Control method and system for operating electrical components

Abstract

To provide a control method and system for controlling operations of electrical components coupled to an electrical supply network to enable a plurality of components to simultaneously participate in two different power consuming and generating means.SOLUTION: A component is either a generator or a load of a type capable of storing energy internally and is operable over a range of power levels from 0 to a maximum power CD, (a) the component is operated for a certain period of time by adjusting the rate of energy transfer between the component and the supply network according to a reference value power function B(t), and (b) a high-speed operation power function F(t) is superimposed on the reference value power function and the component is simultaneously operated for a certain period of time so as to adjust the rate of energy transfer between the component and the supply network according to the high-speed operating power function F(t).SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a method and system adapted to control the operation of electrical components (both generators and loads) coupled to an electricity grid, in order to allow multiple components to participate simultaneously in two different electricity consumption or generation means. It is particularly relevant to types of electrical loads that have some mechanism for storing energy during peak price periods or periods where penalties are incurred for some other reason, and thus tend to operate to minimize energy consumption. It is also relevant to generators that are able to store and hold power, and that make it possible to schedule the time when the generated power is fed into the grid in order to maximize revenue. The present invention allows such implementations to be compatible with the provision of services that help regulate the supply and demand of electricity. [Background technology]

[0002] In any electrical power distribution network, it is important to coordinate the power fed into the network with the power taken from the network. Any imbalance, even in the short term, can result in a variety of problems ranging from reduced efficiency in power transmission and unpredictable fluctuations in power supply to more serious effects such as power outages.

[0003] Current electricity distribution networks, such as the UK's National Grid, adjust with sufficient precision to the second. Any imbalance between the amount of electricity generated and the amount of electricity consumed by the grid loads is evident in the deviation of the operating characteristics of the power supply from its intended setpoint. Monitoring such a parameter makes it possible to detect and therefore correct the imbalance. Typically, the parameter in question is the frequency of the electricity supplied by the grid, which is monitored for this purpose. In the UK, electricity mains are supplied at 50 Hz. If the collection of grid loads takes more power than is supplied, the frequency drops. Roughly speaking, this effect can be seen when an increase in the load on a generator causes it to run (spin) slower. Conversely, if the imbalance is caused by excess generation, the frequency will rise above the nominal value of 50 Hz.

[0004] There are several factors that can cause imbalances in the power grid. These factors can be on the supply side (e.g., a generator experiencing technical problems) or on the demand side (e.g., a demand surge during a televised sporting event). In either case, a correction can be applied by adjusting the power generated or consumed. Typically, to correct on the supply side, the grid has available a backup system of assets (generators) that can be brought online or disconnected from the grid as required. On the demand side, the grid also has a mechanism that allows at least one selected load on the grid to operate at a different power level. The response can be dynamic, operating in response to second-by-second changes in operating frequency, or non-dynamic, typically a separate service triggered in response to a predetermined frequency deviation. In the UK, the grid's conditioning service provider monitors frequency deviations itself. In other jurisdictions, the grid operator itself monitors imbalance signals and communicates with the provider, directly indicating the required conditioning service. Financial incentives are offered by network operators to generators and loads that are ready or able to provide regulation services.

[0005] With the transition from 100% coal-fired generation to renewable energy, the challenge of maintaining predictable supply levels becomes more difficult. Generally, only relatively unlikely faults will take power plants offline, but renewable resources are fundamentally less reliable. When there is little wind, wind farms are less productive. In particular, strong winds may require wind farms to be shut down. When it is cloudy, solar power production decreases, and hydroelectric power is also weather-dependent. Supply levels, like demand, become more difficult to predict. This drives increased demand for regulation provided by the demand side of the grid. Overall, there is a need for adaptive regulation systems that respond rapidly to imbalances.

[0006] At a practical level, power grids need to incorporate various mechanisms for power regulation. In the context of the present invention, a distinction is made between components of such mechanisms that are capable of operating with fully variable power consumption. Components include, for example, batteries and their binary elements (such as refrigeration units that can operate at only two power levels (on and off)). Both of these components can provide regulation services to the power grid, but the mechanisms are different.

[0007] One power regulation mechanism using a battery system is described in U.S. Patent Application Publication No. 2016 / 0013676. To address deviations in grid frequency from a nominal value, a battery increases or decreases the power supplied to (or drawn from) the grid. The battery has the capacity to store charge, and therefore energy, internally. Thus, measurements of the battery's state of charge (SoC) provide an indication of the battery's availability in response to grid imbalances. Generally, batteries participating in a frequency response service provide this service continuously if the battery's state of charge does not vary outside a predetermined range.

[0008] Another mechanism can be provided by groups of loads that can operate at reduced power for short periods without significant degradation in performance. Typically, the aggregate response of subgroups of available responsive loads contributes to supply and demand coordination. The subgroups are selected so that reduced power is required from individual loads only for limited periods. This mechanism allows a portfolio of electrical assets to participate in responsive services as a group, but the individual assets themselves do not meet the technical criteria set by the grid to provide such services.

[0009] Exemplary methods for facilitating the participation of binary loads in frequency response services are described in GB 2361118, WO 2006 / 128709, and WO 2013 / 017896 ("Open Energi's FFR Algorithm"). All of these documents describe selection algorithms that can be used to determine which loads from a large group of binary loads will provide the required response at any given time. The group of loads is controlled by an autonomous device that operates independently of external control. When the control device detects an imbalance in grid frequency, the control device varies the power consumed by the aggregated electrical loads to address the detected imbalance across the grid. By allowing the energy stored in the loads to vary, power consumption varies. That is, each load has an energy storage capacity that varies with the load's duty cycle. The energy storage capacity of the electrical loads allows for occasional adjustment of the load's energy consumption without significantly impairing the electrical load's operational performance. For each electrical load, a threshold of energy storage capacity applies when the electrical load is available to contribute to responsive load service.

[0010] Using the above prior art algorithm, the selection of electrical loads for delivering responsive load service in response to the detection of a grid imbalance is performed by using frequency triggering thresholds assigned to each load by a control device. The thresholds may vary smoothly or individually over time and may be determined taking into account several factors, such as the amount of energy stored in the load at the time a response is required, the amount of time elapsed since the load was last required to provide responsive load service, and the nature of recent fluctuations in grid frequency.

[0011] As the number and variety of electrical components available to participate in regulating services increases, a better balance can be struck between providing service to the grid and minimizing any adverse impact on the normal operating behavior of each component. Additionally, there are many industrial processes that draw large amounts of power from the grid, and where these could present opportunities for providing responsive load services, grid benefits would be significant.

[0012] However, there are serious economic barriers to implementing frequency response services for some components that may be connected to the grid, which arise from electricity pricing strategies. For example, these strategies generally make it more costly for loads to consume electricity at certain peak times. This encourages many loads, types that are capable of storing energy reserves, to engage in "peak price avoidance" or "revenue stacking" behavior. That is, the strategies increase their power consumption just before peak periods to allow stored energy to increase toward its maximum allowable value. During peak periods, these energy reserves are taken from the grid rather than electricity. The reduction in electricity consumption during peak periods allows for an overall reduction in the economic cost of operating the loads.

[0013] Similar considerations arise with respect to the type of generator that can store and store varying amounts of energy. For example, it is often desirable to utilize the energy storage capacity of a battery to ensure that power is delivered to the grid at a time when it is most financially advantageous to do so, and as with loads, this may be considered incompatible with frequency response service operation.

