Field of the Invention

Responsive loads are described, particularly refrigeration (including air-conditioning) loads, and methods by which they may be controlled to take advantage of fluctuations in energy availability and/or to facilitate operation of an electrical grid or network.

Background to the Invention

Typically, a refrigeration load, e.g. a compressor within a refrigerator or air conditioning unit, operates in an ON/OFF cycle to keep temperature within an allowable range between a maximum temperature and a minimum temperature. A refrigerator or refrigeration plant or air-con unit or building or building complex may have many such compressors. When a compressor is in an OFF state and the refrigerator temperature (monitored by a thermometer) rises and approaches the maximum operating temperature, the compressor is switched ON to reduce the temperature. When the compressor is in an ON state and the temperature falls and approaches the minimum operating temperature, the compressor is switched OFF to allow the temperature to rise.

Compressors consume electric power. It is known that the cost of power may depend on the demand and supply of power which can vary with time, for example during different times of a day. It is also a problem that electricity generators have to provide power capacity to meet peak demand, and provision of capacity is expensive. Subject to such power pricing schemes or constraints on peak available power, it may be advantageous to provide a responsive load, such as a refrigerator, in which it is possible manipulate the ON/OFF cycles of compressors, in order to enhance synchronisation between their ON/OFF behaviour and temporal variations in energy availability. For example, in the case of a peak in the supply of energy and hence the availability of less costly energy, the temperature of a refrigerator may be driven down by turning on the compressors, as long as the refrigerator is not cooled to below its minimum operating temperature. Similarly, in the case of a peak in the demand of energy, the temperature of a refrigerator may be allowed to rise by turning off the compressors as long as the refrigerator does not warm up above its maximum operating temperature. Therefore, refrigeration loads with responsive capabilities are preferred. Peaks and troughs in power supply or demand may or may not be predicted or well known in advance. In the case where prior knowledge is available, it may be advantageous, for instance, to drive the temperature down ahead of a predicted trough in energy supply or a predicted peak in energy demand, during which power cost may be higher. Similarly, it may be advantageous to let the temperature rise ahead of a predicted peak in supply of energy or a predicted trough in energy demand, during which energy price may be lower.

A responsive load such as a refrigerator containing one or more refrigeration loads such as compressors may be controlled by a central server. Each individual compressor may be controlled by means of a local control loop or local controller, and may be monitored by a local temperature sensor (e.g. a thermometer or thermostat). In addition, one or more routers may be present in the system, to communicate with the refrigeration loads and the server; the server may issue instructions, based on high-level aims, and override the local control loop(s); for instance, the router(s) may receive instructions from the server and cause the local control loop(s) to respond. In an embodiment, local controllers may communicate via the router(s) to coordinate their behaviour on a peer-to-peer basis.

Aspects of a power distribution control system including responsive loads are discussed in previous patent applications GB 1414724.3 and PCT/GB2014/052542.

In the course of operation, a typical compressor in a refrigerator has "dead times" during which it is unresponsive and cannot be turned on or off. For example, once a compressor is powered off, it may not be possible to turn it on for a period of time immediately afterwards. This is, for example, because of the need for the pressure in the system to settle down, in order to avoid damage to the compressor or the system or otherwise. Similarly, once a compressor is turned on, it may not be possible to turn it off for a period of time immediately afterwards. Other thermal loads such as a combined heat and power (CHP) plant or a heat pump can similarly experience non-responsive dead times (typically immediately after turning on or turning off). Summary of the Invention

According to the present invention, a method of providing a responsive load is provided that comprises: providing first and second thermal loads, each having a dead time during which it is not responsive; controlling the thermal loads by means of local control loops in response to at least one local thermostat; storing, at a server, information relating to the dead times for the thermal loads; and enabling the server to intervene in the operation of the responsive load and override the local control loop of the first thermal load so as to advance or delay the dead time of the first thermal load and thereby avoid, reduce or mitigate overlapping dead times. In one aspect, enabling the server to override the local control loop causes the first thermal load to advance or delay the dead time of the first thermal load so that it does not entirely coincide with the dead time of the second thermal load. In another aspect, the step intervening may comprise causing a battery to provide power to the load or to draw power from a source supplying the load selectively during overlapping dead times.

