NRG-X-Change

A Novel Mechanism for Trading of Renewable Energy in Smart Grids

Mihail Mihaylov

1,2

, Sergio Jurado

1,3

, Kristof Van Moffaert

1,2

, Narc

ıs Avellana

and Ann Now

R&D Department, Sensing & Control, Barcelona, Spain

Computer Science Department, Vrije Universiteit Brussel, Brussels, Belgium

Software Deparment, Universitat Politecnica Catalunya, Barcelona, Spain

Keywords:

Renewable Energy, Smart Grids, Energy Trading, Virtual Currency, NRG-X-Change, Bitcoins, NRGcoins.

Abstract:

In this position paper we propose a novel trading paradigm for buying and selling locally produced energy

in the smart grid. Unlike recently proposed techniques that rely on predictions and a day-ahead market, here

prosumers are billed by the distribute system operator according to their actual usage and rewarded based on

their actual energy input, similar to the current state of affairs. Our mechanism achieves demand response by

providing incentives to prosumers to balance their production and consumption out of their own self-interest.

All rewards and payments are carried out using NRGcoin — a new decentralized digital currency similar to

Bitcoin, that we introduce in this paper. Prosumers exchange NRGcoins with ﬁat currency on an exchange

market for proﬁt, or for paying their energy bills. We study the advantages of our proposed currency over

traditional monetary payment and explore its beneﬁts for all parties in the smart grid.

1 INTRODUCTION

Trading of locally produced renewable energy is ad-

dressed in literature from a market perspective where

prosumers and consumers (or collectively: agents)

participate in a double auction and trade energy on a

day-ahead basis (Olson et al., 1999; Kok et al., 2005;

Vytelingum et al., 2008; Vytelingum et al., 2010;

Kok et al., 2012; Mockus, 2012). Buy and sell or-

ders for energy are submitted to a public orderbook

and orders are matched either in a continuous fashion

(Vytelingum et al., 2008; Vytelingum et al., 2010),

or at discrete market closing times using the equilib-

rium price (Kok et al., 2012; Mockus, 2012). The

advantages of this market-based control concept are

that it achieves close to optimal allocation, neatly bal-

ances supply and demand and aligns the preferences

of self-interested agents. However, bidding for energy

ahead of time relies heavily on predictions of future

supply or demand, the inaccuracy of which translates

to higher costs for both buyers and sellers. In addi-

tion, agents need to rely on advanced trading strate-

gies in order to maximise proﬁt (or minimise costs).

For example, prosumers unfamiliar with the market

may unintentionally set a too high sell price, resulting

in an unmatched order for their energy. Since there is

no buyer at the time when they produce and inject the

energy into the grid, prosumers make zero proﬁt, un-

less they invest in batteries that can store the untraded

energy. Those agents can then inject the energy at the

time they ﬁnd a buyer. Lastly, separate energy bal-

ancing mechanisms need to be employed (Kok et al.,

2012) to cope with real-time demand response.

Market-based energy trade reduces the depen-

dency of agents on the Distribute System Operator

(DSO), as energy supply and demand is matched di-

rectly between individual agents, resulting in a more

decentralized and competitive environment. How-

ever, locally produced energy nowadays covers only

a small percentage of all consumption and therefore

the DSO still needs to supply a large portion of the

energy to cover the total demand. Thus, considering

the role of the DSO in a trading mechanism allows

for easier implementation of that mechanism on top

of the current infrastructure and state of affairs and

thus a faster transition to a smart grid setting.

In this paper we propose NRG-X-Change — a

novel mechanism for trading of locally produced re-

newable energy that does not rely on an energy mar-

ket or matching of orders ahead of time. In our model

locally produced energy is continuously fed into the

grid and payment is received based on actual usage,

rather than predicted, as consumption is measured by

the DSO and billed in near real-time. Thus, our mech-

101

Mihaylov M., Jurado S., Van Moffaert K., Avellana N. and Nowé A..

NRG-X-Change - A Novel Mechanism for Trading of Renewable Energy in Smart Grids.

