Gas Balancing Variable PowerGeneration

(1)

DEGREE PROJECT, IN ELECTRIC POWER ENGINEERING , SECOND LEVEL STOCKHOLM, SWEDEN 2014

Gas Balancing Variable Power Generation

A SYSTEMIC CASE STUDY

BAPTISTE BORTOT

KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ELECTRICAL ENGINEERING

(2)

Gas balancing variable power generation

A systemic case study

BAPTISTE BORTOT

2013-2014

Thesis conducted at EDF R&D, France

(3)

Abstract

With the increasing share of variable renewable generation, balancing electric power systems could become a major concern for system operators because of their variable and hardly predictable nature. However, gas technologies appear as a solution to provide this flexibility, but the impacts on the gas power system have hardly been investigated.

In this thesis, consulting reports on the subject matter, regulator suggestions and gas-electricity interaction models in scientific literature are studied and four sources are identified to be used for balancing: linepack, storage facilities, liquefied natural gas and intraday gas supply from adjacent areas. Then, a gas-electricity model for flexibility supply is designed and three case studies are simulated in order to analyze both gas and electric power systems’ behaviors. In these case studies, electricity generation, contribution of gas sources and costs are analysed.

The study concludes that critical situations on gas market that can occur, e.g. in cases of large variation in the net electricity demand and limited availability of linepack and storage facilities, the need of intraday modulation can exceed the possibilities to provide for it. Then, gas cannot be supplied to power plants during peak periods, and more gas than necessary is used during off-peak periods. The case studies also show that day-ahead forecast errors in variable renewable generation can be handled much easier than variations by the gas system but leads to higher costs.

(4)

Acknowledgements

First of all, I would like to express my deepest appreciation to my supervisor at EDF, Alvaro Andaluz-Alcazar, who supported me during all the thesis. He helped to over- come issues during my thesis by sharing his insight, giving me advices and directions for my work. He was also heavily involved in my good integration at EDF.

I would like also to express my gratitude to all EFESE group for answering my questions and making this experience very interesting and enjoyable. Special thanks to Laurent Gilotte and Vera Silva who gave me precious advices to design my model.

I also would like to thank my supervisor at KTH, Richard Scharff, who helped me a lot with my thesis writing and also gave me advices to move forward in my work.

(5)

Nomenclature

vRES : Variable renewable energy sources. They consist mainly of wind and solar and run-of-river hydro power. Wave and tidal, even if they are still marginal, are also variable renewable sources.

Ramping : Power variation per time unit.

Intraday modulation : Contribution of gas sources to modulate consumption around its daily average value. The need of intraday modulation is due to flat gas injections whereas consumption is varying during a day.

Modulated volume : Sum in absolute values of the difference between the hourly gas consumption and the daily average consumption.

Daily imbalances (on the gas network) : Difference between what was injected and withdrawn on the gas network during a day.

CCGT: Combined-cycle gas turbine. This is the most recent technology for generating electricity from gas, thus its efficiency is the highest of gas power plants.

OCGT: Open-cycle gas turbine. This is a classical turbine operate with gas. Its effi- ciency is lower than a CCGT.

Net electricity demand: Electricity demand minus vRES generation.

(8)

Chapter 1 Introduction

1.1 Background

Since 2000, the share of renewable energy in electricity generation is increasing considerably in Europe and is expected to increase even more in the next decades. One of the European 2020 targets is to raise the share of renewable energy to 20 percent of gross final energy consumption. Since hydrological resources are almost already fully exploited in Europe, other renewable sources like wind, solar and biomass are expected to grow significantly in the next year. In some scenarios for 2030 wind generation is expected to triple compare to 2010 [10, pp. 62-63].

In a large part of European countries variable renewable energy sources (vRES) are non-dispatchable. Indeed, a directive from European commission requires EU members to give priority to renewable energy sources for electricity generation. Wind and solar energy are vRES depending on weather conditions and therefore difficult to forecast.

This difficulty to forecast vRES generation leads to numerous and larger forecast errors. The total accumulated forecast errors in a balancing area have to be balanced with total consumption in real-time.

With an increasing share of renewable energy appears a new need: an additional flexible generation which can quickly compensate the random variation of renewable generation and thus keep the balance between production and consumption. This role is mainly performed by the conventional fleet: hydro, coal, gas and, in some countries, nuclear. Hydro power appears like an efficient driver at low cost to provide the flexibility needed on the electricity system due to its technical characteristics but water reserves are not always available, especially during periods of drought. Thus there are times when hydro cannot provide flexibility and thermal generation is called. For some nuclear power plants, such as the French ones, it is possible to provide flexibility within some limits. Today, coal technology might be preferred compare to gas to provide flexibility since it has cheaper operation costs than gas in Europe. But coal power plants have usually low flexibility and gas generation is often called to ensure, with hydro power, balance on electricity system. This need for flexibility affects the gas system and thus electricity actors could have to pay operators of gas network for the service.

(9)

Only a few studies were conducted about the interaction between gas and electricity systems, especially in Europe. Gas and electricity are intended to be more and more connected in the future and energy actors are trying to find a way to create synergies between the two systems [10].

1.2 Problem definition

The study considers a case with a large share of vRES in parallel with classical power generation. Variable generation induces larger errors in production forecast that causes balancing issues in the electricity system. For now, the electricity system is balanced using conventional generation where gas has a large contribution. This study focuses on balancing through gas generation and the effects cause intraday flexibility issues on gas system.

Flexibility sources in gas system are linepack, storage facilities, liquefied natural gas and some flexible contracts in the gas spot market. These different sources of flexibility have different technical characteristics, costs and available reserves. Depending on which source of flexibility is used, services of flexibility have different costs and these costs affect the bid prices of power plants on the electricity market.

This interaction between electricity and gas market needs to be further understood;

especially if the gas power plants in the near future will be the main source of flexibility due to a decreasing use of coal, in the event of a reinforcement of conditions for CO2 emission. The aim of this study is to further understand and explain the linkage between electricity and gas, especially in relation with flexibility services.

