit designed to stay "on" once it is turned on even though the alarm signal might clear by itself.

The latch circuit consists of a set of contacts that parallel or bypass the alarm contacts and are linked to the output contacts.

Once the output contacts close, the bypass contacts also close and maintain a complete circuit regardless of the state of the alarm.

To finish the "first out" logic contacts linked to the other alarm/shutdown, outputs for the particular unit are added to the right of the latch. These contacts a re normally closed so that when the first alarm is received, the circuit is unbroken and the latch is activated.

Now, the next lines or rungs of logic contain contacts linked to the energized output contact but which are normally closed. When a signal does go into alarm, though, these normally closed contacts switch open.

The open contact prevents these circuits from being completed. Thus, any alarms coming in after the initial alarm cannot be recognized or registered by the program.

Pushing the reset button interrupts the "latch" and clears the alarm logic.

Sequencing the pumps off on a high-pressure alarm required the use of the PLCs' internal timers and counters. Because three of the four main pumps required a manual restart, minimizing the number of pumps shut down on a short-term upset would, in turn, minimize the number of operator callouts while maximizing production.

The pump sequencer used one self-resetting timer with a 5-min preset and one counter.

Once the high-pressure signal was detected, Main Pump I was shut down and the 5-min time delay started. If the high-pressure condition persisted long enough to cause the timer to expire, a logic (internal to the PLC only) signal incremented the counter.

Each time the counter incremented, a logic contact operated. This triggered a physical output, which shut down the next engine in sequence.

If the high-pressure condition corrected itself, the timer and counter logic was disabled with only the minimum number of pumps offline.

START-UP

After commissioning the system, two minor defects were discovered.

The first was in the LACT unit alarm wired directly into the PLC program. The PLC detected and reported on all bad oil signals, no matter how short in duration. The bug generated a number of nuisance alarms.

Installing an on-delay timer programmed for a short delay prevented small slugs of bs&w from being reported.

If the alarm signal persisted until the timer finished its cycle, the alarm was judged to be "real" and reported via the radio alarm system as well as diverting the oil back through the battery.

This fix was installed and operational in less than 1 hr thanks to the programmability of the controller.

The second bug was the inability to detect when an engine died due to poor fuel quality. Abnormal conditions such as overpressure, overspeed, and over-temperature were recognized and set up in the alarm system. But the engine simply dying due to poor fuel quality or other problems had not been initially considered.

Even though it would be possible to infer an engine failure from other signals, a specific alarm was requested.

To correct this problem, a proximity switch and control relay were installed on each engine.

The proximity switch was mounted on the pump to detect passing gear teeth. A control relay dedicated to the proximity switch picked up the switch's signal. If the relay did not receive a signal from the switch, the control relay's contact operated a spare digital input and tripped the alarm logic in the PLC.

Due to concern over reliably reading the proximity switch's low-level signal at the PLC, the dedicated relay was selected over programming the same function in the PLC.

POST MORTEM

The project was successful. The control system has functioned as expected and has since been modified by contract personnel to replace one of the gas-engine drivers with a second electric driver.

Using the various computer programs to design and document the project was definitely a positive factor. Revisions were made much more quickly on the computer than by hand.

On site input during the program design was very valuable in assuring the initial program met the functional requirements and minimizing "on the fly" revisions during commissioning.

Operator acceptance of the system has been generally good although more training time should have been provided.

The size of the system while adequate for the initial scope, has become a limiting factor in future work. Once the operators and engineers became comfortable with the system, additional functions were suggested but could not easily be accommodated in the existing panel.

Additional input and output modules can be added as well as additional programming in the existing memory.