DOI: 10.5220/0004960201010106

In Proceedings of the 3rd International Conference on Smart Grids and Green IT Systems (SMARTGREENS-2014), pages 101-106

ISBN: 978-989-758-025-3

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

anism maintains the important role that the DSO cur-

rently plays in the energy market. The second com-

ponent of our mechanism, and a novel contribution

to the energy trading paradigm, is the introduction of

a new decentralized digital currency for energy ex-

change, called NRGcoin. All payments by consumers

and to producers are carried out in NRGcoins, instead

of ﬁat money. The currency can then be exchanged on

an independent open market for its monetary equiva-

lent, e.g. Euro, Dollar, Pound, etc. In Section 2 we

outline the advantages of our mechanism and elabo-

rate on the beneﬁts of the proposed currency. In Sec-

tion 3 we summarize our approach and propose direc-

tions for future work.

2 THE NRG-X-CHANGE

MECHANISM

In Figure 1 we present an example scenario that will

help visualize the complete cycle of energy and cur-

rency exchange in our mechanism for each 15-minute

period (called time slot). Note that since substations

are only informed of the injected energy and not the

produced energy, henceforth by produced energy we

mean energy injected in the grid. Similarly, energy

consumption is the energy delivered to the household

from the power line, excluding the consumption of the

own produced energy.

A given prosumer P generates certain amount of

renewable energy and feeds x of it to the grid, simul-

taneously broadcasting this information to all nodes in

the NRGcoin network. These nodes then update the

public record of P with f (x) NRGcoins, as deﬁned

by the decentralized NRGcoin protocol. Thus, func-

tion f (x) is responsible for generating the NRGcoins

which then enter circulation. In addition, the local

substation S to which P is connected to has measured

the total energy production t

and total consumption

in that slot. That substation then publicly transfers

g(x, t

, t

) NRGcoins from the balance of the DSO

to that prosumer, where g(·) is the production price

function deﬁned by the DSO. Thus, for injected en-

ergy x prosumer P obtains NRGcoins from both the

NRGcoin protocol (to ensure new money is generated

in the system) and the local substation (to align con-

sumption with production). Note that the amount of

received currency only depends on functions f and g

(where f  g) and is not linked to the monetary value

of NRGcoins on the market.

At any chosen time slot, P joins an exchange mar-

ket with an order to sell m of her NRGcoins in ex-

change for Euro. Similarly, a given consumer C de-

cides to buy n NRGcoins with Euro and places a buy

Figure 1: Example scenario.

order in the market. If according to the market regula-

tions these two bids are matched, the seller P releases

n ≤ m NRGcoins to the buyer and receives her asked

Euro from the market, while the buyer C releases the

offered Euro amount to the seller and receives his n

NRGcoins. He then pays h(y, t

, t

) NRGcoins for his

energy consumption of y to local substation S. Note

that P and C need not belong to the same substa-

tion. In analogy to the price function for production

g(·), the price function for consumption h(·) is de-

ﬁned by the DSO. Below we elaborate on the two

components of our mechanism — the trading of en-

ergy using NRGcoins (left part of Figure 1) and ex-

changing the latter for ﬁat currency (right half of the

ﬁgure). We also describe how the above price func-

tions balance energy supply and demand, how agents

can use learning mechanisms to increase their proﬁts

and what the beneﬁts of the new currency are.

2.1 Buying and Selling Energy

Each prosumer may use her produced energy to cover

her own demand ﬁrst. The excess energy is then fed

into the grid. Information of local energy production

and consumption is sent from smart meters of pro-

sumers to the street-level low voltage energy substa-

tions of the DSO at 15-minute intervals. This infor-

mation is then used to determine the rates at which

prosumers are rewarded for their produced energy

and consumers are billed for their withdrawn energy.

