Modelling the Emergence of Functioning Natural Gas Wholesale Markets G. Bas - [email protected]

Abstract Tue European Commission considers liberalized and liquid natural gas wholesale markets as a priority. Agent-based models are considered suited to provide insight in what market conditions contribute to the emergence of functioning natural gas wholesale markets. We present an abstraction of the natural gas market to explore the emergence of functioning wholesale markets. For this abstraction we apply graph theory, supplier selection theory, auction theory, and insights from finance to the natural gas market. The abstraction is developed into a simulation model, of which different aspects have been validated with economic logic. In this research we demonstrate that the applied theories and methods can form a feedback loop that allows us to explore the evolution of marketplaces; and that the abstraction is valid in the context of natural gas markets. Minor adjustments to the abstraction allow it to be used to explore other markets (e.g. biogas). The abstraction can be extended by considering more than 1 wholesale market, including more aspects of the physical infrastructure, and extent the behaviour of market participants by considering strategic behaviour.

Keywords: agent-based modelling · bilateral negotiation · contract selection · natural gas · socio-technical system · wholesale market

1

Introduction

The European Commission considers centralized trading of natural gas at wholesale markets important for realizing a real competitive natural gas market (Spanjer, 2008). Traditionally natural gas is traded via long-term gas supply contracts, of which the terms are negotiated bilaterally between a producer and a customer (Stern, 2007). In contrast to trading at wholesale markets, trading via long-term gas supply contracts is decentralized and therefore is considered to limit competition. Multiple wholesale markets have been established in Europe, but apart from the NBP in Britain and the TTF in the Netherlands1 , the liquidity of none of these hubs is considered high enough to provide a reliable price signal (Heather, 2012). The liquidity of these wholesale markets increases when more market participants decide to trade natural gas at the wholesale market, rather than via long-term gas supply contracts. However, regulation of gas networks cannot oblige market participants to enter a market, it can only create create conditions 1 A minimum churn rate (the measure of liquidity commonly used) is at least 10. The churn rates of the NBP and TTF are respectively around 21 and 14, and thus meet the minimum required churn rate. However, in comparison to the Henry Hub in the United States, which has a churn rate of around 377 (Konoplyanik, 2011), the European natural gas hubs are only barely liquid.

1

that make it more likely that they are willing to do so (Glachant, 2011). To explore what market conditions contribute to the emergence of functioning natural gas wholesale markets, we propose the development of a simulation model. For the development of the simulation model, we consider the natural gas market from a socio-technical system perspective (cf. Ottens, Franssen, Kroes, & van de Poel, 2006). The natural gas market consists of a technical subsystem (of physical artefacts for the production, transportation and consumption of natural gas) and a social subsystem (of market participants and their relationships for the change in ownership and the transportation of natural gas). Nikolic and Kasmire (2012) argue that “through modelling these [socio-technical] systems in light of the principles of complex adaptive systems, we can better understand the specific systems and how to interact with them in order to achieve goals.” Of the existing modelling paradigms, agent-based modelling is considered the best suited to represent these socio-technical system (Nikolic & Kasmire, 2012). Agent-based models (ABMs) are models that consist of a number of agents which interact both with each other and with their environment, and can make decisions and change their actions as a result of this interaction (Ferber, 1999). This individual decision making and interaction among agents makes that this modelling paradigm is particularly suited to represent the trading of natural gas. In this article we present an agent-based model of a single natural gas market. In this market the market participants have to decide whether to trade on a wholesale market or via a long-term contract, and they negotiate with each other to determine the quantity to trade and the price to trade that quantity for. In their decisions they take into account the perceived historical risk and price of both ways of trading natural gas. Aggregate behaviour, such as the liquidity of the wholesale market, emerges through these decisions and interactions. This allows us to research how the market conditions affect the behaviour of market participants and, through their individual behaviour, the condition of the wholesale market. First, we present a conceptual model of a natural gas system, which later is developed into a simulation model. In section 2 we present an overview of all components (market participants, physical infrastructure, and contracts) considered in our system. In this section also the activities undertaken by market participants are introduced. The behaviour of the market participants is more extensively discussed in section 3. In section 4 we discuss the implementation of the conceptual model into a simulation model. In section 5 we discuss the validation of the simulation model with economic logic. After this, we present the conclusions from this paper.

