WO2008014562A1 - Distributed energy management - Google Patents

Abstract

A distributed energy management system (100) includes a broker agent (102) and a plurality of distributed energy resource (DER) agents (104). Each DER agent (104) is operatively associated with a corresponding energy resource. A shared repository (106) is accessible to the broker agent (102) and the DER agents (104) for posting and retrieval of information and messages by the agents. The broker agent (102) is typically a software component adapted to effect the steps of: receiving (502) periodic supply information of an energy market; and posting (506) corresponding control information to the shared repository. Each DER agent (104) is typically a software component adapted to effect the steps of: retrieving (702) control information from the shared repository; and computing (704) and posting (706) to the shared repository periodic energy forecast information of the corresponding energy resource of the DER agent (104) based upon requirements of the energy resource and at least said control information. The invention advantageously employs the uncoordinated activities of autonomous agents to provide a decentralised, scalable, flexible and adaptable system for managing distributed energy resources within a complex supply system.

Description

DISTRIBUTED ENERGY MANAGEMENT FIELD OF THE INVENTION

The present invention relates to the operation and management of energy supply systems, and more particularly, to the aggregation and/or distributed management of resources within an energy supply network. BACKGROUND OF THE INVENTION

Energy supply networks, and in particular electricity supply networks, are increasing in size and complexity. As a result, the energy industry in many countries faces a number of new pressures which may encourage users to look towards local generation and management solutions. For example, it is now technically feasible for energy users and supply companies to place relatively small generators close to load centres, and/or to participate in the energy market on the "demand side", for example by voluntarily reducing energy consumption in response to rising prices or reduced excess capacity during peak demand periods. Installing and operating systems for responding to price signals and network constraints through local generation or load reduction is becoming more economically viable for businesses requiring greater supply reliability, flexibility, and lower operating costs.

Power generation technologies and demand reduction technologies located locally to sites of energy consumption are commonly known as Distributed Energy Resources (DERs). The introduction of DERs may provide an alternative to, or an enhancement of, the traditional energy distribution system. Small-scale power generation capability installed at a local commercial, industrial, or domestic facility is typically known as Distributed Generation (DG). Many commercial and industrial facilities, and even some domestic facilities, also have the capability to temporarily reduce their demand for energy either in response to local operating conditions (eg reduced demand) or external conditions of the energy distribution network (eg increased prices and/or reduced excess capacity). However, the owners and operators of both DG and load reduction systems typically lack the sophisticated controls and software that would be necessary to optimise the performance of these systems to achieve network benefits. As a result, the benefits of such systems may remain largely unrealised, thereby reducing their economic viability. With improved control and management, however, it is anticipated that substantial improvements in reliability and flexibility of the entire energy supply network may be achieved, while simultaneously delivering lower energy and operating costs to the end consumers.

At the same time, the energy generation and distribution industry in many countries has undergone significant restructuring. Traditionally, the energy supply industry was structured as regulated monopoly. Under this model, generators were typically "dispatched" (Ze mobilised to deliver power into the grid) by the monopoly supplier in response to the load arising within the monopoly service area. The monopoly operator had complete control over the energy resources within a specified service area. However, in accordance with the restructuring occurring in many markets, the dispatch function may be transferred to an independent operator tasked with operating an energy market over the transmission system through which all generators (which may have a number of competitive owners) deliver energy to end consumers. For example, across a large proportion of the Australian market encompassing New South Wales, Queensland, South Australia, Tasmania and Victoria, this function is performed by the National Electricity Market Management Company (NEMMCO).

The NEMMCO operates a wholesale "spot market" in electricity where supply and demand are instantaneously matched in real-time through a centrally coordinated dispatch process. In accordance with this scheme, generators offer to supply the market with specific amounts of electricity and particular prices. Such offers are submitted every five minutes of every day. From all offers submitted, NEMMCO's systems determine the generators required to produce electricity based on a principle of meeting prevailing demand in the most cost-efficient way. The NEMMCO then dispatches the selected generators into production. Under the NEMMCO system, wholesale electricity prices are available to industrial users and retail suppliers every five minutes. Forecast prices are available over the next 24 hours with varying accuracy. This information potentially enables users to regulate their demand in accordance with expected availability, capacity and pricing. However, as previously mentioned, the difficulty lies in implementing DER systems that are able to optimise energy generation and utilisation so as to provide benefits sufficient to justify the cost of deployment and operation. Prior art optimisation methods and systems have generally relied upon a centralised control strategy. In this case, the problem of coordinating a group of DERs may be characterised as combining information such as predicted environmental conditions, models for the constraints and behaviour of load and generator resources, supply- and demand-based pricing information, and system constraints such as a cap on the total energy supply to the group of DERs, to calculate optimised energy utilisation plans for each resource for a given period into the future. As will be appreciated, true global optimisation of such a complicated system is generally an NP-hard problem, and accordingly is not at all scalable to large numbers of DERs.

Accordingly, sub-optimal techniques may be employed, such as simulated annealing, genetic optimisation, and linear programming. Such techniques may produce near-optimum results, utilising processing resources that grow approximately linearly with the number of DERs to be coordinated. Even so, the capacity of a central processing resource is ultimately finite and does not scale naturally and automatically with increasing system complexity and numbers of DERs. Furthermore, the ultimate coordination solutions arrived at, for example within each five-minute time interval, may need to take into account local constraints at each DER, as well as the global system constraints. Accordingly, the complexity of the system, and the processing resources required to reach a solution, may increase quite significantly with the addition of further DERs. Additionally, a centralised solution is limited in its ability to adapt to local changes. For example, if there is a sudden change in the requirements or constraints at one particular DER, for example due to a sudden environmental change, or unexpected additional local demand, a centralised system could only respond by re-optimising the entire system.

According to one proposed method for improving the efficiency of DERs, price-based control is employed, whereby human "owners" of each resource are asked to respond to a varying, broker-determined price for power. However, this approach suffers from a number of disadvantages. For example, human owners may not exist for some resources, or may not be able or willing to respond when asked. Furthermore, there is no guarantee of the level of service to resources. The process may also lead to customer dissatisfaction, since it requires the application of human effort in order to participate, and consumers are thus required to choose between cost and comfort or convenience.