These rates (or functions) are designed to incentivize

agents to balance supply and demand, as well as lower

production and consumption peaks. For example,

during times of low demand or high production, the

cost of energy consumption is low, and analogously,

low supply or high demand drive the prices up. Thus,

every 15 minutes each street-level substation deter-

mines the rates for energy consumption and for pro-

SMARTGREENS2014-3rdInternationalConferenceonSmartGridsandGreenITSystems

102

duction using the following functions. The price func-

tion g(·) for paying producers is shaped as a bell curve

and deﬁned as:

g(x, t

, t

) =

x · q

−t

)

(1)

where q

is the maximum rate at which produc-

ers are rewarded for their input energy x when total

supply t

matches total demand t

and it is deﬁned

by the DSO; and a is a scaling factor for the case

where t

6= t

. When total energy production com-

pletely covers total consumption, the function is at its

peak and simpliﬁes to g = x·q

. On the other hand,

when t

 t

or t

 t

, producers are paid at a rate of

g → 0 NRGcoins. The price function h(·) according

to which consumers pay for their withdrawn energy y

is deﬁned as:

h(y, t

, t

) =

y · r

t

·t

(2)

where r

t

is the maximum cost of energy delivered

by the DSO when the energy supply by prosumers

is low. When production matches consumption, on

the other hand, the substation charges consumers with

t

per kWh. Lastly, when t

 t

then h → 0 and

thus the cost of consumed energy during overproduc-

tion is close to 0, motivating consumers to shift their

energy usage to periods of overproduction.

2.2 Earning and Exchanging NRGcoins

Similarly to Bitcoin (Nakamoto, 2008), NRGcoin is

not issued or controlled by any central authority and

its monetary value is determined solely by trading the

currency on an open exchange market – higher de-

mand for NRGcoins increases their monetary value,

while a large number of sells drives their value down.

However, unlike Bitcoins, which are generated by

sheer computing power and hence energy expenditure

(in a process called “mining”), NRGcoins are gener-

ated by injecting locally produced renewable energy

to the grid. The rate f (x) at which NRGcoins are

generated depends only on the amount x of renew-

able energy fed into the grid. This amount is broad-

cast

by the smart meter of the producer to all other

smart meters running the NRGcoin protocol, allow-

ing all participants in the NRGcoin network to keep

track of the earnings of each smart meter and their

transactions. Note that although transaction informa-

tion is associated with smart meters, the latter are

not publicly linked to actual prosumers and therefore

all earnings and transactions are anonymous as far

It is assumed here that security mechanisms are in place

to prevent tampering with the smart meter.

as agents are concerned. The process of generating

NRGcoins draws parallels to the process of mining

in the Bitcoin protocol and similarly, the bookkeep-

ing of earnings and transactions resembles the Bitcoin

blockchain (Nakamoto, 2008). NRGcoins are earned

according to function f (·) deﬁned as:

f (x) = b · x (3)

where b is a constant specifying the rate at which

NRGcoins are rewarded to prosumers for their in-

jected energy x and is deﬁned by the NRGcoin pro-

tocol, running on all smart meters.

As mentioned in Section 2.1, in addition to the

NRGcoins generated by injecting energy to the grid,

the local substation rewards prosumers based on cur-

rent energy supply and demand at that substation.

Note that the DSO does not issue the currency, but

simply collects and distributes payments, based on the

consumption and production price functions.

To procure or sell NRGcoins agents participate in

an online currency exchange market. An agent who

needs NRGcoins (e.g. in order to pay for his en-

ergy consumption) can place a buy bid on the mar-

ket, and analogously, an agent with excess amount of

NRGcoins can submit a sell bid. Each buy bid con-

tains the requested amount of currency and the price

at which the agent is willing to buy. In addition, the

bid contains order conﬁgurations (Ilic et al., 2012),

such as whether the agent prefers partial or full match

of her bid, and whether the bid needs to be discarded

if not matched immediately, or can stay in the order-

book and possibly be matched at a later time. For ease

of exposition, in the remainder of this section we as-

sume that all bids can be matched partially and remain

in the orderbook if not matched immediately. When a

buy bid is submitted to the market, all sell orders with

a price lower than the buy price are matched (lowest

sell orders ﬁrst) until the buy quantity is fulﬁlled. Any

remaining unmatched buy quantity is added to the

orderbook. All sell bids are processed in analogous

fashion, starting with the highest buys ﬁrst. Thus, or-

ders are matched only if the buy price is higher than

or equal to the sell price. The buyer pays the price he

has speciﬁed in his bid and the seller — her speciﬁed

sell price. The owner of the market earns proﬁt from

the difference between matched buy and sell bids, as

well as a possible commission fee to keep the market

running.

