Eight Agents for Energy

The Energy Mashup Lab (The Lab) is developing open source software for agents that will enable systems that use, produce, or store energy to self-assemble into microgrids. These microgrids can be standalone or grid-attached. If grid-attached, they present a single market or OpenADR interface to the grid, and that interface reveals only the net market position of the microgrid.

The microgrid is operated by a micromarket, trading in availability over time. The Lab uses open standards to transact between agents. Each system or group of systems being represented by an autonomous merchant agent that buys or sells Power for those systems. The software for this agent is Open Source and can be freely downloaded for use in products.

While there is a simplicity in a single Agent, we think there are benefits to creating more than one type of agent. While a single agent running a single set of code could encompass all behaviors could be created, agents that are optimized for specific types of market behavior can be smaller and more secure. Naming similar market behaviors across systems makes it easier for the integrator to understand how introducing an additional system will affect an existing micromarket/microgrid. We name these the Agent Personalities.

The descriptions below refer to electric power for clarity and brevity. The agent behaviors apply to any resource micromarket.

The Simple Agent Personalities

Each Agent Personality denotes a common set of market behaviors.

Homeostasis Agent

A homeostasis agent represents a system that consumes power episodically to support it’s a purpose external to the resource market. A homeostatic agent schedules power purchases to support providing a service external to the grid.

Two examples of systems that would use a Homeostatic Agent are an air conditioning system and a refrigerator. Each of them buys power to support processes that support a service external to the grid. Neither wants to run unless it is able to buy the entire power curve it needs for its next cycle. Each could advance or delay its purchases to some, or even skip a cycle, without harming the service it provides.

Preconsumption Agent

A pre-consumption agent is similar to the homeostatic agent, but it provides an asynchronous server and therefore has a bias to buying only when the price is low. The system is able to increase consumption in the short term to enhance its ability to provide service at a future time. If the refrigerator is a homeostatic agent, the ice-maker may be a pre-consumption agent. There may be overrides to the behavior, i.e., fill up before the party, or high priority when less than a quarter full.

Base Consumer

Base Consumer uses power continuously when the system it represents is providing a service. An example is a light which is either lit and consuming power, or is unlit and not consuming power. An agent representing one or many lightbulbs on a circuit changes in scale only. A base consumer is almost always a high-priority purchaser in the market.

Tiered Consumer

A Tiered Consumer differs from a Base Consumer in that it may be able to reduce power consumption by providing a lower level of services. An example is a dimmable light. More power might provide a better service, or a different service. Using for example the dimmable light again, a low level of light might support movement, a high level of light support reading, and a higher level of light support personal grooming.

Base Supplier

A Base Supplier supplies power continuously. A Base Supplier might include any controllable generator with a long cycle time. Long cycle time is situationally defined.

Market-Driven Supplier

A Market Driven Supplier supplies power intermittently, based on interactions within the microgrid.

Intermittent Supplier

An Intermittent Market Supplier supplies power intermittently, based upon inputs external to the microgrid. An example is a photovoltaic system, which generates power when the sun shines.

Storage Agent

A Storage Agent is able to consume resources later supply the same resource. It stores power. This is similar to a system able to pre-consume, but it is able to bring some portion of its pre-consumption back to the market at a later time.

The Platform Agents

Any of the Agents Personalities named above can in principal interact with any other agent through bilateral transactions. Some markets might be set up with all tenders going to a single entity who manages all transactions.

Broker

The Broker acts as an agent by executing public orders. It may operate a double auction. The Broker does not itself have a position in any trade. (Transactions to power the broker are an exception). In the home, a home router may act as a broker.

Market Maker

A Market Maker acts as a Broker by executing public orders left. It Market Maker further maintains an orderly resource market with a responsibility to buy for its own account in the absence of public buy orders, and sell from its own account in the absence of public sell orders. The market Maker personality may be associated with Storage or with external market sales and purchases. External market sales and purchases are not part of the internal maker that operates the microgrid.