[0014] In a prior art battery storage system, U.S. Patent Application Publication No. 2016 / 0013676 describes an improvement to basic frequency response service that increases the likelihood that the battery state of charge will be maintained within a range in which response service can be provided. This is achieved by allowing the battery to be charged / discharged when the battery's state of charge leaves the central range and by making frequency response adjustments to the power level of the charging or discharging operation. However, this is an internal management system that is limited in its application to storage batteries and does not consider the impact of providing frequency response service on ideal state of charge management. [Prior art documents] [Patent documents]

[0015] [Patent Document 1] US Patent Application Publication No. 2016 / 0013676 [Patent Document 2] GB Patent Application Publication No. 2361118 [Patent Document 3] International Publication No. 2006 / 128709 [Patent Document 4] International Publication No. 2013 / 017896 Summary of the Invention [Problem to be solved by the invention]

[0016] At the same time, a need is seen for alternative implementations of systems that respond to network imbalances that allow for the involvement of components that operate according to some other power consumption scheme. [Means for solving the problem]

[0017] The present invention provides a method for controlling, depending on circumstances, the operation of a component connected to a power grid, the component being either a generator or a load of a type capable of storing energy therein. The method comprises the steps of: (a) operating a component over a period of time, the period of time adjusting the rate at which energy is transferred between the component and the grid in accordance with a reference power function B(t); (b) simultaneously operating the component over a period of time to adjust the rate at which energy is transferred between the component and the grid in accordance with a fast action power function F(t), such that the fast action power function F(t) is superimposed on the baseline power function; The reference power function B(t) is derived from the slow-operation power function D(t), and when adjusting the rate at which energy is transmitted between the component and the grid in accordance with the slow-operation power function D(t), the component will benefit economically from fluctuations in the grid's current electricity prices or fluctuations in values ​​that can be derived directly or indirectly from the grid's power supply; and Adjustments made to the rate at which energy is transferred between components and the grid in accordance with the fast acting power function F(t) are adjustments that respond to, for example, imbalances between power generation and power consumption through the power grid.

[0018] Using the present invention, a fast-acting power function F(t) is superimposed on a reference power function B(t), which is in turn derived from a slow-acting power function D(t). This allows a component (generator or load) to simultaneously follow two different power consumption measures, and determines the reference function B(t) such that the cost of deviating from the slow-acting power function D(t) is acceptable. Specifically, the slow-acting power function D(t) is determined by monetary forces within the electricity market. For example, its value can be the current price, leading to measures such as peak price avoidance. Alternatively, its value can be based on market derivatives, such as the amount of fines imposed on load operators who violate their network contractual obligations. Load operators then attempt and limit their actions to avoid or reduce the exposure of the fine.

[0019] 0~Maximum power C DFor variable speed generators or loads that are operable over a range of power levels, regulation in response to grid imbalances may be performed to maximize power C. D The reference power function B(t) can be set equal to the slow operating power function D(t), provided there is a high probability that the reference power function B(t) is significantly smaller than the maximum power C D Any range of δC D If not within (δ<<1), then B(t) can be set equal to the slow operating power function D(t), and if D(t) is within this range, then B(t) is set equal to δC D or (1-δ)C D It is set to one of the following.

[0020] Alternative embodiments of the method of the present invention are either "off" or "on" and thus have two power levels (0 and C D ) are intended for implementation with binary elements operable only at the low speed operating power function D(t) and also at the high speed operating power function F(t). Each of the binary elements is one of a group of components that work together to provide service responsive to network imbalances, and the responsiveness is distributed around the group so that each component provides only the responsive service intermittently with a high probability. In this embodiment, the baseline power function B(t) is derived from both the slow operating power function D(t) and also from past values ​​of the fast operating power function F(t).

[0021] Using this embodiment, the period preferably includes a series of consecutive decision periods, and for each decision period k=1 to N, the average value D of the slow operating power function D(t) over the decision period is calculated. k Following the determination of , the reference value function B(t) is derived using multiple steps, the multiple steps being: (a) Decision period T k dividing t into a series of consecutive subintervals; (b) Subinterval period s m (m=1~M), each of the subinterval periods s m At the start of the step, the component is not responsive to network imbalances, (i) The average total power transfer P between the component and the grid before the start of the subinterval period.k and determining (ii) The average total power transfer P determined in (i) of step (b) before the subinterval. k is the average value D of the slow operating power function D(t) over the decision period. k Compare with (iii)P k >D k If , set the reference value function B(t) to zero in that subinterval, and P k <D k In this case, the reference function B(t) is set to the maximum power C D and performing the steps of not responding.

[0022] In this way, the impact of F(t) is mitigated by determining the value of B(t) for each successive subinterval in a manner that takes into account the power consumption contributions of all past subintervals. Thus, over the determination period, deviations from the ideal slow operating power function are limited to those exhibited by a single subinterval.

[0023] Furthermore, the subinterval period s m At the beginning of the subinterval s, if the component responds to a network imbalance, m The reference power function B(t) is preferably m-1 The reference value of the power function B(t) is set to the same value as the reference value of the power function B(t).

[0024] In a second aspect, the present invention provides a method for operating a component connected to a power grid, the method comprising adjusting a rate of energy transfer between the component and the power grid in accordance with a power function P(t), the power function comprising two components: a fast operating power function F(t) and a reference power function B(t), the reference function B(t) being derived by the following steps: (a) Decision period T k The average value D of the low-speed operation power function D(t) over the k and deriving (b) Decision period T k a series of consecutive subintervals s m(m=1 to M) (c) Subinterval period s m (m=1~M), (i) Subinterval period s m If F(t) ≠ 0 at the start of the subinterval, then keep the value of B(t) unchanged (B m =B m-1 ), otherwise, (ii) Partial interval period T k Within the range and its subinterval period s m Before the start of k and determining (iii) subinterval s m Before the average total power transfer P determined in step (c)(ii) k is the average value D of the slow operating power function D(t) over the decision period. k Compare with (iv)P k >D k In this case, the reference value function B m Set to zero and P k <D k In this case, the reference value function B m Maximum power C D otherwise, leave the criterion function unchanged in the subinterval (B m =B m-1 ) steps.

[0025] 1. A method of controlling the operation of a non-battery component connected to a power grid, the component being either a generator or a load of a type capable of storing energy therein, the method comprising: (a) monitoring a physical parameter of the component indicative of the amount of energy stored therein; (b) deriving a proxy variable φ, where 0≦φ≦1, from the measured physical parameters, where the proxy variable φ represents the fraction of stored energy held in the energy reservoir at any one time; (c) controlling the operation of the component in accordance with a slow operating power function D(t) when the proxy variable φ has a value between an upper limit value and a lower limit value, wherein adjusting the rate at which energy is transferred between the component and the grid in accordance with the slow operating power function D(t) will cause the component to benefit economically from fluctuations in the grid's current electricity prices or in values ​​that can be derived directly or indirectly from the grid's power supply; (d) if the proxy variable φ has a value outside the range of the upper limit value or the lower limit value, operating the component at an operating power that returns the parameter to a value between the upper limit value and the lower limit value.

[0026] Using a proxy as an indicator of the energy stored in the load allows for the adjustment of component power consumption independently of adjustments based on detailed knowledge of the component's operation. In this way, the proxy is analogous to a battery's state of charge. This provides excellent flexibility in applying the power consumption function to all kinds of generators and loads. Furthermore, the proxy model is flexible when applied to systems where two or more parameters, appropriately, should provide an indication of the component's energy storage. In such cases, the overall proxy can be defined as the product of the proxy for each individual parameter.

[0027] In an alternative aspect, the present invention provides a method of controlling the operation of a component connected to a power grid, the component being either a generator or a load of a type capable of storing energy therein, the method comprising: (a) monitoring a physical parameter of the component indicative of the amount of energy stored therein; (b) controlling the operation of the component when the parameter has a value between the upper and lower limits according to the method described above; (c) if the parameter has a value outside the range of the upper limit or the lower limit, operating the component at an operating power that is in accordance with restoring the parameter to a value between the upper limit and the lower limit.

[0028] In this aspect, the present invention provides a convenient approach to ensuring that simultaneous operation of a component according to two power consumption measures does not cause the component to operate outside its acceptable parameters.

[0029] Preferably, the method is based on the use of a proxy variable φ (0≦φ≦1) derived for the component, the proxy variable representing the fraction of storable energy held in the energy reservoir at any one time and derived from the measured physical parameter, whereby limits for component operation are defined in terms of the proxy variable.

[0030] In another aspect, the present invention provides a system for controlling the operation of a component connected to an electrical power grid, the component being either a generator or a load of a type capable of storing energy therein, the system comprising: an industrial process controller configured to exercise direct control of the operation of the components and monitor process parameters; a central demand server configured to derive a slow operating power function D(t), wherein when adjusting a rate at which energy is transferred between the component and the grid in accordance with the slow operating power function D(t), the component will economically benefit from fluctuations in the grid's current electricity prices or fluctuations in values ​​that can be derived directly or indirectly from the grid's power supply; an indicator configured to provide a signal indicative of an imbalance in power supplied through the grid; a local device controller associated with the industrial process controller and the component, receiving a signal indicating an imbalance in the power supplied from the indicator via the network and receiving a slow operating power function D(t) from a central demand server; Deriving a reference power function B(t) from the slow operating power function D(t); Deriving a fast acting power function F(t) from a signal indicative of an imbalance in power provided through the network; and a local device controller configured to provide instructions to the industrial process controller to operate the component and adjust the rate at which energy is transferred between the component and the grid according to a fast acting power function F(t) superimposed on a baseline power function B(t).