In an alternative proposed approach, agent-based market-oriented algorithms are applied, using real or virtual currency, wherein one or more broker agents carry out a negotiation process with each resource agent to fix usage and price. However, this method lacks scalability, since market-based algorithms require hierarchies of brokers to negotiations with very large numbers of resources leading to potentially fragile structures. Furthermore, market-based algorithms require adaptation or replication to account for interdependence of resource control actions at different times due to their intrinsic time constants. Additionally, although the efficiency of market-based algorithms may be quantified, there remains no guarantee of an adequate level of service at either the resource or system level.

Accordingly, there remains a need for an alternative approach to optimisation of distributed generation systems and distributed energy resources which has improved scalability, flexibility and adaptability. SUMMARY OF THE INVENTION

In one aspect, the present invention provides a distributed energy management system including: a broker agent; a plurality of distributed energy resource (DER) agents, each operatively associated with a corresponding energy resource; and a shared repository accessible to the broker agent and the DER agents for posting and retrieval of information and messages by said agents, wherein the broker agent is adapted to effect the steps of: receiving periodic supply information of an energy market; and posting corresponding control information to the shared repository, and wherein each DER agent is adapted to effect the steps of: retrieving control information from the shared repository; and computing and posting to the shared repository periodic energy forecast information of the corresponding energy resource of the DER agent based upon requirements of the energy resource and at least said control information.

Accordingly, the present invention provides for the distributed and substantially independent operation of a plurality of DER agents, each of which is autonomously and asynchronously responsible for planning the energy consumption and/or generation of an associated energy resource. Rather than attempting to optimise the system centrally, or by direct communication and negotiation amongst all of the DER agents, neither of which approaches are generally scalable, flexible or adaptable to rapid changes in local circumstances, the invention utilises a single shared information "space", /e the repository, for sharing of data and messages amongst agents. Advantageously, each of the DER agents is substantially similar in operation, and each is essentially "selfish" in that it operates primarily to satisfy local goals. The broker agent serves a specific function in the system, most notably in that it is responsible for obtaining supply information, for example pricing information and/or supply caps, and for posting this information or equivalent control information derived therefrom, to the shared repository. Since each DER agent will take into account the control information in making energy forecasts, ie in seeking to satisfy the local goals of the associated energy resource, the control information effectively acts as a "signal" which will influence the behaviour of the DER agents.

The invention thus employs the uncoordinated activities of autonomous agents to provide a decentralised, scalable, flexible and adaptable system for managing distributed energy resources within a complex supply system. In accordance with the invention, through the exchange of information and messages in the shared repository a seemingly coordinated result (improved or substantially optimised management of distributed energy resources) emerges from the relatively uncoordinated activities of the independent agents. The emergence of such order through indirect communication is known as "stigmergy". Many examples of stigmergic activity occur in nature, and have served as the inspiration for the present invention. One such example is the emergence of seemingly coordinated collective activity from the independent actions of the ants of a colony, as each ant leaves pheromone in its environment, and responds to pheromone and the conditions in its immediate vicinity in a simple, predictable way.

In the present specification, "agent" refers to an executable software program that executes autonomously and provides specific functionality and interfaces in accordance with its role within the overall system. For example, each DER agent is operatively associated with its corresponding energy resource such that it is able to monitor and maintain records of current operating conditions and requirements of the resource, in order to compute appropriate energy forecasts, and typically also to control the resource in order to implement a forecast energy utilisation plan. The broker agent is configured to retrieve supply information, for example from a wholesale energy market operator, and additional network information, for example from an energy retailer or an energy network operator, and format that information, and/or calculate derived information such as corresponding retail prices, for provision to the DER agents. All agents include functionality for posting and retrieval of information and messages to and from the shared repository.

In another aspect, the invention provides a method implemented by a broker agent within a distributed energy management system, the method including the steps of: receiving periodic supply information from an energy market; and posting control information corresponding with said supply information to a shared repository accessible to the broker agent and to a plurality of DER agents operatively associated with corresponding energy resources.

In a preferred embodiment, the broker agent further effects the steps of: retrieving periodic energy forecast information of the energy resources from the shared repository, said forecast information being posted by said DER agents; and periodically purchasing energy from said energy market in accordance with said energy forecast information. It will be appreciated, however, that in alternative embodiments the steps of retrieving forecast information and making a corresponding purchase of energy from the energy market may be carried out separately from the broker agent, for example by an autonomous purchasing agent. In a further aspect, the invention provides a method implemented by each of a plurality of DER agents associated with corresponding energy resources within a distributed energy management system, the method including the steps of: retrieving periodic control information from a shared repository accessible to the DER agents and a broker agent, wherein the control information is posted to the repository by a broker agent; computing periodic energy forecast information of a corresponding energy resource based upon requirements of the energy resource and at least said periodic control information; and posting the energy forecast information to the shared repository. In preferred embodiments the broker and DER agents are software agents, which are implemented within a computer system including a processor and associated storage, the storage including executable program instructions which, when executed by the processor, cause the processor to effect the corresponding steps of the method implemented by the agent.

Accordingly, in further aspects the present invention encompasses computer software products embodied in computer-readable media, including computer executable instruction code adapted to cause a computer to effect the methods implemented by the broker agents and DER agents in accordance with the foregoing aspects of the invention.

The energy resources associated with each of said DER agents may include loads and/or generators. One example of a load that may be managed by a DER agent in accordance with embodiments of the invention is a cooling system. In this example, the agent may be enabled to monitor the temperature of the system, to calculate the predicted future temperature variation corresponding with different levels of energy consumption, and to modify levels of energy consumption according to control information retrieved from a shared repository, thereby to compute corresponding energy forecast information, and to control the operation of the cooling system in order to implement a computed forecast plan.

Similar or corresponding functionality of agents associated with other types of load and with various types of energy generators can be devised by skilled persons having knowledge of such other relevant energy resources, e.g. loads and/or generators. It is generally envisaged that embodiments of the present invention are most readily applicable to electricity supply systems, however it will be understood that the general principles may be applied to other forms of energy supply systems, such as gas supplies. The supply information preferably includes pricing information and/or information defining a supply cap. The supply information is preferably time-dependent, covering periodic intervals, for example five-minute intervals. The supply information posted to the shared repository by the broker agent may in each instance cover a predetermined future time-span, for example 30 minutes.