2

System overview

The system presented in this section is the basis of the agent-based model. It consists of different types of components, that all have individual characteristics, and are related to each other. The components that represent agents also have behavioural rules. In this section we present an overview of the components considered in our system, and also will we briefly introduce the activities undertaken by the agents.

2

The system considered consists of market participants (producers, customers, and traders) of which the first two are physically connected to an entry-exit area2 via pipelines. The capacity of these pipelines determines the quantity of gas that the market participants can either put into or take out the entryexit area. Transmission capacity within the entry-exit area is assumed to be unlimited, so that natural gas can flow unhindered from an entry point to an exit point. This allows us to exclude the transmission of natural gas from the system. The gas exchange, which functions as an abstraction of the wholesale market, is located at a virtual point in the entry exit area. Since the transmission of natural gas in the entry-exit area is unhindered, the virtual point is considered to be the entry-exit area, and therefore the gas exchange is located at the intersection of all pipelines. All 3 market participants have an interest in a certain quantity of natural; producers have a stable production capacity (MWh/mo) that they want to sell to generate the highest profit possible; customers have a monthly fluctuating demand (MWh/mo) for their business activities that they want to procure for the lowest price possible.; and traders have a certain capital that they want to invest to generate the highest profit possible. The behaviour of the market participants is aimed at securing these interests. Natural gas is traded by means of contracts, which specify who will deliver what quantity of gas to whom, at what price and at what time this will occur. In this system we consider only two types of contracts, which represent the “traditional” bilateral way of trading and the “new” centralized way of trading. • The first type is the bilateral contract, which is comparable to the longterm gas export contract (discussed by Konoplyanik (2010)). This contract is signed to ensure a flow of natural gas for a period of time (between the start and end date of the contract). The quantity of the contract is delineated by a take-or-pay clause3 with a possible deviation of 25%. The price of 1 MWh delivered under the bilateral contract is negotiated during a bilateral negotiation, and the destination clause ensures that the delivered gas cannot be resold. • The second type of contract is the structured contract. This is a standardized contract, which is traded at the exchange. The quantity is fixed at 1 MWh, and that quantity will be delivered in a single month, which is indicated by the expiration date of the contract. The destination of a structured contract is always the gas exchange, and the origin is a market participant. Since the destination of every structured contract is the gas exchange these contracts can easily be re-traded, which entails that gas can be traded multiple times before it is actually delivered. The (high-level) activities undertaken by the market participants to trade natural gas, and thereby secure their interests, are presented in figure 1. These activities can be divided into two phases; selection and execution4 . In the selection phase the producers and customers determine how they want to respectively sell or procure natural gas. These activities are only performed with a certain interval, which differs per market participant. 2 3 4

For a discussion of entry-exit areas, see Kema Nederland (2011). For a discussion of the take-or-pay clause, see Konoplyanik (2010) Sections ?? and ?? extensively discuss the activities that make up these phases.

3

In the execution phase, the market participants try to execute the intends determined in the selection phase. The first activity in this phase is that the producers and customers, that want to sell or buy natural gas via bilateral contracts, perform bilateral negotiations to determine the terms of the bilateral contracts. Once the bilateral negotiations are finished, all market participants bid for or offer the structured contracts they want to buy or sell. Every bid or offer that is submitted to the exchange is directly processed by the gas exchange, resulting in a signed structured contract or adding the order to the exchange’s limit order book.

