The present invention relates to a controller for a rotating machine, such as a pump or a compressor.

Today, turbo machinery uses numerous independent control systems to analyse and control high speed rotating machinery, such as an electrical motor compressor supported on magnetic bearings. Such systems typically include an active magnetic bearing controller, an anti-surge valve controller, a condition monitoring data controller, a motor power analyser, and one or more process controllers (e.g. for controlling a cooling gas valve setting). Whilst the use of independent control systems is computationally simple, it restricts the ability of these controllers to operate effectively in response to unexpected, transient situations. Such a control system may also not respond optimally in other situations such as during start-up or even during steady state operation.

WO 2015/158734 proposes an alternative control system for a compressor, in which the active magnetic bearing controller also controls the anti-surge valve. In this system, the anti-surge controller and its associated sensors may be omitted, such that the anti-surge valve is based on the amount of current required to maintain the rotor in the desired position. However, this control system still processes data in a reactive manner and would still react ineffectively in response to an unexpected, transient situation.

The present disclosure seeks to provide improved control of such a gas compressor in transient situations. It will be appreciated that the techniques described herein may also be applicable to other rotating machinery, such as a multiphase compressors and/or to single or multiphase pumps.

Viewed from a first aspect, the present invention provides a method of protecting a rotating machine, such as a compressor or a pump, from damage whilst in a transient state, comprising: monitoring at least one property of the rotating machine; detecting a change in the at least one property of the rotating machine; determining that the change is indicative of a potential future transient state of the rotating machine; and initiating at least one protective measure to prevent or reduce damage to the rotating machine from the determined potential future transient state.

In accordance with this method, the rotating machine can be protected from damage by using predictive control, i.e. wherein an action is taken in anticipation of a future transient state, as opposed to reactive control as used in existing systems, in which the rotating machine is controlled only responsive to deviation of its current state of operation from a desired state of operation.

The rotating machine may be a compressor, such as a dry gas compressor or a multi-phase compressor (e.g. a compressor for compressing a well stream or the like containing produced hydrocarbon fluid), or a pump, such as a single-phase pump or a multi-phase pump.

Preferably, the at least one protective measure is initiated before the rotating machine enters the transient state. Thus, if the rotating machine does enter the transient state, then the protective measure is already in place to minimise damage to the rotating machine.

The potentially transient situation may be one that cannot be avoided. That is to say, the protective measure may not prevent the transient situation itself, but rather seeks to minimise damage to the rotating machine if and when the transient situation does occur.

In one embodiment, the transient state occurs responsive to a total loss of electrical power to a motor of the rotating machine. For example, the transient state may comprise spin down of the rotating machine from an operational speed to zero. Typically, in such conditions, the rotating machine would go into surge shortly after entering the transient state, which could cause significant damage if there were to be a hard touchdown of the rotor of the rotating machine against the stator of the rotating machine. In previous systems, that use reactive control, it would not have been possible to prevent this because the rotating machine goes very rapidly from normal conditions to surge conditions, such that conventional anti-surge techniques could not react fast enough.

In one embodiment, the transient state occurs responsive to a change in a composition of a fluid received at the rotating machine. The change in the composition of the fluid may be a change in a phase composition of the fluid to be received at the rotating machine (commonly measured as a liquid-to-gas (L/G) ratio or a volume percentage of the water or gas within the fluid).

Thus, the at least one monitored property may include a composition, preferably a phase composition, of the fluid to be supplied to the rotating machine. The composition of the fluid is preferably monitored upstream of the rotating machine. In one embodiment, the composition of the fluid is monitored at least 0.5 seconds upstream of the rotating machine, more preferably at least 1 second upstream of the rotating machine, and yet more preferably at least 2 seconds upstream of the rotating device. Such fluid composition data may be monitored using a multiphase flow meter or any other suitable means.

An increase in liquid within the fluid received at the rotating machine may result in a transient state in which undesirable vibration of rotating components can occur, which in turn can cause damage within the machine. Thus, the protective measure may comprise an anti-vibration protective measure.

A decrease in the liquid within the fluid received at the rotating machine may result in an effective increase of operating volume within the rotating machine, which may reduce the compression ratio of the compressor sufficient to result in surge. Typically, in such conditions, the rotating machine would go into surge shortly after entering the transient state, which could cause significant damage if there were to be a hard touchdown of the rotor of the rotating machine against the stator of the rotating machine. Thus, the protective measure may comprise an anti surge protective measure and/or an anti-touchdown protective measure.

The transient state may be detected responsive to a change in the phase composition greater than a predetermined threshold. The threshold may represent a predetermined change optionally within a predetermined period of time.