In preferred embodiments, the system also incorporates additional auxiliary agents. For example, one or more sum agents may be provided, each of which is adapted to effect the steps of: retrieving periodic energy forecast information of the DER agents from the shared repository; totalling forecast energy requirements; and posting total forecast energy requirements to the repository.

Advantageously, the task of summarising total forecast requirements of all

DERs is thus delegated to further independent agents, such that the summary information is available to all other agents without the need for duplicated computations throughout the entire system. As will be appreciated, auxiliary agents, such as sum agents, may be provided in order to compute and/or otherwise obtain any information that may be generally useful to a large number of agents within the system, in order to avoid duplication of effort and to distribute such computations efficiently to available processing resources which may be geographically dispersed in the energy distribution network.

Additionally, a settlement agent may be provided which is adapted to effect the steps of: retrieving actual energy consumption of each energy resource from the shared repository corresponding with past time intervals; and computing an energy cost to be charged in relation to each energy resource based upon the actual energy consumption. Advantageously, the settlement agent may therefore assume responsibility for accounting for usage of energy resources by all DERs throughout the system.

The broker agent may be further adapted to compute a customer energy price based upon supply pricing information and energy forecast information posted by the DER agents. The customer energy price may additionally be based upon a supply cap and may include a predetermined operating cost and/or profit margin of the broker agent. The DER agents are preferably further adapted to take customer energy price into account when computing energy forecast information. For example, the DER agents may be adapted to seek to meet energy requirements of the associated energy resource, while substantially minimising or reducing energy cost, and avoiding exceeding a posted supply cap so as to avoid the imposition of energy restrictions. The broker agent may also be adapted to periodically post a reward policy to the shared repository. For example, a reward policy may include bonuses and/or discounts for increasing energy usage during "off peak" periods, and/or for reducing energy usage during peak periods. Preferably, the DER agents are adapted to retrieve the rewards policy from the shared repository, and to take the rewards policy into account when computing energy forecast information, so as to maximise the rewards obtained within the other constraints of the associated energy resource. It will thus be appreciated that through the posted pricing, supply cap information, and rewards policy, the broker agent is effectively enabled to deploy control signals into the system that will effect the subsequent behaviour of the DER agents.

According to preferred embodiments, the DER agents are further adapted to effect the steps of: retrieving energy forecast information of other DER agents from the shared repository; and taking into account said retrieved energy forecast information of other DER agents when computing periodic energy forecast information of the corresponding energy resource. In doing so, each DER agent may retrieve summary information posted by auxiliary agents, eg total forecast energy requirements posted by one or more sum agents, in order to avoid the need to retrieve and locally total the corresponding individual energy forecast information. In computing energy forecast information, the DER agents may also take into account explicit or implicit rewards for their contribution to meeting overall system objectives, eg maintaining total energy demand below a posted supply cap. Explicit rewards may be embodied in reward policies, while implicit rewards may include avoiding an excessive demand situation in which a supply cap is exceeded, whereby all DER agents risk a restriction of supply leading to a failure to satisfy local constraints of an associated energy resource.

According to a preferred embodiment, the steps of retrieving information from the shared repository and computing corresponding energy forecast information may be repeated iteratively by all DER agents until all local and system constraints are satisfied within a present time period. According to a preferred embodiment the DER agents operate asynchronously, however it would also be possible to synchronise agents by the provision of a suitable timing mechanism implemented, for example, using communication links between the agents.

In still another aspect, the invention provides a method of managing a distributed energy system including a plurality of energy resources, the method including the steps of: obtaining periodic supply information of an energy market; posting control information corresponding with said supply information to a shared repository; independently computing periodic energy forecast information of each of said plurality of energy resources based upon requirements of each said energy resource and at least said periodic supply information; posting said periodic forecast information to the shared repository; and periodically purchasing energy from said market in accordance with the energy forecast information.

Further preferred features and advantages of the present invention will be apparent to those skilled in the art from the following description of preferred embodiments of the invention, which should not be considered to be limiting of the scope of the invention as defined in any of the preceding statements, or in the claims appended hereto. BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the invention are described with reference to the accompanying drawings, in which like reference numerals refer to like features and wherein: Figure 1 shows an exemplary system illustrating a preferred embodiment of the present invention;

Figure 2 is a schematic diagram illustrating an exemplary microprocessor-based apparatus for implementing a broker agent according to a preferred embodiment of the invention; Figure 3 is a schematic diagram illustrating an exemplary microprocessor-based apparatus for implementing a DER agent according to a preferred embodiment of the invention;

Figure 4 is a schematic diagram illustrating an exemplary microprocessor-based apparatus for implementing an auxiliary agent according to a preferred embodiment of the invention;

Figure 5 is a flowchart illustrating a method implemented by an exemplary broker agent;

Figure 6 is a flowchart illustrating a method implemented by an exemplary DER agent; and Figure 7 is a flowchart illustrating a method implemented by an exemplary auxiliary agent. DETAILED DESCRIPTION OF PREFERRED EMBODIMENT

Figure 1 illustrates an exemplary system 100 embodying the present invention. The system 100 includes a number of software agents, including a broker agent 102 and a plurality of distributed energy resource (DER) agents 104. The system 100 also includes a shared repository 106, which is accessible to all of the software agents, including the broker agent 102 and the DER agents 104. The system 100 may also include additional auxiliary agents, eg a plurality of sum agents 112 and a settlement agent 114. It will be understood that the system 100 is a computer-implemented system of software components executing on suitable hardware, which is "overlayed" on an existing energy distribution network, such as an electricity supply grid (not shown). The energy distribution network includes generation apparatus, providing power to the grid, and loads which draw power from the grid. In general, each generator or load is an energy resource, and these are distributed throughout the network. Accordingly, each DER agent 104 is operatively associated with a corresponding energy resource (eg a generator or load), and the agents 104 each have some capability to monitor and control the associated energy resource. The present invention is not concerned with the general operation of the energy supply network, or with the details of individual energy resources, but rather with the operation and coordination of the distributed energy management system 100 which is overlayed on the supply network and resources. The object of the system 100 is to manage and control the distributed energy resources of the energy distribution network in order to achieve improved supply reliability, flexibility, adaptability and lower costs to end consumers.