3 3.1

Agent behaviour Selection phase

The selection phase consists of two activities; select quantity to procure through each type of contract conducted by customers, and select quantity to sell through each type of contract conducted by producers. 3.1.1

Quantity to buy

To determine the quantity to procure through each type of contract, the customers determine the quantity to procure through structured contracts, by balancing risk and return of buying natural gas through structured contracts. The remaining demand, then, is procured through bilateral contracts. To balance risk and return, the customers use the constant absolute risk aversion (CARA) utility. The risk involved with buying natural gas via structured contracts is that the customer is not able to procure the natural gas desired, and the return is the difference in price of bilateral and structured contracts5 . Equation 1 show how the quantity to procure through bilateral contracts (qb ) is determined. In this equation qavg is the average monthly demand in the current planning period. The CARAexchange is calculated in Equation 2. The numerator indicates how much more expensive natural gas procured through bilateral contracts is, by comparing the costs of procuring natural via bilateral contracts (cb ) with the costs of procuring natural gas via structured contracts (cs ). The denominator indicates the risk involved with procuring natural gas through structured contracts. For this the customer considers the chance that an order submitted to the exchange results in a signed structured contract (successexchange ) and its individual risk aversion (λ). successexchange is based on considering a subset (due to bounded rationality) of the orders submitted to the exchange, and determining the percentage that has resulted in a signed structured contract. It thereby represents the perceived chance that a submitted order results in a signed structured contract. qb = (1 − CARAexchange )qavg

(1)

5 The customer’s return is positive when the price of bilateral contracts is higher than that of structured contracts. This implies that it is possible that the return is negative.

4

Figure 1 – High level overview of activities

CARAexchange

  0 = 1  

(cb/cs )−1

λ/success

3.1.2

exchange

if CARAexchange ≤ 0; if CARAexchange ≥ 1; else.

(2)

Quantity to sell

Just as customers consider the security of supply when they determine how to procure natural gas, the producers consider the security of demand when they determine how to sell natural gas (van der Linde & Stern, 2004). In order to balance the risk and return involved with selling natural gas through structured contracts, the producers also calculate the CARA utility. The risk in this case is that the customer is not able to sell the desired quantity of natural gas. For producers the return is the difference in price of bilateral and structured contracts, but differs from the customers in that the producer’s return is positive when the price of structured contracts is higher than that of bilateral contracts. Equation 3 indicates how the producers determine what quantity to sell through bilateral contracts (qb ). In this equation capprod represents the maximum monthly production capacity of the producer, being the total quantity of natural gas that is available for sale. Equation 4 shows how the CARA utility is calculated. For producers the numerator of the calculation indicates how much higher the profit from selling natural gas through structured contracts (πs ) is than the profit from selling natural gas through bilateral contracts (πb ). The denominator indicates how the producer perceives the risk of selling natural gas through structured contracts, which is calculated in the same way as the customer calculates its perceived risk. qb = (1 − CARAexchange )capprod

5

(3)

CARAexchange

  0 = 1  

(πs/πb )−1

λ/success

3.2

exchange

if CARAexchange ≤ 0; if CARAexchange ≥ 1; else.

(4)

Execution phase

In the execution phase the market participants undertake activities to execute the intends determined in the selection phase. The three activities making up this phase are bilteral negotiation, submit desired orders to the exchange, and process orders. 3.2.1

Bilateral negotiation

Bilateral negotiations are the negotiations between a single producer and customer on the terms of bilateral contracts. The bilateral negotiations are abstracted as ascending clock auctions undertaken by producers. The ascending clock auctions6 are linked to each other through the customers considering which auction allows them to procure natural gas for the lowest price. The steps undertaken for a bilateral negotiation are: 1. All producers that have available capacity to sell through bilateral contracts communicate a range of prices7 for which they want to know the quantity that the customers are willing to procure. The shaded area in figure 2 indicates the range of prices communicated by a producer. 2. The customers determine the quantity they are willing to procure for the prices in the range, and communicate this (bid) to the producer of the specific range. The customer uses its reservation price to determine what quantity it wants to procure. For a price lower than its reservation price it wants to procure its demand; for a price higher than its reservation price it wants to procure nothing. In figure 2 this bid is represented by the bold part of the demand curve in the upper 3 graphs. For example, customer 2 has a stable demand for all prices in the range, while at a certain price (its reservation price) in the range the demand of customer 1 drops to 0. 3. The producers aggregate all bids they receive and thereby determine the quantity of natural gas they can sell for all prices in the range they specified in step 1. The aggregated demand is presented by the bottom graph in figure 2. This graph clearly indicates that when the price increases, the demand for natural gas decreases. 4. The producers determine whether there is a price in the range for which demand equals supply, called a clearing price. The clearing price is where production capacity (vertical striped line) and demand (bold line) cross. 6