The smart meters of agents can employ learning

techniques that automatically determine the optimal

quantity of NRGcoins to trade in the market and an

acceptable bidding price. The learning mechanism se-

lects a bid quantity that aims to minimize the amount

of excess currency, i.e. the difference between the

current amount of NRGcoins the prosumer owns and

NRG-X-Change-ANovelMechanismforTradingofRenewableEnergyinSmartGrids

103

the amount it is expected to need in the future. In ad-

dition, the bid price is determined by observing the

inside-market, i.e. the difference between the lowest

outstanding sell and the highest outstanding buy in the

orderbook, and taking into account the risk preference

of the prosumer. For example, placing a very high

selling price bears a high return, but also a high risk,

meaning that the probability of ﬁnding an appropriate

match with a consumer agent is low. Thus the learn-

ing mechanism aims to maximize the revenue of the

agent, considering its preferences.

Bidding strategies for trading agents have been a

hot research topic for the last decade. For example,

the Power Trading Agent Competition (TAC)

(Ketter

et al., 2011) is a yearly competition simulating future

retail electric power markets. The agents in the mar-

ket act as retail brokers in a local dwelling, purchas-

ing power from a retail market as well as from local

sources, such as homes and businesses with solar pan-

els, and selling power to local customers and into the

wholesale market. Retail brokers use learning mech-

anisms for their bidding strategies in order to make

proﬁt, while balancing supply and demand (Reddy

and Veloso, 2011). As the environment involves a

highly dynamic setting with competitive agents, adap-

tive algorithms that learn by observation (Kuate et al.,

2013) have proven to be very successful at this com-

petition. The difference between our approach and

Power TAC is that the latter involves self-interested

brokers that aim at making proﬁt through offering

electricity tariffs to customers and trading energy in

the wholesale market. The brokers attempt to contract

consumers, prosumers and electric vehicle customers

by offering speciﬁc tariffs and by negotiating individ-

ual contracts. The brokers balance the ﬂuctuating en-

ergy demands of their contracted power consumers

against the actual output of their contracted energy

producers. In our NRG-X-Change model, there is no

need for brokers or long-term contracts, as NRGcoins

are traded between consumers and prosumers directly.

2.3 Beneﬁts of the NRGcoin currency

NRGcoins offer a number of advantages over tradi-

tional money and other digital currencies. According

to our mechanism, locally produced renewable energy

is continuously “converted” to NRGcoins. Their ad-

vantage over ﬁat currency is that they serve as the

right to receive an equivalent quantity of energy in the

future independent of NRGcoin market value. There-

fore, what this new “green currency” brings for agents

is security towards increasing energy prices, by for

example purchasing NRGcoins at low prices and then

http://www.powertac.org/

spending them on energy when prices are high. Thus,

the currency can be spent to buy renewable energy

at a later point in time, or traded for ﬁat money on

a market, whichever is more proﬁtable for the agent.

In this way NRGcoins act as a form of efﬁcient and

inﬁnite battery for agents, in the form of green certiﬁ-

cates for companies, or simply as a business of buying

and selling the currency for proﬁt. In general, it gives

individual agents accessible means to not only sup-

port renewable energy generation, but also invest in

the energy market as a whole — something that is not

trivial in the current state of affairs. The DSO, on the

other hand, beneﬁts from using NRGcoins as a “debt

instrument” with high liquidity, allowing it to quickly

convert this currency to cash. Paying prosumers with

NRGcoins instead of ﬁat currencies enables the DSO

to focus a larger portion of its cash assets on invest-

ments, rather than rely on bank credits and pay their

associated interest rates.