The shared repository 106 is, in general terms, a data store such as a centralised or distributed database which is shared and accessible by all agents, eg 102, 104, in the system 100. In one exemplary embodiment, the repository 106 may be implemented using standard Internet communications technologies, and may be hosted by a web server in the form of a bulletin board or similar addressable data store. Accordingly, each software agent, eg 102, 104 is provided with a corresponding network data link, eg 108, 110, via which the repository 106 may be accessed. In alternative embodiments, the repository 106 may utilise the distributed memory resources of one or more agents, such as the sum agents 112, in which case it will be understood that the arrangement 100 illustrated in Figure 100 represents a logical, rather than physical, structure. Advantageously, a distributed implementation of the repository may provide for improved scalability of the system.

The precise implementation of the repository 106 and corresponding data links 108, 110 is not critical, so long as each agent, eg 102, 104, is able to post information and messages to the repository 106, and retrieve information and messages posted to the repository 106 by other agents. It is a particular feature of the present invention that there is no direct inter-agent communication between the broker agent 102 and DER agents 104 or between DER agents 104, and that all messaging is carried out via the repository 106. In contrast to prior art systems and methods, embodiments of the present invention avoid centralised control and optimisation, and instead rely upon the autonomous and asynchronous activities of independent agents, eg 102, 104 in order to achieve an emergent coordinated result via the process of stigmergy.

As will described in greater detail below with reference to Figures 2 and 5 respectively, the broker agent 102 is generally adapted to receive supply information of the energy market corresponding with the underlying supply grid, and to post corresponding control information to the shared repository 106. Each DER agent 104 is adapted to retrieve the control information from the shared repository 106, and to generate and post forecast energy usage information based upon local requirements and constraints in view of the control information, as described in greater detail below with reference to Figures 3 and 6 respectively. As previously mentioned, the system 100 may also include additional auxiliary agents, eg 112, 114, which are also enabled to access the repository 106 via data links 116, 118, and which perform additional autonomous processing functions using the information posted to the repository 106 by other agents.

Figure 2 illustrates schematically a number of key components of an exemplary microprocessor-based apparatus for implementing the broker agent 102. It will be appreciated, however, that the figure does not show all peripherals, interfaces and components of the microprocessor system 200, which are well-known in the art but which are not relevant to the present discussion. The apparatus 200 is a computer system which includes a microprocessor 202, interfaced in a conventional manner to a network interface 204. The network interface 204 provides access to a data network, such as the Internet, via which the repository 106 and other networked information systems may be accessed. The processor 202 is also interfaced to one or more memory or storage devices, eg 206. The memory or storage device 206 relevantly contains program instructions for execution by the processor 202, for carrying out various operations of the computer apparatus 200, including those related to the execution of the broker agent 102. As will be appreciated, the memory or storage device 206 will also contain program instructions for execution by the processor 202 for performing a variety of other supporting functions, including various operating system functions of the computer system 200. The broker agent 102 hosted on the computer system 200 is also able to access and retrieve periodic supply information, such as pricing details, from an energy market 208 associated with the underlying supply grid. Supply information may be available, for example, via the Internet through the network interface 204. Alternatively, supply information may be provided to the broker agent 200 via a separate, proprietary channel. The exact mechanism by which the broker agent 102 hosted by computer system 200 is able to receive supply information is not critical, however the invention requires that periodic updates of the supply information are available to enable continuous real-time, or near real-time management of the distributed system 100.

Figure 3 illustrates schematically a number of key components of an exemplary microprocessor-based apparatus 300 which may be used to implement a DER agent according to an embodiment of the present invention. The microprocessor-based apparatus 300 includes a microprocessor 302, interfaced to a network interface 304 and one or more memory or storage devices 306. As with the broker agent system 200, the network interface 304 provides access to the shared repository 106 via a suitable data communications network, such as the Internet. The memory or storage device 306 relevantly contains program instructions for execution by the processor 302, including those related to the execution of the DER agent 104. The DER agent 104 is operatively associated with a local energy resource 308, via a suitable interface between the energy resource 308 and the microprocessor-based apparatus 300. Various mechanisms are available for providing such an association. For example, the microprocessor-based apparatus 300 may be integrated into the energy resource 308, and be provided with direct interfaces enabling operation of the energy resource 308 to be monitored and/or controlled. Alternatively, the microprocessor-based system 300 may be a separate computer which is able to access the energy resource 308 via a suitable network or data link. For example, a standard serial or parallel data interface may be utilised between a computer system 300 and an energy resource 308, or the energy resource 308 may be Internet-enabled so as to be accessible via the network interface 304 of the computer 300. Generally, the DER software agent hosted by the microprocessor-based apparatus 300 is adapted to retrieve periodic control information from the shared repository 106, which has been posted by the broker agent 102, and to calculate and post back to the repository 106 forecast information of the energy resource 308, based upon local requirements of the energy resource known to the DER agent 104 in view of the control information retrieved from the repository 106.

Figure 4 illustrates schematically a number of key components of a microprocessor-based apparatus 400 suitable for implementing an auxiliary agent, eg 112, 114, according to an embodiment of the invention. The apparatus 400 includes a microprocessor 402, a network interface 404 and memory or storage device 406, and as with the broker agent and DER agent host systems 200, 300, may be embodied using standard computer hardware. The memory or storage device 406 relevantly contains program instructions for execution by the processor 402 for implementing functions of an auxiliary agent, eg 112, 114. Exemplary auxiliary agents are characterised by the ability to retrieve information from the shared repository 106 via the network interface 404, perform various processing operations on the retrieved information, and post the results of such processing back to the shared repository 106. Examples of auxiliary agents include a sum agent and a settlement agent, which are described in greater detail below.

Figure 5 shows a flowchart 500 illustrating a method implemented by a broker agent 102 according to a preferred embodiment of the invention. At step 502 the broker agent retrieves periodic supply information of a relevant energy market corresponding with the underlying supply grid. The supply information may include details such as wholesale energy prices corresponding with a particular supply period, along with any overall system constraints, such as a maximum available supplied power (energy cap). In an exemplary system, supply information covers fixed intervals, such as five-minute intervals, over a specified duration, such as 30 minutes. As will be appreciated, the exact details of the supply information may be determined by the properties of the particular energy market from which the information is received. In general, however, periodic supply information will be provided in order to enable continuous real-time, or near real-time, management of the distributed energy system 100.