How to use the Agents

Each of the simple agent personalities could characterize a single node or a collection of nodes. Microgrids can be characterized just as nodes are characterized. This point is fundamental to considering interactions within aggregations of microgrids, as to considering the dis-aggregation if a node into smaller component systems.

A system or device developer will be able to select the personality that he desires to represent his technology, and download it.

A set of agents sufficient to support systems with each of these characteristics is able to support all systems potentially within a microgrid. Such a set does not rule out potential hybrid systems, in which two or more of these characteristics coexist within a single system—such a system is a natural outcome of a microgrid at one level being a node at a higher level.

Launching The Energy Mashup Lab

After nearly two years of work, The Energy Mashup Lab is creaking into public operation. Since Dave Cohen and I initially sketched out its activities in the Fall of 2013, we have been putting the pieces together. William Cox joined us in 2014. Our goal is open source software agents for self-assembling microgrids.

Much of the work is routine business. We are incorporated as a 501C3 non-profit. We have a working web site, www.TheEnergyMashupLab.org, We have a merchant account to process memberships. We have set up secure private forums and list servers. Dave has contributed his award-winning software for energy-aware agents for home and office systems. William is creating the stub architecture and is days away from the secure code repository in which to nurture the work.

Some of it is refining the mission. The Lab is aimed at the microgrid owned and operated by the owner and operator of the site is supports. A microgrid that is operated by its inhabitant, personal or commercial, should have no interaction with its containing grid other than economic negotiations over supply and demand. For microgrids to be everywhere, you must be able to put them together easily. Our goal is that a homeowner can pick up a major appliance on a Saturday and have it fully integrated into his home microgrid before dinner.

Such a microgrid is a necessary enabler of rapid technology innovation. By limiting interactions to standards-based economic communications, the larger grid has no need to see or understand the technology within the microgrid. The owner can readily adopt new technology (for use, recycling, generation, conversion, storage, ...) of energy, because the larger grid can see only the net effect, i.e., the negotiations over supply and demand.

As the most valuable loads (lovely name the utilities have for their customers) are smooth and predictable, these economic interactions reward the microgrid that internally smooths its load curve (no jitter) with storage, whether up or down. They reward the microgrid that can predict (or even better commit) to a particular load at a particular time. The transactive agent is able to get better terms for being a better partner. Smooth, well-chosen forward transactions for particular load shapes augmented by the ability to take advantage of spot markets to adjust position up or down are most rewarding.

That’s where the agents come in. The agents are participants in an internal market negotiation over time. If you consider the matter, the laptop, or the tablet you reading this with fits almost every definition of microgrid. It can disconnect from the grid and continue to operate. It deals with supply and demand internally, adjusting operating parameters based on whether power is available or not, and whether they are in active use. It has internal energy storage. What it lacks, though, is an economic interface on the plug. In the same way, the consumption-only Heat Pump microgrid and the consumption-only Refrigerator microgrid need know nothing about each other—they need only know not to buy at the same time from the home’s internal market. As supply and demand align, the load curve is smoothed and the aggregate market position is improved.

An economic interface is the lightest path to integration. The Lab’s agents will need to go beyond the published OASIS specification Energy Interoperation to discover the market it is in. These standards-based economic behaviors will be grafted onto the energy-aware autonomous agents.

Math and Power and the System with No Name

Every once in a while you run into something that just does not fit into any categories. The world welcomes a better mousetrap, but won’t even consider a mouse dispatcher that sends the mice outside to mow the lawn. We all want things that fit the categories we know. It is hard for a new category to make our purchasing lists.

For the last year, I have been talking to a company that manages energy based on math. The founder created new math to understand how dolphins process signals over time. We create three dimensional models based on what is effectively instant access to shadows and shapes. Dolphins assemble three dimensions based on time-delays in echoes—sound is much slower than light. The founder then applied this math to digital processing of power.

Normal complex-instruction set computers are slow to do certain kinds of math. We all use special-purpose vector-processing CPUs to do graphics (“graphics chips”). In a similar way, this mathematician had to come up with special-purpose CPUs to analyze and fix power in real time. But what does it mean to digitally fix power?