The new currency also resembles tradable green

certiﬁcates (TGCs) (Schaeffer et al., 2000; Morthorst,

2003) as a measure of produced renewable energy

and as a way to purchase its environmental attributes

by consumers. As such it can serve as a form of

competition between prosumers or an indication of

prestige, where companies can be valued for their

green energy production. Whereas TGCs only beneﬁt

producers by imposing purchase obligation to some

consumers, NRGcoins are, among others, a form of

investment and thus of potential advantage to both

types of agents, as described above. Moreover, unlike

TGCs, NRGcoins can be traded across countries and

serve as an international currency for green energy.

Similarly to other decentralized digital currencies,

NRGcoin is not regulated by any bank or central au-

thority and it is not tied to the stock market or ﬁat

currencies. However, it is generated by producing re-

newable energy, as opposed to any other digital cur-

rency that is mined by computing power and hence

energy expenditure. Although NRGcoin is not regu-

lated, it is dependent on its community. Therefore its

trade value can have large ﬂuctuations as a result of

market speculations.

It should be noted that the NRGcoin currency is an

added value to the NRG-X-Change mechanism and

not designed to be an indivisible part of it. The trading

of energy is also possible using ﬁat currency instead

of NRGcoins by modifying price functions g and h

to consider the value of the ﬁat currency, and drop-

ping function f . Nevertheless, detailed investigations

need to be carried out to determine to what extend

NRGcoins can be replaced by standard currency in an

initial phase, e.g. to simplify deployment.

SMARTGREENS2014-3rdInternationalConferenceonSmartGridsandGreenITSystems

104

2.4 Balancing Supply and Demand

Price functions g and h are designed to align the ob-

jectives of agents. Since the rates at which substa-

tions pay prosumers depend on local supply and de-

mand, different prosumers may earn different number

of NRGcoins for the same amount of injected energy

at different locations of the smart grid. Again, these

rates are independent from the current market value of

the NRGcoins. The difference in the rates is related to

the balance of local energy production and consump-

tion that the DSO strives to achieve, as well as for

ﬂattening supply and demand peaks. For example,

the value of generated energy in a neighborhood full

of producers will be much lower than the NRGcoins

that a single producer will earn in a neighborhood full

of consumers. Thus, the value difference imposed by

the DSO may stimulate consumers to install renew-

able energy generators and become producers, while

at the same time discourage excess production or con-

sumption that overload the transmission lines. Simi-

larly, consumers are motivated to shift their consump-

tion away from demand peaks and towards production

peaks, as that will lower their energy bill.

The more energy supply matches demand, the

more NRGcoins producers receive from the substa-

tion and the fewer coins are paid by consumers to the

substation, as the additional energy it needs to supply

to that neighborhood is low. In this way agents strive

to balance supply and demand, i.e. achieve demand

response, out of their own self-interest. Prosumers

are motivated to feed just enough renewable energy

to the grid, while consumers minimize their costs by

shifting their consumption pattern towards time slots

of higher production. Note that the parameters q

and r

t

of price functions 1 and 2 need to be care-

fully conﬁgured to ensure that the proﬁt of the DSO is

always positive and covers the costs of energy trans-

mission.

Learning techniques can help agents maximize

their revenue using the payments from the substation

as a feedback signal. For example, the learning mech-

anism can switch off (some of) the renewable energy

generators of the prosumer during times of overpro-

duction in order to maximize the proﬁt according to

Equation 1, while taking into account the agent’s own

consumption. Similarly, the energy bill of consumers

can be reduced by learning to shift the consumption

pattern to periods of high production, while preserv-

ing the comfort level of the agent. For example, the

learning mechanism can learn to shift the operation

of the washing machine to time periods when energy

is the cheapest, taking into account the requirements

of the agent that the operation should be completed

by a particular moment in time. The learning mech-

anism would have to advice the agent on how much

to consume at each time slot in order to minimize the

price it is expected to pay while taking into account

the electricity needs of the occupants. Inherently, this

is a scheduling problem with multiple objectives. Var-

ious machine learning algorithms have been proposed

to address multi-objective scheduling, such as rein-

forcement learning (Aissani et al., 2009), evolution-

ary algorithms (L

opez-Ib

nez et al., 2005) and local

search (Dubois-Lacoste et al., 2011).