1.3 Objectives

This study will describe and model behaviors of gas and electricity systems, regarding flexibility issues, in a typical day with random errors in electricity generation and consumption forecast. The model operates the different sources of flexibility in order to minimize expected variable costs, including the costs of gas flexibility sources. Different scenarios are considered and combined: regular or tight situations on electricity market combined with regular or tight situations on gas market. Given the results of the model and a costs analysis is conducted, describing the share of flexibility in total costs and who are paying these.

1.4 Overview of the report

First of all, an overview of gas and electricity systems will be described and some existing model for gas-electricity interaction will be summarized. Then, a joint modeling of electricity and gas systems involving gas flexibility sources will be proposed. Finally,

(10)

results from different case studies will be analyzed. Appendix provides GAMS code of the model.

(11)

Chapter 2 Theoretical background

2.1 On the electric side

2.1.1 Day-ahead and intraday markets

In most European countries, retailers and producers can buy or sell electricity by send- ing production and demand bids to the market operator at different closing dates. The most significant market in terms of volume, excluding the financial market, is usually the day-ahead market. Deviations from their bids’ volume when the bids are accepted cause penalty fees to bidders. In order to avoid these fees, producers and retailers have to forecast their consumption and/or production to be close as possible to the reality of the following day. Then, the bids are gathered in the market, called spot market, and a price, day-ahead price, is derived from these bids.

An intraday market also usually exists but it is, for now, much less liquid, i.e. less production and consumption bids are offered, than the day-ahead market. It takes place usually a few hours before the real date, and it is a means to correct forecast errors of day-ahead market. Advantages of using this market are today still limited since programmable power plants represents substantially all generation means. Therefore it is not really liquid but in a close future with a lot more renewable intermittent generation, the intraday market should play a greater role in the electricity market design.

2.1.2 Balancing mechanism description

In real time, the transmission system operator (TSO) observes production and consumption and ensures the physical balance between them. To do that it has several ways to proceed. Two automatic mechanisms ensure power balance and frequency stability in short timescales from the seconds to the minutes, primary and secondary control. Then, in case of large forecast deviation in day-ahead and/or to replenish the primary and secondary reserve, tertiary reserve can be activated. The TSO receives balancing bids, to raise or to decrease power plants production, in order to balance the system. The variation of generation needed (upper or lower) is provided by flexible power plants. Flexible generation consists usually of conventional generation but

(12)

Table 2.1: Estimated growth of RES installed capacity in the member states of the European Union according to the NREAPs for 2010 and 2020 [3].

Installed capacity [GW] 2010 2020

Wind 85 214

Solar 26 91

Hydro and biomass 140 179

Total 215 484

technology differs depending on the countries. In France, where a large part of generation fleet is nuclear technology, nuclear power plants can offer flexibility services on balancing market [11]. Hydro and thermal generation such as gas, coal and oil can also provide flexibility. Finally, fast starting power plants can also be called to compensate a lack of energy, there are not called flexible generation but back up generation and they mainly consist of open cycle gas turbines (OCGT) which can start very quickly but have very high marginal generation costs.

2.1.3 Variable renewable generation

Variable renewable sources (vRES) are creating a need for flexibility due to their variable and hardly predictable nature. vRES are wind power, solar, wave energy, run-of- river hydro and tidal. While wave and tidal are still very marginal and do not have a big impact on the electricity system, wind and solar are growing very fast and is expected to grow even faster in the next decades.

National Renewable Energy Action Plans (NREAPs), submitted by each EU member state in 2010, set out the targets, the technology mix they expect to use, the trajectory they will follow and the measures and reforms they will undertake to develop renewable energy. According to 2010 data and targets by 2020 of NREAPs, it can be seen that wind installed capacity is expected to increase of 150 % in ten years, and 250 % for solar as shown in table 2.1. Then, vRES are expected to reach 60 % of renewable sources (RES) installed capacity in 2020. However wind power does not produce electricity at maximal capacity all the time, so that its capacity factor¹ is usually from 20 % to 40 % [16]. Then vRES are expected to produce only 50 % of total RES electricity generation as it can be seen in table 2.2.

The two particularities of wind and solar generation are high variability and difficulty to forecast power generation, as shows figure 2.1. Wind forecasts are usually less accurate than solar forecasts, as it can be seen in figure 2.2. As stated in [17], this can be explained by the following reasons:

• Solar plants are located in sunny areas which have less irradiation variability than

1It corresponds to the annual energy produced divided by annual energy produced if wind turbines was producing at maximal capacity during the year

(13)

Table 2.2: Estimated growth of RES electricity generation in the member states of the European Union according to the NREAPs for 2010 and 2020 [3].

Electricity generation (TWh) Member states NREAP 2010 Member states NREAP 2020

Wind 164 496

Solar 21 107

Non variable RE 461 604

Total 646 1207

Figure 2.1: Wind onshore generation over a year. Figure taken from [6].

(14)

Figure 2.2: Cumulated generation of wind and solar power in whole Germany during the last week of May 2011. Figure taken from [6].

(15)

Figure 2.3: Difference between real production and day-ahead forecasts at every hour at Horns Rev during September 2006. Figure taken from [6].

wind variability in windy area.

• Solar generation can be approximated as a linear function of solar radiance whereas wind generation is a cube function of wind speed. Thus a relative error in wind speed forecast would have a larger impact than the same relative error in solar radiance.

• Cloud coverage, which is the dominant factor of variability for solar generation, can be observed accurately with satellites whereas wind speed cannot be observed with the same accuracy.

To compare wind and solar forecasts, they evaluate the monthly mean absolute (MAE) error, in percentage of capacity, of wind and solar forecasts in USA. They find a MAE of 11.8 % for wind and only 6.9 % for solar (taking into consideration only daylight hours).

Firmness of vRES - capacity than can be relied upon most of the time - is then low, typical firmness according to the Spanish TSO is 5 to 10 %. That means that 5 to 10 % of total vRES installed capacity can be considered reliable for generation at any time [6].

Then, despite geographical smoothing effect, a major introduction of vRES will still result in a high variability in generation and large forecast errors. This will have two impacts which will challenge the balancing mechanism: (1) larger ramping rate and (2) larger deviation between forecasts and real generation which will need to be compensated by conventional power. The need for flexibility impacts all timescales from minute to weeks. Currently the main part of generation is scheduled in the day- ahead market but the deviation between day-ahead wind power forecast and actual generation is quite high. It can be seen on figure 2.3 that the hourly errors for a wind farm can sometimes reach more than 60 % the installed capacity (160 MW).