[0031] In another aspect, the present invention provides a local device controller associated with an industrial process controller configured to exercise direct control of the operation of a component connected to an electrical power grid, the component being either a generator or a load of a type capable of storing energy therein, the local device controller comprising: receiving a signal indicating an imbalance in power supplied from the indicator via a network and receiving a slow operating power function D(t) from a central server; Deriving a reference power function B(t) from the slow operating power function D(t); Deriving a fast acting power function F(t) from a signal indicative of an imbalance in power provided through the network; providing instructions to an industrial process controller to operate the component and adjusting the rate at which energy is transferred between the component and the grid according to a fast action power function F(t) superimposed on a baseline power function B(t); A slow operating power function D(t) is derived so that the component benefits economically from fluctuations in the price of electricity distributed by the grid over the course of a day when the component adjusts the rate at which energy is transmitted between the component and the grid according to this function D(t).

[0032] The invention will now be described in detail, by way of example only, with reference to the accompanying drawings, in which: [Brief explanation of the drawings]

[0033] [Figure 1] 1 illustrates a power supply system incorporating a reactive load that can be operated with a power consuming means according to the present invention; [Figure 2a]1 is a representative graph of power consumption versus time for a collection of load devices or groups of devices that provide frequency response services to a grid. [Figure 2b] 1 is a representative graph of power consumption versus time for a load device implementing peak price avoidance. [Figure 3] 1 is a flow chart depicting the process steps involved in implementing dual power consumption measures for an electrical load. DETAILED DESCRIPTION OF THE INVENTION

[0034] An electrical power supply system, generally designated 10, is shown in Figure 1. The electrical power supply system 10 includes one or more electrical power generators 12 and a plurality of electrical loads 14, 16. The electrical power generators 12 provide electrical energy to the electrical loads 14, 16 via an electrical power distribution grid 18 (hereinafter "grid 18").

[0035] In this particular embodiment, the generator is not considered to provide service in accordance with the present invention. This is not because the generator is useless, but simply for convenience. Those skilled in the art will appreciate that the present invention can be implemented with a generator if it meets appropriate performance criteria. However, the following description focuses on load operation.

[0036] Although depicted together in FIG. 1 , the several power generators 12 are not considered to be the same type. One may be a coal-fired power plant, while another may be a wind farm, a hydroelectric generator, or any of several known systems capable of generating electricity and supplying it to the grid. Generally, each generator works with a grid manager to supply a set amount of electricity to the grid. This set amount may be adjusted according to the generator's contractual requirements to provide some regulation to the power grid.

[0037] Furthermore, electrical loads 14, 16 are inherently different. Generally, they draw power from the grid on demand, and only a limited subset are capable of providing responsive service and addressing grid imbalances. However, capable electrical loads may adapt, either individually or as a collection of multiple loads operating collectively to provide responsive load service, which adapts to help balance grid supply and demand.

[0038] With respect to electrical loads, the present invention is primarily concerned with loads that consume significant amounts of electricity. This includes, for example, industrial processes, electric vehicles, and heating and cooling systems. However, the present invention does not address the technical reasons why loads that consume less electricity cannot be implemented, making it unlikely that the economic benefits of such loads would outweigh the costs.

[0039] There is clearly a financial incentive for electrical loads to minimize the power taken from the grid when costs are high. As a result of increased consumption just prior to peak periods, loads with some form of internal energy storage capacity can utilize their accumulated energy reserves during peak price periods. A simple example would be a battery, allowing for fluctuations in the charge level it holds while preserving its ability to supply power to the grid. Another example would be a building management system (BMS) that maintains a building temperature within a range around a nominal set point. The system can therefore be thought of as storing energy when the building temperature is above the minimum value of this range.

[0040] The loads 14, 16 shown in FIG. 1 are industrial process loads having the capacity to store energy. Of course, other loads and load types may also be connected to the grid, but these are not shown. The industrial process loads 14, 16 are all associated with local industrial process controllers 20, 22. The industrial process controllers 20, 22 are generally responsible for controlling and monitoring the operation of the loads. To this end, the industrial process controllers 20, 22 include or are in communication with appropriately positioned detectors 24 that are configured to monitor parameters indicative of the amount of energy stored within the loads. This allows the process controller to ensure that the energy storage is maintained within acceptable limits throughout the process.

[0041] The details of the detector 24 depend on the particular industrial process serving the load 14. The detector 24 is necessarily adapted to measure a parameter that depends on the load characteristics, specifically the form in which energy is stored in the load. Using the above example, a battery would have a controller adapted to infer the instantaneous state of charge (SoC) from observing the voltage that charges or discharges the battery. On the other hand, the detector 24 in a building management system would include a thermometer, thermocouple, or similar device that measures the temperature of the building being controlled. With other industrial process loads where more than one parameter needs to provide a reliable indication of energy storage, it may be appropriate to have more than one detector 24 associated with each of the industrial process controllers 20, 22.

[0042] The difference between the two types of loads and controllers shown in FIG. 1 is that the first type 14, 20 is adapted to participate in the responsive load service being provided to the network 18. The other loads 16, 22 are not. In this embodiment, the responsive industrial load 14 is assumed to be a binary load within a group of responsive loads that collectively provide service. The loads 14 within a group can be the same or different. Regardless of the details of the group membership, whenever a response to an imbalance is required, a single load or subgroup of loads is requested to provide the response on behalf of the group. There are many selection algorithms known in the prior art that can be employed to determine which particular load or subgroup of loads will provide the response at any given time. A non-limiting example of the above is provided by Open Energi's FFR algorithm. Selection can be made based on past response requests made to each load, internal load parameters, and the degree of imbalance detected on the network. The details of these algorithms are not critical to the present invention, as they affect it. Most attempt to distribute response demands evenly among a group of loads, ensuring that each load incurs a similar cost in terms of degradation of normal operating performance. In such situations, the typical period of a response load "switch," i.e., the time required for a particular load in the portfolio to adjust its consumption as a result of a network imbalance, can take several minutes.

[0043] An algorithm for binary load operation suitable for use with the present invention is one that results in the impact of low utilization of each individual load within any "decision" period, as an inference of an even distribution of demand. In this regard, the decision period is longer than that of a typical responsive load switch, but significantly shorter than one day (24 hours). For example, a 30-minute decision period is appropriate for FFR. For a portfolio of binary loads spanning a long period (many decision periods), the requests made to the group as a whole for response service will always be a small fraction of the group's total electrical consumption. The present invention, which allows for the implementation of response service together with several other slow-acting electrical consumers, further requires that the fraction of time within any decision period during which an individual load is required to provide response service be short. This is true for the algorithms described in the prior art mentioned above, but is not typical.

[0044] For non-binary or variable speed loads, the requirements are different. Such loads continuously adjust their operating power to address network imbalances. Therefore, the loads are not part of a portfolio of loads where each load is only required to provide response intermittently. Rather, the underutilization condition in this case is to always provide response service using only a small fraction of the load's total electrical consumption.

[0045] Each of the responsive loads 14 and the industrial process controller 20 is associated with a local responsive load controller 26. The responsive load controller 26 is connected to and communicates with the process controller 20, and they work together. The local responsive load controller 26 includes or is connected to a frequency monitor 28, which is adapted to monitor the power frequency provided by the network. The term "local" is used to indicate that the controller 26 is physically near or even integral with the load 14 being controlled. With components intended to respond to network 18 imbalances in seconds, it is important to minimize communication delays. In some embodiments (not shown), multiple electrical loads 14 may be controlled collectively as a group by a single load controller 26.

[0046] While monitoring temporary fluctuations in frequency is currently the preferred approach to detecting power supply imbalances across the grid 18, alternative monitors adapted to detect fluctuations in other characteristics of the power grid 18 indicative of imbalances in the power delivered across the grid 18 are envisioned and may replace the frequency monitor 28. In alternative embodiments appropriate for markets such as the PJM grid in the United States and the AEMO grid in Australia, the frequency monitor 28 may be replaced by a receiver. Both of these grids self-monitor to detect imbalances, as determined by the grid operator. The algorithm for determining the degree of imbalance, or at least the degree of correction required, may be more complex than simply monitoring how much the grid frequency deviates from a nominal value. For example, it may include factors that consider the rate of change of frequency. However, once an imbalance is detected, the grid responds by sending a signal to any loads that require them to adjust their power consumption. The present invention allows for any responsiveness algorithm to be implemented that ensures, with a high probability, that any individual load will only be required to provide responsive service for a small fraction of the time within a determined period.