The change threshold may be a relative threshold, for example greater than a 50% relative increase and/or a 50% relative decrease in the quantity of liquid (or gas) within the fluid and in some embodiment greater than a 100% relative increase. For example, the transient state may be detected where the liquid content in the fluid increases from 10% by volume to 20% by volume. Such a change in liquid content may be indicative of a‘slug’ that will shortly be received at the rotating machine.

The change threshold may be an absolute threshold, for example the threshold may comprise at least a 5% by volume increase and/or a 5% by volume decrease in the quantity of liquid (or gas) within the fluid, and in some embodiments the threshold may comprise at least a 10% by volume increase and/or a 10% by volume decrease.

The at least one property may comprise a shaft speed of the rotating machine. The at least one property may also or alternatively comprise a current and/or a voltage supplied to a motor of the rotating machine. These properties are particularly useful for providing predictive control where the transient state is caused by power fluctuations, as small deviations in these values can indicate potential future problems in the supply of power to the rotating machine.

The at least one property may comprise a position signal of a shaft of the rotating machine. The at least one property may comprise a coil current of a bearing of the rotating machine. The at least one property may comprise a flux measurement from adjacent a bearing of the rotating machine. The at least one property may comprise a flow rate of a fluid through the rotating machine. The at least one property may comprise a differential pressure across the rotating machine. The at least one property may comprise fibreoptic strain data from upstream and/or inside the rotating machine. The at least one property may comprise multiphase meter data from upstream and/or inside the rotating machine, as discussed above.

The at least one property may comprise acoustic sound generated by the rotating machine, which may be recorded by an acoustic sensor positioned within the rotating machine.

Preferably the at least one protective measure is to prevent or reduce damage that would be caused by contact between a static component of the rotating machine and a rotating component of the rotating machine during the transient state.

In one embodiment, a controller may monitor a lateral position of the shaft and control a flux supplied by an active magnetic bearing of the rotating machine to maintain a shaft of the rotating machine at a neutral position.

The at least one preventative measure may comprise changing a control algorithm of the controller for how much flux is supplied by the active magnetic bearing. For example, the controlling may include at least a proportional component based on a deviation between the position of the shaft and the neutral position, and the at least one preventative measure may comprise changing a gain of the proportional component.

By controlling the active magnetic bearing in this manner, the“stiffness” of the active magnetic bearings can be adjusted so as to minimise the risk of damage to the rotating machine. For example the bearings may be stiffened, e.g. to prevent contact of the rotor or stator. Alternatively, the bearings may be softened, e.g. to prevent vibrational damage and/or unwanted noise within the rotational machine.

Alternatively or additionally, the at least one preventative measure may comprise adjusting the neutral position, for example by applying an offset to the neutral position such that the neutral position does not align with a centreline of the rotating machine. By setting a new neutral position, an airgap between the magnetic bearing and the shaft may be reduced so as to mitigate a surge impulse.

The at least one protective measure may, alternatively or additionally, comprise at least partially opening a valve to recirculate fluid from an outlet of the rotating machine to an inlet of the rotating machine. The valve may be an anti surge valve associated with the rotating machine.

The at least one protective measure may, alternatively or additionally, comprise adjusting a surge line and/or a surge margin used for controlling an anti surge valve associated with the rotating machine. The system may be optimized for more efficient compressor control (e.g. a narrow surge margin) or for more efficient transient response (e.g. a larger surge margin), as required. For example, the rotating machine could be controlled to operate in a state further away from a true surge line than would normally be required, thereby providing more time for the rotating machine to react should a transient state cause the rotating machine rapidly approach a surge condition.

The at least one protective measure may, alternatively or additionally, comprise adjusting an axial force that is applied to counter an axial thrust generated by the rotating machine. For example, the axial force may be adjusted by at least partially closing a valve in a line connecting a discharge outlet of the rotating machine to a balance piston of the rotating machine, wherein the balance piston applies the axial force to the rotating machine.

When a rotating machine, such as a compressor, goes into surge, the thrust force generated by the rotating machine increases. However, the force applied by the balance piston may not account for the reversed flow condition. Rapid loss of power (e.g. prior to total power loss) could result in a significant axial force being applied to the compressor by the gas, which could cause damage. By neutralizing the gas forces inside the machine, a more dynamic response can be expected from the active magnetic bearing.

The method may further comprise detecting a further change in the at least one property of the rotating machine; determining that the further change in the property of the rotating machine is indicative of the transient state of the rotating machine occurring; and initiating at least one further protective measure to prevent or reduce damage to the rotating machine from the transient state. Thus, in addition to the reactive measures discussed above, the rotating machine may take responsive actions also. In some embodiments, these responsive actions could be taken earlier than would normally be the case. For example, when the controller has already predicted that the transient state is likely to occur, it may use more sensitive tolerances for properties that indicate that the transient state is occurring. Thus, the rotating machine can respond more rapidly to entering the transient state.