(16)

Table 2.3: Flexibility characteristics of conventional generators [6].

Nuclear Hard coal CCGT OCGT Pumped storage

Start up cold (h) 40 6 <2 0.4 0.1

Start up warm (h) - 3 >1.5 0.4 0.1

Load gradient

5 % 2 % 4 % 6 % 40 %

(% of installed capacity/min) Minimal load

50 % 40 % <50 % 33 % 15 %

(% of installed capacity/min)

2.1.4 How to provide flexibility?

To manage high variability, the electricity system will require solutions with high power ramping possibilities, which may be planned one or several days before, whereas to manage forecast errors the system will need solutions able to change their generation planning in a very short term manner.

From a technical point of view, there are different options which offer reserve or ramping possibilities. The most obvious one is to back up vRES with conventional flexible generation, such as hydro, gas, coal and sometimes nuclear power plants. This is the subject of interest in the following paragraphs. Other technical options can be implemented such as demand-side participation or storage, other than hydro, but these are long term solutions since these technologies are not yet profitable.

From a system point of view, interconnection can contribute to provide the flexibility needed by using flexibility sources on adjacent markets, for example countries with large hydro potential, through market coupling. This solution is expected to be implemented in whole Europe in the near future, but today it is still in the development phase. In addition, new market tools need to be implemented in order to ensure that the existing technical flexibility is used optimally, such as improvements of intraday and balancing markets.

Flexibility characteristics of conventional power plants are gathered in table 2.3.

Hydro (pumped storage) is the most responsive technology, almost instantaneous with a start-up time of 6 minutes; it has also the fastest load gradient: 40 % of installed capacity/min and a very low marginal cost. The only drawback is the minimal shut down time of 10 hours needed to refill the reservoir.

Flexible nuclear power plants have also a good load gradient considering the usual installed capacities close to 1 GW with 5 % of installed capacity in one minute. But they take a long time to start so they can only provide flexibility when started.

Coal is the least responsive in terms of warm start-up time and gradient. So they

(17)

will not be chosen for large ramping needs or large forecasts error. But they will be preferred to gas power plants when their specifications allow it since they have lower marginal cost than gas power plant, given the current coal, gas and CO2 prices.

OCGTs have a design consisting of a jet engine which rotates when the burning gas goes through it and drives a generator. Then by rotating the electrical generator, it produces electrical power. The efficiency of OCGTs is usually quite poor, between 35 and 42 %, since energy from resulting burning gas is not exploited. But OCGTs have a great flexibility potential [15]. Their start-up time until maximal power is low as well as minimal up and down time. That is why these power plants are used in period of extreme peaks which require a high level of flexibility or when all other power plants are already producing at maximal capacity. Moreover, these plants are quite easy to build and thus have the lowest fixed costs, but also generally the highest marginal costs.

CCGTs are a more recent technology. One part of a CCGT works just as an OCGT, but the difference is that the burning gas, after going through the turbine, is recovered to heat up water, turning it into steam and then a regular steam cycle operates to rotate another turbine and produce electricity. This technology can raise the overall efficiency up to almost 60 % which reduces fuel costs but increases fixed costs [15]. However, the more advanced turbine design requires increased fixed costs. Thus CCGT are usually used for load following operation thanks to a good load gradient, 4 %, and a fair start-up time, one and a half hour if the power plant is still warm, two hours if it is cold.

In conclusion, hydro will be the favorite source of flexibility, in terms of ramping and fast changes in generation planning, but unfortunately it cannot provide the whole flexibility need and European hydro resources are almost fully exploited in most countries. Then, nuclear in some countries can provide high ramping generation rate but has more difficulties to change their planning in a short-term manner. Among thermal technologies gas appears as the most flexible one, in term of ramping possibilities (CCGTs) and also in terms of back up (OCGTs) but it will often be used in the last resort since it has the highest fuel cost.

2.2 On gas side

As seen previously flexibility can be provided by different sources. But unlike other energy sources (e.g. coal), gas cannot be stored to great extent close to power stations at competitive costs. Tank storage close to power stations are very costly and dedicated for gas shortage periods to ensure electricity generation. Thus gas is usually continuously supplied from pipelines and short-term balancing actions on the electricity side can have impacts on the gas system.

2.2.1 Gas market

Gas power plants produced 21.1 % of European electricity in 2011 and they represented 14.5 % of total installed capacity [4]. In 2010, 60 % of European gas demand was imported. Only Netherlands and Norway export significant volume of natural gas

(18)

Figure 2.4: Inputs and Outputs on gas network.

to the rest of Europe, the rest is imported by pipelines, 90 % of imports, from Russia and Algeria or by ship as liquefied gas from all over the globe, 10 % of imports [8, pp. 247-253]. On land, local gas networks for transmission are operated by gas TSOs.

For each local network corresponds a price area and a TSO operates in one or several price areas. Its role is to ensure transport of gas within its network and the proper network operation, i.e. that the system is kept between physical limits. Local networks can be connected to other adjacent networks at interconnection points. Moreover non adjacent areas can also be connected by pipelines across the seas. This is the case of Nord-Stream and South-Stream projects, which will connect, respectively, Russia to Germany through Baltic sea and Russia to south Europe through Black sea. Then, the injection points into the network consists of interconnection points (or cross border pipelines) for imports, connections to production fields, LNG terminals and storage facilities (generally during winter). Whereas withdrawal points are connections to gas distribution networks, to some industrial customers, to storage facilities (generally during summer) and interconnection points for exports. Injections and withdrawals on gas network are summarized in figure 2.4.