[0047] The local load controller 26 is adapted to send commands regarding load power consumption to the industrial process controller 20. If it requires an adjustment in power consumption, for whatever reason, this may be implemented by the process controller 20, provided that the load energy storage is maintained within an acceptable operating range. If the load energy storage is outside this range or if the adjustment in power consumption requested by the local load controller 26 is found to be outside this range, the industrial process controller 20 will not respond to the signal of the local load controller 26 and will then operate the load 14 independently. Furthermore, it will signal the local load controller 26 that the load 14 is not available to provide responsive load service.

[0048] The local load controller 26 is also adapted to monitor the network frequency as detected by a frequency monitor 28 or otherwise.

[0049] In this embodiment of the present invention, loads provide their response service according to an algorithm that assigns a trigger frequency to each load in the portfolio. The distribution of demand response is ensured by the uniform reallocation of trigger frequencies. Accordingly, the local load controller 26 is also adapted to maintain information regarding the trigger frequency at which the load 14 switches from normal operation to operation responsive to a network frequency deviation. If the network frequency deviates from its nominal value to an extent that exceeds the load's trigger frequency, the local load controller 26 sends a signal to the industrial process controller 20 indicating that response service is required and (if necessary) the level of response. Similarly, the local load controller determines when the responsive load service will not require and notify the industrial process controller 20 to revert to a previous level of operation. The industrial process controller 20 is adapted to report load processing parameters (e.g., stored energy reserves) to the local load controller 26.

[0050] The local load controller 26 includes a communication interface that accommodates bidirectional communication, enabling data exchange over the network between the local load controller 26 and the remote central demand server 30. The central demand server 30 may communicate with multiple local load controllers 26. The data signals transmitted by the central demand server 30 relate to power management according to relatively slowly changing criteria. In this embodiment of the invention, it relates to peak price avoidance measures. Unlike responsive load services, power consumption adjustments made in connection with peak price avoidance are known days, weeks, or even months in advance. In comparison, any time delay in transmitting information over the network is negligible. This means that a single central demand server 30 can provide information regarding power adjustments over the network to many local load controllers 26. The requested adjustment patterns are tailored to the specific operating requirements of each of the loads 14, as described in more detail below. The data provided by the local load controllers 26 to the central demand server 30 may include, for example, information regarding the instantaneous total power consumed by the loads. Over time, this allows the central demand server 30 to record and analyze load power consumption patterns, thus verifying the success of responsive load and peak price avoidance measures, etc.

[0051] Figure 2 graphically illustrates the variation of load power consumption patterns over several hours for (a) a demand-side frequency response service and (b) a peak price avoidance scheme during operation. Graphs 34 and 36 each plot the power consumption by a variable-speed load device (or aggregate group of binary loads) against the time of day.

[0052] Referring to FIG. 2a, it is noted that many distribution system operators in Europe and elsewhere require stabilization measures to be initiated as soon as the grid frequency deviates by more than 0.01 Hz from a target frequency of 50 Hz. In system 10, the grid frequency is monitored by frequency monitor 28. If the grid frequency rises to 50.01 Hz or falls to 49.99 Hz, this threshold is noted and at least one of load controllers 26 responds by adjusting the power consumption of each of the loads 14 to address the frequency deviation. In the UK, providers of responsive load services are required to meet several specific performance criteria. For example, to meet stable frequency response (FFR) requirements, a provider is required to come online within 10 or 30 seconds of detecting a frequency deviation and have a minimum of 1 MW of responsive energy available to provide it. Improved frequency response (EFR) is a service that can provide full power regulation within 1 second (or less) of registering a frequency deviation. When grid frequency is monitored on a second-by-second basis, UK FFR and EFR providers potentially vary their aggregated load power consumption on similar timescales. This is reflected in the rapid fluctuations in the total power consumption of aggregated devices operating FFR services, as observed in Figure 2a.

[0053] Referring to FIG. 2b, the power consumption of an exemplary load operating with peak price avoidance is shown. For this grid system, electricity consumed between 4:00 PM and 7:00 PM is more costly than at other times during the day. The normal power consumption levels for this load, for which the system operates up to specification, are those shown between 2:00 PM and 3:00 PM and between 7:00 PM and 8:00 PM. However, at 3:00 PM, the device operates at maximum power for an hour. This allows the device to store sufficient reserves of energy to provide for zero-power operation during the peak period between 4:00 PM and 7:00 PM. The load then returns to normal operating power. Of course, this is merely an example of behavior that enables peak price avoidance. The specific details depend on the associated load. For example, the load may operate at maximum power for more than an hour, requiring the entire energy reserve to be accumulated. The energy reserve may not be sufficient to maintain zero-power operation throughout the entire peak period, thus requiring some form of reduced-power operation. Additionally, the load may operate at a higher power consumption after the peak period to augment depleting energy stores. Regardless of the specific details, it can be seen that peak price avoidance results in a slowly time-varying power consumption profile 36.

[0054] Peak price avoidance is not the only behavior that slow-acting variable power consumption generates. In this context, "slow-acting" typically means that it should change only slightly every 30 minutes, while "fast-acting" means that it can change from second to second. Most distribution networks offer multiple markets for purchasing power and rights to request flexible responses from both generators and loads. Providers are incentivized and penalized to encourage operation within contractual limits. While responsive frequency markets require rapid responses, some other markets only require slow-acting patterns of power consumption. For example, to participate in day-ahead optimization, power is purchased at a price set one day in advance of use, and the operating power level is then set to match the purchased amount. Adjustment mechanism services provide manually commanded responses that correct for generation or demand forecast errors and unexpected losses. These types of fluctuations are signaled sometime in advance. Other means for requesting slow-acting power consumption profiles include Project TERRE, imbalance tracking, and DNO control signals, among many others known to those skilled in the art.

[0055] In the past, it was not considered possible for loads involved in peak price avoidance to simultaneously participate in frequency response services. This prevented the operators of such loads from enjoying the financial benefits of these markets. It also meant that the loads tended to use their capacity to provide other services to the grid that were compatible with the slow-acting function, reducing their ability to provide services such as FFR and ultimately compromising the quality of the regulating services they could provide to the grid. Similarly, some generators may operate with variable capacity to input power to the grid according to the slow-acting market function. Until now, such generators were also unable to simultaneously participate in frequency response services such as FFR.

[0056] However, with the present invention, such loads and generators (especially those with the capacity to store energy in some form) can simultaneously engage in both implementations. The present invention has different embodiments depending on whether the component (load or generator) is capable of operating with fully variable power consumption or whether the component is a binary element. Implementations of the present invention for both situations are described in detail below.

[0057] However, broadly speaking, a load is free to engage in practices such as FFR that result in fast-acting fluctuations in power consumption F(t), while simultaneously aiming to comply with slow-acting measures, provided that small deviations from the ideal slow-acting power consumption are permitted. To ensure that deviations remain small, the slow-acting reference power consumption B(t) should be derived such that, on average, the total power consumed by assets implementing FFR matches the total power that would likely be consumed if FFR were not implemented and slow-acting measures were not complied with alone.

[0058] The above reference to "high probability" means that for a load that provides an FFR for a fixed time, the probability tends to 1 as that time period tends to infinity.

[0059] Using the function B(t), the instantaneous total power consumed at time t by the load undergoing FFR is: It can be expressed as P(t)=B(t)+F(t).

[0060] That is, instantaneous power can be modeled as the superposition of two functions: the slow-operating reference value function B(t) and the fast-operating FFR term F(t). This model is valid, subject to a correction algorithm for deriving B(t), subject to various constraints on both the average power and the instantaneous power fluctuations of a binary asset involved as a variable-speed asset providing FFR or as one of a group of assets providing reactive load service.