These novel CPUs are now built into special purpose computer systems that take the normal dirty power we all get and make it look as much as possible like the idealized model of a three dimensional power wave. This redefines what we mean by dirty power. Normal power conditioning creates “trapezoids”, power shapes that only mimic a sine wave. True digital power quality is something quite different. And we don’t have a name for what it is.

I have written here before that the effects of digital power quality can be pronounced. Florescent lights stop humming. Motor vibration is reduced and heat generation is reduced. A closer look shows subtler effects. Power output of motors is increased. Impedance is reduced and harmful power harmonics are reduced. Outside the device, power factor tends toward one, which may reduce power bills.

A large facility with a motor load reduced power requirements by 20%, according to a 3^rd party engineer monitoring a trial. Data centers have seen power requirement reduction of 10%, as harmonic stresses are reduced. High-rises with digital power conditioning on each floor may not need to upgrade neutral throughout the building.

Sites close to the Carolinas, where 3DFS is headquartered and can monitor installations closely, can experience something new. But 3DFS is a startup, and their product is not a better mousetrap. It is something else. It is power conditioning based on novel advanced math. And for too many of us, we start the day hoping there will be no math required.

RoSy outlook for distributed autonomy within systems

I feel I must be one of the last people to discover the open source Robotic Operating Systems (RoS). RoS is more of a framework than an operating system. The framework could be atop any operating system. In practice, for now, it is on Linux. (There are some interesting DotNet / Mono extensions, but those appear incomplete). ROS is providing the base for open source robotics, and the effect of robotics on all our lives will expand because of it.

RoS engages my imagination because it is inherently distributed. “Service Oriented Robotics” as a phrase that is used. Replacing the step-by-step commands that have ruled robotic manufacturing, ROS developers aim at tasks such as “Go upstairs, go to my room, find my stapler on my desk, and bring it back”. This must be decoupled into applications for climbing stairs, navigating a floor plan, identifying a stapler, and picking that stapler up.

Just as smart energy looks to fractal dis-assembling of power grids, RoS looks to fractal dis-assembly of robotic tasks. There are multiple ROS services for a robotic hand, decoupling the technology and the mechanics from the request. As ROS-capable systems get smaller and cheaper, there will likely be RoS applications for each knuckle on a hand. A RoS-enabled knuckle can more easily incorporate advanced features such as haptic feedback leading to a “gentle touch”. Gentle touch and heavy lifting can be different limbs responding to the same command.

Robotics is outside of my wheel-house. Service enabling of the internet of things is in. Service oriented energy is in. Fractal microgrids as described by the Galvin Initiative seem natural, and they will have their decision-making local, where they can respond to the needs of site, and the owner, and the situation.

Robotics started out with fixed activities under direct control. In the larger systems, one can still see the single control even as they grow more autonomous. The future is distributed service oriented robotics. In the same way smart grids started with planned sequence to control transmission. It evolved into fixed sequences to control energy consumption, centrally operated, by OpenADR and by EnerNOC and by Constellation. It is slowly evolving into centrally orchestrated DR services.

Even Microgrids are often simply the old architecture, and the old protocols, but just a little bit of isolation. Duke is pushing microgrids barely distinguishable from their distribution networks. Oncor salutes service orientation while extending the old technologies. The real advances are among those building those “smart hands”, autonomous microgrids that make their own decisions and technology choices. >Eventually, just as in the smart knuckles, the same service orientations will arrive in the end appliances and systems of the end nodes.

When I was young, in my Dinosaur age, I was fascinated by the Stegosaurus, and its hind-brain bigger than its fore-brain. That was settled science then, although controversial now. I enjoyed imagining a slow placid creature able to defend itself with some nimble, precise tail-bludgeoning.

The microgrids of the future will leverage distributed energy and local storage to for some precise tail-bludgeoning in the smart building—the far away head will not even be sure what is going on.