For a discussion on ascending clock auctions, see Cramton (1998). In the first round of an ascending clock the lower bound of the range is equal to the marginal costs of the producer, and the upper bound is 10% higher. In the rounds that follow, the producer considers the quantity demand for in the previous round. If that demand is higher than supply, the range becomes higher; if that demand is lower than supply, the range becomes lower. If a range becomes higher, the upper bound of the old range becomes the lower bound of the new range and the new upper bound is 10% higher than the new lower bound. If the range becomes lower, the old lower bound becomes the new upper bound and the new lower bound is 90.9% of the new upper bound. 7

6

(a) If this is not the case, the producers update their range of prices (higher if demand exceeded supply, and lower if supply exceeded demand) and start at step 1 again. (b) If there is a clearing price, the producer sends an offer to the customers that are willing to pay the clearing price or more. The quantity offered to the customers is represented by the horizontal position of the customer’s bid. Customer 2 gets offered all the quantity it has bid for, while customer 1 only gets offered a part of what it has bid for, and customer 3 does not get offered anything. The price of the offer is the clearing price; in the case of customer 1 this is the price it was willing to pay for the natural gas, while for customer 2 this is less than it was willing to pay. 5. The customers compare the offers they have received and signs contracts with the producer(s) that is (are) willing to accept the lowest price. 6. Unless all quantity is sold or the producer believes there is no more demand, the producer starts a new round (at step 1). 3.2.2

Submit orders

To submit orders to the exchange, the market participants first have to determine what orders they want to submit. The way in which this occurs differs per market participant. However, the behaviour of producers and customers (with regard to submitting orders) is comparable and thus is discussed together. The behaviour of traders differs significantly and thus is discussed separately. For determining which orders to submit producers (customers) identify for every remaining month in their planning period what quantity to sell (procure) through structured contracts. This quantity is determined through subtracting the already contracted quantity or the quantity reserved for bilateral contracts (depending on which one is higher) from the production capacity (demand) in a particular month. A positive quantity implies that the producer (customer) wants to sell natural gas, while a negative quantity implies that the producer (customer) wants to buy natural gas. The limit price of the order depends on whether it is an offer (to sell) or a bid (to buy). When an offer is submitted, the limit price of that offer is equal to the marginal costs of the producer (customer); when a bid is submitted, the limit price of that bid is equal to the reservation price of the producer (customer). Before the producer (customer) actually submits its new orders, it removes its old orders, so that it is not possible that the producer (customer) duplicates orders it has submitted previous months. Since traders are not concerned with physical gas, but merely want to benefit from price fluctuations, they have another way of determining what orders to submit to the exchange. A trader starts with determining what the profit will be from either buying or selling a specific natural gas contract. Equation 5 and 6 present how the trader calculates the expected profit from respectively buying “long” or selling “short” a risky natural gas contract with expiration date i. In these equations h is the length of the planning period of the trader, pexp,i is the expected price of a structured contract at its expiration date, and pmkt,i is the

7

Figure 2 – Illustration of the processing of bids for the bilateral negotiation

current market price of a structured contract with expiration date i. Erisk,i =

pexp,i h−i r pmkt,i

(5)

pmkt,i h r pexp,i

(6)