At step 504, the broker agent retrieves from the repository 106 a total planned power consumption of the distributed energy resources over the next relevant interval, eg five minutes. As described below with reference to Figures 6 and 7, the total planned power consumption retrieved by the broker agent 102 at step 504 may be derived from forecasts generated by all of the DER agents and posted to the repository 106, which are subsequently totalled and summarised through the operation of an auxiliary agent known as a sum agent 112. It is, however, of little consequence to the broker agent 102 how the total planned power consumption is derived, so long as it is ultimately made available for retrieval from the shared repository 106.

The broker agent 102 then computes control information based upon the supply information received from the energy market and the total planned power consumption, and posts the control information to the shared repository 106 at step 506. The control information may include the raw supply information provided by the energy market, however is more preferably control information that is more readily utilised by DER agent 104. For example, the control information may include: a grid supply cap, eg corresponding with five-minute intervals over a 30-minute period; a particular reward policy in force over the relevant period of time, providing incentives for DER agents 104 to modify their energy usage; and/or consumer pricing that is derived from the wholesale pricing taking into accounts constraints such as the grid supply cap, and any costs or profit margin that are required to be recovered by the broker agent 102. Optionally, on the basis of the planned power consumption, the broker agent is able to purchase the required power from the energy market at step 508. It is not, however, essential that the broker agent perform any financial function, and in alternative embodiments the purchasing process may be implemented through the action of a separate auxiliary agent, or by any other suitable mechanism.

The process executed by the broker agent 102 then recommences for the next relevant, eg five-minute, interval. Figure 6 shows a flowchart 600 illustrating a method implemented by an exemplary DER agent 104. By way of example, in the remainder of this specification a relatively simple DER agent is described. The simple DER agent corresponds with a cool room, which may be simply modelled as a load (Ze power drawing energy resource) which has the local objective of maintaining an internal temperature within specified constraints, eg upper and lower boundary temperatures. In a simple implementation, cool room refrigeration units are switched on and off in order to control internal temperature, and draw a fixed power when active. The corresponding DER agent is thus required to monitor the internal temperature of the cool room, compare this temperature with the boundary constraints, and plan periods of activity and inactivity of the refrigeration units in order to maintain the internal temperature within the boundary constraints throughout a forecast period (eg 30 minutes). In order to do so, the cool room DER agent also requires a mathematical model of the internal temperature variation of the cool room during periods of activity and/or inactivity of the refrigeration units, and it will be appreciated by those skilled in the relevant art that various suitable physical models are available for this purpose.

In accordance with the exemplary method illustrated by the flowchart 600, at step 602 the cool room DER agent measures the internal and external temperatures of the cool room. At step 604 the agent retrieves the control information posted to the repository 106 by the broker agent 102, and also the available forecast information provided by other DER agents, typically in the form of the total planned power consumption. The cool room DER agent then seeks to compute its own forecast energy requirements during the upcoming relevant period (eg 30 minutes) based upon the control information and demands of the other DER agents in the system 100. In general, the cool room DER agent may seek to operate selfishly, but is constrained and controlled by the need to maintain the temperature of the cool room within the target range, by a general interest in minimising the cost of energy utilisation, by the overall supply network constraints, and by any reward policy that may have been posted by the broker agent 102 in order to influence the decision-making of the DER agents 104. More specific strategies that may be employed by the cool room DER agent are set out in the examples below. Whatever strategy is employed, the result is that at step 608 the cool room DER agent will compute a forecast energy usage plan, and post this plan to the shared repository 106.

At the decision step 610 the cool room DER agent will perform a further check to determine whether all relevant constraints are satisfied. It may be that all constraints are not satisfied, because the autonomous actions of other DER agents seeking to satisfy their own local requirements have resulted in changes to the other forecasts posted to the shared repository 106. As a result it may be, for example, that an overall system cap is violated for some interval during the relevant forecast period. A further iteration may therefore be required in order for all DER agents to simultaneously satisfy their local requirements, while remaining with overall system capacity constraints. This may involve, for example, the cool room DER agent shifting its refrigeration unit active periods to other intervals within the forecast period. That is, the cool room DER agent may take action such as pre-cooling the cool room during an earlier time interval in order to avoid the need to operate the refrigeration units during a later, higher demand, time interval. It is found that the asynchronous, autonomous activities of all of the DER agents 104 will result, within the available time interval, in an overall set of forecast power consumption plans that satisfy all local and system constraints via a process of stigmergy without the need for centralised optimisation and control. Accordingly, the cool room DER agent iterates the steps 604, 608, 610, until all constraints are satisfied, at which time the procedure returns to step 602 in preparation for repeating the process during the next relevant time interval.

Figure 7 is a flowchart illustrating a method implemented by a generic auxiliary agent according to an embodiment of the present invention. As illustrated by the flowchart 700, the general operation of such an auxiliary agent is to retrieve information from the shared repository 106, at step 702, to compute some form of summary or derived information from the retrieved information, at step 704, and to post the derived information back to the repository 106 at step 706. The process then subsequently repeats, in order to maintain current summary or derived information in the event of change in the input information, which may be posted by other agents in the system 100.

One simple auxiliary agent utilised in preferred embodiments of the invention is a "sum agent". The function of the sum agent is to retrieve the planned power requirements of all DER agents in the system for a relevant set of intervals, eg 5 minutes, over a specified forecast period, eg 30 minutes. The sum agent then simply totals all of the power requirements of all DER agents, and posts the resulting summary back to the repository 106. Accordingly, the sum agent ensures that total power forecasts are available to any agent requiring this information, without the need for all agents in the system to retrieve all relevant forecast plans and perform the summation operation for themselves, which would duplicate effort throughout the system 100 and thus be wasteful of processing resources. A plurality of sum agents 112 may be used to distribute the summation operation efficiently to available processing resources which may be geographically dispersed in the energy distribution network.