According to a second aspect of the invention, a responsive load system is provided comprising a plurality of thermal loads, each having a dead time during which it is not responsive; means for controlling the plurality of thermal loads by means of local control loops in response to at least one local thermostat; a server for storing information relating to the dead times for the plurality of thermal loads; and means for enabling the server to intervene in the operation of the responsive load and override the local control loop of a first of the plurality of thermal loads so as to advance or delay the dead time of the first thermal load and thereby avoid, reduce or mitigate overlapping dead times. In one aspect, the means for enabling the server to intervene in the operation of the responsive load and override the local control loop causes the first thermal load to advance or delay the dead time of the first thermal load so that it does not entirely coincide with the dead time of a second of the plurality of thermal loads. In another aspect, the means for avoiding, reducing or mitigating overlapping dead times may comprise a battery under control of the server.

According to a third aspect of the invention, a method of providing a responsive load is provided comprising: monitoring, at a server, future times tl and t2 between which a responsive load will be required; providing a thermal load (e.g. a refrigeration load or a heat pump) having a compressor that has a dead time during which it is not responsive; controlling the compressor by means of a local control loop in response to a local thermostat; storing, at the server, information relating to the dead times for the compressor; and avoiding, reducing or mitigating anticipated reduction in responsiveness of the load between times tl and t2. In one aspect, the step of avoiding, reducing or mitigating may comprise enabling the server to override the local control loop so as to advance or delay a dead time until it is outside the time tl to t2. In another aspect, the step of avoiding, reducing or mitigating may comprise switching a battery to provide power to the load or to draw power from a source supplying the load between times tl to t2.

The parameters in relation to this so-called dead time (for example its duration) that takes place immediately after turning a thermal load on or off, are known to the server controlling the responsive load and the thermal loads. They may be programmed into the server, or the server may be provided with means for discovering them. For example, if the dead time for a compressor is not already known, it may be detected or measured, and subsequently stored at the server. In most cases, the dead time is constant in duration for each respective compressor, but it may be necessary from time to time to detect, measure and update the dead time.

In view of the presence of dead times in the ON/OFF cycle of a typical thermal load in normal operation, and in view of variations in energy cost in time, it is advantageous for a responsive load, such as a refrigerator, to remain responsive ahead of predicted energy supply peaks or troughs. In other words, it is advantageous to make sure some or all of its compressors are not within their respective dead times when the refrigerator is required to respond to a peak or trough in electric power supply. Therefore, the ability to manipulate the cyclical ON/OFF behaviour of thermal loads improves overall flexibility of the system. In one aspect, the manipulating can be achieved by delaying or advancing switching on or off one or more compressors such that one or more dead times do not coincide with a time during which the responsive load may be needed. In one aspect, this entails enabling the server to override the local control loop so as to advance or delay the dead time.

In an alternative embodiment, it is possible to fill compressor dead times with short times of battery operation if battery assets are available in the system.

Optionally, especially in the case where the precise times of power supply peaks or troughs are not well known in advance and in the case where there are a plurality of compressors (thermal loads) in the responsive load, it is advantageous to control and coordinate the ON/OFF times of the plurality of thermal loads such that their dead times do not coincide, and some or all of the thermal loads remain responsive at any time or at a specific time. In one embodiment, this can be achieved by staggering the ON/OFF cycles of the thermal loads. As long as the operating thresholds of the refrigerator (or CUP plant or air conditioner or heat pump or other thermal load) are not exceeded or violated, i.e. the temperature is kept within a range between the pre-determined maximum and minimum values, the presently disclosed system of a responsive load with thermal loads and method of providing such a system can take better advantage of variations in power supply or demand and power cost by allowing improved responsiveness through reducing or eliminating the impact of dead times of thermal loads.