[0061] For example, the power consumption profile required by a slow-acting measure to avoid consumption during peak prices is called the slow-acting demand profile and is denoted D(t). A simple approach is to set a reference value function B(t) equal to the slow-acting demand profile D(t), resulting in the total power consumed by the load being D(t) + F(t). If the deviation from the ideal demand profile is significant, the cost becomes significantly larger relative to any benefits resulting from implementing the measure according to F(t). Furthermore, for binary loads, the error becomes significantly larger. Such loads can only be "on" or "off." If both measures simultaneously require the load to be "on," superposition results in the load consuming twice its maximum power, an obviously impossible situation. Therefore, the load is forced to deviate from one of the measures it is following and is most likely to incur a penalty.

[0062] An alternative approach (following the present invention) uses a derived function B(t) as the baseline power consumption profile instead of D(t). B(t) is constrained to provide a high probability that, over a predetermined period (decision period), the total power consumed by the load will be the same as the power that would be consumed if the load operated according to the slow-motion demand profile. The period is selected to be significantly longer than the typical period during which frequency response power adjustments are required, but short compared to the length of time that power consumption adjustments are required for the slow-motion demand profile. The period of frequency response adjustments, of course, depends on the details of the algorithm used, but is typically about 1 to 3 minutes. As a result, the load is allowed to freely participate in (for example) FFR, but must also follow the baseline profile, subject to the above-mentioned constraints. Thus, if a day is divided into a series of consecutive decision periods and this method steps through each decision period throughout the day, the result will be that, although there will be some short-term deviations from the ideal slow-motion demand profile caused by the involvement of FFR, the results will be minimal over each decision period. Thus, the overall cost to the load of engaging both power adjustment methods remains low.

[0063] As noted above, the implementation of a power consumption means that targets minimum cost will differ depending on whether the load is capable of operating with variable power consumption or whether the load simply needs to be either on or off. Examples of algorithms that may be used in accordance with the present invention in these different scenarios are described separately below.

[0064] Variable Speed ​​Demand Assets A variable-speed demand asset is a load that has the ability to consume electricity at a configurable rate, i.e., its instantaneous power consumption is P(t), where 0≦P(t)≦C D , C D is the maximum power consumption of the load.

[0065] Referring to FIG. 1 and the case of a variable-speed load 14, the steps for operating the load 14 so that it can simultaneously participate in two separate power conditioning measures will now be described. During preparation, the central demand server 30 is informed of or accesses a database of stored information related to a slow-speed operating demand profile D(t). With respect to peak price avoidance, this includes the cost of electricity throughout the day. It also includes historical data specific to the type of load or load 14 being operated in accordance with the present invention. These may be obtained from the local industrial process controller 20 and may thus be load-specific or generally defined for the same type of load. Such data includes operational constraints, such as, but not limited to, variations in stored energy reserves as a function of operating power. This is described in more detail below with respect to an exemplary industrial process. From this data, the central demand server calculates an optimized slow-speed operating demand profile D(t) that provides the lowest operating cost subject to the operating constraints of the load 14. This demand profile D(t) is derived assuming that this is the only power conditioning measure involved in the load. The central demand server 30 may be configured to provide optimization of the demand profile D(t) for one or more loads 14 connected to the network 18 .

[0066] If two or more loads are operated under the control of this demand server 30, the optimized demand profile D(t) is transmitted over the network to each of the local responsive load controllers 26 or 26. The local responsive load controllers 26 for the variable speed loads set the reference value function B(t). C D -Δ <D(t)≦C D If B(t)=C D -Δ If 0≦D(t)<Δ, then B(t)=Δ Otherwise, B(t)=D(t).

[0067] Δ is a small constant (≪C D ), the origin of that constant becomes clear below. So far, such a simple superposition of D(t) and F(t) has been thought to be quite expensive; that is, it is somewhat surprising that this approach works. The approach is performed only when D(t) approaches zero or full operating power, which cannot be used directly as a reference function, and makes a small adjustment Δ. If a reactive load demand occurs during off periods or periods of operation at maximum capacity, this small adjustment is only necessary to leave some headroom.

[0068] The local frequency monitor 28 provides an indication of the grid frequency to the reactive load controller 26. The local load controller 26 calculates any adjustments in load power consumption that need to be made according to the detected frequency deviation, as is known in the art of reactive load service. That is, the local load controller 26 calculates F(t).

[0069] The local reactive load controller 26 then commands the local industrial process controller 20 to operate the load according to the profile so that the load draws power from the grid. P(t)=B(t)+F(t)

[0070] This operating method allows the load's baseline consumption to closely follow the slow operation signal while simultaneously allowing full participation of FFR services. Only the baseline consumption deviates from the slow operation demand profile while the load can operate at approximately full power or approximately zero power.

[0071] It should be noted that by operating a load according to the superposition of two power consumption profiles, simultaneous participation in two different services depends on the fulfillment of the following three conditions: The load is a variable speed load and therefore there are no restrictions on the power adjustments that can be made, provided that the consumption is kept within the limits. The fast function F(t) satisfies low utilization conditions. -F(t) satisfies the mean reversion condition.

[0072] Low utilization means that whenever an asset is required to provide frequency response service F(t), the magnitude of the power adjustment request is likely to be a small fraction of the total power available, i.e., |F(t)|≦δC D Here, with high probability, δ<<1.

[0073] In practice, it is recognized that, typically, for variable-speed loads, FFR utilization is approximately 6%, which represents a slightly smaller perturbation in the asset's power consumption profile. That is, the small portion of capacity used to provide the FFR response is typically 6% (δ ∼ 0.06) of the provided FFR availability. Other algorithms may be used to implement FFR (or other frequency response services) with different utilization factors, but with δ << 1, these other algorithms also satisfy the low utilization condition.

[0074] While the low utilization condition applies to instantaneous frequency response power adjustments, the mean reversion condition relates to the behavior of the response F(t) over the long term of operation T. At any given instantaneous response, the F(t) adjustment can be positive (increased load) or negative (increased generation). Thus, over time period T, the result is that on some occasions the response adjustment is positive and on other occasions it is negative. Ideally, over the decision period T, the responsive service is energy neutral, i.e., the average value of F(t) is zero. In practice, this is rarely the case, and the average value of F(t) over this period is bounded by the value ε. The mean reversion condition requires that the average value of the responsive power adjustments made over the decision period T (i.e., ε) should, with high probability, be small, such as a small fraction of the utilized power. That is,

[0075]

number

[0076] That is, the average responsive power adjustment over ε is significantly less than the maximum power adjustment that is likely to be made at any point during the long period T for a load at utilization factor δ.

[0077] Over the course of the decision period T, F(t) is positive at some times and negative at other times, so the mean reversion condition is primarily satisfied.

[0078] Recall that the baseline function B(t) is set equal to the slow-operating demand function D(t) for a variable-speed demand asset when the demand function does not approach the asset's maximum or minimum power consumption. When D(t) approaches these limits, B(t) is displaced from D(t) by an amount Δ, the value of Δ being chosen to leave the asset with enough headroom to provide frequency response service F(t), if required. When the asset satisfies the low-utilization condition, the value of F(t) at any point in time is δC D (δ≪1) is unlikely to be exceeded. Consequently, only δC DIf δ<<1, then Δ<<C D and thus this adjustment of B(t) represents only a small deviation from the ideal demand profile.

[0079] In an alternative embodiment, the reference value function is such that the function is C D or even approaching zero, always follows an ideal demand function. In this scenario, at the point when a load operates near these limits, it needs to be removed from the portfolio of loads available to provide response regulation. That is, when the reference value is C D If , the load cannot provide regulation service when high frequency grid excursions occur, but is still available to reduce power consumption in response to low frequency excursions. Similarly, if the load is operating at zero power reference, it cannot provide response service to low frequency excursions, but is available to participate in the response when the grid frequency becomes significantly higher.

[0080] Which of these alternative embodiments is preferable is determined primarily by economic cost, considering whether the penalty for not utilizing responsive load service exceeds the increased cost of deviation of the amount Δ from the ideal slow-operating demand function.

[0081] Generally, the asset is left free to respond to any demand F(t) over a period T(P T The average power consumption of an asset operating according to a reference function B(t) that follows a slow-operating demand profile over a period of time is given by:

[0082]

number

[0083] As mentioned above, the average value of F(t) over a period T will be less than or equal to a small value ε provided that the mean reversion condition holds.

[0084]

number

[0085] where B T is the average value of the reference term over this period T. B T is the ideal average low-speed operating demand profile D T B to ensure that it matches T When ε is derived, the result is that the cost of operating load 14 according to this algorithm has a small deviation (ε) from the ideal average power consumption profile, which quite often represents an acceptable cost that is outweighed by the benefits involved in responsive load service.