TSOs are not responsible for gas exchange between different price areas. Then shippers, whose role is to convey gas from point A, generally the border of a gas producing country, to point B, generally the border of a consuming country, need to contract pipeline capacity to convey gas from a price area to another at, so called, interconnection points. Pipeline capacity can be bought for different time horizons, from several years to daily contracts. Capacity owners can also sells their unused ca- pacity to other buyers on a secondary market. In addition to pipeline capacity and for transit countries, a premium can be required for gas transportation through a country since shippers can use a local network without connecting any facilities to it and thus they are not paying for the use of the network. This is done for example in countries from East Europe where gas from Russia transits to the Western Europe. Gas market structure is summarized in figure 2.5.

(19)

Figure 2.5: Gas market structure.

Gas contracts

On the gas market, there are two types of gas trades: long-term contracts, from 10 to 25 years, and trade on spot markets. Long-term contracts represent the main part of gas trades, 80 to 90 %. Spot markets, except in England, still play a marginal role in Europe, even if they have developed in recent years thanks to lower prices than the long-term contracts [8] and the pressure of European Commission to reduce long term contracts for a market-based approach. This large share of long-term transactions is due historical reasons when market participants needed to build the transmission net- work. Those contracts have usually a take or pay clause, which means that a buyer contracts a certain amount of gas for a specified period to a producer and the buyer will pay this amount even if it does not use it entirely. However this amount is flexible, from 10 % for Algerian contracts to 30 % for dutch contracts. Gas price of long-term contracts are indexed on a portfolio of oil products [8, pp. 254]. Because the cost of transporting gas from is based on contracted transport capacity and not transited energy, shippers have incentives to deliver gas flatly during the day. That what we can observe in practice. Regarding spot markets, gas market operators trades two types of short-term products, daily and intraday. Unlike day-ahead electricity price on EPEX spot, gas prices are not derived by clearing a gas price from demand and supply bids for the considered period, but transactions of the same day can have different prices.

Then two price indicators can be the end-of-day price or the daily average price. A gas futures market also exists, it trades monthly, quarterly and seasonal products [13].

In most countries², time step for gas trading periods is a day. Then, users of transmission network, have only incentives to be balanced on a daily basis and not half-hourly as it is the case on the French electricity market. Thus, there may be large deviation between injection and withdrawal on gas network in real-time.

2Only Belgium and Netherlands have a hourly time step

(20)

Table 2.4: Different reserve types on gas system [6].

Gas storage Amout of reserve Withdrawal rate Gas injection Costs

Linepack Depends on

Maximal Yes No cost

network size Depleted

Very high Low

Yes Medium-High

hydrocarbon fields (seasonal storage)

Salt caverns Low Quick Yes Small

Aquifer reservoirs Medium Medium-Low Yes High

LNG peak-shaving Very low High No^b Medium

bor very expansive and limited rate

2.2.2 Storage

As seen previously gas contract are usually minimum daily contracts, and gas injections on the network are flat during the day whereas consumption is varying a lot from hour to hour. Then in order to continue to supply consumers, flexibility sources provide the difference between gas injections and withdrawals. One source of flexibility is gas from storage facilities. The different types of gas storage are gathered in table 2.4 with their characteristics.

• Linepack represents the amount of gas stored in pipelines which can be used while ensuring continuity of supply. In most cases, it has no variable cost³ and is instantaneous. That is why it represents the favorite flexibility source for TSO.

Linepack quantity depends obviously on pipelines and network sizes, however it depends also on external factors such as temperature and gas flow. As shown on figure 2.6, pipelines have a maximal operating pressure (MOP), a minimal pressure (Ps) required to ensure continuity of supply and a decrease of pressure along the pipeline (from P1 to Ps) which depends on the gas flow. As long as the pressure stands between MOP and Ps everywhere in the pipeline, the gas can be delivered normally. Then the amount of gas which corresponds to a decrease from MOP to P1, in yellow on the figure, can be used for other use such as to provide flexibility on gas network. Other storage can also be used for balancing when linepack is not enough. Although fixed costs depends a lot of storage type and size, variable costs of other storage facilities are almost identical.

• Depleted hydrocarbons fields are the most common form of storage. Fixed cost are high but they can stored a very large gas quantity. For all these reasons they are covering base load need.

• Salt caverns are artificial structures in underground salt domes. They can store considerably less natural gas but they have higher withdrawal and injection rates.

3gas TSOs can operate compressors in order to increase linepack quantity

(21)

Figure 2.6: Available linepack, figure translated from [7, pp. 15].

Figure 2.7: Example of utilization of gas storage facilities [14].

Therefore they provide gas usually during peak load.

• Aquifer reservoir characteristics is between the above two with medium storage capability, and a slightly higher withdrawal rate than depleted fields. However they have very high fixed cost compare to their storage size, that’s why depleted fields are more common.

• LNG peak-shaving is the smallest type of gas storage but has the fastest withdrawal rate. It provides also gas during peak periods [6].

An example of utilization of different types of gas storage facilities during a year is given on figure 2.7. It distinguishes storage facilities by their injection and withdraw rates, from slow to very fast, which corresponds to large storage and smaller facilities.

We can see that large storage facilities are used for seasonal storage (withdraw in winter

(22)

and injection in summer), whereas smaller storage facilities are used to meet residual demand, which is varying much faster.

Another difference between storage facilities is the time required to change the scheduled operation. Linepack is an internal source of flexibility which can provide flexibility instantaneously without TSO notification. Other need to be notified. Con- ventional storage facilities can usually change their injection or withdrawal during the day, LNG need to be noticed at least one day in advance and it cannot always provide flexibility services.

2.2.3 Gas intraday mechanisms

The gas TSO of a given area is responsible for the reliable operation of the network, i.e that the network stays inside its physical limits. On the other side the TSO has to respect its engagement to the shipper for gas transportation. In order to guarantee these two requirements, the TSO needs to be informed of the forecast injections and withdrawals in its area. The TSO requires network users to send their planning on daily basis. In most cases TSO requests information about the quantity injected and withdrawn within the day and injections and withdrawals locations.

Flexibility is used for two different purposes; to handle real-time deviation between injection and withdrawal as well as to handle daily imbalances of network’s users.

• Usually injection on gas network comes from shippers or producers and is flat during the day whereas gas consumption follows gas need of consumers. This creates a need for intraday modulation to compensate the difference between injections and withdrawals, from month to day-ahead. This need is forecasts from months to days before the actual day assuming that network users are daily balanced. Thus intraday modulation balance is zero at the end of the day.