Since locally produced energy nowadays covers

only a small percentage of consumption within neigh-

borhoods, the DSO still needs to produce electricity

to cover the total demand. Bunn and Farmer (Bunn

and Farmer, 1985) pointed out that a 1% decrease in

forecasting error implied £10 million savings in oper-

ating costs. Therefore, reliable forecasting techniques

are needed to improve the energy supply planning of

the DSO and decrease its costs. Several prediction

techniques can be applied here, such as autoregressive

methods (Contreras et al., 2003), artiﬁcial neural net-

works (Khamis et al., 2011), support vector machines

(Tan et al., 2010), etc. In addition to global energy

prediction, the DSO can aggregate predictions of in-

dividual local substations. The advantages of predict-

ing demand at substation level are twofold: on the one

hand individual local predictions can improve the ac-

curacy of the global prediction model by employing

weighted aggregation; while on the other hand these

predictions allow the DSO to exert better control over

the load of individual transmission lines and thus im-

prove the quality and robustness of the electric power

infrastructure.

3 SUMMARY AND OUTLOOK

In summary, instead of relying on a day-ahead energy

market to sell or purchase their energy, prosumers

simply inject to or draw from the grid, as is the cur-

rent state of affairs, but at prices that depend on mea-

sured supply and demand of energy. Payment is in

the form of NRGcoins, the value of which is deter-

mined based on trades in an open currency exchange

market. Using concepts from the rising in popularity

Bitcoin phenomenon, our novel mechanism creates a

microeconomic ecosystem that allows prosumers to

trade locally produced renewable energy at competi-

tive prices. At the same time agents are incentivized

to balance energy supply and demand out of their own

self-interest and thus ﬂatten production and consump-

tion peaks. Lastly, our proposed approach is scalable

— newly joining agents do not increase the complex-

NRG-X-Change-ANovelMechanismforTradingofRenewableEnergyinSmartGrids

105

ity of the energy trade thanks to the local substations,

or of the currency exchange, as the NRGcoin protocol

is decentralized.

As this concept is still work-in-progress, extensive

simulations need to be carried out, backed up by mi-

croeconomic theories, to determine the parameters of

the price functions of the DSO and the rate at which

NRGcoins are generated in the network. Last but not

least, special attention needs to be paid to the privacy

and security aspects of the NRGcoin protocol and in

the design of the smart meter middleware.

ACKNOWLEDGEMENTS

We would like to thank Ildefons Magrans de Abril

for many fruitful discussions. In addition, we would

like to acknowledge the comments and suggestions of

the anonymous referees. The research presented in

this article is funded by the FP7 framework’s Marie

Curie Industry-Academia Partnerships and Pathways

(IAPP) project SCANERGY, grant agreement num-

ber 324321.

REFERENCES

Aissani, N., Beldjilali, B., and Trentesaux, D. (2009).

Dynamic scheduling of maintenance tasks in the

petroleum industry: A reinforcement approach. Eng.

Appl. Artif. Intell., 22(7):1089–1103.

Bunn, D. and Farmer, E. (1985). Comparative models for

electrical load forecasting. Wiley.

Contreras, J., Espinola, R., Nogales, F. J., and Conejo, A. J.

(2003). Arima models to predict next-day electric-

ity prices. Power Systems, IEEE Transactions on,

18(3):1014–1020.

Dubois-Lacoste, J., L

opez-Ib

nez, M., and St

utzle, T.

(2011). A hybrid TP+PLS algorithm for bi-objective

ﬂow-shop scheduling problems. Computers & Oper-

ations Research, 38(8):1219–1236.