A further auxiliary agent implemented in preferred embodiments of the invention is a "settlement agent". The settlement agent reads actual posted power consumption of each DER agent for a current relevant interval, eg the present five-minute interval, from the repository 106. The settlement agent further reads the current customer price posted to the repository 106 by the broker agent 102. From this information, the settlement agent computes the corresponding cost to each DER agent for the current interval. The settlement agent may also read a reward policy posted by the broker agent 102, which may be used to modify the actual costs charged to each DER agent. The settlement agent thereby calculates each DER agent's total costs for a given past period of time based on actual power consumption, price per interval, and any relevant reward policy. The total cost details may be posted back to the repository 106, from which they may be retrieved for the purposes of actually charging each corresponding end consumer.

Example 1 : Coordination for Constant Grid Supply Cap

In accordance with a first example, agent methods are implemented according to the following rules:

• Based on a national market electricity price, broker agent 102 calculates and posts supply cap (Cap_sys) to repository 106.

• Based on a cool room physical model, temperature constraints and external temperature prediction, each cool room DER agent 104 plans its own future half hour power needs and posts the mean power on 5 minute intervals (indPowJ) to repository 106.

• A sum agent 112 keeps a running of total planned power needs (totalPow) and posts to repository 106.

• Each DER agent 104 retrieves total power needs and supply cap from repository 106. If the supply cap is not satisfied for certain intervals, each agent 104 updates its power needs by deferring or advancing the "on" action during those intervals, while ensuring that the updated actions satisfy its own internal temperature constrains. This process is iterated until supply cap is satisfied.

• Once supply cap is satisfied, the broker agent 102 purchases power for the next 5 minutes.

• The process is repeated for every 5 minutes.

In exemplary simulated implementations, the following results were obtained for the number of iterations and total processing time required to satisfy a constant supply cap, demonstrating the stability and scalability of the system.

Example 2: Pre-Cooling and 2-Level Coordination Strategies for Severe Variable Cap

In accordance with a second example, a variable supply cap is envisaged, in which severe restrictions on power (eg no supply) may be imposed in specific time intervals. Agent methods are implemented according to the following rules:

• Each cool room DER agent 104 retrieves control (cap) information from the repository 106.

• If the retrieved cap is a severe variable: s Pre-cool the room before the cap drop down (supply restriction) by reducing the upper temperature boundary. s 2-Level coordination

> Each DER agent 104 calculates a local supply cap based on its power consumption proportion within the system

pt_i ^{= EL}ψ ; Cap _i = Cap _sys x pt _i

∑totalPow int v=l

> First level coordination to satisfy local supply cap, i.e., each DER agent 104 plans its future half hour power needs based on both local supply cap and temperature constraints, then posts the mean power of 5 minute intervals on stigspace.

> Second level coordination to satisfy system supply cap, i.e., each agent updates its own power needs to maximise its own return, which is a combination of satisfying internal constraints and maximising external rewards for satisfying system supply cap. This process is iterated until system supply cap is satisfied.

• Each resource agent detects the supply cap to see whether it is a flat cap.

• If the cap is a severe variable one

S Pre-cooling the room before the cap drop down by reducing the upper temperature boundary. s 2-Level coordination

> Each resource agent calculates its local supply cap based on its power consumption proportion within the system

6

∑indPow _i p_{t ^ i =} ^mv=^\ . Cap _i = Cap _sys x pt _i

∑totalPow int v=l

> First level coordination to satisfy local supply cap, i.e., each resource agent plans its future half hour power needs based on both local supply cap and temperature constraints, then places the mean power on 5 minute intervals to repository 106. > Second level coordination to satisfy system supply cap, i.e., each DER agent 104 updates its own power needs to maximise its own return, which is a combination of satisfying internal constraints and maximising external rewards for satisfying system supply cap. This process is iterated until system supply cap is satisfied.

• Once supply cap is satisfied, the broker agent 102 purchases power for the next 5 minutes.

• The process is repeated for every 5 minutes.

In exemplary simulated implementations, the following results were obtained for the number of first and second level iterations required to satisfy a constant supply cap under a variable cap represented by [1.2, 1.2, 0, 1.2, 1.2, 1.2]^* N (five minutes intervals over 30 minutes), demonstrating the stability and scalability of the system under the two level method.

Example 3: Broker Pricing Control

In accordance with a third example, the broker agent 102 calculates and posts to the repository 106 control information in the form of customer pricing.

The broker agent 102 sets customer electricity price based on national market (NEM) price, predicted power usage and other factors, as described by:

f_B(NEM, Predictedilsage, OtherFactors) The function f_B can be designed in different ways and can be optimized to find the best way to set the price. In one embodiment, a four level electricity price scheme is employed, wherein the broker agent 102 changes the price based on the market price only:

LowPrice NEMPrice < a

MediumPήce a < NEMPrice < b

Λ = HighPήce b ≤ NEMPrice < c

CriticalPrice NEMPrice ≥ c

In the present example, LowPrice = 5c, MediumPrice = 10c, HighPrice = 15c and CriticalPrice = 35c per kWh. NEM price thresholds are a = $50, b = $110, c = $170 per MWh. All prices are valid over 5 minute intervals.

After a certain period T, e.g. one day or one month, based on NEM price, customer electricity price and total customer power usage, the broker agent 102 calculates its profit, actProfit. In order to get its expected profit, broker calculates proportion value to customer price:

T prop = 1 - {actProfit - expectedProfit) I ^ customerCost(i)

I=I

So the real customer electricity price is the proportion of fβ, ie, f_B = prop^* f_B. The customer cost is:

T T

V real _ customerCost{ϊ) — ^T prop x customerCost{ϊ)

1=1 1=1

Each DER agent 104 optimizes its actions based on customer electricity price, its own predicted power usage, total predicted power usage and other factors, as described by:

1 _R (Price, Own Usage, Total U sage, OtherFactors)

As for the broker pricing function f_B, agent function f_R can be designed in different ways. For comparative purposes, three alternative agent functions for, f_R2, and f_R3 have been considered.

Under f_R1, each cool room DER agent 104 optimizes its actions to minimize its cost based on customer electricity price only as follows: • Retrieve next half hour price from repository 106, determine if it is constant price or variable price

• If constant price, get action based on room mode, external temperature and DER agent internal temperature constraints. • If variable price,

- locate the lowest price time

- generate provisional forecast plan with on during lowest price period and off in remaining.