• Daily imbalances due to forecasts errors stand usually on consumptions side. It can be due to forecast errors due to climatic reasons such as a cold spell but it also can be caused by perturbations on the electricity system. Gas power plants for electricity generation forecast their gas demand a day-ahead, but in case of large errors in electricity demand or generation, these can be needed to change their generation planning and thus their gas consumption which causes intraday imbalances on gas network.

Then to ensure the network’s physical balance the TSO can use different flexibility sources:

• The linepack is the favorite flexibility source since it is instantaneous and has no cost. It provides a large part of intraday modulation and a part of daily gas imbalances but linepack needs to be refilled quite fast to keep its role of providing intraday modulation.

• Access to conventional storage and LNG peak-shaving facilities. It provides intraday modulation when linepack is not sufficient and it can balance forecasts errors but it tends to disappear in favor of intraday market.

(23)

• Trade on intraday market, temporal and locational products. It tends to be the preferred gas source for imbalances management. For example in France, TSO trades gas on within day market to balance injections and withdrawals at the end of the day. French TSO also trades on day-ahead market by forecasting imbalances but tends to disappears.

• Flexibility can also be provided by adjacent network’s operator. For example in France, the TSO GRTgaz contracted a flexibility service to the south-west TSO TIGF [7].

2.3 Studies and regulation

Discussions at European level have been conducted in order to harmonize operation of gas and electricity markets and to be able to exploit synergies between both markets.

2.3.1 ENTSOG proposals

The association of gas European TSO, ENTSOG, has come up with a network code proposal supposed to exploit better those synergies in a European level. They suggest to develop new rules and to enhance existing ones. First, they suggest to enhance TSOs operational balancing by creating a balancing market, on which the TSO will provide to balance its network, with different products and the following merit order:

• Locational products: it corresponds to a balancing action (change of gas flow) at a precise entry or exit point to start from a specific period of time. It can be used, for example, to release locational congestion.

• Temporal products: It corresponds to a balancing action when change of gas flow are needed within a specific period of time within the gas day.

• Balancing services: it can be, for example, contracts with a storage manager for providing flexibility to the TSO when needed, it tends to disappear in the near future because this method is not the most economically efficient.

They also suggest developing cross border cooperation by enhancing the market coupling between neighbor areas. To do that they suggest the creation of a joint balancing platform and an exchange of information about shipper cross-border portfolios.

It has already been undertaken between the two french TSOs, GRTgaz and TIGF, as mentioned previously. TIGF offers flexibility services to GRTgaz which gathers all gas power plants in its area and thus have large flexibility needs.

Another measure is to create incentives for shipper to be balanced on daily basis and also within the day. First, it extends to all EU countries the daily imbalance charges system. It consists of charging shippers fees in case of daily imbalances. Then, it suggests within day obligations such as:

• Shippers send their hourly profile a day-ahead or several days ahead.

(24)

• Create hourly nominations as on electricity market and then eliminate the risk of supply for peak hours.

• Charge the intraday modulation services such as access to linepack or storage facilities.

In France those within day obligations already apply as a combination of the first and last suggestions of the network code. The TSO requires special highly modulated site, i.e. for site with more than 0.8 GWh of modulated volume⁴, to send their consumption planning and their maximal and minimal technical flow for each hour of the next three days. Then the TSO inform users about the feasibility of their planning and in case of non-feasible planning, the TSO suggests several alternative planning. French TSOs charge a premium for the flexibility service depending on modulated volume and flow amplitude⁵.

A final idea focuses on the linepack management. The TSOs should offer linepack flexibility services, limited to available linepack and should charge the shippers for linepack utilization. Moreover the shippers should notify in advance the TSO of the use of linepack. That way TSOs could provide an accurate use of the linepack.

2.3.2 KEMA-COWI report

Another important study on the subject matter is KEMA-COWI⁶ report sponsored by the European commission [10]. It studies the synergies between gas and electricity systems in the context of high penetration of intermittent generation such as wind or solar. And from their studies they derive some measures that can exploit those synergies.

• One measure (Measure 7) suggests the development of trades of flexibility products on intraday markets. This measure is close to ENTSOG recommendations and to what French TSO offers through their "highly modulated site" requirements.

• Two other measures (Measure 8 and 9) suggest a better coordinated planning and operation of both network and smart investment on the gas network to optimize gas power plants supply. It suggests in particular optimizing network development location, supply capacity, flexibility capacity given the location and characteristics of gas power plants in the area. It requires more exchange of information between gas and electricity TSOs about forecast demand and supply on both systems, maintenance schedule, intermittent generation forecast, and other local constraints.

• Another measure (Measure 10) suggests a better linepack management, close to ENTSOG recommendations.

4It corresponds to the sum in absolute values of the difference between the hourly consumption and the daily average consumption

5It corresponds to the difference between maximal and minimal gas flows during a day

6KEMA and COWI are two consulting companies

(25)

2.4 Review of gas electricity interaction models

In this section gas-electricity interactions models from the literature are described in order to understand better the linkage between those two systems.

2.4.1 Combining Energy Networks (CEN)

Abrel et al. [1] presents the results of a gas-electricity model and applies the scenarios for European decarbonization in 2030 and 2050. It gives an analysis on security of supply, impact on energy prices and need of infrastructure.

To justify the need of investigating both systems interaction, this study relies on an event that happened February 9th, 2012. At this time a cold wave struck France rising up electricity demand to more than 100GW. Normally imports from Germany supply extra power needed in these cases. Because of the German moratorium on nuclear power after the Fukushima accident, nuclear power plants that generally export electricity to France had been shut down and gas fired plants were called to produce in their stead. But due to a lack of access to gas pipeline capacity in Germany, power exports were limited to France and France had to solve power balance constraint through active demand.

The model considers both systems connected through gas fired power plants. As it can be seen in figure 2.8, the participants on gas markets are producers, traders (shippers), LNG operators, and a pipeline operator (usually TSO’s role) which sells transport capacity. All these participants respond to a natural gas demand market which consists of an exogenous part and an endogenous part due to fuel for electricity generation. The only participant on electricity side is system operator which has to meet electricity demand. Electricity demand is exogenous.