Ilic, D., Da Silva, P. G., Karnouskos, S., and Griesemer,

M. (2012). An energy market for trading electricity in

smart grid neighbourhoods. In 6th IEEE International

Conference on Digital Ecosystems and Technologies

(DEST), pages 1–6. Ieee.

Ketter, W., Collins, J., Reddy, P., Flath, C., and Weerdt, M.

(2011). The power trading agent competition. ERIM

Report Series Reference No. ERS-2011-027-LIS.

Khamis, M., Baharudin, Z., Hamid, N., Abdullah, M.,

and Solahuddin, S. (2011). Electricity forecasting

for small scale power system using artiﬁcial neural

network. In 5th International Power Engineering

and Optimization Conference (PEOCO), pages 54–

59. IEEE.

Kok, J., Warmer, C., and Kamphuis, I. (2005). Power-

Matcher: multiagent control in the electricity infras-

tructure. In Proceedings of the fourth international

joint conference on Autonomous agents and multia-

gent systems, pages 75–82.

Kok, K., Roossien, B., Macdougall, P., and Pruissen, O. V.

(2012). Dynamic Pricing by Scalable Energy Man-

agement Systems - Field Experiences and Simulation

Results using PowerMatcher. In IEEE Power and En-

ergy Society General Meeting, pages 1–8.

Kuate, R. T., He, M., Chli, M., and Wang, H. H. (2013). An

intelligent broker agent for energy trading: An mdp

approach. In Rossi, F., editor, IJCAI. IJCAI/AAAI.

opez-Ib

nez, M., Prasad, T. D., and Paechter, B. (2005).

Optimal pump scheduling: Representation and mul-

tiple objectives. In Savic, D. A., Walters, G. A.,

King, R., and Thiam-Khu, S., editors, Proceedings

of the Eighth International Conference on Computing

and Control for the Water Industry (CCWI 2005), vol-

ume 1, pages 117–122, University of Exeter, UK.

Mockus, J. (2012). On Simulation of the Nash Equilib-

rium in the Stock Exchange Contest. Informatica,

23(1):77–104.

Morthorst, P. (2003). A green certiﬁcate market com-

bined with a liberalised power market. Energy policy,

31(2003):1393–1402.

Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic

cash system. Consulted, pages 1–9.

Olson, M. A., Rassenti, S. S. J., Smith, V. L., Rigdon, M. L.,

and Ziegler, M. J. (1999). Market design and moti-

vated human trading behavior in electricity markets.

In Proceedings of the 32nd Annual Hawaii Interna-

tional Conference on System Sciences, pages 1–27.

Reddy, P. and Veloso, M. (2011). Learned Behaviors of

Multiple Autonomous Agents in Smart Grid Markets.

In Proceedings of the Twenty-Fifth AAAI Conference

on Artiﬁcial Intelligence, pages 1396–1401.

Schaeffer, G., Boots, M., Mitchell, C., Anderson, T., Timpe,

C., and Cames, M. (2000). Options for design of trad-

able green certiﬁcate systems. Netherlands Energy

Research Foundation ECN.

Tan, Q.-L., Yuan, H.-Z., Tan, Y.-K., Zhang, X.-P., and Li,

X.-F. (2010). Energy consumption forecasting us-

ing support vector machines for beijing. In Inter-

national Conference on E-Product E-Service and E-

Entertainment (ICEEE), pages 1–4. IEEE.

Vytelingum, P., Cliff, D., and Jennings, N. (2008). Strate-

gic bidding in continuous double auctions. Artiﬁcial

Intelligence, 172(14):1700–1729.

Vytelingum, P., Ramchurn, S., Voice, T. D., Rogers, A., and

Jennings, N. (2010). Trading agents for the smart elec-

tricity grid. In 9th Int. Conf. on Autonomous Agents

and Multiagent Systems (AAMAS 2010), pages 897–

904.

SMARTGREENS2014-3rdInternationalConferenceonSmartGridsandGreenITSystems

106