It is noted the system will also work, beyond thermal loads such as refrigerators and CUP plants, for any process or plant that has an in-built energy store, such as a building's thermal capacity, coolant stored in an air conditioner's on-site refrigerant pipework, or hot water in a CUP or other heating plant site. Effectively, the presence of the inherent energy store allows the system to manipulate the ON/OFF times of the assets indirectly, without influencing the outcome of the primary process.

Brief Description of the Drawings

Fig. 1 illustrates an example of a system comprising a responsive load 100.

Figs. 2a and 2b are exemplary representations of the ON/OFF cycle of a refrigeration load during operation.

Fig. 2c shows an exemplary representation of the ON/OFF behaviour of a refrigeration load along with variations in power supply and temperature.

Figs. 3a and 3b illustrates the ON/OFF cycles 310-314 of three refrigeration loads 110-114 in a responsive load 100. Fig. 4 illustrates an embodiment in which battery assets are available in the system comprising a responsive load 100. Detailed Description of Preferred Embodiments

Fig. 1 illustrates an example of a responsive load 100. The responsive load, e.g. a refrigerator bank (or air conditioner or heat pump), comprises a plurality of refrigerators 105-107 (or air conditioner units or heat pumps), each having a compressor 110-114, a temperature sensor or thermostat 120-124, and a local control input 130-134. Each compressor is under control by a local control loop comprising the local controller 102 and local area network 104. The temperature sensors feed their measurements to the local controller 102 via the network 104; the local controller 102 controls the inputs of the compressor 110-114 through the network 104. Note that in a single control loop there may be multiple compressors for a given temperature sensor or multiple temperature sensors for a given compressor. The refrigerators 105-107 are illustrated as being under common control of a local controller, but they may be stand-alone, each under control of its own local controller, with each local controller separately communicating with the server 30 through a respective router. There may be additional refrigerators or refrigerator banks allocated to the server 30 via routers 11 and 12. In the case of a CHP plant or other thermal load, there may be other active (driven) elements in place of the compressors 110-114.

Coupled to the local area network 104 is a router 10 that is coupled over a public IP network 20 to server 30. The server is connected to the electrical power grid 40. Compressors 110- 114 are also connected to the power grid 40. In the embodiment illustrated in Fig. 1, the server and router(s) are not part of the refrigerator or responsive load itself, but are part of a larger control system.

In normal operation of the refrigerator 100, the local controller 102 provides commands to the inputs 130-134 to drive the compressors 110-114 based on the temperature readings provided by the sensors 120-124, in order to keep the temperature of the refrigerator within the acceptable operating temperature range. For instance, if the temperature approaches the maximum temperature from below, when the temperature crosses an upper set point (or its rate of temperature rise crosses a threshold, or some combination of both) the local controller 102 switches on one or more compressors to reduce the temperature. Similarly, if the temperature approaches the minimum temperature from above (e.g. when it drops below a lower set point or its rate of fall crosses a threshold), the local controller 102 switches off one or more compressors to allow the temperature to rise. If the allowable temperature range is exceeded, the contents of the refrigerator 100 may be damaged and have to be destroyed. The upper and lower set points and the maximum and minimum temperatures are local constraint parameters. This ON/OFF cyclical pattern in the operation of the compressors 110-114 is a basic and typical local operation of the responsive load 100, and variations on top of this are possible.

The ON/OFF behaviour can be in response to the pattern of supply and demand of electric power and its cost. For example, the local controller 320 may have timing instructions to preferentially drive the refrigerators during periods of off-peak (low-cost) power. The local controller may be pre-programmed to control the ON/OFF behaviour in a certain predetermined way on a regular basis, for instance based on recurring electric power cost variation patterns.

Router 10 receives parameters and/or instructions from server 30 to influence the operation of the responsive load 100. In case of a conflict between these parameters or instructions and the pre-programmed parameters in the local controller 102, the former may override the latter. This may be because of the presence of some higher priorities important to the overall operation of the system, or other high-level aims. In an embodiment, this may involve the arrival of an unexpected peak or trough in energy supply or demand, leading to an unexpected price increase or decrease, which is to be advantageously taken advantage of or avoided.