Erisk,i =

After determining the expected profit for every structured contract with an expiration date in the planning period of the trader, the trader determines every couple of months what percentage of its capital to invest in “risky” natural gas contracts. For this the trader uses the CARA utility function, in which it compares the maximum return of a structured contract with the risk free rate of return. Equation 7 indicates how the percentage to invest in natural gas contracts is calculated. For this, the trader balances the profit above the risk free rate (r) with the perceived risk of trading at the gas exchange. This risk is represented by the variation of the market prices, σ 2 . After this the trader calculates how much structured contracts it can buy (sell) for the capital it is willing to risk. It does this by dividing the capital it wants to risk by the market price of the contracts it wants to buy (sell). Once this is done the trader determines what the limit price of the order will be, which is done by determining at what market price of the best performing contract, its return will be equal to the profits of the second best performing contract. Submitting the orders to the exchange concludes the activities of the trader with regard to the orders to submit to the exchange. x= 3.2.3

Erisk − (1 + r) λσ 2

(7)

Process orders

When an order is submitted to the gas exchange, it directly processes the order by consulting its limit order book to match supply and demand. The steps it 8

performs for this are: 1. The gas exchange receives a bid (offer) with a certain limit price. 2. The exchange checks its order book for offers (bids) with a limit price that is lower (higher) than the limit price of submitted bid (offer). These offers (bids) are called possible matches. (a) If there are no possible matches, the submitted bid (offer) is added to the order book. (b) If there are possible matches, the submitted bid (offer) is matched with the offer (bid) with the lowest (highest) limit price. In that case two contracts are signed; one for delivery to the exchange and one for delivery by the exchange. The price of the contracts is the price of the offer (bid) that was already in the order book. The quantity is either the quantity of the submitted bid (offer) or the quantity of the offer (bid) that was already in the order book; depending on which of the two orders has the lowest quantity. i. If the quantity of the submitted bid (offer) is higher than the quantity of the offer (bid) that was already in the order book, after signing the contract, the offer (bid) is removed from the order book and the bid (offer) is matched with the next best match (see step 2b). This continues until the entire bid (offer) has resulted in signed structured contracts or there are no more possible matches. ii. If the quantity of the submitted bid (offer) is lower than the quantity of the offer (bid) that was already in the order book, after signing the contract, the quantity of the offer (bid) is decreased by the quantity of the contract.

4

Software implementation

The conceptual model, of the system discussed in the 3 previous sections, is implemented in software by using the AgentSpring framework. Using this framework implies that all components8 of the model are stored at the vertices of a graph database, with the relationships among them being abstracted as edges of the graph. Another implication of using the AgentSpring framework is that the agents and their behaviour are decoupled. The behaviour, then, is recorded in scripts, which are “acted” by the agents. The scripts roughly match the activities discussed in section 3. For more information with regard to the implementation, we refer to (Bas, 2012) or to the software code, available from https://github.com/EMLab/.

5

Model validation

The simulation model is validated through comparing the system’s behaviour (in differing scenarios) with economic logic. The validation is performed through 4 different cases, which all highlight a different aspect of the model. However, 8 Producers, customers, traders, exchange, pipelines, bilateral contracts, and structured contracts.

9

in this paper we present a single case, which is concerned with the availability of physical gas. Clingendael International Energy Programme (2008) concludes that when there is a buyer’s market (an oversupply of natural gas) the buyers are able to dictate the terms of natural gas contracts and that therefore more natural gas will be traded at wholesale markets. However, when there is a seller’s market (an undersupply of natural gas) less natural gas will be traded at wholesale markets. For this case we consider 3 different scenarios, which differ through the ratio of physical supply to physical demand. In the oversupply scenario the ratio of supply to demand is 5, in the neutral scenario the ratio of supply to demand is 1, and in the undersupply scenario the ratio of supply to demand is 0.2. For every scenario we determine the extent to which a functioning wholesale market has emerged by considering 2 indicators; 1) the churn rate of the gas exchange, and 2) the market share of structured contracts. Thus when the ratio of supply to demand increases, economic logic dictates that the churn rate of the gas exchange increases and that the market share of structured contracts increases. If the model is valid, it should follow this logic. Figure 3 presents how the churn rate develops over time in the 3 scenarios discussed before. The graph indicates that when the supply to demand ratio increases the churn rate of the gas exchange increases as well. It also indicates that the difference between the scenarios is substantial and sustainable over time.