In one embodiment, the at least one further protective measure may comprise opening a two-state valve to recirculate fluid from an outlet of the rotating machine to an inlet of the rotating machine. Preferably the two-state valve is closed during normal operation of the rotating machine. A two-state valve can respond much more quickly than a valve controlled by a stepper motor or similar valve, as would typically be used for an anti-surge valve. Thus, such a valve can rapidly put the rotating machine into a state in which a significant quantity of fluid can be recirculated to prevent the rotating machine going into surge in the event of the transient state occurring.

In one embodiment, the two-state valve may be arranged in parallel with a controllable valve, such as the anti-surge valve. Thus, the anti-surge valve can be used for precisely controlling the operating conditions of the rotating machine to prevent surge during normal operation, and the two-state valve can be used for preventing the rotating machine going into surge in response to a transient condition that causes a rapid change in operating conditions.

The rotating machine preferably comprises at least one active magnetic bearing. A flux supplied by the active magnetic bearing may be controlled by the controller. The flux may be controlled based on a deviation of a shaft of the rotating machine from a neutral position. The flux may be controlled using at least a proportional component, wherein a gain of the proportional component may be adjusted by the controller.

Viewed from a second aspect, the present invention provides a method of controlling an active magnetic bearing for a rotating machine, such as a compressor or a pump, wherein the rotating machine is located within a larger system, the method comprising: monitoring a first property of the rotating machine, the first property comprising a lateral position of the shaft of the rotating machine;

calculating a lateral deviation between the measured position of the shaft and a neutral position; and calculating a flux to be supplied by an active magnetic bearing including at least a proportional component based on the deviation between the position of the shaft and the neutral position, wherein the method further comprises: monitoring at least one second property of the system, the at least one second property being different from the first property; predicting a change in future behaviour of the system based at least one second property; and based on the predicted change in future behaviour of the system, changing a gain of the proportional component and/or the neutral position.

The at least one second property comprises one or more of: a rotational speed of the shaft; a property of a motor driving the rotating machine; a suction pressure of the rotating machine; a discharge pressure of the rotating machine; an axial position of the shaft; a differential pressure across the rotating machine; a flow rate through the rotating machine; a fluid composition upstream of the rotating machine; and a fluid composition within the rotating machine.

The fluid composition upstream of the rotating machine may be determined using a multiphase flow meter. The fluid composition within the rotating machine may be derived based on recorded acoustic data within the rotating machine.

Viewed from a third aspect, the present invention also provides a system comprising a rotating machine and a controller arranged to perform a method as described above, optionally including any one or more or all of the preferred and optional features thereof.

The system preferably comprises an electric motor configured to drive the rotating machine.

The system may comprise a sensor for detecting a shaft speed of the motor and/or compressor. The system may comprise a sensor for detecting a voltage and/or a current supplied to the motor. The system may comprise a sensor for detecting a fluid flow rate through the rotating machine. The system may comprise a sensor for detecting pressure rise/difference across the rotating machine.

The system may comprise a sensor for detecting pressure rise/difference across a scrubber, de-liquidizer or mixer upstream of the rotating machine. The system may comprise a sensor for detecting a process temperature. The system may comprise a sensor for detecting a temperature of a cooling gas system. The system may comprise a sensor for detecting a temperature of the motor or a magnetic bearing of the rotating machine. The system may comprise a sensor for detecting a surface temperature. The system may comprise a sensor for detecting a bearing touch down condition. The system may comprise a sensor for detecting casing vibration. The system may comprise a sensor for detecting acoustic sound generated by the system. The system may comprise a distance sensor for detecting displacement generated by components in the system. The system may comprise a sensor for micro strain. The system may comprise a PVT computation system for estimating a fluid composition within the system.

The system preferably comprises an anti-surge valve for recirculating fluid from downstream of the rotating machine to upstream of the rotating machine. The anti-surge valve is preferably a controllable valve, i.e. in which the opening of the valve may be controlled to a position between fully open and fully closed.

The system may comprise a second valve for recirculating fluid from downstream of the rotating machine to upstream of the rotating machine, e.g. in parallel with the anti-surge valve. The second valve is preferably two-state valve, i.e. only switchable between fully open and fully closed. The second valve preferably has a response time between a fully open position and a fully closed position of less than 500ms, and preferably between 200ms and 400ms.

The system may comprise a scrubber, de-liquidizer, mixer, e.g. for separation of liquid upstream of the rotating machine.

The rotating machine may comprise a balance piston, e.g. for applying an axial force to counter a thrust generated by operation of the rotating machine. The balance piston preferably receives pressurised fluid from an outlet of the rotating machine. The system may be arranged such that the axial force applied by the balance piston may be adjusted, for example by closing a valve connecting the balance piston to the outlet of the rotating machine.