In the above, server 30 may send parameters and/or instructions to router 10 either directly, or via a local intermediary or proxy, such as a routing server, to act on its behalf; the latter possibility may advantageously help to mitigate lost communications between the central server and the routers of the individual assets. Fig. 2a shows an exemplary representation 200 of the ON/OFF cycle of a compressor (refrigeration load) during typical operation. The compressor may either be in the OFF state or be in the ON state. The duty cycle of the load is the percentage of one period in which the compressor is ON. Its period is the time it takes to complete an on-and-off cycle. At some point in time, the compressor is switched on 202, and there may immediately follow a period of dead time ("ON dead time") 204 during which it cannot be turned off again. At another point in time, the compressor is switched off 206, and there may immediately follow a period of dead time ("OFF dead time") 208 during which it cannot be turned on again. The compressor may only be turned on again 210 after the dead time 208 has ended.

There may be reasons why it is desirable to advance a turning ON time of the compressor in a cycle. This is illustrated in Fig. 2b. The turning ON time is brought forward 220 from the anticipated or previously determined ON time 210 to the new ON time 212. The ON time cannot be advanced 222 to a point in time before the dead time 208, resulting from the previous turning OFF of the compressor, is over.

In steady-state operation, the duty cycle and period of the control loop of a compressor may be constant. The period can be shortened without affecting long-term temperature control, or the phase can be adjusted. The period can be lengthened, but this may give rise to greater temperature fluctuations. Adjusting the phase may cause a short-term change in temperature, but this is acceptable if the change falls within acceptable limits.

Fig. 2c illustrates an example of bringing forward a turning ON time and its effect. An example of a reason for advancing a turning ON time is the presence of a peak 240 in electric power supply which may result in cheaper power. When the compressor is turned on ahead of its anticipated or previously determined ON time, the temperature of the refrigerator is driven below its normal range using this cheaper power; this allows more room for the temperature to rise naturally in a future period of higher-cost power when the compressor is preferably in an OFF state. Overall, this type of control scheme may significantly reduce the running cost of the refrigerator.

As a result of advancing 220 the turning ON time of the compressor, as illustrated, the temperature variation in time changes from exemplary curve 230 to curve 232. As a direct or indirect consequence of this alteration, the server may or may not need to adjust the future turning ON/OFF times of this compressor or other compressors. The refrigerator may, for example, be allowed to return to a steady state under its normal local closed-loop control without intervention from the server. The system makes sure that the temperature does not go below a pre-determined minimum allowed temperature 234 or above a maximum allowed temperature 236. Another consequence of moving the turning ON (or OFF) time of a compressor is its dead time is also moved. This may also require the future turning ON/OFF times of this or other compressors to be adjusted. For example, a one-off phase adjustment may simply bring forward the ON time and the next OFF time by the same amount. Fig. 2c serves to illustrate only one possible "event" (a power supply peak) in response to which the ON time of a compressor is altered. More complex situations are possible.

In another example, a turning ON time of a compressor is delayed by the server and controller in response to a first event, e.g. the detection of a peak in demand for electric power. Along with a phase delay, the ON dead time is also delayed, during which the compressor cannot be turned OFF. This overlap with the ON dead time of another compressor. This may impact of the responsiveness of the refrigerator load in response to a second event, e.g. a sudden increase in power demand, or a sudden decrease in power supply, e.g. due to a technical fault in the power grid, that requires turning off one or more compressors promptly. This demonstrates the advantage of a method to coordinate a plurality of refrigeration loads such that their cycles are not synchronised in terms of their dead times, so that responsiveness and flexibility are enhanced, in particular to the benefit of the power grid operator who is seeking to smooth out peaks and troughs in demand by the deployment of responsive loads.

In effect, when the plurality of compressors or loads are operated such that their ON/OFF cycles are spread out more evenly across time, they can be seen externally as one big compressor with constant power loading, not as a series of compressors being switched on and off on demand. On the other hand the compressors can be completely independent (e.g. belong to independent systems, locations or clients), as long as the electrical power supply to the compressors can be aggregated as a single load. This can be done, for example, by physically connecting the power supplies to a single supply point or by using a server to virtualise and aggregate compressors.