The study perimeter is the 27 countries of the European Union and the model was simulated for data from 2010 to 2050 with a ten years step. The model consists in a nodal representation with 1 to 4 nodes per countries. The data for year 2010 are those given by ENTSO-G and IEA for gas and ENTSO-E for electricity. For future years, data from the Ten Year Network Development Plan for gas and electricity networks are used as well as the Energy Roadmap 2050 from the European commission.

They simulate different scenarios with different C02 emissions: 40 % CO₂ emission reduction in 2050, 80 % and higher share in RES generation and stronger energy efficiency development with a constrained usage of carbon capture and storage (CCS).

And they analyze gas and electricity prices in peak periods, with high wind or high solar penetration and the impact of gas market issues on electricity.

The study concludes that the future network expansion will decrease congestion and that European Union will remain dependent on gas imports. They also conclude that the increase in renewable generation will lower the need of gas fired plants, which is not obvious since flexibility characteristics of gas fired plants can be a great advantage

(26)

Figure 2.8: Model 1 overview

(27)

to support renewable generation. The article do not address short-term issues, such as balancing and flexibility costs on both systems. Instead they focus more on long term perspectives and supply issues.

2.4.2 Modeling Gas-Electricity Coordination in a Competi- tive Market (MGEC)

Duenas et al. [5] studies the behavior of a generation company which owns a set of natural gas power plants. The study takes a U.S perspective where gas power plants are attractive because of their low marginal prices, due to shale gas exploitation, low fixed costs, reduced CO₂ emissions compare to other fossil fuel and power plants flexibility.

The model consists in optimizing gas purchases, pipeline capacity contracting and power plants operation. It considers different wind and solar generation scenarios. The capacity contracts have different time periods (hour, week and month). Electricity demand and variable renewable generation are modeled with a net load duration curve derived from the scenarios. Gas price is an affine function of gas volume and the problem uses unit commitment to take into consideration start up and shut down costs from gas power plants.

The article does not mention data used except that those "are inspired by real sys- tems, but not by a specific system". Model outputs are gas and electricity prices, net electricity demand, contracting capacities, and generation by fuels.

The study focuses on security supply and contracting capacity. It shows that the model contracts capacity above the expected gas flow as a consequence of the uncertainty from the wind power generation with 25 % margin. Moreover, because of vRES uncertainty, 10 % of pipeline capacity is traded daily when vRES generation are more accurate. By suppressing the secondary market the margin decreases, which lower the security margin and increases total costs.

This study, unlike the previous one, take into consideration the flexibility characteristics of gas power plants by adding a higher start up and shut down costs to non-flexible plants. But power ramping and minimum up and down times characteristics are still not considered. Moreover hydro power is not included into the model even if it plays a significant role in flexibility supply. Finally, excepted LNG, fast and seasonal gas storage than LNG are neglected in gas modeling which is questionable especially given that the time scope is one year.

2.4.3 Multi-time period combined gas electric network opti- mization (GENO)

Chaudry, Jenkins and Strbac [2] refers to a cost optimization model of a system combining gas and electricity network. It takes into account system’s physical behaviors

(28)

such as the varying nature of gas flow, network support (gas storage and power ramping) and the optimization is performed from an economic point of view, minimizing the total costs.

On gas side, gas flow is modeled with fluid mechanics equations for gas flow.

Linepack is modeled using Boyle’s law⁷. Gas storages are modeled by setting a max- imum and a minimum amount of gas that can be stored and maximal withdraw and injection rates. And finally gas balance constraints force pressure in pipelines to stay between physical limits.

Electric power system have also its own constraints: power balance constraint, power plants generation physical limitations (up and down limits), limitation on power flowing through power lines and generation ramp up and down limitation

The two systems are bound by the relationship between the gas fuel flow and the electrical power generated. The objective function is the sum of costs for purchasing gas, injection into and withdraw from storage cost, extra cost for electrical power generation and cost of load shedding.

Model inputs are the gas demand and price, storage characteristics, electrical demand and coal price which is constant. Outputs are gas storage reserve, flow, pressure in pipeline, linepack reserve and generator power output. Study perimeter is Great Britain and the analyzed time period is one day.

The article analyzes to case studies: a simple combined gas electricity network as shown in figure 2.9 and Great Britain’s network. The simple network analysis concludes that for large storage gas is stored during low price periods and linepack is rather used to alleviate fluctuations. Given that coal is cheaper than gas, CCGs shut down when the gas price is high and when it is possible to compensate with coal. It also shows that a storage outage can cause load shedding, resulting in a much higher cost for the system. In this case gas-fired generation is limited due to pressure constraints in pipelines and cannot meet the demand. In simulations of the British gas network, the loss of storage facilities impacts the share of gas in electricity generation (from 13 % to only 2 %) and higher the costs as in the simple network analysis.

However the model does not take into consideration other flexibility services on gas system such as LNG or balancing market products. Moreover it does not differentiate storage types (salt mines, aquifers, depleted fields). In addition the article gives no information on how other sources than gas generation are managed in the model when all GB is modeled.

7Boyle’s law describes the linkage between gas volume and pressure for a given temperature, PV=constant

(29)

Figure 2.9: Simple combined gas electricity network

2.4.4 Models comparison

Analysis of previous studies gives a good understanding of major aspects of the gas system and its linkage with electricity. Article [1] gives a clear overview of geopolitics of European energy by highlighting each producer weight and location, the structure of European gas and electricity markets and the European energy politics. Moreover it gives an interesting analysis about the linkage between gas and electricity prices, especially in case of supply shortage. While, article [5] gives a clear explanation about operation of the transmission capacity market and analyzes their impacts on security of supply, in particular for secondary market. Finally, article [2] describes accurately the physical behavior of gas and electricity systems for linepack, compressor and gas storage facilities.

As shown in table 2.5, the three models studied are quite different, especially on subject focus, study perimeters and time scale. On the other hand, all articles take into consideration renewable sources impacts on gas system. However no model discuss really gas flexibility issues and advantages in the very short-term. That’s what the following model aims to do.

(30)

Table 2.5: Comparison of articles models.