[0086] Non-variable speed demand assets For non-variable speed demand assets, the power drawn by the load can be one of only two values: zero or full operating capacity. Examples of such loads are compressors (for refrigeration) such as bitumen tanks and relay-operated industrial heaters. These loads cannot effectively participate in both peak price avoidance and frequency response services (such as FFR) by simple superposition of both power consumption profiles. The asset is either on or off; thus, when the asset is on and responding to an FFR request, F(t) is equal to C D is equal to.

[0087] As detailed above, the algorithm by which the baseline function B(t) was derived for variable speed assets required a high speed operating function F(t) to satisfy low utilization conditions. |F(t)|≦δC D , δ≪1.

[0088] Obviously, for non-variable speed assets, this condition does not hold. If a responsive switch is required, then δ = 1. Furthermore, it is not possible to derive the reference function B(t) separately from the unpredictable F(t). Since the asset cannot operate at a power greater than full power, the condition B(t) + F(t) ≤ C Dmust be satisfied, and thus B(t) must respond to changes in F(t). That is, a more complex algorithm is required to derive the criterion function B(t).

[0089] However, for non-variable speed assets, the mean reversion condition holds. This is satisfied primarily as a result of two factors: -If, over the course of a decision period T, F(t) is positive at some times and negative at other times, then -If a single load is involved in an element of a responsive group in an FFR (or similar) market, this load will almost certainly only be required to provide a response intermittently, and this, in turn, means that F(t) will be non-zero for only a small fraction of the period T (provided that the FFR algorithm demonstrating group behavior ensures a sufficiently even distribution around the responsive group).

[0090] For loads 14 that are non-variable speed assets, the method begins with deriving a reference value function B(t), similar to the variable speed case. The central demand server 30 calculates an optimization of the low speed operating demand profile D(t) that provides the lowest operating cost subject to the operating constraints of the loads 14. This demand profile D(t) is derived assuming that this is the only power conditioning measure involving the load. The central demand server 30 may be configured to provide the optimization of the demand profile D(t) for one or more loads 14 connected to the network 18.

[0091] A flowchart illustrating the steps involved in implementing a power consumption measure that allows both engagement of a fast acting service F(t) (e.g., FFR) and compliance with a slow acting demand profile D(t) derived by the central demand server 30 is set forth in Figure 3. For ease of reference, it is assumed that the fast acting service F(t) is an FFR, but this is by way of example only and not limitation.

[0092] In a first step 40, the day is divided into N decision periods (starting at time tk (k=1,2,.....N,N+1)(t N+1 The decision periods are not required to be of equal length, but in practice a 30 minute decision period is valid for FFR and peak price avoidance, in accordance with the rules of the UK electricity market.

[0093] In a second step 42, for each decision period, the central demand server 30 calculates the average power consumption at which the loads 14 would operate if they were to follow only the slow operating demand profile. That is, over the kth decision period, the average demand power consumption D k is given by the following formula:

[0094]

number

[0095] This average demand profile D k ∀k is sent to the local reactive load controller 26 .

[0096] Over the next series of steps, the local responsive load controller 26 adjusts the load P k The average power consumption of k Based on the requirement that the length of the decision period T k For each decision period, function B k As shown above, the load 14 is calculated by the quantity C D power consumption or off. That is, at any given instant t, P(t) is either 0 or C D Therefore, D k To limit the average power consumption over the kth decision period up to the kth decision period, the load 14 k That is, over the kth decision period, B k =D k ,∀ k Derive B(t) so that

[0097] The load is therefore driven in a modulated on / off pattern such that the total duration of the "on" phase matches the desired slow operating demand function over the length of the determined period.

[0098] In this derivation of B(t), we ignore the contribution of F(t) to the total power consumption, as with variable-speed assets. For variable-speed assets, the cost results in a small deviation ε from the ideal demand-power profile, as shown above. This cost has been found to be acceptably low in situations where F(t) satisfies low utilization and mean-reversion conditions. That is, if F(t) consumes a small fraction of the available power, F(t) is as likely to be positive as negative.

[0099] For non-variable speed assets, if F(t) is non-zero, all available power must be consumed. k Deriving it solely from means that over the kth decision period:

[0100]

number

[0101] It is easy to see that the contribution provided by the F(t) term can be significant if the asset is required to provide an FFR response switch during this period. Over the length of the decision period, the average power consumption P k The "on" time is not required to be significantly longer in response to an FFR requirement that increases demand, such that P is significantly greater than the ideal value. Similarly, if a network imbalance requires a decrease in demand on the asset and a decrease in "on" time during the decision period, P k The value of B(t) can be significantly smaller than required to follow an ideal slow-acting demand profile. Second, the asset may not store enough energy throughout the peak price period. Therefore, it is also necessary to derive B(t) in a way that takes into account the consequences of any FFR switches that occur during the decision period.

[0102] In step 44, the local reactive load controller 26 begins its algorithm to calculate B(t) over the kth decision period. Each decision period is divided into modulation subintervals. The modulation subintervals of the kth decision period are j (t k ≦s j <t k+1 ) and time s j+1 (s j j+1 ≦t k+1 ) if we ignore the influence of F(t), e.g., by switching the asset "on" in one subinterval and "off" in the next, switching "on", "off", etc., and by using D for a small fraction of the decision period in which the asset is "on". k We can derive B(t) by setting the duration of the subintervals so that the average consumption of F(t) is ≡ ∑ F(t) = ...

[0103] For the first subinterval s1, B(s1) is C D or 0 (44), and the appropriate command is sent to the industrial process controller 20 (46). This setting is determined by random assignment or k ≧0.5C D If (0 otherwise), then B(s1)=C D ​This can be done by following the first rule, such as: Every I minutes thereafter, the method performs one subinterval assignment loop. That is, the value of B(t) is determined for the next (mth) subinterval. This initiates, in step 48, the local reactive load controller checking whether the load 14 is currently providing reactive load service. If so, nothing changes and the load 14 remains free to participate in FFR in the next subinterval. If the FFR requirement is not set "on" for the load 14, then, in step 50, the average total power consumption of the load over all past subintervals of the kth decision period is either calculated or otherwise obtained by the reactive load controller 26. For the mth subinterval, the average total power consumption up to this subinterval is given by:

[0104]

number

[0105] where λ j is a function that has a value of 1 if the load is "on" in the jth subinterval and a value of 0 otherwise.

[0106] This value of the load's average power consumption over the determination period up to that point can be obtained in several ways. In the case of FFR, the industrial process controller 20 continuously monitors the load's power consumption and provides this actual data to the responsive load controller 26, allowing the responsive load controller 26 to extract the required information. In other embodiments, the responsive load controller 26 is configured to store instruction data that it communicates to the industrial process controller 20. The responsive load controller 26 can then use this stored historical data to calculate the load's average total power consumption over all past subintervals, assuming it is operating according to these instructions. If, for some reason, the industrial process controller 20 disables the responsive load controller 26, for example, if the stored energy reserves are insufficient, this information is communicated by the industrial process controller 20, allowing the responsive load controller 26 to adjust its calculations accordingly.

[0107] In step 52, the slow operating demand profile (D k ), the average power consumption (P k ) is compared to the ideal average power that the load 14 would be consuming. If the load 14 uses too much power, then in the mth subinterval, B(t) is set to 0 in step 54. On the other hand, if the power required by the load 14 is insufficient, then in the mth subinterval, B(t) is set to C D (56) is set to B(t). In either case, the value of B(t) for the subinterval is transmitted by the reactive load controller 26 to the local process controller 20 (58). Over the duration of the mth subinterval, the local process controller 20 ensures that the load 14 operates according to the determined value of B(t), unless this is overridden by any FFR requirements. If, in step 52, the average power draw is found to be equal to the ideal demand function, no action is taken. The process then waits an additional I minutes to complete the subinterval allocation loop 48-58 iteration and, if necessary, determine B(t) for the (m+1)th subinterval and the commanded load 14 operation. This procedure continues until the end of the decision period.

[0108] If B(t) is determined according to this algorithm, then with high probability, P k is maintained close to its target value. The length of the subinterval is selected according to the decision period and possibly also according to the requirements of the particular load 14.

[0109] Clearly, as with variable speed assets, enabling engagement of fast acting response services such as FFR incurs some cost in terms of deviation from the slow acting demand function. This cost needs to be determined. If the cost is significant, loads that regularly operate measures such as peak price avoidance may not intend to employ this technology to enable engagement of both this measure and FFR at the same time.