Fig. 3a illustrates the ON/OFF cycles 310-314 of the three refrigeration loads or compressors 110-114 in the refrigerator 100 in Fig. 1, as an example of how they can be managed to reduce, minimise or eliminate the impact of dead times. Situations with a plurality of more than three such refrigeration loads are possible, e.g. 10 to 20 such loads in a string under control of a common server. For clarity of illustration, only the dead times resulting from turning off each compressor ("OFF dead times", shaded) are illustrated; the dead times resulting from turning on each compressor are omitted. Also, the duration of unaltered ON/OFF states and the duration of OFF dead times are shown as being identical among the three compressors, but this is not necessarily the case.

It is noted that in Fig. 3a, the operation ON/OFF cycles 310-314 of the three compressors 110-114 are staggered, in terms of their ON/OFF times and OFF dead times. In other words, an OFF dead time of any compressor does not overlap or coincide with an OFF dead time of any other compressor in the given set that is illustrated. (Note that these compressors may be part of a larger string of compressors under control of a common server, in which there may be compressors having overlapping dead times, but the server will operate to distribute the dead times to reduce or minimize overlapping.) Therefore, in a case where it is desirable to drive the temperature of the refrigerator 100 lower, for example due to an unexpected peak in power supply and lower power cost, the refrigeration load as a whole remains responsive, since at least one compressor is not within an OFF dead time and hence is able to be turned on. In the example illustrated in Fig. 3a, at time Tl, even though compressor 110 is in an OFF dead time and hence cannot be turned on, and compressor 112 is already in an ON state, compressor 114 is in an OFF state and not within a dead time, and so it can be turned on to drive the temperature of the refrigerator 100 lower if required.

Altering one or more of the turning OFF times of one or more compressors affects how many compressors are in an ON state, and OFF state or within an unresponsive dead time. Dead times may overlap. The server may further adjust the ON/OFF behaviours of the compressors to allow them as a whole to remain responsive in the future. Fig. 3b illustrates a different set of ON/OFF cycles 310-314 of compressors 110-114 in relation to their ON dead times. Again, for clarity of illustration, OFF dead times are omitted, and the duration of unaltered ON/OFF states and the duration of OFF dead times are identical among the three compressors.

As for Fig. 3a which relates to the OFF dead times, the operating cycles illustrated in Fig. 3b are staggered such that an ON dead time of each compressor does not overlap or coincide with an ON dead time of any other compressor in the given set. In other words, in the case where it is permissible for individual refrigerators in the refrigerator bank 100 to warm up (permissible within the constraints of the local control loop), the refrigeration load as a whole remains responsive, as at least one compressor is not within an ON dead time and hence is able to be turned OFF. In the example illustrated in Fig. 3b, at T2, compressors 110 and 112 are in an ON state and out of an ON dead time and are able to be turned OFF responsively.

As mentioned above, this method of enhancing flexibility and responsiveness in a plurality of refrigeration loads within a responsive load such as a refrigerator bank can lead to cost saving in terms of being highly responsive to fluctuating in power in the power grid. A significant advantage of employing coordinated ON/OFF cycles for compressors concerns the structural cost in obtaining a high level of flexibility and responsiveness. Without such coordination (e.g. by a server 30), a greater number of resources may be required to allow the refrigeration load as a whole to be responsive (to a given level) to the fluctuations of power demand and supply. For example, 20 compressors may be collected in a string, to create a responsive load, but using the method currently disclosed to attain a given level of responsiveness, the required number of compressors may be reduced, as one can now be confident there will be a portion of the compressors not locked in a dead time when required to respond to an unexpected event.

The central server 30 issues commands or parameters to the router and local controller(s) to control the ON/OFF cycles of the compressors, based on information it stores. This includes the ON/OFF states and dead times (for example their durations) of individual compressors. It may include readings from sensors 120-124 and other temperature parameters such as the maximum and minimum allowed operating temperatures and the upper and lower set points.