Figure 3 – The churn rate at the gas exchange for different supply to demand ratios

The second indicator, presented in figure 4, confirms this. When the supply to demand ratio increases, also the market share of structured contracts increases. As is the case with the churn rate, the difference between the scenarios is both substantial and sustainable. These observations lead to the conclusion that when there is a buyer’s market the wholesale market prospers, and when there is a seller’s market the development of the wholesale market is hampered. This means that with regard to the availability of physical natural gas the behaviour of the model is in line with 10

Figure 4 – The market share of structured contracts for different supply to demand ratios

economic logic and is therefore validated.

6

Conclusions

In this paper we present an agent-based model of trade in the natural gas market to explore the emergence of functioning wholesale markets. In order to model the actual behaviour of market participants, we applied supplier selection theory, auction theory, graph theory, and insights from finance to the natural gas market. In combination with discussions of the natural gas markets (e.g. Heather, 2012), these theories and methods were developed into an abstraction of the behaviour of market participants in the natural gas market. The behaviour of market participants is “neo-classical” utility maximizing, which implies that strategic behaviour was excluded from the abstraction. In this research we have demonstrated that the applied theories and methods can be used to create a feedback loop that allows us to explore the evolution of marketplaces. We have validated the abstraction in the context of the natural gas market, by comparing the system’s behaviour with economic logic. This abstraction may also be applied to other markets, where wholesale markets have not emerged yet (e.g. biogas). To further research the emergence of functioning natural gas wholesale markets, we recommend to extend the presented abstraction. Potential extensions for the abstraction are 1) considering more than 1 wholesale markets, and connecting these markets, 2) include the physical infrastructure in the abstraction, and 3) extend the behaviour of market participants by including strategic behaviour.

11

References Bas, G. (2012). Modelling the Emergence of Functioning Natural Gas Wholesale Markets. Clingendael International Energy Programme. (2008). Pricing Natural Gas: the outlook for the European market. Cramton, P. (1998, May). Ascending auctions. European Economic Review , 42 (3-5), 745–756. doi: 10.1016/S0014-2921(97)00122-0 Ferber, J. (1999). Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence. Boston: Addison-Wesley Longman Publishing Co., Inc. Glachant, J.-M. (2011). A Vision for the EU Gas Target Model: The MECO-S Model. Heather, P. (2012). Continental European Gas Hubs : Are they fit for purpose ? (No. June). Kema Nederland. (2011). Dutch gas value chain: A high level description of gas the value chain and roles and responsibilities of different stakeholders (Tech. Rep.). Konoplyanik, A. A. (2010). Pricing gas: evolution not revolution. Energy Economist(349), 6–8. Konoplyanik, A. A. (2011). How market hubs and traded gas in European gas market dynamics will influence European gas prices and pricing (No. February). Nikolic, I., & Kasmire, J. (2012). Theory. In K. van Dam, I. Nikolic, & Z. Lukszo (Eds.), Agent-based modelling of socio-technical systems (pp. 11–72). Ottens, M., Franssen, M., Kroes, P., & van de Poel, I. (2006). Modelling infrastructures as socio-technical systems. International Journal of Critical Infrastructures, 2 (2-3), 133–145. Spanjer, A. (2008). Structural and regulatory reform of the European natural gas market Does the current approach secure the public service obligations? Unpublished doctoral dissertation. Stern, J. (2007). Is there a rationale for the continuing link to oil product prices in continental European long-term gas contracts? International Journal of Energy Sector Management(April). van der Linde, C., & Stern, J. (2004). The future of gas: Will reality meet expectation? Background Paper for the 9th International Energy Forum.

12