- input provisional plan to DER agent to get real actions, which both satisfy resource agent internal temperature constraints and follow the plan actions as much as possible.

Under f_R2, each cool room DER agent 104 optimizes its actions to minimize its cost based on customer electricity price and individual consumption constraints as follows: • Responding to customer electricity price, each DER agent 104 generates forecast to minimize its costs as fø_? .

• Each DER agent 104 sets its individual consumption cap based on customer electricity price and responds to the cap. o If customer electricity price is constant in plan time span, no further adaptation required. o Otherwise s calculate possible maximum cap

M

THxCaP₁ = maxOWiOw )

where / = 1, ..., N, N is number of resource agents, M = H/intv, H is plan span, e.g. 30 minutes, intv \§ interval, e.g., 5 minutes

S set individual cap

M mxCap, if price _j = max(/?πce₇ )

M indCap_t = 0 if price = min(price )

YHxCaP₁ x rat otherwise where raf=0.5, or 0.25 or 0.125, ... for 2^nd, 3^rd, 4^th, ... largest price interval.

S Each DER agent 104 adapts forecast plan to meet individual cap, eg using "CordCap" algorithm described below, and calculates forecast based on minimum cost actions.

Under f_R3, each cool room DER agent 104 optimizes its actions to minimize its cost based on customer electricity price and system consumption constraints as follows. Note that according to this embodiment, additional broker actions are required to derive and post system constraints (supply cap) to repository 106. • Responding to customer electricity price, each DER agent 104 generates forecast to minimize its costs as f^.

• Each DER agent 104 posts its power, indPow_{lι P} to repository 106, / = 1, ... N, j = 1, ..., M, N is number of resource agents, M = H/intv, H is plan span, e.g. 30 minutes, intv is interval, e.g., 5 minutes • Broker: o calculates total power in each interval

totalPoW_j = ^T WdPoW_{1 j} , j = \, - - - ,M ι=l o calculates possible maximum power needs in one interval

N M mxCap = ∑maxiindPow, _J )

o calculates total power in plan span H

J=M

HPower - ^ totalPoW_j

7=1 o checks customer electricity price S Constant price

Broker calculates and posts constant consumption constraints, i.e., constant supply cap

Cap _j = max(mxCap, 2x HPower /M), j = \, - - ,M

S Variable price

Based on customer price, broker calculates and posts variable consumption constraints, i.e., variable supply cap. This variable cap must satisfy the following conditions: a) There is no redundant power supplied in each interval:

M max(Cap ) < mxCap

7=1 b) Guarantee there is enough power for resource agents:

M

∑Cap_j = HPower

J=I c) Zero supply in high price interval as much as possible.

• Each DER agent 104 adapts forecast plan to meet system supply cap, eg using "CordCap" algorithm described below.

• Actions with minimum error are defined as best coordination actions, and used as final resource agent actions.

M err - ^ max(totalPoW_j - Cap_j ,0)

;=i

As will be understood from the foregoing, DER agents 104 may be required to adapt an initial provisional or preferred usage plan in order to meet local or system constraints. In many cases, this will involve shifting energy consumption from high demand intervals, where capacity may be exceeded, into lower demand intervals in which capacity is available, while still meeting other local constraints (such as cool room temperature limits). One exemplary algorithm for achieving a desired outcome in this respect, referred to herein as "CordCap", operates as follows.

The basic CordCap algorithm shifts energy consumption in each interval that overuses energy into its left and right-hand neighbours. The process includes three steps, and is carried out for all offending intervals:

• locate a random point t_x in the interval;

• shift all energy usage in the interval on the left and right of t_x into the left and right-hand neighbouring intervals respectively; and • revise the planned switching status to satisfy the resource agent's temperature constraints.

In exemplary simulated implementations, the following results were obtained for total costs for different numbers of DER agents implementing each of the foregoing functions /^₁ f_R2, and fR₃. For comparison, an example in which a constant price is set was also simulated. In all cases, the broker agent 102 is "fair", and operates at break-even point, ie no net profit. The results clearly demonstrate the potential to reduce overall costs, with f_R2 achieving the best result of a 29.57% reduction.

Claims

CLAIMS:

1. A distributed energy management system including: a broker agent; a plurality of distributed energy resource (DER) agents, each operatively associated with a corresponding energy resource; and a shared repository accessible to the broker agent and the DER agents for posting and retrieval of information and messages by said agents, wherein the broker agent is adapted to effect the steps of: receiving periodic supply information of an energy market; and posting corresponding control information to the shared repository, and wherein each DER agent is adapted to effect the steps of: retrieving control information from the shared repository; and computing and posting to the shared repository periodic energy forecast information of the corresponding energy resource of the DER agent based upon requirements of the energy resource and at least said control information.

2. The distributed energy management system of claim 1 wherein the broker agent is further adapted to effect the steps of: retrieving periodic energy forecast information of the energy resources corresponding with each of said DER agents from the shared repository; and periodically purchasing energy from said energy market in accordance with said energy forecast information.

3. The distributed energy management system of claim 1 or claim 2 wherein the supply information includes one or more of pricing information and information defining a supply cap.

4. The distributed energy management system of any one of the preceding claims further including at least one sum agent adapted to effect the steps of: retrieving periodic energy forecast information of the DER agents from the shared repository; totalling forecast energy requirements; and posting total forecast energy requirements to the repository.

5. The distributed energy management system of any one of the preceding claims further including a settlement agent adapted to effect the steps of: retrieving actual energy consumption of each energy resource from the shared repository corresponding with past time intervals; and computing an energy cost to be charged in relation to each energy resource based upon the actual energy consumption.

6. The distributed energy management system of any one of the preceding claims wherein the broker agent is further adapted to compute a customer energy price based upon supply pricing information and energy forecast information posted by the DER agents.

7. The distributed energy management system of claim 6 wherein the DER agents are further adapted to take the customer energy price into account when computing energy forecast information.

8. The distributed energy management system of claim 7 wherein the DER agents are adapted to seek to meet energy requirements of the associated energy resource, while substantially minimising or reducing energy cost, and avoiding exceeding a posted supply cap so as to avoid the imposition of energy restrictions.