The server may have knowledge of the current periods and duty cycles of the compressors and/or it may be able to control the behaviour of the compressors by monitoring their states (ON, OFF or within a dead time), the temperature readings (such as the rate of change) and other parameters.

For example, the server 30 may tell a compressor to "carry on" in its current ON or OFF state and delay turning OFF or ON respectively, overriding the instructions of the local controller. This may be in response to a recent event in which one or more compressors controlled and monitored by the server has changed its state or behaviour, e.g. because a refrigerator has moved from its steady state because a door has been temporarily opened. Note that if a compressor is made to switch between an ON or OFF state at a lower frequency (longer cycle), the impact of dead times on the responsiveness and flexibility of the system as a whole is reduced, since dead times are typically constant in duration.

Where the server stores or has access to higher-level information relating to the parameters of the compressors, such as the length of their ON/OFF cycle, the server can plan and modify the behaviour of the compressors in advance, rather than making control decisions on a cycle- by-cycle basis.

The server may be aware of an anticipated or previously determined ON time or OFF time for a compressor or may predict such times, or it may simply operate to stagger the cycles of the compressors in the present. For example, if (as illustrated in Figs. 3a and 3b) the compressors have similar periods, the server may simply cause those periods to be staggered - i.e. in the event of one compressor being caused to switch ON or OFF at the same time as another compressor (or when another compressor is within its corresponding dead time), the server may override the local control loop and delay the switching of the second compressor. Alternatively, the server may predict the next transitions of the compressors (knowing their periods and duty cycles) and, in anticipation that a first compressor and a second compressor will have co-incident dead times, it may select one of them to have its switching delayed or advanced. If the server knows the temperatures and temperature limits of a first and a second refrigerator (or air conditioning units or the like), it can select which one to override. It can, for example, select the one that has greater margin for re-adjusting temperature to its steady state condition.

It has been already described that it is advantageous to reduce, minimise or eliminate the impact of compressor dead times. The server may further adjust the ON/OFF behaviours of the compressors to allow them as a whole to remain responsive when there is an unexpected peak or trough in demand or supply, for instance. One way to implement this is to simply stagger the ON/OFF cycles of different compressors to avoid or reduce overlapping of individual dead times. It is noted that another way to implement this (although it may appear contradictory on the face of it) is "lump" the dead times of two or more compressors together in one period of time. This may bring about the (immediate) onset of multiple overlapping compressor dead times making the system less responsive or non-responsive, but as a result of this, after this set of overlapping dead times are over, these compressors may not have to experience another dead time in a certain subsequent period of time. For example, if it is already known that there is a length of 30 minutes when there is definitely going to be no need for responsiveness, the ON/OFF cycles of compressors can be manipulated, using the current invention, to make them less responsive in the immediate or short term, but more responsive afterwards. This can be advantageous when one knows a specific time at/during which a specific response is needed or desirable. This alternative arrangement also achieves the technical effect of reducing, minimising or mitigating the impact of compressor dead times, though taking effect at a later time instead of immediately.

Fig. 4 illustrates an alternative embodiment of the responsive load system. In this embodiment, one or more battery assets 50 are available in the system. The one or more battery assets 50 are connected to the power grid 40 via a switch 51 under control of the server 30. This arrangement allows the possibility that, during the dead times of one or more compressors 110-114, the battery assets 50 can be operated, so as to: selectively provide power to the load during OFF dead times or during multiple overlapping OFF dead times; or draw power from the energy grid supplying the load, to charge the battery, during ON dead times or during multiple overlapping ON dead times. This may be in the form of filling the compressor dead times with short times of battery operation; this would apply when there is visibility of compressor breaks.

The above description of embodiments and examples is given by way of example only. Various aspects and embodiments of the invention can be combined. Various aspects and embodiments can be modified in accordance with other aspects and embodiments. The scope of the invention is not to be limited by details of the embodiments, but is defined in the appended claims.