The rotating machine preferably comprises at least one active magnetic bearing. A flux supplied by the active magnetic bearing may be controlled by the controller. The flux may be controlled based on a deviation of a shaft of the rotating machine from a neutral position. The flux may be controlled using at least a proportional component, wherein a gain of the proportional component may be adjusted by the controller.

Certain preferred embodiments of the present invention will now be described in greater detail by way of example only and with reference to the drawings, in which:

Figure 1 illustrates a compression portion of a conventional gas processing system including a gas compressor with an anti-surge controller;

Figure 2 illustrates the inputs and outputs of the anti-surge controller; and Figure 3 illustrates a rotor shaft of the gas compressor and an active magnetic bearing controller;

Figure 4 illustrates the inputs and outputs of the active magnetic bearing controller; and

Figure 5 illustrates a motor for driving the gas compressor and a motor power analyser;

Figure 6 illustrates the inputs and outputs of a motor cooling gas valve controller; and

Figure 7 illustrates a combined controller for the gas compressor.

Figure 1 illustrates a compression portion 2 of a gas processing system between a first location 4 and a second location 6. The compressor portion 2 comprises, in series between the first location 4 and the second location 6, a scrubber 8, a gas compressor 10 and a cooler 12. The scrubber 8 is upstream of the gas compressor 10 and the cooler 12 is downstream of the gas compressor 10. By way of example, the compressor may compress a gas from about 1 bar to about 1000 bar.

Also illustrated is an anti-surge line 14 including an anti-surge valve 16 controlled by an anti-surge controller 18. The anti-surge line 14 functions to recirculate high-pressure fluid from downstream of the compressor 10 to upstream of the compressor 10 to prevent compressor surge. In this embodiment, the anti surge line 14 is shown extending between a location downstream of the cooler 12 and a location upstream of the scrubber 8.

The anti-surge line 14, in typically arrangements, may have a length of between about 30m and 50m and a diameter of between 0.2m and 0.5m. The anti surge valve 16 is usually driven by a stepper motor or similar such that the quantity of recirculated fluid can be precisely controller. The anti-surge valve 16 would typically have a response time of between 1000ms and 1500ms between its fully closed position and fully open position.

The anti-surge controller 18 is connected to a plurality of sensors monitoring the input and output of the gas compressor 10. These sensors include:

• a flowmeter 20 for monitoring a flowrate Q through the compressor;

• upstream and downstream pressure sensors 22a, 22b for monitoring upstream and downstream temperatures P ₂ of the fluid being compressed; and • (optionally) upstream and downstream temperature sensors 24a,

24b for monitoring upstream and downstream temperatures T^ T ₂ of the fluid being compressed.

The inputs 26 and outputs 28 of the anti-surge controller 18 are illustrated in Figure 2. The anti-surge controller 18 receives, as an input 26, the flow rate Q through the compressor 10 (often measured as a pressure difference dP across a flow element of the flowmeter 20), the upstream and downstream pressures P ₂ of the fluid being compressed, and optionally the upstream and/or downstream temperature T i , T ₂ of the fluid being compressed. An anti-surge controller 18 may have a relatively low sample rate, such as about 1 to 10 Hz.

The anti-surge controller 18 provides, as an output 28, a command to control the anti-surge valve 16. Based on the inputs 26 received from the sensors 20, 22, 24, the anti-surge controller 18 controls the anti-surge valve 16 to recycle fluid from downstream of the compressor 10 to upstream of the compressor 10.

The anti-surge controller 18 controls the anti-surge valve 16 so as to prevent compressor surge. The anti-surge controller 18 is programmed with a pre-defined operation envelope and, if the flow rate leaves the envelope, then the anti-surge valve 18 will be opened in a controlled manner by an amount sufficient to bring the flow rate back within the envelope to avoid aerodynamic surge of the compressor 10.

Figure 3 illustrates a rotor shaft 30 of the gas compressor 10. The rotor shaft 30 is held in place by active magnetic bearings 32, 34 so as to permit contact- free rotation. The active magnetic bearings 32, 34 are controlled by an active magnetic bearing controller 36. The active magnetic bearings controller 36 regulates the current supplied to a plurality of electromagnets 33a, 33b, 35a, 35b forming the active magnetic bearings 32, 34.

Adjacent to each of the magnetic bearings 32, 34 is provided a position sensor 38, 40. Each position sensor 38, 40 is arranged to measure the position of the shaft 30, which is supplied to the bearings controller 36.

The inputs 42 and outputs 44 of the bearings controller 36 are illustrated schematically in Figure 4. The bearings controller 36 receives, as an input 42, for each bearing 32, 34, at least the position of the shaft 30 at the bearing 32, 34 and the electrical current that is currently being supplied to the coils 33a, 33b, 35a, 35b of the bearing 32, 34. The bearings controller 36 may optionally also receive the rotational speed of the rotor shaft 30. The bearing controller 36 may optionally directly measure the bearing flux applied by the respective bearing 32, 34.