9. The distributed energy management system of any one of the preceding claims wherein the broker agent is further adapted to periodically post a reward policy to the shared repository.

10. The distributed energy management system of claim 9 wherein the reward policy includes bonuses and/or discounts for increasing energy usage during "off peak" periods, and/or for reducing energy usage during peak periods.

11. The distributed energy management system of claim 9 or claim 10 wherein the DER agents are further adapted to retrieve the rewards policy from the shared repository, and to take the rewards policy into account when computing energy forecast information, so as to maximise the rewards obtained within the other constraints of the associated energy resource.

12. The distributed energy management system of any one of the preceding claims wherein the DER agents are further adapted to effect the steps of: retrieving energy forecast information of other DER agents from the shared repository; and taking into account said retrieved energy forecast information of other DER agents when computing periodic energy forecast information of the corresponding energy resource.

13. The distributed energy management system of any one of the preceding claims wherein, in effecting the step of computing energy forecast information, the DER agents are adapted to take into account explicit or implicit rewards for their contribution to meeting overall system objectives.

14. The distributed energy management system of any one of the preceding claims wherein the DER agents are adapted to repeat iteratively the steps of retrieving information from the shared repository and computing corresponding energy forecast information until all local and system constraints are satisfied within a present time period.

15. A method implemented by a broker agent within a distributed energy management system, the method including the steps of: receiving periodic supply information from an energy market; and posting control information corresponding with said supply information to a shared repository accessible to the broker agent and to a plurality of DER agents operatively associated with corresponding energy resources.

16. The method of claim 15 further including the steps of: retrieving periodic energy forecast information of the energy resources corresponding with each of said DER agents from the shared repository; and periodically purchasing energy from said energy market in accordance with said energy forecast information.

17. The method of claim 15 or claim 16 further including the step of computing a customer energy price based upon supply pricing information and energy forecast information posted by the plurality of DER agents.

18. The method of any one of claims 15 to 17 further including the step of periodically posting a reward policy to the shared repository.

19. The method of claim 18 wherein the reward policy includes bonuses and/or discounts for increasing energy usage during "off peak" periods, and/or for reducing energy usage during peak periods.

20. A method implemented by each of a plurality of DER agents associated with corresponding energy resources within a distributed energy management system, the method including the steps of: retrieving periodic control information from a shared repository accessible to the DER agents and a broker agent, wherein the control information is posted to the repository by a broker agent; computing periodic energy forecast information of a corresponding energy resource based upon requirements of the energy resource and at least said periodic control information; and posting the energy forecast information to the shared repository.

21. The method of claim 20 wherein the step of computing periodic energy forecast information takes into account a customer energy price based upon supply pricing information and energy forecast information posted to the shared repository by the plurality of DER agents.

22. The method of claim 21 wherein the step of computing periodic energy forecast information includes seeking to meet energy requirements of the associated energy resource, while substantially minimising or reducing energy cost, and avoiding exceeding a posted supply cap so as to avoid the imposition of energy restrictions.

23. The method of any one of claims 20 to 22 further including the steps of: retrieving energy forecast information of other DER agents from the shared repository; and taking into account said retrieved energy forecast information of other DER agents when computing periodic energy forecast information of the corresponding energy resource.

24. The method of any one of claims 20 to 23 wherein the step of computing periodic energy forecast information includes taking into account explicit or implicit rewards for the contribution of the DER agent to meeting overall system objectives.

25. The method of any one of claims 20 to 24 wherein at least the steps of retrieving information from the shared repository and computing periodic energy forecast information are repeated iteratively until all local and system constraints are satisfied within a present time period.

26. A method of managing a distributed energy system including a plurality of energy resources, the method including the steps of: obtaining periodic supply information of an energy market; posting control information corresponding with said supply information to a shared repository; independently computing periodic energy forecast information of each of said plurality of energy resources based upon requirements of each said energy resource and at least said periodic supply information; posting said periodic forecast information to the shared repository; and periodically purchasing energy from said market in accordance with the energy forecast information.

27. The method of claim 26 wherein the supply information includes one or more of pricing information and information defining a supply cap.

28. The method of claim 26 or claim 27 further including the steps of: totalling forecast energy requirements posted to the shared repository; and posting total forecast energy requirements to the repository.

29. The method of any one of claims 26 to 28 further including the step of computing an energy cost to be charged in relation to each energy resource based upon an actual energy consumption of each said energy resource during past time intervals.

30. The method of any one of claims 26 to 29 further including the step of computing a customer energy price based upon supply pricing information and said energy forecast information posted to the shared repository.

31. The method of any one of claims 26 to 30 wherein the step of independently computing periodic energy forecast information of each energy resource takes into account explicit or implicit rewards for a contribution of the said energy resource to meeting overall system objectives.

32. The method of claim 31 further including the step of periodically posting a reward policy to the shared repository, and wherein the reward policy includes bonuses and/or discounts for increasing energy usage during "off peak" periods, and/or for reducing energy usage during peak periods.

33. The method of claim 32 wherein the step of independently computing periodic energy forecast information of each energy resource includes maximising rewards obtained subject to other constraints of the said energy resource.

34. The method of any one of claims 26 to 33 wherein at least the steps of independently computing periodic energy forecast information and posting said periodic forecast information to the shared repository are repeated iteratively until all local and system constraints are satisfied within a present time period.

35. A broker agent apparatus including: a microprocessor; a network interface operatively coupled to the microprocessor for accessing a shared repository via an associated data network; and at least one memory device operatively coupled to the microprocessor, wherein the memory device includes executable program instructions which, when executed by the microprocessor, cause the apparatus to implement a method in accordance with any one of claims 15 to 19.

36. A DER agent apparatus including: a microprocessor; a network interface operatively coupled to the microprocessor for accessing a shared repository via an associated data network; an energy resource interface operatively coupling the microprocessor with a corresponding energy resource for monitoring and/or controlling the energy resource; and at least one memory device operatively coupled to the microprocessor, wherein the memory device includes executable program instructions which, when executed by the microprocessor, cause the apparatus to implement a method in accordance with any one of claims 20 to 25.