[0110] As noted above, low utilization conditions are maintained for non-variable speed assets. However, for a group of assets involved in FFR, e.g., according to the algorithm described in WO 2006 / 128709, the time elapsed since a particular asset was required to adjust its energy consumption in response to a grid frequency deviation is taken into account in determining the likelihood of response to a potential future deviation. This results in the FFR response representing a relatively small perturbation with respect to normal operating behavior over a sufficiently long period S. Consequently, the resulting change in total power consumption caused by the FFR response over the period S is also small. That is,

[0111]

number

[0112] There exists a value of S such that δ<<1. This weak low utilization condition, which is obeyed when FFT cutoffs to the reference value occur irregularly, is satisfied by the fact that for each subinterval, {0;C DThis means that the dynamic allocation of} values ​​allows B(t) to be derived in such a way that errors introduced in one (or more) subintervals by the requirement to provide frequency response services are corrected in subsequent subintervals with high probability.

[0113] Without considering F(t), the allocation rule described in Fig. 3 is followed for the decision period T k The average power consumption over a period of time is D k For example, if the ideal demand is 0.7C, D Consider the situation where: The algorithm operates the load according to the modulation pattern set out in Table 1 below.

[0114] [Table 1]

[0115] To make it easier to understand, P k While the value of P can vary significantly over the first few modulation subintervals, its value tends to fluctuate around the ideal value of 70% in subsequent subintervals. When the operating power of the load in each subinterval represents a small fraction of the total power consumed, subsequent subintervals are k This results in a small percentage change in the increase.

[0116] At any point, if the load is required to initiate an FFR switch and F(t) becomes non-zero, the algorithm modulates the load's behavior to compensate. Again, as the number of subintervals following the subinterval where the switch occurred increases, D k Improved recovery from

[0117] For example, 0.7C as shown above D Consider a situation where a load attempting to follow the average demand profile of is required to undergo an FFR switch to zero power for half the duration of the sixth subinterval s6. The load behavior is then described in Table 2 below.

[0118] [Table 2]

[0119] As can be seen, the average power is different after the switch, but according to the seventh subinterval, which is the first subinterval that can be compensated, the average power consumption remains within 6% of its ideal value.

[0120] It is understood that this is just one particular example given to demonstrate the behavior of this algorithm in compensating loads that provide responsive load service in addition to following a slow-acting demand profile. However, the general principle applies: the F(t) switches are compensated within a small number of subintervals. The more subintervals that exist in a decision period, the better the convergence of the average power consumed in that period to the mean value of the slow-acting demand function.

[0121] Average reference value function B in the kth decision period k Returning to the derivation of , this gives the slow operating demand function D over this period. k However, if the ideal slow-operating demand profile is set to be equal to the average value of 0 or C over this period, D If the demand function D(t) approaches one of the limits, there is no capacity remaining to participate in FFR (or at least, not in one direction). For variable-speed assets, if the demand function D(t) approaches one of the limits, B(t) will increase by a value δC away from the limit. D For non-variable speed assets, B k We use a similar approach when deriving B. However, in this case, k Since is the mean value of the reference function that is of primary interest, a more appropriate adjustment factor to use in this case is the frequency response function F over this decision period. k This is the quantity ε derived above. Under the condition that the mean reversion condition holds, ε is a relatively small value (<C D )

[0122] The decision period used in the calculations above is set to 30 minutes. This is primarily because that decision period complies with UK energy market regulations. Longer or shorter periods may be used as well, provided the period remains significantly longer than the typical FFR switch length. There is no impact to making the decision period longer, as it takes longer to correct the reference function in response to the FFR, but it also corrects for more FFR switches. The worst-case scenario of an FFR switch occurring at the end of the decision period occurs with similar frequency. Also, a shorter decision period has little impact unless its length approaches the typical FFR switch length. In this situation, there is insufficient time to correct for the effects of any FFR switch.

[0123] Power generation assets While the above algorithms are described with respect to application to demand assets, it will be apparent to those skilled in the art that the algorithms can be utilized for generators as well. There are times when it is financially advantageous for a generator to feed energy into the grid, which can provide a slow operating demand profile. A generator that can step up or down its output can simultaneously participate in response services, provided that the generator follows the algorithms described herein, with appropriate adjustments to the direction of electricity flow.

[0124] Load Energy Storage Example When implementing the above algorithms (both variable speed and non-variable speed), it is important to track the amount of energy reserve stored in the asset. If the asset has sufficient reserve to operate with non-ideal energy consumption for the required period, it may be possible for the asset to only engage in any variable consumption.

[0125] 1 , the industrial process controller 20 includes or is in communication with a detector 24 that is configured to monitor a parameter indicative of the amount of stored energy stored within the load. That is, for each generator or load system operated in accordance with the present invention, it is necessary to derive a measurable parameter of the system that serves as a “proxy” for the energy stored within the system and retained to enable operation in accordance with the variable power consumption measures. This proxy variable is monitored by the detector and communicated to the industrial process controller 20. If the proxy variable exceeds a safe limit, the industrial process controller 20 disables the responsive load controller 26 and returns the load (or generator) to its ideal operating power level.

[0126] Some non-limiting examples of proxies are now described for different systems that may be included in a portfolio of assets operating in accordance with the present invention. By analogy with a battery energy storage system and measuring the state of charge (SoC) of that system, the proxies for each alternative system are real numbers between 0 and 1 that represent a small fraction of the storable energy currently stored in the system.

[0127] A bitumen tank is an insulated cylindrical container that is partially filled with liquid bitumen. For safety and quality assurance reasons, the temperature of the tank must be maintained below a lower threshold T min and the upper threshold T max Bitumen tanks are generally highly insulated, which allows them to store energy for long periods by converting this energy into heat, but if the temperature of the bitumen tank exceeds T min and T max In this case, the tank temperature and the threshold temperature can be used to define a suitable proxy variable φ.

[0128]

number

[0129] Building management systems typically use a temperature setpoint, T, that corresponds to the ideal temperature to maintain the building at. set In practice, the temperature is allowed to drop or rise by an amount ΔT below or above this set point without the system changing behavior. set -ΔT to T set +ΔT, which results in an effective deadband that lengthens to +ΔT, allowing the energy stored in the system to fluctuate. If the heating system is "on," it will increase the temperature so that the heating system enters the deadband, and the heating will be turned off until the temperature is high enough to move out of the deadband. Thus, given the current building temperature, T, and the setpoint, T, set , and the width of the dead zone 2ΔT are used to define a proxy variable φ of this system.

[0130]

number

[0131] Activated sludge process (ASP) is a type of industrial process used in wastewater treatment. The process involves the biological breakdown of organic content and contaminants in sewage by bacteria and protozoa. Air or oxygen is bubbled through the raw sewage to remove bacteria. Various parameters are monitored and controlled during the process, including: -Dissolved oxygen levels must be maintained to support bacteria. - The sludge volume index provides an indication of the settling characteristics of the sludge. - Ammonia levels, a water contaminant, provide an early indication of the need for any process adjustments due to environmental sensitivity of nitrogen-combining bacteria.

[0132] Arguably, the ammonia level is the most important variable, as violation of the authorized level can lead to substantial fines. For this reason, the simplest proxy variables for the use of variable power operation of the ASP are the current ammonia level L and the maximum allowable limit L. max It is based on the following.

[0133]

number

[0134] However, dissolved oxygen D is also important in these processes. If dissolved oxygen D falls to a low enough value, bacteria can die, and lower concentrations of bacteria reduce the efficiency of the process. If the variable demand case is based solely on ammonia levels, a situation can arise where ammonia levels are low, allowing the process to be turned off for a period of time. However, during this "off" time, oxygen is not being blown into the sludge, and as the oxygen is used up by the bacteria, this results in a drop in oxygen levels. This obviously risks killing the bacteria, with serious consequences for the water treatment process.

[0135] If there are two (or more) critical parameters that need to be considered when defining safe operating limits for a process, a composite surrogate must be defined. In this case, similarly, only a surrogate for dissolved oxygen level should be derived. The dissolved oxygen level D must be below its minimum threshold D to reflect the rate at which low dissolved oxygen levels can become damaging to bacteria. min Construct a proxy function so that the function approaches zero rapidly as the oxygen level approaches the threshold. When the oxygen level is far from the threshold, the proxy will be close to 1. A suitable function is

[0136]

number

[0137] where C is a constant used to parameterize how aggressively the function should approach zero.