The bearings controller 36 is arranged to provide, as an output 44, a flux command indicating the current to be supplied to each of the coils 33a, 33b, 35a, 35b of the bearing 32, 34.

The bearings controller 36 operates at a comparatively high frequency compared to the speed of rotation of the shaft 30. For example, for a typical compressor 10 operable at speeds of between 5,000 rpm and 10,000 rpm (one revolution every 6ms to 12ms), the bearings controller 36 may check and control the shaft position every 0.1 ms (a sample rate of 10 kHz). Such a controller might be suitable for compressors operable at a maximum speed of up to about 20,000 rpm (one revolution every 3ms). Higher speeds would require a higher rate of checking and controlling.

For each control operation, the bearings controller 36 determines the deviation of the shaft position from a reference position and adjusts the flux commands to the coils 33a, 33b, 35a, 35b to bring the shaft back to the neutral position. The bearings controller 36 may use PI control, for example, to maintain the shaft position in the neutral position. The bearings controller 36 may also be arranged to counter specific, known vibration behaviours based on the rotational speed of the shaft 30.

The compressor 10 is driven by a motor 50, as illustrated in Figure 5. The motor 50 would usually receive power from a three-phase power source 52. A motor drive 54 may use pulse-width modulation to regulate the power supplied from the power source 52 to the motor 50.

The operation of the motor 50 is usually monitored by a motor power analyser 56. The motor power analyser 56 may monitor the current and voltage supplied from the power source 52 and/or the modulated current and voltage supplied from the motor drive 54, as illustrated in Figure 5.

A dynamometer 58 may also be connected to an output shaft of the motor 50 to directly measure the torque and speed of the motor 50. The motor power analyser 56 may monitor the torque and speed of the motor 50, e.g. using the dynamometer 58.

The motor current, motor voltage, motor torque and motor speed of the motor 50 would typically be recorded by the power analyser 56 for analysis in the event of malfunction. In previous systems, this data would not normally have been supplied to any of the other control systems within the gas processing system for control purposes during operation.

The gas compressor 10 may comprise a cooling system (not shown) for regulating the temperature of the motor 50 of the compressor 10, and particularly the stationary parts of the motor 50. In one embodiment, the cooling system may be a gas cooling system. Such cooling systems are often relatively simple and are controlled by operation of a motor cooling (gas) valve. This valve would normally be controlled by a plant control system (SAS) controller.

Figure 6 illustrates the inputs 60 and outputs 62 for an exemplary cooling gas valve controller 64. The controller 64 receives, as an input 60, a temperature of a stator of the motor 50. The controller 64 is programmed with a pre-defined operating envelope and, if the motor stator temperature leaves the envelope, then the cooling gas valve will be opened or closed, as appropriate, to increase or decrease cooling of the motor 50 such that the motor 50 is maintained within the operating temperature envelope.

In addition to the illustrated controllers 18, 36, a typical gas compressor 10 may be provided with many other types of controller and monitoring system. In the case of a subsea gas compressor 10, each controller or monitoring system would usually be provided within a separate subsea pod, i.e. as a container housing the electronic components of the respective system.

As will be appreciated, the described anti-surge controller 18, the bearings controller 36 and the cooling valve controller 64 are purely reactive. That is to say, they assess the current situation of their specific element of the gas processing system and adjust their output responsive to that situation. Such reactive control systems operate sufficiently well when the gas processing system is in steady state conditions. However, these existing control systems typically respond poorly to transient situations as they cannot respond sufficiently quickly.

The present inventors have identified these failings and propose an improved controller 100 using predictive control to regulate the gas compressor 10 based on events occurring elsewhere in the gas processing system.

Particularly, the inventors propose a combined controller 100 that is a multiple in, multiple out (Ml MO) controller. The combined controller 100 may utilise tailored fuzzy control rules to process the inputs, which may be established using a “machine learning” algorithm. The combined controller 100 is faster and more accurate than existing individual system because it is able to analyse the behaviour of the compressor system as a whole, thereby permitting predictive control in addition to reactive control. Figure 7 illustrates exemplary inputs and outputs for such a combined controller 100.

As a first input 102, the combined controller 100 may receive data that would normally be sent to an active magnetic bearing controller.

The first input 102 may include a measurement of the current bearing flux generated by one or more or each of the active magnetic bearings. The bearing flux may be measured by one or more fluxmeters. Each fluxmeter may be adjacent a corresponding bearing.

The first input 102 may include a position of the shaft. The position of the shaft is preferably the position of the shaft in a plane substantially perpendicular to the axis of the shaft. The position of the shaft may be measured by a position sensor. The position sensor is preferably adjacent to an active magnetic bearing, e.g. an active magnetic bearing controlled by the combined controller 100.