Model Simulation

Perimeters Focus on Conclusions Uncertainty time scale

CEN Long-term Europe Energy prices Network and supply investments No MGEC Long and

Virtual Capacity market Secondary

Short-term market value Yes

GENO Short-term GB Gas storage

Storage value No impacts

(31)

Chapter 3 Model

The model described in this part distinguishes itself by focusing more on short-term analysis. It aims to provide a better understanding of operation of gas flexibility sources and the linkage to electricity system, especially in a context with large vRES penetration. Another objective is to give a rough estimation of flexibility costs and find out who are paying these costs.

3.1 Specifications

The model aims to simulate operation of gas flexibility sources in a costs minimization way for one day. Gas flexibility sources are operated in order to change electricity generation to meet net electricity demand¹. This need in caused by large variation on electricity system and random errors in generation forecasts of the previous day. Thus the model needs to be stochastic to take into consideration the random forecasts errors.

Unlike the models described previously, this model should operate different sources of gas flexibility (Storage, market, linepack) to emphasize their different technical constraints and costs. The model should have several decision periods, at least two (day ahead and intraday), with decreasing forecasts errors the closer it gets to real-time.

The model simulates the global system, consisting of the electricity and gas systems. It neglects transmission networks issues in order to simplify the study.

3.2 Assumptions

The model makes different assumptions in order to simplify results analysis and focus on the most significant parameters. The hypothesis are :

• System point of view only. The model does not take into consideration possible strategies of market participants.

1It corresponds to the electricity demand minus vRES generation

(32)

Figure 3.1: Model overview

• No simulation of gas and electricity transmission networks and then, no simulation of gas capacity market. This is done to simplify the model. Transmission networks can have impacts on flexibility supply in very specific cases (congestions, maximal or minimal power flow) that are not studied by the model.

• Only gas technologies to meet the net electricity demand. Other technologies such as hydro, coal or nuclear are not simulated. This is a big assumptions since the other technologies can also provide flexibility. Only the model does not aim to compare the different technologies but it aims to study impacts on gas system in a case where gas provides a part of the flexibility for the electricity system.

• Two decision times are considered: day ahead when gas nomination for the next day is decided and intraday when use of flexibility sources on the gas system and electricity generations by power plants for every hour are decided.

• Gas price for long-term contracts are supposed to be similar to the average price on day-ahead spot market. Moreover shippers are considered to use gas in order to stay inside the range of gas purchases allowed by producers (see "take or pay"

clause in 2.2.1) so that any day-ahead nominations of gas can be made regardless the amount of gas purchased the previous days.

3.3 Overview

As shown by figure 3.1, the model takes as inputs different scenarios of net electricity demand with different probability and different gas prices on intraday market. The model has two decision times, day-ahead and intraday. Day-ahead, the system chooses

(33)

a level for day-ahead gas nominations which applies to all scenarios. Intraday, the model considers all parameters known. Then, the electricity system chooses the most efficient planning program given the electricity generators considered, their technical constraints, the day-ahead gas nominations and the availability of gas sources for flexibility. The flexibility are then called in a cost efficient way to provide the intraday modulation needed, i.e. gas consumption minus day-ahead nominations for each hour, and daily imbalances due to forecasts errors on electricity side. The flexibility sources on the gas system are linepack, storage facilities, access to intraday gas market. Those sources are operated in accordance with their technical and economical constraints described in the following section. If gas sources are not enough to provide the flexibility, then the electricity system can do load shedding. To prevent the system to do it, a very high cost is associated to load shedding.

The sets of the model are then:

• h : hour of the day h = h₁..h₂₄

• i : power plant indice i = pp₁..pp₂₈

• ω : scenario ω = ω₁..ω₃ The model’s control variables are:

• g_i,h,ω: generation of power plant i (including a virtual load shedding power plant), hour h and scenario ω.

• u_i,h,ω: unit commitment for hour h, power plant i, scenario ω. u_i,h,ω is a binary variable equal to 1 if power plant i is turned on, 0 otherwise.

• n₀ day-ahead nominations for gas, flat.

• λ_h,ω linepack contribution at hour h and for scenario ω. λ_h,ω ≤ 0 for injection and ≥ 0 for withdrawal on linepack.

• q_ω quantity traded on intraday gas market at hour h and for scenario ω. q_ω ≤ 0 for purchasing and ≥ 0 for selling.

• σ_h,ω access to storage at hour h and for scenario ω. σ_h,ω ≤ 0 for injection and

≥ 0 for withdrawal.

3.4 Objective function

The model proceeds an optimization by minimizing expected total costs represented by the following objective function:

min C^DA+

ωn

X

ω=ω1

p_ω(C_ω^m+

24

X

h=1

C_h,ω^su + C_h,ω^addsto− V_h,ω^sto+ C_h,ω^LoL) (3.1)

(34)

With

• C_h,ω^LoL: costs for load shedding, i.e. costs for disconnecting electrical loads.

C_h,ω^LoL = g_LoL,h,ωVoLL (3.2)

where g_LoL,h,ω is the amount of demand which is not served and VoLL is the value of lost load.

• C_h^DA: costs of day-ahead nomination for gas.

C_h^DA = p^DAn0 (3.3)

where p^DA day-ahead price for gas.

• C_ω^m: costs or revenues for trading on intraday gas market.

C_ω^m = p^intra_ω q_ω (3.4)

where p^intra_ω intraday gas price.

• V_h,ω^sto: value of stored gas.

V_h,ω^sto= p^DAσ_h,ω (3.5)

• C_h,ω^addsto: additionnal costs for injecting or withdrawing gas from storage facilities.

C_h,ωâddsto = Câddinjσ_h,ωînj + Câddwitσ^wit_h,ω (3.6) where Câddinj and Câddwit additionnal costs per MWh for, respectively, injection and withdrawal. And σ_h,ωînj and σ^wit_h,ω, respectively, injected and withdrawn quantity on storage facilities.

• Start-up costs C_h,ω^su.

C_h,ω^su =

N

X

i=1

s⁺_i,h,ωC_i⁺ (3.7)

where C_i⁺ costs for starting up power plant i.