[0138] A proxy value representing the fraction of energy stored in the overall system as a fraction of the energy that is available is then set to the product of the two proxy functions.

[0139]

number

[0140] In several examples, it is demonstrated how a proxy variable for the state of charge can be derived to provide an indicator of the instantaneous capacity of an industrial system and regulate its power consumption, enabling the involvement of measures that take advantage of electricity market conditions. Obviously, the details of the indicator depend on the specifics of the relevant industrial process. What governing variable can be used to indicate the processing capacity to use the system's own resources when not drawing power from the grid? In systems such as ASPs, the variable used is the product of two different processing parameters. This approach can be extrapolated to processes where two, three, or even four or more indicators need to be maintained between determined upper or lower thresholds.

[0141] The state-of-charge proxy φ is a parameter that may prove useful in many elements of the above-described embodiments of the present invention. This single parameter provides a ready-to-use indicator of the amount of stored energy reserves, and thus a ready-to-use indicator of the load's 14 ability to respond to power adjustments required to address grid imbalances. It allows a particular load's 14 state-of-charge proxy φ to be mapped to its power consumption as it varies over time. This allows the central demand server 30 to tailor the slow-motion demand profile provided to each load based on the requirement that that load's state-of-charge proxy φ be maintained within a defined range. This, in turn, ensures that the load operates within its normal operating range despite any power adjustments made according to grid conditions. Furthermore, monitoring the state-of-charge proxy φ enables the industrial process controller 20 to determine when a load 14 is unavailable for responsive load service, i.e., when the industrial process controller 20 should override commands from the local load controller 26. Finally, it provides a single parameter that allows responsive load service to be performed according to a single algorithm, regardless of the specifics of the industrial process operated by each load in the portfolio.

[0142] While the above-described embodiments of the present invention are all implemented using loads that adjust their power consumption in response to market conditions, embodiments of the present invention equally apply to generators that are capable of adjusting the power they feed into the grid in response to market conditions. For example, a water treatment plant based on the anaerobic digestion of sewage produces biogas. This biogas is converted into electrical and thermal energy by a gas engine, and any electrical energy obtained in excess of the energy required to operate the plant is sold to the grid. Many such generators are fitted with buffering technology, such as gas bags. The amount of buffering gas is an indicator of the generator's potential to provide electricity. Furthermore, the buffering gas is required to be maintained below a pressure threshold that limits the energy storage capacity. This parameter can therefore be used as a proxy for the state of charge of the system. Thus, a process can be monitored when it simultaneously operates a power generation means in accordance with the present invention.

Claims

1. A method for controlling the operation of a component connected to a power grid, said component being either a generator or a load of a type capable of storing energy internally, and having a power range of 0 to a maximum power C Ｄ and the method is operable over a range of power levels, (a) operating the component over a period of time, adjusting the rate at which energy is transferred between the component and the grid within the period of time in accordance with a reference power function B(t); (b) simultaneously operating the component over the time period to adjust the rate at which energy is transferred between the component and the grid in accordance with a fast action power function F(t), the fast action power function F(t) being superimposed on the baseline power function; The reference power function B(t) is derived from a slow operating power function D(t), and the reference power function B(t) is set to be equal to the slow operating power function D(t), provided that D(t) is 0 or the maximum power C Ｄ Any range δC Ｄ Provided that (δ<<1), the equality holds and the D(t) is within the range, the B(t) is equal to or less than δC Ｄ or (1-δ)C Ｄ and adjusting the rate at which energy is transferred between the component and the grid in accordance with the slow operating power function D(t), the component will economically benefit from fluctuations in the current electricity prices of the grid or fluctuations in values ​​that can be derived directly or indirectly from the power supply of the grid; and The adjustments made to the rate at which energy is transferred between the component and the grid in accordance with the fast operating power function F(t) are adjustments that respond to and address imbalances between the amount of power generated and consumed through the grid, and these adjustments are made to the maximum power C Ｄ A method that is set under the condition that there is a high probability that σ is significantly smaller than σ.

2. A method for controlling the operation of a component connected to a power grid, said component being either a generator or a load of a type capable of storing energy internally, and having a power range of 0 to a maximum power C Ｄ and the method is operable over a range of power levels, (a) monitoring a physical parameter of said component indicative of the amount of energy stored therein; (b) controlling operation of the component in accordance with the method of claim 1 when the parameter has a value between an upper limit value and a lower limit value; (c) if the parameter has a value outside the range of the upper limit or the lower limit, operating the component at an operating power that causes the parameter to return to a value between the upper limit and the lower limit; A method comprising:

3. 3. The method of claim 2, wherein a proxy variable φ, where 0≦φ≦1, is derived for the component, the proxy variable representing the amount of a portion of stored energy held in the energy reservoir at any one time, the proxy variable being derived from the measured physical parameter, whereby limits for component operation are defined in terms of the proxy variable.

4. The method according to any one of claims 1 to 3, wherein the component is an electrical generator.

5. The method according to any one of claims 1 to 3, wherein the component is a load.

6. A system for controlling the operation of a component (14) connected to a power grid (18), said component (14) being either a generator or a load of a type capable of storing energy internally, and having a power range of 0 to a maximum power C Ｄ and the system is operable over a range of power levels, an industrial process controller (20) configured to exercise direct control of the operation of the component (14) and to monitor process parameters; a central demand server (30) configured to derive a slow operating power function D(t), wherein when adjusting a rate at which energy is transferred between the component (14) and the grid (18) according to the slow operating power function D(t), the component (14) will economically benefit from fluctuations in the current electricity prices of the power grid or fluctuations in values ​​that can be derived directly or indirectly from the power supply of the power grid (18); an indicator (28) configured to provide a signal indicative of an imbalance in power supplied through the power supply network (18); a local device controller (26) associated with the industrial process controller (20) and the component (14), receiving the signal indicative of an imbalance in power supplied from the indicator (28) through the power grid (18) and receiving the slow operating power function D(t) from the central demand server (30); deriving a reference power function B(t) from the slow operating power function D(t), wherein the reference power function B(t) is set to be equal to the slow operating power function D(t), provided that the D(t) is equal to 0 or the maximum power C Ｄ Any range δC Ｄ Provided that (δ<<1), the equality holds and the D(t) is within the range, the B(t) is equal to or less than δC Ｄ or (1-δ)C Ｄ deriving a reference value power function B(t) from the low speed operation power function D(t), the reference value power function B(t) being set to one of deriving a fast acting power function F(t) from the signal indicative of an imbalance in power supplied through the power distribution network (18); a local device controller (26) configured to provide instructions to the industrial process controller (20) to operate the component (14) and adjust the rate at which energy is transferred between the component and the grid according to the fast operating power function F(t) superimposed on the baseline power function B(t); A system comprising:

7. A local device controller (26) associated with an industrial process controller (20) configured to exercise direct control of the operation of a component (14) connected to a power grid (18), the component (14) being either a generator or a load of a type capable of storing energy internally, and having a power range of 0 to a maximum power C Ｄ and the local device controller (26) is operable at a range of power levels of: receiving a signal indicative of an imbalance in power supplied from an indicator (28) through the power grid (18) and receiving a slow operating power function D(t) from a central server (30); deriving a reference power function B(t) from the slow operating power function D(t), wherein the reference power function B(t) is set to be equal to the slow operating power function D(t), provided that the D(t) is equal to 0 or the maximum power C Ｄ Any range δC Ｄ Provided that (δ<<1), the equality holds and the D(t) is within the range, the B(t) is equal to or less than δC Ｄ or (1-δ)C Ｄ deriving a reference value power function B(t) from the low speed operation power function D(t), the reference value power function B(t) being set to one of deriving a fast acting power function F(t) from the signal indicative of an imbalance in power supplied through the power distribution network (18); providing instructions to the industrial process controller (20) to operate the component (14) and adjusting the rate at which energy is transferred between the component and the grid according to the fast operating power function F(t) superimposed on the baseline power function B(t); The slow operating power function D(t) is derived such that the component (14) economically benefits from fluctuations in the price of electricity delivered by the power grid over the course of a day when adjusting the rate at which energy is transferred between the component and the grid in accordance with the function D(t).

8. The controller of claim 7, wherein the component (14) is a generator.

9. The controller of claim 7, wherein the component (14) is a load.

Images