The first input 102 may include the current supplied to an electromagnet of an active magnetic bearing, e.g. an active magnetic bearing controlled by the combined controller 100. Preferably the first input 102 includes the currents supplied to each electromagnet of the active magnetic bearing. The active magnetic bearing may be the same active magnetic bearing to which the position sensor is adjacent. Preferably the current supplied to the/each electromagnet is measured by a current sensor.

The first input 102 may include a rotational speed of the shaft. The rotational speed may be measured by a speed sensor.

The first input 102 may include an indication of whether a touchdown has occurred. A touchdown occurs when a rotating part of the compressor 10 comes into contact with a static part of the compressor 10. A touchdown may be detected using a contact sensor.

As a second input 104, the combined controller 100 may receive data that would normally be sent to an anti-surge controller.

The second input 104 may include an upstream pressure corresponding to a pressure of a fluid upstream of the gas compressor, preferably at a location immediately upstream of the gas compressor. The second input 104 may include a downstream pressure P ₂ corresponding to a pressure of a fluid downstream of the gas compressor, i.e. after compression, preferably at a location immediately upstream of the gas compressor. The second input 104 may include an upstream temperature T,

corresponding to a temperature of a fluid upstream of the gas compressor, preferably at a location immediately upstream of the gas compressor. The second input 104 may include a downstream temperature T ₂ corresponding to a pressure of a fluid downstream of the gas compressor, i.e. after compression, preferably at a location immediately upstream of the gas compressor.

The second input 104 may include a flow rate of a fluid through the gas compressor. The flow rate is preferably at least one of a volumetric flow rate of the fluid and a mass flow rate of the fluid. The flow rate may be input as a pressure difference dP across a flow element.

As a fourth input 106, the combined controller 100 may receive data that would normally be sent to motor power analyser 56.

The third input 106 may comprise a measurement of electrical power supplied to the motor 50. For example the fourth input 106 may comprise one or more measurement of a voltage and/or a current supplied to the motor 50. The voltage and/or current may be supplied as three-phase power. The

measurement(s) may comprise a line voltage and/or current, and/or a line-to-line voltage and/or current for each phase. The measurements may comprise one or more measurements of a regulated voltage and/or current supplied to the motor 50, for example where the power supply is pulse-width modulated.

The third input 106 may comprise a measurement of the mechanical power output by the motor 50. The measurement may comprise a measurement of a motor torque and/or a measurement of a motor speed.

The third input 106 may include other variable speed drive (VSD) parameters from the motor 50.

As a fourth input 108, the combined controller 100 may receive data that would normally be sent to motor cooling gas valve controller 64. The fourth input 108 may comprise a temperature of the motor 50. Particularly, the temperature may be a temperature of a stator of the motor 50.

As a fifth input 110, the combined controller 100 may receive other condition data representing the compressor 10 and/or the gas processing system. Such condition data may include one or more of: data from one or more accelerometers (e.g. measured in g), data from one or more velocity probes (e.g. measured in mm/s), data from one or more dynamic pressure probes (e.g. measured in bar), data from one or more flow meters (e.g. measured in Am ³ /h or kg/s), data from one or more hydrophone, and data from one or more microphone. The condition data may include fibre data or data from a proximity probe.

The fifth input 110 may include fluid phase data, such as a liquid/gas ratio of a fluid upstream of or inside the compressor 10.

As a first output 112, the combined controller 100 may output a flux command to one or more electromagnetic coils 33a, 33b, 35a, 35b of an active magnetic bearing 32, 34. The flux command may be controller so as to maintain a shaft, such as the shaft 30 of the gas compressor 10, in a neutral position.

In a conventional active magnetic bearing controller 36, the control loop gain would be fixed. However, as a second output 114, the combined controller 100 may set a (variable) control loop gain. The second output 114 may not be output as a signal from the controller 100, but may instead be merely an internally calculated value. The control loop gain 114 may be used as a parameter in determining the flux command of the first output 112. The control loop gain 114 may be equated to the“stiffness” of the magnetic bearing 32, 34. For example, a high gain results in a stiff bearing, whereas a low gain results in a soft bearing.

As a third output 116, the combined controller 100 may provide a command to a flow control valve. In this embodiment, the flow control valve serves the same function as the anti-surge valve 16 is a conventional system. However, it is referred to by a different name to avoid confusion with the hot bypass valve, which is described below and also provides an anti-surge function.

The flow control valve may be controlled by comparing compressor operating conditions against a surge line on a compressor map. A surge line is a line indicating the compressor conditions when compressor surge would be expected to occur. A typical compressor map used for this purpose may plot compressor pressure ratio against flow rate. However, other related properties may also be measured. Typically, a surge margin is applied, such that if the compressor operating conditions fall within the margin, action will be taken to avoid surge, such as opening the flow control valve.