(35)

3.5 Constraints

The model is subject to several constraints both on the gas system and the electricity system. On the electricity side, the constraints regarding power plants operation are the following:

• Maximal (P_i^max) and minimal (P_i^min) power generation for power plant i, hour h and scenario ω.

g_i,h,ω ≤ P_i^maxu_i,h,ω, h = h₁..h₂₄, i = pp₁..pp₂₇, ω = ω₁..ω₃ (3.8)

g_i,h,ω ≥ P_i^minu_i,h,ω, h = h₁..h₂₄, i = pp₁..pp₂₇, ω = ω₁..ω₃ (3.9)

• Commitment, start-up and shut-down relationship.

u_i,h,ω− u_i,h−1,ω = s⁺_i,h,ω− s⁻_i,h,ω, h = h₁..h₂₄, i = pp₁..pp₂₇, ω = ω₁..ω₃ (3.10) with s⁻_i,h,ω and s⁺_i,h,ω shut-down variable and start-up variable of power plant i, hour h, scenario ω; t⁻_i minimal down time of power plant i. s⁻_i,h,ω and s⁺_i,h,ω are

binary variables. s⁻_i,h,ω is equal to 1 if power plant i has just switched off, 0 otherwise. s⁺_i,h,ω is equal to 1 if power plant i has just turned on, 0 otherwise.

• Start-up time for power plant i (only CCGT), hour h and scenario ω.

s⁻_i,h,ω+

h+t⁻_i

X

t=h

s⁺_i,t,ω ≤ 1, h = h₁..h₂₄, i = pp₁..pp₇, ω = ω₁..ω₃ (3.11)

• Maximal and minimal power ramping for power plant i (only CCGT), hour h and scenario ω.

g_i,h,ω− g_i,h−1,ω≤ R^up_i + s⁺_i,h,ω(P_i^min− R^up_i ), h = h₁..h₂₄, i = pp₁..pp₇, ω = ω₁..ω₃ (3.12)

g_i,h,ω−g_i,h−1,ω ≥ −R^dn_i + s⁻_i,h,ω(R^dn_i −P_i^min), h = h₁..h₂₄, i = pp₁..pp₇, ω = ω₁..ω₃ (3.13)

(36)

with R^up_i ≥ 0 and R^dn_i ≥ 0, respectively, maximal ramping up and ramping down for power plant i and P_i^min minimal power of power plant i if turned on.

These constraints force power plant i to keep its generation to P_i^min when started and to decrease its generation to P_i^min before stopping. When power plant i is already committed for more than one hour, ramping values are limiting generation change.

• Balance between electricity demand and generation for hour h, scenario ω.

N

X

i=1

g_i,h,ω = D^elec_h,ω, h = h₁..h₂₄, ω = ω₁..ω₃ (3.14)

with D^elec_h,ω electricity demand for hour h, scenario ω and N the number of power plants.

The constraints on gas system are :

• Gas demand which has to be equal to intraday gas flow n_h,ω.

n_h,ω =

N

X

i=1

g_i,h,ωα_i+ D^gas_h , h = h₁..h₂₄, ω = ω₁..ω₃ (3.15)

with D_h^gas conventional gas demand² for hour h and α_i reverse efficiency of power plant i.

• Flexibility supply by linepack, storage and intraday gas market.

nh,ω− n0 = λ_h,ω+ q_ω/24 + σh,ω, h = h1..h24, ω = ω1..ω3 (3.16) Intraday market supply is flat during the day such as day-ahead nominations.

• Maximal injection and withdrawal rates of storage facilities for hour h and sce- nario ω.

2It corresponds to the industrial (excluding gas power plants) plus the domestic demand

(37)

σh,ω ≥ D_injection^max , h = h₁..h₂₄, ω = ω₁..ω₃ (3.17)

σ_h,ω ≤ D^max_withdrawal, h = h₁..h₂₄, ω = ω₁..ω₃ (3.18) with D^max_withdrawal ≥ 0 and D^max_injection ≤ 0 maximal injection and withdrawal rates of storage facilities.

• Limited trades on intraday gas market for each scenario.

−V_intraday ≥ q_ω ≥ V_intraday, ω = ω₁..ω₃ (3.19) with V_intraday maximal volume that can be traded on intraday gas market.

• Maximal and minimal linepack level for hour h and scenario ω.

Lph,ω ≤ Lp^max, h = h1..h24, ω = ω1..ω3 (3.20)

Lp_h,ω ≥ Lp^min, h = h₁..h₂₄, ω = ω₁..ω₃ (3.21) with Lp_h,ω linepack level at hour h, scenario ω; Lp^max and Lp^min maximal and minimal linepack levels.

• Daily linepack contribution is zero, so that linepack are used only for providing intraday modulation.

Lp_24,ω = Lp₀, ω = ω₁..ω₃ (3.22) with Lp₀ initial linepack value.

Gas Balancing Variable PowerGeneration

Gas Balancing Variable Power Generation

A SYSTEMIC CASE STUDY

Gas balancing variable power generation

A systemic case study

BAPTISTE BORTOT

2013-2014

Thesis conducted at EDF R&D, France

Abstract

Acknowledgements

Contents

Nomenclature

Chapter 1 Introduction

1.1 Background

1.2 Problem definition

1.3 Objectives

1.4 Overview of the report

Chapter 2

Theoretical background

2.1 On the electric side

2.1.1 Day-ahead and intraday markets

2.1.2 Balancing mechanism description

2.1.3 Variable renewable generation

2.1.4 How to provide flexibility?

2.2 On gas side

2.2.1 Gas market

2.2.2 Storage

2.2.3 Gas intraday mechanisms

2.3 Studies and regulation

2.3.1 ENTSOG proposals

2.3.2 KEMA-COWI report

2.4 Review of gas electricity interaction models

2.4.1 Combining Energy Networks (CEN)

2.4.2 Modeling Gas-Electricity Coordination in a Competi- tive Market (MGEC)

2.4.3 Multi-time period combined gas electric network opti- mization (GENO)

2.4.4 Models comparison

Chapter 3 Model

3.1 Specifications

3.2 Assumptions

3.3 Overview

3.4 Objective function

3.5 Constraints