The combined controller 100 may adjust the surge margin (or in some embodiments may modify the surge line directly) based on conditions monitored elsewhere in the system. The surge margin may be optimized for more efficient compressor control (e.g. a narrow surge margin) or for more efficient transient response (e.g. a larger surge margin), as required. For example, the compressor could be controlled to operate in a state further away from the surge line than would normally be required, thereby providing more time for the rotating machine to react should a transient state cause the rotating machine rapidly approach a surge condition.

In addition to the flow control valve, the present embodiment may be provided with a hot bypass valve (not shown). The hot bypass valve may be provided in parallel with the flow control valve, but connecting to a location downstream of the compressor 10 and upstream of the cooler 12 at one end and to a location downstream of the scrubber 8 and upstream of the compressor 10 at the other end. The hot bypass valve may be a fast-acting valve having a fully closed position and a fully open position. The hot bypass valve may have a response time from fully closed to fully open of less than 500ms, for example in the region of 200ms to 400ms. The hot bypass valve may have a diameter smaller than that of the flow control valve, for example between 0.05m and 0.15m. The purpose of the hot bypass valve may be to allow a bypass line to be rapidly opened, thereby permitting full or partial recycling of the compressor output.

As discussed above, an anti-surge valve 16 is precisely controllable to maintain the compressor 10 within the operating envelope. However, it might have a response time from fully closed to fully open 1000ms to 1500ms. In the case of a severe axial surge, such as might occur in the event of complete loss of compressor power, this response time would not be sufficient to prevent surge of the compressor 10. In such situations, the hot bypass valve may be opened to prevent surge of the gas compressor 10.

Thus, as a fourth output 118, the combined controller 100 may provide a command to the hot bypass valve.

As a fifth output 120, the combined controller 100 may provide a command to a motor cooling valve, such as the motor cooling gas valve described above.

A conventional gas (or multiphase) compressor 10 uses a“balance piston” to minimize the net axial force generated by components within the compressor 10. A balance piston is often located at the discharge end of the compressor shaft 30 and includes a rotating element with a predetermined area and an extended rim for sealing. One side of the balance piston is exposed to discharge pressure and the other side of the balance piston is vented, normally to suction pressure. The differential pressure across the balance piston acts to develop a thrust force opposite that generated by the impellers. By careful selection of the area of the balance piston, the thrust force can be sized to negate the axial force generated by the impellers. However, it is difficult to design this balance piston for all operating conditions.

The inventors propose to provide a flow control valve that controls a back pressure in a balance piston line. This valve may change the back pressure and hence may vary the axial force, ideally down to 0 net force. In some embodiments, the valve may be controlled to preload the shaft in a direction opposite to any expected impulse during a transient. Thus, as a sixth output 122, the combined controller 100 may provide a thrust pressure command to control the back pressure of the balance piston.

Figures 8 and 9 illustrate an exemplary transient situation in which a total loss of electrical power to the motor 50 of the gas compressor 10 occurs. In this situation, the compressor 10 will steadily lose power and the pressure differential between the upstream and downstream fluid can cause a low, zero or even negative flow rate through the gas compressor 10, which will cause it to go into surge.

Figure 8 shows the vertical displacement of a rotor shaft 30 of the gas compressor 10 with time. As can be seen, in the illustrated example, a significant transient occurs 800ms after the loss of power when the compressor 10 goes into surge. However, there is little to no warning of this transient based solely on measurement of the position of the rotor 10. If no preventative action is taken, then this transient could cause the compressor shaft 30 to touch down, making direct contact with the magnetic bearings 32, 34 or other parts of the housing. This can cause significant damage, requiring costly repairs and incurring delays before the compressor 10 can be restarted.

Figure 9 shows the vertical displacement of a rotor shaft 30 of the gas compressor 10 as well as the speed of the motor 50 of the gas compressor 10, each varying with time during the same transient event. Figure 9 shows the second immediately preceding the loss of power and the second immediately following the loss of power, with the time on the x-axis being shown in seconds relative to the loss of power event.

In this Figure, an under voltage occurs initially about 600ms before the loss of power event, causing the compressor rotor speed to decrease by about 50rpm over this period. Then, total loss of power occurs, whereupon the compressor rotor speed begins to rapidly reduce. As shown, the compressor rotor speed decreases by about 1400rpm to about 5100rpm over a period of about 800ms. When it reaches this speed, the gas compressor 10 goes into surge causing a transient, which causes a deflection of the shaft 30 of the gas compressor 10.

In the situation illustrated in Figures 8 and 9, a conventional active bearing controller 36 would likely be unable to constrain the rotor shaft 30 against the transient, and so contact of the rotor with the stator would likely occur, which could cause damage to the compressor 10.

Furthermore, although the under voltage first occurs almost 1400ms before the transient, it would likely be detected by a conventional anti-surge controller 18 much later because the conventional controller 18 behaves only reactively based on flow rate and pressure rise across the compressor. For example, a conventional anti-surge controller might only start to attempt to counteract the surge some time after the flow rate and pressure begins to rapidly drop. However, this occurs only 800ms before the transient and a conventional anti-surge valve 18 cannot respond sufficiently quickly to prevent the surge.

When using the combined controller 100, however, contact between the rotor and stator might be avoided. First, by employing predictive control, the combined controller 100 can identify the under voltage by directly monitoring the power supplied to the rotor. This allows the combined controller 100 to take preparatory action in case a total loss of power occurs. For example, the flow control valve may be opened greater than required in preparation for a potential loss of power and/or the gain of the active magnetic bearing may be increased to minimise deflection of the rotor shaft 30.

Furthermore, when the total loss of power occurs, the combined controller 100 can quickly react by opening the hot bypass valve to rapidly open the anti surge line 14 and prevent the compressor 10 going into surge.

Thus, as will be seen, by employing predictive control, the combined controller 100 can take effective steps to minimise the risk of damage to the compressor 10 from transient behaviour of the gas compressor system.

The above arrangement combines multiple control function into a single, combined controller 100 to thereby obtain a system with higher availability, less complex, lower cost, synchronous data collection, quicker response time, and reduced failure modes. Particularly, the combined controller 100 provides a high sampling, synchronous data algorithm with a fast output that allows the controller 100 to react to an abnormal situation. Within an implemented algorithm combining all relevant information, it is possible to predict the future behaviour of the gas compressor 10 and to send pre emptive output signals.

Furthermore, the above control system is particularly advantageous for compression of multiphase fluids because there are a greater number of parameters that must be accounted for to maintain smooth operation of the gas compressor 10.

Also, particularly in the context of a subsea gas processing system, the use of a single combined controller 100 reduces the number of subsea pods required for control systems, which are expensive components in terms of both installation and maintenance.

The above example illustrates one transient situation where the illustrated combined controller 100 could protect a compressor 10 from damage. However, it will be appreciated that other predictive and reactive control schemes may be implemented to assist in different transient situation.

In addition to transient situations, such as that illustrated above, the described combined controller 100 may provide advantageous control of the compressor 10 even during steady state operation.

The following example illustrates possible ways in which the combined controller 100 may be used to improve the control of an active magnetic bearing of the compressor 10.

As discussed previously, active magnetic bearings 32, 34 are typically controlled using a feedback control loop. The feedback control loop adjusts the flux supplied by each of the bearings 32, 34 based on at least a component proportional to a lateral deviation between the measured position of the shaft 30 and a neutral position. In a conventional system control that acts in a purely responsive manner, the neutral position and the gain used by the proportional component are typically predetermined properties.

The combined controller 100 advantageously controls the active magnetic bearings 32, 34 based on the behaviour of the system as a whole, thereby allowing control based on changes detected elsewhere in the system. Such control would not have been possible using the previous, localised bearings controller 36. The following examples illustrate ways in which the combined controller 100 may control the active magnetic bearings 32, 34. ln a first example, a speed sensor may detect variation in the speed of the compressor 10. In a second example, the speed sensor may detect a rapid speed reduction of the compressor 10. In a third example, a motor sensor may detect an external power trip. In a fourth example, a position sensor may detect a large axial displacement of the shaft 30. In a fifth example, a change may be detected in a differential pressure across the compressor 10, which may be indicative of a change in flow rate through the compressor 10.

The above changes may be indicative of a change in use of the compressor 10. Responsive to this change, the controller 100 may change the shaft neutral position, increase the control loop gain, or decrease of control loop gain, as appropriate.

In a sixth example, a change in the suction and discharge pressure may be detected. In this situation, the controller 100 may increase the control loop gain or decrease of control loop gain, as appropriate.

In a seventh example, a current sensor may detect a change in the coil current of the active magnetic bearings 32, 34. In response to an unexpected change in coil current, the controller 100 may change between a symmetric coil current in the bearing 32, 34 to an asymmetric coil current in the bearing 32, 34.

In practice, it is expected that the combined controller 100 will incorporate “machine learning” algorithms. That is to say, the combined controller 100 will be pre-programmed with initial relationships and boundary conditions, and the machine learning algorithms in the combined controller 100 will alter these as“experience” is obtained throughout the lifetime of the compressor system.

In use, the combined controller 100 will record data in real time from all of the inputs 102-110. It will then apply the machine learning algorithm to analyse the data. This algorithm will continuously assess the optimum control strategy for the compressor system based on its current operating conditions and environment.

This control may protect the compressor 10 from damage, such as illustrated above, as well as improving overall efficiency of the compressor by analysing system-wide performance instead of performing highly localised control.